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 * IPv6|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 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 set in 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 or 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 in skb->csum. This means 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, but it 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 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, it should treat the packet as if CHECKSUM_NONE were 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 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 that there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, so 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, and both csum_start and
215 * csum_offset are set to refer to the outermost checksum being offloaded
216 * (two offloaded 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 ahash_request;
242 struct net_device;
243 struct scatterlist;
244 struct pipe_inode_info;
245 struct iov_iter;
246 struct napi_struct;
247 struct bpf_prog;
248 union bpf_attr;
249 struct skb_ext;
250
251 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
252 struct nf_bridge_info {
253 enum {
254 BRNF_PROTO_UNCHANGED,
255 BRNF_PROTO_8021Q,
256 BRNF_PROTO_PPPOE
257 } orig_proto:8;
258 u8 pkt_otherhost:1;
259 u8 in_prerouting:1;
260 u8 bridged_dnat:1;
261 __u16 frag_max_size;
262 struct net_device *physindev;
263
264 /* always valid & non-NULL from FORWARD on, for physdev match */
265 struct net_device *physoutdev;
266 union {
267 /* prerouting: detect dnat in orig/reply direction */
268 __be32 ipv4_daddr;
269 struct in6_addr ipv6_daddr;
270
271 /* after prerouting + nat detected: store original source
272 * mac since neigh resolution overwrites it, only used while
273 * skb is out in neigh layer.
274 */
275 char neigh_header[8];
276 };
277 };
278 #endif
279
280 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
281 /* Chain in tc_skb_ext will be used to share the tc chain with
282 * ovs recirc_id. It will be set to the current chain by tc
283 * and read by ovs to recirc_id.
284 */
285 struct tc_skb_ext {
286 __u32 chain;
287 __u16 mru;
288 };
289 #endif
290
291 struct sk_buff_head {
292 /* These two members must be first. */
293 struct sk_buff *next;
294 struct sk_buff *prev;
295
296 __u32 qlen;
297 spinlock_t lock;
298 };
299
300 struct sk_buff;
301
302 /* To allow 64K frame to be packed as single skb without frag_list we
303 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
304 * buffers which do not start on a page boundary.
305 *
306 * Since GRO uses frags we allocate at least 16 regardless of page
307 * size.
308 */
309 #if (65536/PAGE_SIZE + 1) < 16
310 #define MAX_SKB_FRAGS 16UL
311 #else
312 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
313 #endif
314 extern int sysctl_max_skb_frags;
315
316 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
317 * segment using its current segmentation instead.
318 */
319 #define GSO_BY_FRAGS 0xFFFF
320
321 typedef struct bio_vec skb_frag_t;
322
323 /**
324 * skb_frag_size() - Returns the size of a skb fragment
325 * @frag: skb fragment
326 */
skb_frag_size(const skb_frag_t * frag)327 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
328 {
329 return frag->bv_len;
330 }
331
332 /**
333 * skb_frag_size_set() - Sets the size of a skb fragment
334 * @frag: skb fragment
335 * @size: size of fragment
336 */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)337 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
338 {
339 frag->bv_len = size;
340 }
341
342 /**
343 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
344 * @frag: skb fragment
345 * @delta: value to add
346 */
skb_frag_size_add(skb_frag_t * frag,int delta)347 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
348 {
349 frag->bv_len += delta;
350 }
351
352 /**
353 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
354 * @frag: skb fragment
355 * @delta: value to subtract
356 */
skb_frag_size_sub(skb_frag_t * frag,int delta)357 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
358 {
359 frag->bv_len -= delta;
360 }
361
362 /**
363 * skb_frag_must_loop - Test if %p is a high memory page
364 * @p: fragment's page
365 */
skb_frag_must_loop(struct page * p)366 static inline bool skb_frag_must_loop(struct page *p)
367 {
368 #if defined(CONFIG_HIGHMEM)
369 if (PageHighMem(p))
370 return true;
371 #endif
372 return false;
373 }
374
375 /**
376 * skb_frag_foreach_page - loop over pages in a fragment
377 *
378 * @f: skb frag to operate on
379 * @f_off: offset from start of f->bv_page
380 * @f_len: length from f_off to loop over
381 * @p: (temp var) current page
382 * @p_off: (temp var) offset from start of current page,
383 * non-zero only on first page.
384 * @p_len: (temp var) length in current page,
385 * < PAGE_SIZE only on first and last page.
386 * @copied: (temp var) length so far, excluding current p_len.
387 *
388 * A fragment can hold a compound page, in which case per-page
389 * operations, notably kmap_atomic, must be called for each
390 * regular page.
391 */
392 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
393 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
394 p_off = (f_off) & (PAGE_SIZE - 1), \
395 p_len = skb_frag_must_loop(p) ? \
396 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
397 copied = 0; \
398 copied < f_len; \
399 copied += p_len, p++, p_off = 0, \
400 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
401
402 #define HAVE_HW_TIME_STAMP
403
404 /**
405 * struct skb_shared_hwtstamps - hardware time stamps
406 * @hwtstamp: hardware time stamp transformed into duration
407 * since arbitrary point in time
408 *
409 * Software time stamps generated by ktime_get_real() are stored in
410 * skb->tstamp.
411 *
412 * hwtstamps can only be compared against other hwtstamps from
413 * the same device.
414 *
415 * This structure is attached to packets as part of the
416 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
417 */
418 struct skb_shared_hwtstamps {
419 ktime_t hwtstamp;
420 };
421
422 /* Definitions for tx_flags in struct skb_shared_info */
423 enum {
424 /* generate hardware time stamp */
425 SKBTX_HW_TSTAMP = 1 << 0,
426
427 /* generate software time stamp when queueing packet to NIC */
428 SKBTX_SW_TSTAMP = 1 << 1,
429
430 /* device driver is going to provide hardware time stamp */
431 SKBTX_IN_PROGRESS = 1 << 2,
432
433 /* device driver supports TX zero-copy buffers */
434 SKBTX_DEV_ZEROCOPY = 1 << 3,
435
436 /* generate wifi status information (where possible) */
437 SKBTX_WIFI_STATUS = 1 << 4,
438
439 /* This indicates at least one fragment might be overwritten
440 * (as in vmsplice(), sendfile() ...)
441 * If we need to compute a TX checksum, we'll need to copy
442 * all frags to avoid possible bad checksum
443 */
444 SKBTX_SHARED_FRAG = 1 << 5,
445
446 /* generate software time stamp when entering packet scheduling */
447 SKBTX_SCHED_TSTAMP = 1 << 6,
448 };
449
450 #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
451 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
452 SKBTX_SCHED_TSTAMP)
453 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
454
455 /*
456 * The callback notifies userspace to release buffers when skb DMA is done in
457 * lower device, the skb last reference should be 0 when calling this.
458 * The zerocopy_success argument is true if zero copy transmit occurred,
459 * false on data copy or out of memory error caused by data copy attempt.
460 * The ctx field is used to track device context.
461 * The desc field is used to track userspace buffer index.
462 */
463 struct ubuf_info {
464 void (*callback)(struct ubuf_info *, bool zerocopy_success);
465 union {
466 struct {
467 unsigned long desc;
468 void *ctx;
469 };
470 struct {
471 u32 id;
472 u16 len;
473 u16 zerocopy:1;
474 u32 bytelen;
475 };
476 };
477 refcount_t refcnt;
478
479 struct mmpin {
480 struct user_struct *user;
481 unsigned int num_pg;
482 } mmp;
483 };
484
485 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
486
487 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
488 void mm_unaccount_pinned_pages(struct mmpin *mmp);
489
490 struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
491 struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
492 struct ubuf_info *uarg);
493
sock_zerocopy_get(struct ubuf_info * uarg)494 static inline void sock_zerocopy_get(struct ubuf_info *uarg)
495 {
496 refcount_inc(&uarg->refcnt);
497 }
498
499 void sock_zerocopy_put(struct ubuf_info *uarg);
500 void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
501
502 void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
503
504 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
505 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
506 struct msghdr *msg, int len,
507 struct ubuf_info *uarg);
508
509 /* This data is invariant across clones and lives at
510 * the end of the header data, ie. at skb->end.
511 */
512 struct skb_shared_info {
513 __u8 __unused;
514 __u8 meta_len;
515 __u8 nr_frags;
516 __u8 tx_flags;
517 unsigned short gso_size;
518 /* Warning: this field is not always filled in (UFO)! */
519 unsigned short gso_segs;
520 struct sk_buff *frag_list;
521 struct skb_shared_hwtstamps hwtstamps;
522 unsigned int gso_type;
523 u32 tskey;
524
525 /*
526 * Warning : all fields before dataref are cleared in __alloc_skb()
527 */
528 atomic_t dataref;
529
530 /* Intermediate layers must ensure that destructor_arg
531 * remains valid until skb destructor */
532 void * destructor_arg;
533
534 /* must be last field, see pskb_expand_head() */
535 skb_frag_t frags[MAX_SKB_FRAGS];
536 };
537
538 /* We divide dataref into two halves. The higher 16 bits hold references
539 * to the payload part of skb->data. The lower 16 bits hold references to
540 * the entire skb->data. A clone of a headerless skb holds the length of
541 * the header in skb->hdr_len.
542 *
543 * All users must obey the rule that the skb->data reference count must be
544 * greater than or equal to the payload reference count.
545 *
546 * Holding a reference to the payload part means that the user does not
547 * care about modifications to the header part of skb->data.
548 */
549 #define SKB_DATAREF_SHIFT 16
550 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
551
552
553 enum {
554 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
555 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
556 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
557 };
558
559 enum {
560 SKB_GSO_TCPV4 = 1 << 0,
561
562 /* This indicates the skb is from an untrusted source. */
563 SKB_GSO_DODGY = 1 << 1,
564
565 /* This indicates the tcp segment has CWR set. */
566 SKB_GSO_TCP_ECN = 1 << 2,
567
568 SKB_GSO_TCP_FIXEDID = 1 << 3,
569
570 SKB_GSO_TCPV6 = 1 << 4,
571
572 SKB_GSO_FCOE = 1 << 5,
573
574 SKB_GSO_GRE = 1 << 6,
575
576 SKB_GSO_GRE_CSUM = 1 << 7,
577
578 SKB_GSO_IPXIP4 = 1 << 8,
579
580 SKB_GSO_IPXIP6 = 1 << 9,
581
582 SKB_GSO_UDP_TUNNEL = 1 << 10,
583
584 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
585
586 SKB_GSO_PARTIAL = 1 << 12,
587
588 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
589
590 SKB_GSO_SCTP = 1 << 14,
591
592 SKB_GSO_ESP = 1 << 15,
593
594 SKB_GSO_UDP = 1 << 16,
595
596 SKB_GSO_UDP_L4 = 1 << 17,
597
598 SKB_GSO_FRAGLIST = 1 << 18,
599 };
600
601 #if BITS_PER_LONG > 32
602 #define NET_SKBUFF_DATA_USES_OFFSET 1
603 #endif
604
605 #ifdef NET_SKBUFF_DATA_USES_OFFSET
606 typedef unsigned int sk_buff_data_t;
607 #else
608 typedef unsigned char *sk_buff_data_t;
609 #endif
610
611 /**
612 * struct sk_buff - socket buffer
613 * @next: Next buffer in list
614 * @prev: Previous buffer in list
615 * @tstamp: Time we arrived/left
616 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
617 * for retransmit timer
618 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
619 * @list: queue head
620 * @sk: Socket we are owned by
621 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
622 * fragmentation management
623 * @dev: Device we arrived on/are leaving by
624 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
625 * @cb: Control buffer. Free for use by every layer. Put private vars here
626 * @_skb_refdst: destination entry (with norefcount bit)
627 * @sp: the security path, used for xfrm
628 * @len: Length of actual data
629 * @data_len: Data length
630 * @mac_len: Length of link layer header
631 * @hdr_len: writable header length of cloned skb
632 * @csum: Checksum (must include start/offset pair)
633 * @csum_start: Offset from skb->head where checksumming should start
634 * @csum_offset: Offset from csum_start where checksum should be stored
635 * @priority: Packet queueing priority
636 * @ignore_df: allow local fragmentation
637 * @cloned: Head may be cloned (check refcnt to be sure)
638 * @ip_summed: Driver fed us an IP checksum
639 * @nohdr: Payload reference only, must not modify header
640 * @pkt_type: Packet class
641 * @fclone: skbuff clone status
642 * @ipvs_property: skbuff is owned by ipvs
643 * @inner_protocol_type: whether the inner protocol is
644 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
645 * @remcsum_offload: remote checksum offload is enabled
646 * @offload_fwd_mark: Packet was L2-forwarded in hardware
647 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
648 * @tc_skip_classify: do not classify packet. set by IFB device
649 * @tc_at_ingress: used within tc_classify to distinguish in/egress
650 * @redirected: packet was redirected by packet classifier
651 * @from_ingress: packet was redirected from the ingress path
652 * @peeked: this packet has been seen already, so stats have been
653 * done for it, don't do them again
654 * @nf_trace: netfilter packet trace flag
655 * @protocol: Packet protocol from driver
656 * @destructor: Destruct function
657 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
658 * @_nfct: Associated connection, if any (with nfctinfo bits)
659 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
660 * @skb_iif: ifindex of device we arrived on
661 * @tc_index: Traffic control index
662 * @hash: the packet hash
663 * @queue_mapping: Queue mapping for multiqueue devices
664 * @head_frag: skb was allocated from page fragments,
665 * not allocated by kmalloc() or vmalloc().
666 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
667 * @active_extensions: active extensions (skb_ext_id types)
668 * @ndisc_nodetype: router type (from link layer)
669 * @ooo_okay: allow the mapping of a socket to a queue to be changed
670 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
671 * ports.
672 * @sw_hash: indicates hash was computed in software stack
673 * @wifi_acked_valid: wifi_acked was set
674 * @wifi_acked: whether frame was acked on wifi or not
675 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
676 * @encapsulation: indicates the inner headers in the skbuff are valid
677 * @encap_hdr_csum: software checksum is needed
678 * @csum_valid: checksum is already valid
679 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
680 * @csum_complete_sw: checksum was completed by software
681 * @csum_level: indicates the number of consecutive checksums found in
682 * the packet minus one that have been verified as
683 * CHECKSUM_UNNECESSARY (max 3)
684 * @dst_pending_confirm: need to confirm neighbour
685 * @decrypted: Decrypted SKB
686 * @napi_id: id of the NAPI struct this skb came from
687 * @sender_cpu: (aka @napi_id) source CPU in XPS
688 * @secmark: security marking
689 * @mark: Generic packet mark
690 * @reserved_tailroom: (aka @mark) number of bytes of free space available
691 * at the tail of an sk_buff
692 * @vlan_present: VLAN tag is present
693 * @vlan_proto: vlan encapsulation protocol
694 * @vlan_tci: vlan tag control information
695 * @inner_protocol: Protocol (encapsulation)
696 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
697 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
698 * @inner_transport_header: Inner transport layer header (encapsulation)
699 * @inner_network_header: Network layer header (encapsulation)
700 * @inner_mac_header: Link layer header (encapsulation)
701 * @transport_header: Transport layer header
702 * @network_header: Network layer header
703 * @mac_header: Link layer header
704 * @tail: Tail pointer
705 * @end: End pointer
706 * @head: Head of buffer
707 * @data: Data head pointer
708 * @truesize: Buffer size
709 * @users: User count - see {datagram,tcp}.c
710 * @extensions: allocated extensions, valid if active_extensions is nonzero
711 */
712
713 struct sk_buff {
714 union {
715 struct {
716 /* These two members must be first. */
717 struct sk_buff *next;
718 struct sk_buff *prev;
719
720 union {
721 struct net_device *dev;
722 /* Some protocols might use this space to store information,
723 * while device pointer would be NULL.
724 * UDP receive path is one user.
725 */
726 unsigned long dev_scratch;
727 };
728 };
729 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
730 struct list_head list;
731 };
732
733 union {
734 struct sock *sk;
735 int ip_defrag_offset;
736 };
737
738 union {
739 ktime_t tstamp;
740 u64 skb_mstamp_ns; /* earliest departure time */
741 };
742 /*
743 * This is the control buffer. It is free to use for every
744 * layer. Please put your private variables there. If you
745 * want to keep them across layers you have to do a skb_clone()
746 * first. This is owned by whoever has the skb queued ATM.
747 */
748 char cb[48] __aligned(8);
749
750 union {
751 struct {
752 unsigned long _skb_refdst;
753 void (*destructor)(struct sk_buff *skb);
754 };
755 struct list_head tcp_tsorted_anchor;
756 };
757
758 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
759 unsigned long _nfct;
760 #endif
761 unsigned int len,
762 data_len;
763 __u16 mac_len,
764 hdr_len;
765
766 /* Following fields are _not_ copied in __copy_skb_header()
767 * Note that queue_mapping is here mostly to fill a hole.
768 */
769 __u16 queue_mapping;
770
771 /* if you move cloned around you also must adapt those constants */
772 #ifdef __BIG_ENDIAN_BITFIELD
773 #define CLONED_MASK (1 << 7)
774 #else
775 #define CLONED_MASK 1
776 #endif
777 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
778
779 /* private: */
780 __u8 __cloned_offset[0];
781 /* public: */
782 __u8 cloned:1,
783 nohdr:1,
784 fclone:2,
785 peeked:1,
786 head_frag:1,
787 pfmemalloc:1;
788 #ifdef CONFIG_SKB_EXTENSIONS
789 __u8 active_extensions;
790 #endif
791 /* fields enclosed in headers_start/headers_end are copied
792 * using a single memcpy() in __copy_skb_header()
793 */
794 /* private: */
795 __u32 headers_start[0];
796 /* public: */
797
798 /* if you move pkt_type around you also must adapt those constants */
799 #ifdef __BIG_ENDIAN_BITFIELD
800 #define PKT_TYPE_MAX (7 << 5)
801 #else
802 #define PKT_TYPE_MAX 7
803 #endif
804 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
805
806 /* private: */
807 __u8 __pkt_type_offset[0];
808 /* public: */
809 __u8 pkt_type:3;
810 __u8 ignore_df:1;
811 __u8 nf_trace:1;
812 __u8 ip_summed:2;
813 __u8 ooo_okay:1;
814
815 __u8 l4_hash:1;
816 __u8 sw_hash:1;
817 __u8 wifi_acked_valid:1;
818 __u8 wifi_acked:1;
819 __u8 no_fcs:1;
820 /* Indicates the inner headers are valid in the skbuff. */
821 __u8 encapsulation:1;
822 __u8 encap_hdr_csum:1;
823 __u8 csum_valid:1;
824
825 #ifdef __BIG_ENDIAN_BITFIELD
826 #define PKT_VLAN_PRESENT_BIT 7
827 #else
828 #define PKT_VLAN_PRESENT_BIT 0
829 #endif
830 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
831 /* private: */
832 __u8 __pkt_vlan_present_offset[0];
833 /* public: */
834 __u8 vlan_present:1;
835 __u8 csum_complete_sw:1;
836 __u8 csum_level:2;
837 __u8 csum_not_inet:1;
838 __u8 dst_pending_confirm:1;
839 #ifdef CONFIG_IPV6_NDISC_NODETYPE
840 __u8 ndisc_nodetype:2;
841 #endif
842
843 __u8 ipvs_property:1;
844 __u8 inner_protocol_type:1;
845 __u8 remcsum_offload:1;
846 #ifdef CONFIG_NET_SWITCHDEV
847 __u8 offload_fwd_mark:1;
848 __u8 offload_l3_fwd_mark:1;
849 #endif
850 #ifdef CONFIG_NET_CLS_ACT
851 __u8 tc_skip_classify:1;
852 __u8 tc_at_ingress:1;
853 #endif
854 #ifdef CONFIG_NET_REDIRECT
855 __u8 redirected:1;
856 __u8 from_ingress:1;
857 #endif
858 #ifdef CONFIG_TLS_DEVICE
859 __u8 decrypted:1;
860 #endif
861
862 #ifdef CONFIG_NET_SCHED
863 __u16 tc_index; /* traffic control index */
864 #endif
865
866 union {
867 __wsum csum;
868 struct {
869 __u16 csum_start;
870 __u16 csum_offset;
871 };
872 };
873 __u32 priority;
874 int skb_iif;
875 __u32 hash;
876 __be16 vlan_proto;
877 __u16 vlan_tci;
878 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
879 union {
880 unsigned int napi_id;
881 unsigned int sender_cpu;
882 };
883 #endif
884 #ifdef CONFIG_NETWORK_SECMARK
885 __u32 secmark;
886 #endif
887
888 union {
889 __u32 mark;
890 __u32 reserved_tailroom;
891 };
892
893 union {
894 __be16 inner_protocol;
895 __u8 inner_ipproto;
896 };
897
898 __u16 inner_transport_header;
899 __u16 inner_network_header;
900 __u16 inner_mac_header;
901
902 __be16 protocol;
903 __u16 transport_header;
904 __u16 network_header;
905 __u16 mac_header;
906
907 /* private: */
908 __u32 headers_end[0];
909 /* public: */
910
911 /* These elements must be at the end, see alloc_skb() for details. */
912 sk_buff_data_t tail;
913 sk_buff_data_t end;
914 unsigned char *head,
915 *data;
916 unsigned int truesize;
917 refcount_t users;
918
919 #ifdef CONFIG_SKB_EXTENSIONS
920 /* only useable after checking ->active_extensions != 0 */
921 struct skb_ext *extensions;
922 #endif
923 };
924
925 #ifdef __KERNEL__
926 /*
927 * Handling routines are only of interest to the kernel
928 */
929
930 #define SKB_ALLOC_FCLONE 0x01
931 #define SKB_ALLOC_RX 0x02
932 #define SKB_ALLOC_NAPI 0x04
933
934 /**
935 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
936 * @skb: buffer
937 */
skb_pfmemalloc(const struct sk_buff * skb)938 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
939 {
940 return unlikely(skb->pfmemalloc);
941 }
942
943 /*
944 * skb might have a dst pointer attached, refcounted or not.
945 * _skb_refdst low order bit is set if refcount was _not_ taken
946 */
947 #define SKB_DST_NOREF 1UL
948 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
949
950 /**
951 * skb_dst - returns skb dst_entry
952 * @skb: buffer
953 *
954 * Returns skb dst_entry, regardless of reference taken or not.
955 */
skb_dst(const struct sk_buff * skb)956 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
957 {
958 /* If refdst was not refcounted, check we still are in a
959 * rcu_read_lock section
960 */
961 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
962 !rcu_read_lock_held() &&
963 !rcu_read_lock_bh_held());
964 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
965 }
966
967 /**
968 * skb_dst_set - sets skb dst
969 * @skb: buffer
970 * @dst: dst entry
971 *
972 * Sets skb dst, assuming a reference was taken on dst and should
973 * be released by skb_dst_drop()
974 */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)975 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
976 {
977 skb->_skb_refdst = (unsigned long)dst;
978 }
979
980 /**
981 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
982 * @skb: buffer
983 * @dst: dst entry
984 *
985 * Sets skb dst, assuming a reference was not taken on dst.
986 * If dst entry is cached, we do not take reference and dst_release
987 * will be avoided by refdst_drop. If dst entry is not cached, we take
988 * reference, so that last dst_release can destroy the dst immediately.
989 */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)990 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
991 {
992 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
993 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
994 }
995
996 /**
997 * skb_dst_is_noref - Test if skb dst isn't refcounted
998 * @skb: buffer
999 */
skb_dst_is_noref(const struct sk_buff * skb)1000 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1001 {
1002 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1003 }
1004
1005 /**
1006 * skb_rtable - Returns the skb &rtable
1007 * @skb: buffer
1008 */
skb_rtable(const struct sk_buff * skb)1009 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1010 {
1011 return (struct rtable *)skb_dst(skb);
1012 }
1013
1014 /* For mangling skb->pkt_type from user space side from applications
1015 * such as nft, tc, etc, we only allow a conservative subset of
1016 * possible pkt_types to be set.
1017 */
skb_pkt_type_ok(u32 ptype)1018 static inline bool skb_pkt_type_ok(u32 ptype)
1019 {
1020 return ptype <= PACKET_OTHERHOST;
1021 }
1022
1023 /**
1024 * skb_napi_id - Returns the skb's NAPI id
1025 * @skb: buffer
1026 */
skb_napi_id(const struct sk_buff * skb)1027 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1028 {
1029 #ifdef CONFIG_NET_RX_BUSY_POLL
1030 return skb->napi_id;
1031 #else
1032 return 0;
1033 #endif
1034 }
1035
1036 /**
1037 * skb_unref - decrement the skb's reference count
1038 * @skb: buffer
1039 *
1040 * Returns true if we can free the skb.
1041 */
skb_unref(struct sk_buff * skb)1042 static inline bool skb_unref(struct sk_buff *skb)
1043 {
1044 if (unlikely(!skb))
1045 return false;
1046 if (likely(refcount_read(&skb->users) == 1))
1047 smp_rmb();
1048 else if (likely(!refcount_dec_and_test(&skb->users)))
1049 return false;
1050
1051 return true;
1052 }
1053
1054 void skb_release_head_state(struct sk_buff *skb);
1055 void kfree_skb(struct sk_buff *skb);
1056 void kfree_skb_list(struct sk_buff *segs);
1057 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1058 void skb_tx_error(struct sk_buff *skb);
1059
1060 #ifdef CONFIG_TRACEPOINTS
1061 void consume_skb(struct sk_buff *skb);
1062 #else
consume_skb(struct sk_buff * skb)1063 static inline void consume_skb(struct sk_buff *skb)
1064 {
1065 return kfree_skb(skb);
1066 }
1067 #endif
1068
1069 void __consume_stateless_skb(struct sk_buff *skb);
1070 void __kfree_skb(struct sk_buff *skb);
1071 extern struct kmem_cache *skbuff_head_cache;
1072
1073 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1074 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1075 bool *fragstolen, int *delta_truesize);
1076
1077 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1078 int node);
1079 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1080 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1081 struct sk_buff *build_skb_around(struct sk_buff *skb,
1082 void *data, unsigned int frag_size);
1083
1084 /**
1085 * alloc_skb - allocate a network buffer
1086 * @size: size to allocate
1087 * @priority: allocation mask
1088 *
1089 * This function is a convenient wrapper around __alloc_skb().
1090 */
alloc_skb(unsigned int size,gfp_t priority)1091 static inline struct sk_buff *alloc_skb(unsigned int size,
1092 gfp_t priority)
1093 {
1094 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1095 }
1096
1097 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1098 unsigned long data_len,
1099 int max_page_order,
1100 int *errcode,
1101 gfp_t gfp_mask);
1102 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1103
1104 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1105 struct sk_buff_fclones {
1106 struct sk_buff skb1;
1107
1108 struct sk_buff skb2;
1109
1110 refcount_t fclone_ref;
1111 };
1112
1113 /**
1114 * skb_fclone_busy - check if fclone is busy
1115 * @sk: socket
1116 * @skb: buffer
1117 *
1118 * Returns true if skb is a fast clone, and its clone is not freed.
1119 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1120 * so we also check that this didnt happen.
1121 */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1122 static inline bool skb_fclone_busy(const struct sock *sk,
1123 const struct sk_buff *skb)
1124 {
1125 const struct sk_buff_fclones *fclones;
1126
1127 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1128
1129 return skb->fclone == SKB_FCLONE_ORIG &&
1130 refcount_read(&fclones->fclone_ref) > 1 &&
1131 fclones->skb2.sk == sk;
1132 }
1133
1134 /**
1135 * alloc_skb_fclone - allocate a network buffer from fclone cache
1136 * @size: size to allocate
1137 * @priority: allocation mask
1138 *
1139 * This function is a convenient wrapper around __alloc_skb().
1140 */
alloc_skb_fclone(unsigned int size,gfp_t priority)1141 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1142 gfp_t priority)
1143 {
1144 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1145 }
1146
1147 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1148 void skb_headers_offset_update(struct sk_buff *skb, int off);
1149 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1150 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1151 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1152 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1153 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1154 gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1155 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1156 gfp_t gfp_mask)
1157 {
1158 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1159 }
1160
1161 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1162 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1163 unsigned int headroom);
1164 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1165 int newtailroom, gfp_t priority);
1166 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1167 int offset, int len);
1168 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1169 int offset, int len);
1170 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1171 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1172
1173 /**
1174 * skb_pad - zero pad the tail of an skb
1175 * @skb: buffer to pad
1176 * @pad: space to pad
1177 *
1178 * Ensure that a buffer is followed by a padding area that is zero
1179 * filled. Used by network drivers which may DMA or transfer data
1180 * beyond the buffer end onto the wire.
1181 *
1182 * May return error in out of memory cases. The skb is freed on error.
1183 */
skb_pad(struct sk_buff * skb,int pad)1184 static inline int skb_pad(struct sk_buff *skb, int pad)
1185 {
1186 return __skb_pad(skb, pad, true);
1187 }
1188 #define dev_kfree_skb(a) consume_skb(a)
1189
1190 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1191 int offset, size_t size);
1192
1193 struct skb_seq_state {
1194 __u32 lower_offset;
1195 __u32 upper_offset;
1196 __u32 frag_idx;
1197 __u32 stepped_offset;
1198 struct sk_buff *root_skb;
1199 struct sk_buff *cur_skb;
1200 __u8 *frag_data;
1201 };
1202
1203 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1204 unsigned int to, struct skb_seq_state *st);
1205 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1206 struct skb_seq_state *st);
1207 void skb_abort_seq_read(struct skb_seq_state *st);
1208
1209 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1210 unsigned int to, struct ts_config *config);
1211
1212 /*
1213 * Packet hash types specify the type of hash in skb_set_hash.
1214 *
1215 * Hash types refer to the protocol layer addresses which are used to
1216 * construct a packet's hash. The hashes are used to differentiate or identify
1217 * flows of the protocol layer for the hash type. Hash types are either
1218 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1219 *
1220 * Properties of hashes:
1221 *
1222 * 1) Two packets in different flows have different hash values
1223 * 2) Two packets in the same flow should have the same hash value
1224 *
1225 * A hash at a higher layer is considered to be more specific. A driver should
1226 * set the most specific hash possible.
1227 *
1228 * A driver cannot indicate a more specific hash than the layer at which a hash
1229 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1230 *
1231 * A driver may indicate a hash level which is less specific than the
1232 * actual layer the hash was computed on. For instance, a hash computed
1233 * at L4 may be considered an L3 hash. This should only be done if the
1234 * driver can't unambiguously determine that the HW computed the hash at
1235 * the higher layer. Note that the "should" in the second property above
1236 * permits this.
1237 */
1238 enum pkt_hash_types {
1239 PKT_HASH_TYPE_NONE, /* Undefined type */
1240 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1241 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1242 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1243 };
1244
skb_clear_hash(struct sk_buff * skb)1245 static inline void skb_clear_hash(struct sk_buff *skb)
1246 {
1247 skb->hash = 0;
1248 skb->sw_hash = 0;
1249 skb->l4_hash = 0;
1250 }
1251
skb_clear_hash_if_not_l4(struct sk_buff * skb)1252 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1253 {
1254 if (!skb->l4_hash)
1255 skb_clear_hash(skb);
1256 }
1257
1258 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1259 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1260 {
1261 skb->l4_hash = is_l4;
1262 skb->sw_hash = is_sw;
1263 skb->hash = hash;
1264 }
1265
1266 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1267 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1268 {
1269 /* Used by drivers to set hash from HW */
1270 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1271 }
1272
1273 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1274 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1275 {
1276 __skb_set_hash(skb, hash, true, is_l4);
1277 }
1278
1279 void __skb_get_hash(struct sk_buff *skb);
1280 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1281 u32 skb_get_poff(const struct sk_buff *skb);
1282 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1283 const struct flow_keys_basic *keys, int hlen);
1284 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1285 void *data, int hlen_proto);
1286
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto)1287 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1288 int thoff, u8 ip_proto)
1289 {
1290 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1291 }
1292
1293 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1294 const struct flow_dissector_key *key,
1295 unsigned int key_count);
1296
1297 struct bpf_flow_dissector;
1298 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1299 __be16 proto, int nhoff, int hlen, unsigned int flags);
1300
1301 bool __skb_flow_dissect(const struct net *net,
1302 const struct sk_buff *skb,
1303 struct flow_dissector *flow_dissector,
1304 void *target_container,
1305 void *data, __be16 proto, int nhoff, int hlen,
1306 unsigned int flags);
1307
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1308 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1309 struct flow_dissector *flow_dissector,
1310 void *target_container, unsigned int flags)
1311 {
1312 return __skb_flow_dissect(NULL, skb, flow_dissector,
1313 target_container, NULL, 0, 0, 0, flags);
1314 }
1315
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1316 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1317 struct flow_keys *flow,
1318 unsigned int flags)
1319 {
1320 memset(flow, 0, sizeof(*flow));
1321 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1322 flow, NULL, 0, 0, 0, flags);
1323 }
1324
1325 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)1326 skb_flow_dissect_flow_keys_basic(const struct net *net,
1327 const struct sk_buff *skb,
1328 struct flow_keys_basic *flow, void *data,
1329 __be16 proto, int nhoff, int hlen,
1330 unsigned int flags)
1331 {
1332 memset(flow, 0, sizeof(*flow));
1333 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1334 data, proto, nhoff, hlen, flags);
1335 }
1336
1337 void skb_flow_dissect_meta(const struct sk_buff *skb,
1338 struct flow_dissector *flow_dissector,
1339 void *target_container);
1340
1341 /* Gets a skb connection tracking info, ctinfo map should be a
1342 * map of mapsize to translate enum ip_conntrack_info states
1343 * to user states.
1344 */
1345 void
1346 skb_flow_dissect_ct(const struct sk_buff *skb,
1347 struct flow_dissector *flow_dissector,
1348 void *target_container,
1349 u16 *ctinfo_map,
1350 size_t mapsize);
1351 void
1352 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1353 struct flow_dissector *flow_dissector,
1354 void *target_container);
1355
1356 void skb_flow_dissect_hash(const struct sk_buff *skb,
1357 struct flow_dissector *flow_dissector,
1358 void *target_container);
1359
skb_get_hash(struct sk_buff * skb)1360 static inline __u32 skb_get_hash(struct sk_buff *skb)
1361 {
1362 if (!skb->l4_hash && !skb->sw_hash)
1363 __skb_get_hash(skb);
1364
1365 return skb->hash;
1366 }
1367
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1368 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1369 {
1370 if (!skb->l4_hash && !skb->sw_hash) {
1371 struct flow_keys keys;
1372 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1373
1374 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1375 }
1376
1377 return skb->hash;
1378 }
1379
1380 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1381 const siphash_key_t *perturb);
1382
skb_get_hash_raw(const struct sk_buff * skb)1383 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1384 {
1385 return skb->hash;
1386 }
1387
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1388 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1389 {
1390 to->hash = from->hash;
1391 to->sw_hash = from->sw_hash;
1392 to->l4_hash = from->l4_hash;
1393 };
1394
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1395 static inline void skb_copy_decrypted(struct sk_buff *to,
1396 const struct sk_buff *from)
1397 {
1398 #ifdef CONFIG_TLS_DEVICE
1399 to->decrypted = from->decrypted;
1400 #endif
1401 }
1402
1403 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1404 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1405 {
1406 return skb->head + skb->end;
1407 }
1408
skb_end_offset(const struct sk_buff * skb)1409 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1410 {
1411 return skb->end;
1412 }
1413 #else
skb_end_pointer(const struct sk_buff * skb)1414 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1415 {
1416 return skb->end;
1417 }
1418
skb_end_offset(const struct sk_buff * skb)1419 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1420 {
1421 return skb->end - skb->head;
1422 }
1423 #endif
1424
1425 /* Internal */
1426 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1427
skb_hwtstamps(struct sk_buff * skb)1428 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1429 {
1430 return &skb_shinfo(skb)->hwtstamps;
1431 }
1432
skb_zcopy(struct sk_buff * skb)1433 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1434 {
1435 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1436
1437 return is_zcopy ? skb_uarg(skb) : NULL;
1438 }
1439
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1440 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1441 bool *have_ref)
1442 {
1443 if (skb && uarg && !skb_zcopy(skb)) {
1444 if (unlikely(have_ref && *have_ref))
1445 *have_ref = false;
1446 else
1447 sock_zerocopy_get(uarg);
1448 skb_shinfo(skb)->destructor_arg = uarg;
1449 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1450 }
1451 }
1452
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1453 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1454 {
1455 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1456 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1457 }
1458
skb_zcopy_is_nouarg(struct sk_buff * skb)1459 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1460 {
1461 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1462 }
1463
skb_zcopy_get_nouarg(struct sk_buff * skb)1464 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1465 {
1466 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1467 }
1468
1469 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy)1470 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1471 {
1472 struct ubuf_info *uarg = skb_zcopy(skb);
1473
1474 if (uarg) {
1475 if (skb_zcopy_is_nouarg(skb)) {
1476 /* no notification callback */
1477 } else if (uarg->callback == sock_zerocopy_callback) {
1478 uarg->zerocopy = uarg->zerocopy && zerocopy;
1479 sock_zerocopy_put(uarg);
1480 } else {
1481 uarg->callback(uarg, zerocopy);
1482 }
1483
1484 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1485 }
1486 }
1487
1488 /* Abort a zerocopy operation and revert zckey on error in send syscall */
skb_zcopy_abort(struct sk_buff * skb)1489 static inline void skb_zcopy_abort(struct sk_buff *skb)
1490 {
1491 struct ubuf_info *uarg = skb_zcopy(skb);
1492
1493 if (uarg) {
1494 sock_zerocopy_put_abort(uarg, false);
1495 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1496 }
1497 }
1498
skb_mark_not_on_list(struct sk_buff * skb)1499 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1500 {
1501 skb->next = NULL;
1502 }
1503
1504 /* Iterate through singly-linked GSO fragments of an skb. */
1505 #define skb_list_walk_safe(first, skb, next_skb) \
1506 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1507 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1508
skb_list_del_init(struct sk_buff * skb)1509 static inline void skb_list_del_init(struct sk_buff *skb)
1510 {
1511 __list_del_entry(&skb->list);
1512 skb_mark_not_on_list(skb);
1513 }
1514
1515 /**
1516 * skb_queue_empty - check if a queue is empty
1517 * @list: queue head
1518 *
1519 * Returns true if the queue is empty, false otherwise.
1520 */
skb_queue_empty(const struct sk_buff_head * list)1521 static inline int skb_queue_empty(const struct sk_buff_head *list)
1522 {
1523 return list->next == (const struct sk_buff *) list;
1524 }
1525
1526 /**
1527 * skb_queue_empty_lockless - check if a queue is empty
1528 * @list: queue head
1529 *
1530 * Returns true if the queue is empty, false otherwise.
1531 * This variant can be used in lockless contexts.
1532 */
skb_queue_empty_lockless(const struct sk_buff_head * list)1533 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1534 {
1535 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1536 }
1537
1538
1539 /**
1540 * skb_queue_is_last - check if skb is the last entry in the queue
1541 * @list: queue head
1542 * @skb: buffer
1543 *
1544 * Returns true if @skb is the last buffer on the list.
1545 */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1546 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1547 const struct sk_buff *skb)
1548 {
1549 return skb->next == (const struct sk_buff *) list;
1550 }
1551
1552 /**
1553 * skb_queue_is_first - check if skb is the first entry in the queue
1554 * @list: queue head
1555 * @skb: buffer
1556 *
1557 * Returns true if @skb is the first buffer on the list.
1558 */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1559 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1560 const struct sk_buff *skb)
1561 {
1562 return skb->prev == (const struct sk_buff *) list;
1563 }
1564
1565 /**
1566 * skb_queue_next - return the next packet in the queue
1567 * @list: queue head
1568 * @skb: current buffer
1569 *
1570 * Return the next packet in @list after @skb. It is only valid to
1571 * call this if skb_queue_is_last() evaluates to false.
1572 */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1573 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1574 const struct sk_buff *skb)
1575 {
1576 /* This BUG_ON may seem severe, but if we just return then we
1577 * are going to dereference garbage.
1578 */
1579 BUG_ON(skb_queue_is_last(list, skb));
1580 return skb->next;
1581 }
1582
1583 /**
1584 * skb_queue_prev - return the prev packet in the queue
1585 * @list: queue head
1586 * @skb: current buffer
1587 *
1588 * Return the prev packet in @list before @skb. It is only valid to
1589 * call this if skb_queue_is_first() evaluates to false.
1590 */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1591 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1592 const struct sk_buff *skb)
1593 {
1594 /* This BUG_ON may seem severe, but if we just return then we
1595 * are going to dereference garbage.
1596 */
1597 BUG_ON(skb_queue_is_first(list, skb));
1598 return skb->prev;
1599 }
1600
1601 /**
1602 * skb_get - reference buffer
1603 * @skb: buffer to reference
1604 *
1605 * Makes another reference to a socket buffer and returns a pointer
1606 * to the buffer.
1607 */
skb_get(struct sk_buff * skb)1608 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1609 {
1610 refcount_inc(&skb->users);
1611 return skb;
1612 }
1613
1614 /*
1615 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1616 */
1617
1618 /**
1619 * skb_cloned - is the buffer a clone
1620 * @skb: buffer to check
1621 *
1622 * Returns true if the buffer was generated with skb_clone() and is
1623 * one of multiple shared copies of the buffer. Cloned buffers are
1624 * shared data so must not be written to under normal circumstances.
1625 */
skb_cloned(const struct sk_buff * skb)1626 static inline int skb_cloned(const struct sk_buff *skb)
1627 {
1628 return skb->cloned &&
1629 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1630 }
1631
skb_unclone(struct sk_buff * skb,gfp_t pri)1632 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1633 {
1634 might_sleep_if(gfpflags_allow_blocking(pri));
1635
1636 if (skb_cloned(skb))
1637 return pskb_expand_head(skb, 0, 0, pri);
1638
1639 return 0;
1640 }
1641
1642 /**
1643 * skb_header_cloned - is the header a clone
1644 * @skb: buffer to check
1645 *
1646 * Returns true if modifying the header part of the buffer requires
1647 * the data to be copied.
1648 */
skb_header_cloned(const struct sk_buff * skb)1649 static inline int skb_header_cloned(const struct sk_buff *skb)
1650 {
1651 int dataref;
1652
1653 if (!skb->cloned)
1654 return 0;
1655
1656 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1657 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1658 return dataref != 1;
1659 }
1660
skb_header_unclone(struct sk_buff * skb,gfp_t pri)1661 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1662 {
1663 might_sleep_if(gfpflags_allow_blocking(pri));
1664
1665 if (skb_header_cloned(skb))
1666 return pskb_expand_head(skb, 0, 0, pri);
1667
1668 return 0;
1669 }
1670
1671 /**
1672 * __skb_header_release - release reference to header
1673 * @skb: buffer to operate on
1674 */
__skb_header_release(struct sk_buff * skb)1675 static inline void __skb_header_release(struct sk_buff *skb)
1676 {
1677 skb->nohdr = 1;
1678 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1679 }
1680
1681
1682 /**
1683 * skb_shared - is the buffer shared
1684 * @skb: buffer to check
1685 *
1686 * Returns true if more than one person has a reference to this
1687 * buffer.
1688 */
skb_shared(const struct sk_buff * skb)1689 static inline int skb_shared(const struct sk_buff *skb)
1690 {
1691 return refcount_read(&skb->users) != 1;
1692 }
1693
1694 /**
1695 * skb_share_check - check if buffer is shared and if so clone it
1696 * @skb: buffer to check
1697 * @pri: priority for memory allocation
1698 *
1699 * If the buffer is shared the buffer is cloned and the old copy
1700 * drops a reference. A new clone with a single reference is returned.
1701 * If the buffer is not shared the original buffer is returned. When
1702 * being called from interrupt status or with spinlocks held pri must
1703 * be GFP_ATOMIC.
1704 *
1705 * NULL is returned on a memory allocation failure.
1706 */
skb_share_check(struct sk_buff * skb,gfp_t pri)1707 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1708 {
1709 might_sleep_if(gfpflags_allow_blocking(pri));
1710 if (skb_shared(skb)) {
1711 struct sk_buff *nskb = skb_clone(skb, pri);
1712
1713 if (likely(nskb))
1714 consume_skb(skb);
1715 else
1716 kfree_skb(skb);
1717 skb = nskb;
1718 }
1719 return skb;
1720 }
1721
1722 /*
1723 * Copy shared buffers into a new sk_buff. We effectively do COW on
1724 * packets to handle cases where we have a local reader and forward
1725 * and a couple of other messy ones. The normal one is tcpdumping
1726 * a packet thats being forwarded.
1727 */
1728
1729 /**
1730 * skb_unshare - make a copy of a shared buffer
1731 * @skb: buffer to check
1732 * @pri: priority for memory allocation
1733 *
1734 * If the socket buffer is a clone then this function creates a new
1735 * copy of the data, drops a reference count on the old copy and returns
1736 * the new copy with the reference count at 1. If the buffer is not a clone
1737 * the original buffer is returned. When called with a spinlock held or
1738 * from interrupt state @pri must be %GFP_ATOMIC
1739 *
1740 * %NULL is returned on a memory allocation failure.
1741 */
skb_unshare(struct sk_buff * skb,gfp_t pri)1742 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1743 gfp_t pri)
1744 {
1745 might_sleep_if(gfpflags_allow_blocking(pri));
1746 if (skb_cloned(skb)) {
1747 struct sk_buff *nskb = skb_copy(skb, pri);
1748
1749 /* Free our shared copy */
1750 if (likely(nskb))
1751 consume_skb(skb);
1752 else
1753 kfree_skb(skb);
1754 skb = nskb;
1755 }
1756 return skb;
1757 }
1758
1759 /**
1760 * skb_peek - peek at the head of an &sk_buff_head
1761 * @list_: list to peek at
1762 *
1763 * Peek an &sk_buff. Unlike most other operations you _MUST_
1764 * be careful with this one. A peek leaves the buffer on the
1765 * list and someone else may run off with it. You must hold
1766 * the appropriate locks or have a private queue to do this.
1767 *
1768 * Returns %NULL for an empty list or a pointer to the head element.
1769 * The reference count is not incremented and the reference is therefore
1770 * volatile. Use with caution.
1771 */
skb_peek(const struct sk_buff_head * list_)1772 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1773 {
1774 struct sk_buff *skb = list_->next;
1775
1776 if (skb == (struct sk_buff *)list_)
1777 skb = NULL;
1778 return skb;
1779 }
1780
1781 /**
1782 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1783 * @list_: list to peek at
1784 *
1785 * Like skb_peek(), but the caller knows that the list is not empty.
1786 */
__skb_peek(const struct sk_buff_head * list_)1787 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1788 {
1789 return list_->next;
1790 }
1791
1792 /**
1793 * skb_peek_next - peek skb following the given one from a queue
1794 * @skb: skb to start from
1795 * @list_: list to peek at
1796 *
1797 * Returns %NULL when the end of the list is met or a pointer to the
1798 * next element. The reference count is not incremented and the
1799 * reference is therefore volatile. Use with caution.
1800 */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)1801 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1802 const struct sk_buff_head *list_)
1803 {
1804 struct sk_buff *next = skb->next;
1805
1806 if (next == (struct sk_buff *)list_)
1807 next = NULL;
1808 return next;
1809 }
1810
1811 /**
1812 * skb_peek_tail - peek at the tail of an &sk_buff_head
1813 * @list_: list to peek at
1814 *
1815 * Peek an &sk_buff. Unlike most other operations you _MUST_
1816 * be careful with this one. A peek leaves the buffer on the
1817 * list and someone else may run off with it. You must hold
1818 * the appropriate locks or have a private queue to do this.
1819 *
1820 * Returns %NULL for an empty list or a pointer to the tail element.
1821 * The reference count is not incremented and the reference is therefore
1822 * volatile. Use with caution.
1823 */
skb_peek_tail(const struct sk_buff_head * list_)1824 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1825 {
1826 struct sk_buff *skb = READ_ONCE(list_->prev);
1827
1828 if (skb == (struct sk_buff *)list_)
1829 skb = NULL;
1830 return skb;
1831
1832 }
1833
1834 /**
1835 * skb_queue_len - get queue length
1836 * @list_: list to measure
1837 *
1838 * Return the length of an &sk_buff queue.
1839 */
skb_queue_len(const struct sk_buff_head * list_)1840 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1841 {
1842 return list_->qlen;
1843 }
1844
1845 /**
1846 * skb_queue_len_lockless - get queue length
1847 * @list_: list to measure
1848 *
1849 * Return the length of an &sk_buff queue.
1850 * This variant can be used in lockless contexts.
1851 */
skb_queue_len_lockless(const struct sk_buff_head * list_)1852 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1853 {
1854 return READ_ONCE(list_->qlen);
1855 }
1856
1857 /**
1858 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1859 * @list: queue to initialize
1860 *
1861 * This initializes only the list and queue length aspects of
1862 * an sk_buff_head object. This allows to initialize the list
1863 * aspects of an sk_buff_head without reinitializing things like
1864 * the spinlock. It can also be used for on-stack sk_buff_head
1865 * objects where the spinlock is known to not be used.
1866 */
__skb_queue_head_init(struct sk_buff_head * list)1867 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1868 {
1869 list->prev = list->next = (struct sk_buff *)list;
1870 list->qlen = 0;
1871 }
1872
1873 /*
1874 * This function creates a split out lock class for each invocation;
1875 * this is needed for now since a whole lot of users of the skb-queue
1876 * infrastructure in drivers have different locking usage (in hardirq)
1877 * than the networking core (in softirq only). In the long run either the
1878 * network layer or drivers should need annotation to consolidate the
1879 * main types of usage into 3 classes.
1880 */
skb_queue_head_init(struct sk_buff_head * list)1881 static inline void skb_queue_head_init(struct sk_buff_head *list)
1882 {
1883 spin_lock_init(&list->lock);
1884 __skb_queue_head_init(list);
1885 }
1886
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)1887 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1888 struct lock_class_key *class)
1889 {
1890 skb_queue_head_init(list);
1891 lockdep_set_class(&list->lock, class);
1892 }
1893
1894 /*
1895 * Insert an sk_buff on a list.
1896 *
1897 * The "__skb_xxxx()" functions are the non-atomic ones that
1898 * can only be called with interrupts disabled.
1899 */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)1900 static inline void __skb_insert(struct sk_buff *newsk,
1901 struct sk_buff *prev, struct sk_buff *next,
1902 struct sk_buff_head *list)
1903 {
1904 /* See skb_queue_empty_lockless() and skb_peek_tail()
1905 * for the opposite READ_ONCE()
1906 */
1907 WRITE_ONCE(newsk->next, next);
1908 WRITE_ONCE(newsk->prev, prev);
1909 WRITE_ONCE(next->prev, newsk);
1910 WRITE_ONCE(prev->next, newsk);
1911 list->qlen++;
1912 }
1913
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)1914 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1915 struct sk_buff *prev,
1916 struct sk_buff *next)
1917 {
1918 struct sk_buff *first = list->next;
1919 struct sk_buff *last = list->prev;
1920
1921 WRITE_ONCE(first->prev, prev);
1922 WRITE_ONCE(prev->next, first);
1923
1924 WRITE_ONCE(last->next, next);
1925 WRITE_ONCE(next->prev, last);
1926 }
1927
1928 /**
1929 * skb_queue_splice - join two skb lists, this is designed for stacks
1930 * @list: the new list to add
1931 * @head: the place to add it in the first list
1932 */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)1933 static inline void skb_queue_splice(const struct sk_buff_head *list,
1934 struct sk_buff_head *head)
1935 {
1936 if (!skb_queue_empty(list)) {
1937 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1938 head->qlen += list->qlen;
1939 }
1940 }
1941
1942 /**
1943 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1944 * @list: the new list to add
1945 * @head: the place to add it in the first list
1946 *
1947 * The list at @list is reinitialised
1948 */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)1949 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1950 struct sk_buff_head *head)
1951 {
1952 if (!skb_queue_empty(list)) {
1953 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1954 head->qlen += list->qlen;
1955 __skb_queue_head_init(list);
1956 }
1957 }
1958
1959 /**
1960 * skb_queue_splice_tail - join two skb lists, each list being a queue
1961 * @list: the new list to add
1962 * @head: the place to add it in the first list
1963 */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)1964 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1965 struct sk_buff_head *head)
1966 {
1967 if (!skb_queue_empty(list)) {
1968 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1969 head->qlen += list->qlen;
1970 }
1971 }
1972
1973 /**
1974 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1975 * @list: the new list to add
1976 * @head: the place to add it in the first list
1977 *
1978 * Each of the lists is a queue.
1979 * The list at @list is reinitialised
1980 */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)1981 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1982 struct sk_buff_head *head)
1983 {
1984 if (!skb_queue_empty(list)) {
1985 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1986 head->qlen += list->qlen;
1987 __skb_queue_head_init(list);
1988 }
1989 }
1990
1991 /**
1992 * __skb_queue_after - queue a buffer at the list head
1993 * @list: list to use
1994 * @prev: place after this buffer
1995 * @newsk: buffer to queue
1996 *
1997 * Queue a buffer int the middle of a list. This function takes no locks
1998 * and you must therefore hold required locks before calling it.
1999 *
2000 * A buffer cannot be placed on two lists at the same time.
2001 */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)2002 static inline void __skb_queue_after(struct sk_buff_head *list,
2003 struct sk_buff *prev,
2004 struct sk_buff *newsk)
2005 {
2006 __skb_insert(newsk, prev, prev->next, list);
2007 }
2008
2009 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2010 struct sk_buff_head *list);
2011
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)2012 static inline void __skb_queue_before(struct sk_buff_head *list,
2013 struct sk_buff *next,
2014 struct sk_buff *newsk)
2015 {
2016 __skb_insert(newsk, next->prev, next, list);
2017 }
2018
2019 /**
2020 * __skb_queue_head - queue a buffer at the list head
2021 * @list: list to use
2022 * @newsk: buffer to queue
2023 *
2024 * Queue a buffer at the start of a list. This function takes no locks
2025 * and you must therefore hold required locks before calling it.
2026 *
2027 * A buffer cannot be placed on two lists at the same time.
2028 */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)2029 static inline void __skb_queue_head(struct sk_buff_head *list,
2030 struct sk_buff *newsk)
2031 {
2032 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2033 }
2034 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2035
2036 /**
2037 * __skb_queue_tail - queue a buffer at the list tail
2038 * @list: list to use
2039 * @newsk: buffer to queue
2040 *
2041 * Queue a buffer at the end of a list. This function takes no locks
2042 * and you must therefore hold required locks before calling it.
2043 *
2044 * A buffer cannot be placed on two lists at the same time.
2045 */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2046 static inline void __skb_queue_tail(struct sk_buff_head *list,
2047 struct sk_buff *newsk)
2048 {
2049 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2050 }
2051 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2052
2053 /*
2054 * remove sk_buff from list. _Must_ be called atomically, and with
2055 * the list known..
2056 */
2057 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2058 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2059 {
2060 struct sk_buff *next, *prev;
2061
2062 WRITE_ONCE(list->qlen, list->qlen - 1);
2063 next = skb->next;
2064 prev = skb->prev;
2065 skb->next = skb->prev = NULL;
2066 WRITE_ONCE(next->prev, prev);
2067 WRITE_ONCE(prev->next, next);
2068 }
2069
2070 /**
2071 * __skb_dequeue - remove from the head of the queue
2072 * @list: list to dequeue from
2073 *
2074 * Remove the head of the list. This function does not take any locks
2075 * so must be used with appropriate locks held only. The head item is
2076 * returned or %NULL if the list is empty.
2077 */
__skb_dequeue(struct sk_buff_head * list)2078 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2079 {
2080 struct sk_buff *skb = skb_peek(list);
2081 if (skb)
2082 __skb_unlink(skb, list);
2083 return skb;
2084 }
2085 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2086
2087 /**
2088 * __skb_dequeue_tail - remove from the tail of the queue
2089 * @list: list to dequeue from
2090 *
2091 * Remove the tail of the list. This function does not take any locks
2092 * so must be used with appropriate locks held only. The tail item is
2093 * returned or %NULL if the list is empty.
2094 */
__skb_dequeue_tail(struct sk_buff_head * list)2095 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2096 {
2097 struct sk_buff *skb = skb_peek_tail(list);
2098 if (skb)
2099 __skb_unlink(skb, list);
2100 return skb;
2101 }
2102 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2103
2104
skb_is_nonlinear(const struct sk_buff * skb)2105 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2106 {
2107 return skb->data_len;
2108 }
2109
skb_headlen(const struct sk_buff * skb)2110 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2111 {
2112 return skb->len - skb->data_len;
2113 }
2114
__skb_pagelen(const struct sk_buff * skb)2115 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2116 {
2117 unsigned int i, len = 0;
2118
2119 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2120 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2121 return len;
2122 }
2123
skb_pagelen(const struct sk_buff * skb)2124 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2125 {
2126 return skb_headlen(skb) + __skb_pagelen(skb);
2127 }
2128
2129 /**
2130 * __skb_fill_page_desc - initialise a paged fragment in an skb
2131 * @skb: buffer containing fragment to be initialised
2132 * @i: paged fragment index to initialise
2133 * @page: the page to use for this fragment
2134 * @off: the offset to the data with @page
2135 * @size: the length of the data
2136 *
2137 * Initialises the @i'th fragment of @skb to point to &size bytes at
2138 * offset @off within @page.
2139 *
2140 * Does not take any additional reference on the fragment.
2141 */
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2142 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2143 struct page *page, int off, int size)
2144 {
2145 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2146
2147 /*
2148 * Propagate page pfmemalloc to the skb if we can. The problem is
2149 * that not all callers have unique ownership of the page but rely
2150 * on page_is_pfmemalloc doing the right thing(tm).
2151 */
2152 frag->bv_page = page;
2153 frag->bv_offset = off;
2154 skb_frag_size_set(frag, size);
2155
2156 page = compound_head(page);
2157 if (page_is_pfmemalloc(page))
2158 skb->pfmemalloc = true;
2159 }
2160
2161 /**
2162 * skb_fill_page_desc - initialise a paged fragment in an skb
2163 * @skb: buffer containing fragment to be initialised
2164 * @i: paged fragment index to initialise
2165 * @page: the page to use for this fragment
2166 * @off: the offset to the data with @page
2167 * @size: the length of the data
2168 *
2169 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2170 * @skb to point to @size bytes at offset @off within @page. In
2171 * addition updates @skb such that @i is the last fragment.
2172 *
2173 * Does not take any additional reference on the fragment.
2174 */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2175 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2176 struct page *page, int off, int size)
2177 {
2178 __skb_fill_page_desc(skb, i, page, off, size);
2179 skb_shinfo(skb)->nr_frags = i + 1;
2180 }
2181
2182 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2183 int size, unsigned int truesize);
2184
2185 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2186 unsigned int truesize);
2187
2188 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2189
2190 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2191 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2192 {
2193 return skb->head + skb->tail;
2194 }
2195
skb_reset_tail_pointer(struct sk_buff * skb)2196 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2197 {
2198 skb->tail = skb->data - skb->head;
2199 }
2200
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2201 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2202 {
2203 skb_reset_tail_pointer(skb);
2204 skb->tail += offset;
2205 }
2206
2207 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2208 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2209 {
2210 return skb->tail;
2211 }
2212
skb_reset_tail_pointer(struct sk_buff * skb)2213 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2214 {
2215 skb->tail = skb->data;
2216 }
2217
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2218 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2219 {
2220 skb->tail = skb->data + offset;
2221 }
2222
2223 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2224
2225 /*
2226 * Add data to an sk_buff
2227 */
2228 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2229 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2230 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2231 {
2232 void *tmp = skb_tail_pointer(skb);
2233 SKB_LINEAR_ASSERT(skb);
2234 skb->tail += len;
2235 skb->len += len;
2236 return tmp;
2237 }
2238
__skb_put_zero(struct sk_buff * skb,unsigned int len)2239 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2240 {
2241 void *tmp = __skb_put(skb, len);
2242
2243 memset(tmp, 0, len);
2244 return tmp;
2245 }
2246
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2247 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2248 unsigned int len)
2249 {
2250 void *tmp = __skb_put(skb, len);
2251
2252 memcpy(tmp, data, len);
2253 return tmp;
2254 }
2255
__skb_put_u8(struct sk_buff * skb,u8 val)2256 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2257 {
2258 *(u8 *)__skb_put(skb, 1) = val;
2259 }
2260
skb_put_zero(struct sk_buff * skb,unsigned int len)2261 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2262 {
2263 void *tmp = skb_put(skb, len);
2264
2265 memset(tmp, 0, len);
2266
2267 return tmp;
2268 }
2269
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2270 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2271 unsigned int len)
2272 {
2273 void *tmp = skb_put(skb, len);
2274
2275 memcpy(tmp, data, len);
2276
2277 return tmp;
2278 }
2279
skb_put_u8(struct sk_buff * skb,u8 val)2280 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2281 {
2282 *(u8 *)skb_put(skb, 1) = val;
2283 }
2284
2285 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2286 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2287 {
2288 skb->data -= len;
2289 skb->len += len;
2290 return skb->data;
2291 }
2292
2293 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2294 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2295 {
2296 skb->len -= len;
2297 BUG_ON(skb->len < skb->data_len);
2298 return skb->data += len;
2299 }
2300
skb_pull_inline(struct sk_buff * skb,unsigned int len)2301 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2302 {
2303 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2304 }
2305
2306 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2307
__pskb_pull(struct sk_buff * skb,unsigned int len)2308 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2309 {
2310 if (len > skb_headlen(skb) &&
2311 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2312 return NULL;
2313 skb->len -= len;
2314 return skb->data += len;
2315 }
2316
pskb_pull(struct sk_buff * skb,unsigned int len)2317 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2318 {
2319 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2320 }
2321
pskb_may_pull(struct sk_buff * skb,unsigned int len)2322 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2323 {
2324 if (likely(len <= skb_headlen(skb)))
2325 return true;
2326 if (unlikely(len > skb->len))
2327 return false;
2328 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2329 }
2330
2331 void skb_condense(struct sk_buff *skb);
2332
2333 /**
2334 * skb_headroom - bytes at buffer head
2335 * @skb: buffer to check
2336 *
2337 * Return the number of bytes of free space at the head of an &sk_buff.
2338 */
skb_headroom(const struct sk_buff * skb)2339 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2340 {
2341 return skb->data - skb->head;
2342 }
2343
2344 /**
2345 * skb_tailroom - bytes at buffer end
2346 * @skb: buffer to check
2347 *
2348 * Return the number of bytes of free space at the tail of an sk_buff
2349 */
skb_tailroom(const struct sk_buff * skb)2350 static inline int skb_tailroom(const struct sk_buff *skb)
2351 {
2352 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2353 }
2354
2355 /**
2356 * skb_availroom - bytes at buffer end
2357 * @skb: buffer to check
2358 *
2359 * Return the number of bytes of free space at the tail of an sk_buff
2360 * allocated by sk_stream_alloc()
2361 */
skb_availroom(const struct sk_buff * skb)2362 static inline int skb_availroom(const struct sk_buff *skb)
2363 {
2364 if (skb_is_nonlinear(skb))
2365 return 0;
2366
2367 return skb->end - skb->tail - skb->reserved_tailroom;
2368 }
2369
2370 /**
2371 * skb_reserve - adjust headroom
2372 * @skb: buffer to alter
2373 * @len: bytes to move
2374 *
2375 * Increase the headroom of an empty &sk_buff by reducing the tail
2376 * room. This is only allowed for an empty buffer.
2377 */
skb_reserve(struct sk_buff * skb,int len)2378 static inline void skb_reserve(struct sk_buff *skb, int len)
2379 {
2380 skb->data += len;
2381 skb->tail += len;
2382 }
2383
2384 /**
2385 * skb_tailroom_reserve - adjust reserved_tailroom
2386 * @skb: buffer to alter
2387 * @mtu: maximum amount of headlen permitted
2388 * @needed_tailroom: minimum amount of reserved_tailroom
2389 *
2390 * Set reserved_tailroom so that headlen can be as large as possible but
2391 * not larger than mtu and tailroom cannot be smaller than
2392 * needed_tailroom.
2393 * The required headroom should already have been reserved before using
2394 * this function.
2395 */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2396 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2397 unsigned int needed_tailroom)
2398 {
2399 SKB_LINEAR_ASSERT(skb);
2400 if (mtu < skb_tailroom(skb) - needed_tailroom)
2401 /* use at most mtu */
2402 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2403 else
2404 /* use up to all available space */
2405 skb->reserved_tailroom = needed_tailroom;
2406 }
2407
2408 #define ENCAP_TYPE_ETHER 0
2409 #define ENCAP_TYPE_IPPROTO 1
2410
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2411 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2412 __be16 protocol)
2413 {
2414 skb->inner_protocol = protocol;
2415 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2416 }
2417
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2418 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2419 __u8 ipproto)
2420 {
2421 skb->inner_ipproto = ipproto;
2422 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2423 }
2424
skb_reset_inner_headers(struct sk_buff * skb)2425 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2426 {
2427 skb->inner_mac_header = skb->mac_header;
2428 skb->inner_network_header = skb->network_header;
2429 skb->inner_transport_header = skb->transport_header;
2430 }
2431
skb_reset_mac_len(struct sk_buff * skb)2432 static inline void skb_reset_mac_len(struct sk_buff *skb)
2433 {
2434 skb->mac_len = skb->network_header - skb->mac_header;
2435 }
2436
skb_inner_transport_header(const struct sk_buff * skb)2437 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2438 *skb)
2439 {
2440 return skb->head + skb->inner_transport_header;
2441 }
2442
skb_inner_transport_offset(const struct sk_buff * skb)2443 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2444 {
2445 return skb_inner_transport_header(skb) - skb->data;
2446 }
2447
skb_reset_inner_transport_header(struct sk_buff * skb)2448 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2449 {
2450 skb->inner_transport_header = skb->data - skb->head;
2451 }
2452
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2453 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2454 const int offset)
2455 {
2456 skb_reset_inner_transport_header(skb);
2457 skb->inner_transport_header += offset;
2458 }
2459
skb_inner_network_header(const struct sk_buff * skb)2460 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2461 {
2462 return skb->head + skb->inner_network_header;
2463 }
2464
skb_reset_inner_network_header(struct sk_buff * skb)2465 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2466 {
2467 skb->inner_network_header = skb->data - skb->head;
2468 }
2469
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2470 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2471 const int offset)
2472 {
2473 skb_reset_inner_network_header(skb);
2474 skb->inner_network_header += offset;
2475 }
2476
skb_inner_mac_header(const struct sk_buff * skb)2477 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2478 {
2479 return skb->head + skb->inner_mac_header;
2480 }
2481
skb_reset_inner_mac_header(struct sk_buff * skb)2482 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2483 {
2484 skb->inner_mac_header = skb->data - skb->head;
2485 }
2486
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)2487 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2488 const int offset)
2489 {
2490 skb_reset_inner_mac_header(skb);
2491 skb->inner_mac_header += offset;
2492 }
skb_transport_header_was_set(const struct sk_buff * skb)2493 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2494 {
2495 return skb->transport_header != (typeof(skb->transport_header))~0U;
2496 }
2497
skb_transport_header(const struct sk_buff * skb)2498 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2499 {
2500 return skb->head + skb->transport_header;
2501 }
2502
skb_reset_transport_header(struct sk_buff * skb)2503 static inline void skb_reset_transport_header(struct sk_buff *skb)
2504 {
2505 skb->transport_header = skb->data - skb->head;
2506 }
2507
skb_set_transport_header(struct sk_buff * skb,const int offset)2508 static inline void skb_set_transport_header(struct sk_buff *skb,
2509 const int offset)
2510 {
2511 skb_reset_transport_header(skb);
2512 skb->transport_header += offset;
2513 }
2514
skb_network_header(const struct sk_buff * skb)2515 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2516 {
2517 return skb->head + skb->network_header;
2518 }
2519
skb_reset_network_header(struct sk_buff * skb)2520 static inline void skb_reset_network_header(struct sk_buff *skb)
2521 {
2522 skb->network_header = skb->data - skb->head;
2523 }
2524
skb_set_network_header(struct sk_buff * skb,const int offset)2525 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2526 {
2527 skb_reset_network_header(skb);
2528 skb->network_header += offset;
2529 }
2530
skb_mac_header(const struct sk_buff * skb)2531 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2532 {
2533 return skb->head + skb->mac_header;
2534 }
2535
skb_mac_offset(const struct sk_buff * skb)2536 static inline int skb_mac_offset(const struct sk_buff *skb)
2537 {
2538 return skb_mac_header(skb) - skb->data;
2539 }
2540
skb_mac_header_len(const struct sk_buff * skb)2541 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2542 {
2543 return skb->network_header - skb->mac_header;
2544 }
2545
skb_mac_header_was_set(const struct sk_buff * skb)2546 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2547 {
2548 return skb->mac_header != (typeof(skb->mac_header))~0U;
2549 }
2550
skb_unset_mac_header(struct sk_buff * skb)2551 static inline void skb_unset_mac_header(struct sk_buff *skb)
2552 {
2553 skb->mac_header = (typeof(skb->mac_header))~0U;
2554 }
2555
skb_reset_mac_header(struct sk_buff * skb)2556 static inline void skb_reset_mac_header(struct sk_buff *skb)
2557 {
2558 skb->mac_header = skb->data - skb->head;
2559 }
2560
skb_set_mac_header(struct sk_buff * skb,const int offset)2561 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2562 {
2563 skb_reset_mac_header(skb);
2564 skb->mac_header += offset;
2565 }
2566
skb_pop_mac_header(struct sk_buff * skb)2567 static inline void skb_pop_mac_header(struct sk_buff *skb)
2568 {
2569 skb->mac_header = skb->network_header;
2570 }
2571
skb_probe_transport_header(struct sk_buff * skb)2572 static inline void skb_probe_transport_header(struct sk_buff *skb)
2573 {
2574 struct flow_keys_basic keys;
2575
2576 if (skb_transport_header_was_set(skb))
2577 return;
2578
2579 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2580 NULL, 0, 0, 0, 0))
2581 skb_set_transport_header(skb, keys.control.thoff);
2582 }
2583
skb_mac_header_rebuild(struct sk_buff * skb)2584 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2585 {
2586 if (skb_mac_header_was_set(skb)) {
2587 const unsigned char *old_mac = skb_mac_header(skb);
2588
2589 skb_set_mac_header(skb, -skb->mac_len);
2590 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2591 }
2592 }
2593
skb_checksum_start_offset(const struct sk_buff * skb)2594 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2595 {
2596 return skb->csum_start - skb_headroom(skb);
2597 }
2598
skb_checksum_start(const struct sk_buff * skb)2599 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2600 {
2601 return skb->head + skb->csum_start;
2602 }
2603
skb_transport_offset(const struct sk_buff * skb)2604 static inline int skb_transport_offset(const struct sk_buff *skb)
2605 {
2606 return skb_transport_header(skb) - skb->data;
2607 }
2608
skb_network_header_len(const struct sk_buff * skb)2609 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2610 {
2611 return skb->transport_header - skb->network_header;
2612 }
2613
skb_inner_network_header_len(const struct sk_buff * skb)2614 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2615 {
2616 return skb->inner_transport_header - skb->inner_network_header;
2617 }
2618
skb_network_offset(const struct sk_buff * skb)2619 static inline int skb_network_offset(const struct sk_buff *skb)
2620 {
2621 return skb_network_header(skb) - skb->data;
2622 }
2623
skb_inner_network_offset(const struct sk_buff * skb)2624 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2625 {
2626 return skb_inner_network_header(skb) - skb->data;
2627 }
2628
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)2629 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2630 {
2631 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2632 }
2633
2634 /*
2635 * CPUs often take a performance hit when accessing unaligned memory
2636 * locations. The actual performance hit varies, it can be small if the
2637 * hardware handles it or large if we have to take an exception and fix it
2638 * in software.
2639 *
2640 * Since an ethernet header is 14 bytes network drivers often end up with
2641 * the IP header at an unaligned offset. The IP header can be aligned by
2642 * shifting the start of the packet by 2 bytes. Drivers should do this
2643 * with:
2644 *
2645 * skb_reserve(skb, NET_IP_ALIGN);
2646 *
2647 * The downside to this alignment of the IP header is that the DMA is now
2648 * unaligned. On some architectures the cost of an unaligned DMA is high
2649 * and this cost outweighs the gains made by aligning the IP header.
2650 *
2651 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2652 * to be overridden.
2653 */
2654 #ifndef NET_IP_ALIGN
2655 #define NET_IP_ALIGN 2
2656 #endif
2657
2658 /*
2659 * The networking layer reserves some headroom in skb data (via
2660 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2661 * the header has to grow. In the default case, if the header has to grow
2662 * 32 bytes or less we avoid the reallocation.
2663 *
2664 * Unfortunately this headroom changes the DMA alignment of the resulting
2665 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2666 * on some architectures. An architecture can override this value,
2667 * perhaps setting it to a cacheline in size (since that will maintain
2668 * cacheline alignment of the DMA). It must be a power of 2.
2669 *
2670 * Various parts of the networking layer expect at least 32 bytes of
2671 * headroom, you should not reduce this.
2672 *
2673 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2674 * to reduce average number of cache lines per packet.
2675 * get_rps_cpu() for example only access one 64 bytes aligned block :
2676 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2677 */
2678 #ifndef NET_SKB_PAD
2679 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2680 #endif
2681
2682 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2683
__skb_set_length(struct sk_buff * skb,unsigned int len)2684 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2685 {
2686 if (WARN_ON(skb_is_nonlinear(skb)))
2687 return;
2688 skb->len = len;
2689 skb_set_tail_pointer(skb, len);
2690 }
2691
__skb_trim(struct sk_buff * skb,unsigned int len)2692 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2693 {
2694 __skb_set_length(skb, len);
2695 }
2696
2697 void skb_trim(struct sk_buff *skb, unsigned int len);
2698
__pskb_trim(struct sk_buff * skb,unsigned int len)2699 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2700 {
2701 if (skb->data_len)
2702 return ___pskb_trim(skb, len);
2703 __skb_trim(skb, len);
2704 return 0;
2705 }
2706
pskb_trim(struct sk_buff * skb,unsigned int len)2707 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2708 {
2709 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2710 }
2711
2712 /**
2713 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2714 * @skb: buffer to alter
2715 * @len: new length
2716 *
2717 * This is identical to pskb_trim except that the caller knows that
2718 * the skb is not cloned so we should never get an error due to out-
2719 * of-memory.
2720 */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)2721 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2722 {
2723 int err = pskb_trim(skb, len);
2724 BUG_ON(err);
2725 }
2726
__skb_grow(struct sk_buff * skb,unsigned int len)2727 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2728 {
2729 unsigned int diff = len - skb->len;
2730
2731 if (skb_tailroom(skb) < diff) {
2732 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2733 GFP_ATOMIC);
2734 if (ret)
2735 return ret;
2736 }
2737 __skb_set_length(skb, len);
2738 return 0;
2739 }
2740
2741 /**
2742 * skb_orphan - orphan a buffer
2743 * @skb: buffer to orphan
2744 *
2745 * If a buffer currently has an owner then we call the owner's
2746 * destructor function and make the @skb unowned. The buffer continues
2747 * to exist but is no longer charged to its former owner.
2748 */
skb_orphan(struct sk_buff * skb)2749 static inline void skb_orphan(struct sk_buff *skb)
2750 {
2751 if (skb->destructor) {
2752 skb->destructor(skb);
2753 skb->destructor = NULL;
2754 skb->sk = NULL;
2755 } else {
2756 BUG_ON(skb->sk);
2757 }
2758 }
2759
2760 /**
2761 * skb_orphan_frags - orphan the frags contained in a buffer
2762 * @skb: buffer to orphan frags from
2763 * @gfp_mask: allocation mask for replacement pages
2764 *
2765 * For each frag in the SKB which needs a destructor (i.e. has an
2766 * owner) create a copy of that frag and release the original
2767 * page by calling the destructor.
2768 */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)2769 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2770 {
2771 if (likely(!skb_zcopy(skb)))
2772 return 0;
2773 if (!skb_zcopy_is_nouarg(skb) &&
2774 skb_uarg(skb)->callback == sock_zerocopy_callback)
2775 return 0;
2776 return skb_copy_ubufs(skb, gfp_mask);
2777 }
2778
2779 /* 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)2780 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2781 {
2782 if (likely(!skb_zcopy(skb)))
2783 return 0;
2784 return skb_copy_ubufs(skb, gfp_mask);
2785 }
2786
2787 /**
2788 * __skb_queue_purge - empty a list
2789 * @list: list to empty
2790 *
2791 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2792 * the list and one reference dropped. This function does not take the
2793 * list lock and the caller must hold the relevant locks to use it.
2794 */
__skb_queue_purge(struct sk_buff_head * list)2795 static inline void __skb_queue_purge(struct sk_buff_head *list)
2796 {
2797 struct sk_buff *skb;
2798 while ((skb = __skb_dequeue(list)) != NULL)
2799 kfree_skb(skb);
2800 }
2801 void skb_queue_purge(struct sk_buff_head *list);
2802
2803 unsigned int skb_rbtree_purge(struct rb_root *root);
2804
2805 void *netdev_alloc_frag(unsigned int fragsz);
2806
2807 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2808 gfp_t gfp_mask);
2809
2810 /**
2811 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2812 * @dev: network device to receive on
2813 * @length: length to allocate
2814 *
2815 * Allocate a new &sk_buff and assign it a usage count of one. The
2816 * buffer has unspecified headroom built in. Users should allocate
2817 * the headroom they think they need without accounting for the
2818 * built in space. The built in space is used for optimisations.
2819 *
2820 * %NULL is returned if there is no free memory. Although this function
2821 * allocates memory it can be called from an interrupt.
2822 */
netdev_alloc_skb(struct net_device * dev,unsigned int length)2823 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2824 unsigned int length)
2825 {
2826 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2827 }
2828
2829 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)2830 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2831 gfp_t gfp_mask)
2832 {
2833 return __netdev_alloc_skb(NULL, length, gfp_mask);
2834 }
2835
2836 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)2837 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2838 {
2839 return netdev_alloc_skb(NULL, length);
2840 }
2841
2842
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)2843 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2844 unsigned int length, gfp_t gfp)
2845 {
2846 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2847
2848 if (NET_IP_ALIGN && skb)
2849 skb_reserve(skb, NET_IP_ALIGN);
2850 return skb;
2851 }
2852
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)2853 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2854 unsigned int length)
2855 {
2856 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2857 }
2858
skb_free_frag(void * addr)2859 static inline void skb_free_frag(void *addr)
2860 {
2861 page_frag_free(addr);
2862 }
2863
2864 void *napi_alloc_frag(unsigned int fragsz);
2865 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2866 unsigned int length, gfp_t gfp_mask);
napi_alloc_skb(struct napi_struct * napi,unsigned int length)2867 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2868 unsigned int length)
2869 {
2870 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2871 }
2872 void napi_consume_skb(struct sk_buff *skb, int budget);
2873
2874 void __kfree_skb_flush(void);
2875 void __kfree_skb_defer(struct sk_buff *skb);
2876
2877 /**
2878 * __dev_alloc_pages - allocate page for network Rx
2879 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2880 * @order: size of the allocation
2881 *
2882 * Allocate a new page.
2883 *
2884 * %NULL is returned if there is no free memory.
2885 */
__dev_alloc_pages(gfp_t gfp_mask,unsigned int order)2886 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2887 unsigned int order)
2888 {
2889 /* This piece of code contains several assumptions.
2890 * 1. This is for device Rx, therefor a cold page is preferred.
2891 * 2. The expectation is the user wants a compound page.
2892 * 3. If requesting a order 0 page it will not be compound
2893 * due to the check to see if order has a value in prep_new_page
2894 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2895 * code in gfp_to_alloc_flags that should be enforcing this.
2896 */
2897 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2898
2899 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2900 }
2901
dev_alloc_pages(unsigned int order)2902 static inline struct page *dev_alloc_pages(unsigned int order)
2903 {
2904 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2905 }
2906
2907 /**
2908 * __dev_alloc_page - allocate a page for network Rx
2909 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2910 *
2911 * Allocate a new page.
2912 *
2913 * %NULL is returned if there is no free memory.
2914 */
__dev_alloc_page(gfp_t gfp_mask)2915 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2916 {
2917 return __dev_alloc_pages(gfp_mask, 0);
2918 }
2919
dev_alloc_page(void)2920 static inline struct page *dev_alloc_page(void)
2921 {
2922 return dev_alloc_pages(0);
2923 }
2924
2925 /**
2926 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2927 * @page: The page that was allocated from skb_alloc_page
2928 * @skb: The skb that may need pfmemalloc set
2929 */
skb_propagate_pfmemalloc(struct page * page,struct sk_buff * skb)2930 static inline void skb_propagate_pfmemalloc(struct page *page,
2931 struct sk_buff *skb)
2932 {
2933 if (page_is_pfmemalloc(page))
2934 skb->pfmemalloc = true;
2935 }
2936
2937 /**
2938 * skb_frag_off() - Returns the offset of a skb fragment
2939 * @frag: the paged fragment
2940 */
skb_frag_off(const skb_frag_t * frag)2941 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2942 {
2943 return frag->bv_offset;
2944 }
2945
2946 /**
2947 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2948 * @frag: skb fragment
2949 * @delta: value to add
2950 */
skb_frag_off_add(skb_frag_t * frag,int delta)2951 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2952 {
2953 frag->bv_offset += delta;
2954 }
2955
2956 /**
2957 * skb_frag_off_set() - Sets the offset of a skb fragment
2958 * @frag: skb fragment
2959 * @offset: offset of fragment
2960 */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)2961 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2962 {
2963 frag->bv_offset = offset;
2964 }
2965
2966 /**
2967 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2968 * @fragto: skb fragment where offset is set
2969 * @fragfrom: skb fragment offset is copied from
2970 */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)2971 static inline void skb_frag_off_copy(skb_frag_t *fragto,
2972 const skb_frag_t *fragfrom)
2973 {
2974 fragto->bv_offset = fragfrom->bv_offset;
2975 }
2976
2977 /**
2978 * skb_frag_page - retrieve the page referred to by a paged fragment
2979 * @frag: the paged fragment
2980 *
2981 * Returns the &struct page associated with @frag.
2982 */
skb_frag_page(const skb_frag_t * frag)2983 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2984 {
2985 return frag->bv_page;
2986 }
2987
2988 /**
2989 * __skb_frag_ref - take an addition reference on a paged fragment.
2990 * @frag: the paged fragment
2991 *
2992 * Takes an additional reference on the paged fragment @frag.
2993 */
__skb_frag_ref(skb_frag_t * frag)2994 static inline void __skb_frag_ref(skb_frag_t *frag)
2995 {
2996 get_page(skb_frag_page(frag));
2997 }
2998
2999 /**
3000 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3001 * @skb: the buffer
3002 * @f: the fragment offset.
3003 *
3004 * Takes an additional reference on the @f'th paged fragment of @skb.
3005 */
skb_frag_ref(struct sk_buff * skb,int f)3006 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3007 {
3008 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3009 }
3010
3011 /**
3012 * __skb_frag_unref - release a reference on a paged fragment.
3013 * @frag: the paged fragment
3014 *
3015 * Releases a reference on the paged fragment @frag.
3016 */
__skb_frag_unref(skb_frag_t * frag)3017 static inline void __skb_frag_unref(skb_frag_t *frag)
3018 {
3019 put_page(skb_frag_page(frag));
3020 }
3021
3022 /**
3023 * skb_frag_unref - release a reference on a paged fragment of an skb.
3024 * @skb: the buffer
3025 * @f: the fragment offset
3026 *
3027 * Releases a reference on the @f'th paged fragment of @skb.
3028 */
skb_frag_unref(struct sk_buff * skb,int f)3029 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3030 {
3031 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3032 }
3033
3034 /**
3035 * skb_frag_address - gets the address of the data contained in a paged fragment
3036 * @frag: the paged fragment buffer
3037 *
3038 * Returns the address of the data within @frag. The page must already
3039 * be mapped.
3040 */
skb_frag_address(const skb_frag_t * frag)3041 static inline void *skb_frag_address(const skb_frag_t *frag)
3042 {
3043 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3044 }
3045
3046 /**
3047 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3048 * @frag: the paged fragment buffer
3049 *
3050 * Returns the address of the data within @frag. Checks that the page
3051 * is mapped and returns %NULL otherwise.
3052 */
skb_frag_address_safe(const skb_frag_t * frag)3053 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3054 {
3055 void *ptr = page_address(skb_frag_page(frag));
3056 if (unlikely(!ptr))
3057 return NULL;
3058
3059 return ptr + skb_frag_off(frag);
3060 }
3061
3062 /**
3063 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3064 * @fragto: skb fragment where page is set
3065 * @fragfrom: skb fragment page is copied from
3066 */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3067 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3068 const skb_frag_t *fragfrom)
3069 {
3070 fragto->bv_page = fragfrom->bv_page;
3071 }
3072
3073 /**
3074 * __skb_frag_set_page - sets the page contained in a paged fragment
3075 * @frag: the paged fragment
3076 * @page: the page to set
3077 *
3078 * Sets the fragment @frag to contain @page.
3079 */
__skb_frag_set_page(skb_frag_t * frag,struct page * page)3080 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3081 {
3082 frag->bv_page = page;
3083 }
3084
3085 /**
3086 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3087 * @skb: the buffer
3088 * @f: the fragment offset
3089 * @page: the page to set
3090 *
3091 * Sets the @f'th fragment of @skb to contain @page.
3092 */
skb_frag_set_page(struct sk_buff * skb,int f,struct page * page)3093 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3094 struct page *page)
3095 {
3096 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3097 }
3098
3099 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3100
3101 /**
3102 * skb_frag_dma_map - maps a paged fragment via the DMA API
3103 * @dev: the device to map the fragment to
3104 * @frag: the paged fragment to map
3105 * @offset: the offset within the fragment (starting at the
3106 * fragment's own offset)
3107 * @size: the number of bytes to map
3108 * @dir: the direction of the mapping (``PCI_DMA_*``)
3109 *
3110 * Maps the page associated with @frag to @device.
3111 */
skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3112 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3113 const skb_frag_t *frag,
3114 size_t offset, size_t size,
3115 enum dma_data_direction dir)
3116 {
3117 return dma_map_page(dev, skb_frag_page(frag),
3118 skb_frag_off(frag) + offset, size, dir);
3119 }
3120
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3121 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3122 gfp_t gfp_mask)
3123 {
3124 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3125 }
3126
3127
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3128 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3129 gfp_t gfp_mask)
3130 {
3131 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3132 }
3133
3134
3135 /**
3136 * skb_clone_writable - is the header of a clone writable
3137 * @skb: buffer to check
3138 * @len: length up to which to write
3139 *
3140 * Returns true if modifying the header part of the cloned buffer
3141 * does not requires the data to be copied.
3142 */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3143 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3144 {
3145 return !skb_header_cloned(skb) &&
3146 skb_headroom(skb) + len <= skb->hdr_len;
3147 }
3148
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3149 static inline int skb_try_make_writable(struct sk_buff *skb,
3150 unsigned int write_len)
3151 {
3152 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3153 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3154 }
3155
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3156 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3157 int cloned)
3158 {
3159 int delta = 0;
3160
3161 if (headroom > skb_headroom(skb))
3162 delta = headroom - skb_headroom(skb);
3163
3164 if (delta || cloned)
3165 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3166 GFP_ATOMIC);
3167 return 0;
3168 }
3169
3170 /**
3171 * skb_cow - copy header of skb when it is required
3172 * @skb: buffer to cow
3173 * @headroom: needed headroom
3174 *
3175 * If the skb passed lacks sufficient headroom or its data part
3176 * is shared, data is reallocated. If reallocation fails, an error
3177 * is returned and original skb is not changed.
3178 *
3179 * The result is skb with writable area skb->head...skb->tail
3180 * and at least @headroom of space at head.
3181 */
skb_cow(struct sk_buff * skb,unsigned int headroom)3182 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3183 {
3184 return __skb_cow(skb, headroom, skb_cloned(skb));
3185 }
3186
3187 /**
3188 * skb_cow_head - skb_cow but only making the head writable
3189 * @skb: buffer to cow
3190 * @headroom: needed headroom
3191 *
3192 * This function is identical to skb_cow except that we replace the
3193 * skb_cloned check by skb_header_cloned. It should be used when
3194 * you only need to push on some header and do not need to modify
3195 * the data.
3196 */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3197 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3198 {
3199 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3200 }
3201
3202 /**
3203 * skb_padto - pad an skbuff up to a minimal size
3204 * @skb: buffer to pad
3205 * @len: minimal length
3206 *
3207 * Pads up a buffer to ensure the trailing bytes exist and are
3208 * blanked. If the buffer already contains sufficient data it
3209 * is untouched. Otherwise it is extended. Returns zero on
3210 * success. The skb is freed on error.
3211 */
skb_padto(struct sk_buff * skb,unsigned int len)3212 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3213 {
3214 unsigned int size = skb->len;
3215 if (likely(size >= len))
3216 return 0;
3217 return skb_pad(skb, len - size);
3218 }
3219
3220 /**
3221 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3222 * @skb: buffer to pad
3223 * @len: minimal length
3224 * @free_on_error: free buffer on error
3225 *
3226 * Pads up a buffer to ensure the trailing bytes exist and are
3227 * blanked. If the buffer already contains sufficient data it
3228 * is untouched. Otherwise it is extended. Returns zero on
3229 * success. The skb is freed on error if @free_on_error is true.
3230 */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3231 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3232 unsigned int len,
3233 bool free_on_error)
3234 {
3235 unsigned int size = skb->len;
3236
3237 if (unlikely(size < len)) {
3238 len -= size;
3239 if (__skb_pad(skb, len, free_on_error))
3240 return -ENOMEM;
3241 __skb_put(skb, len);
3242 }
3243 return 0;
3244 }
3245
3246 /**
3247 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3248 * @skb: buffer to pad
3249 * @len: minimal length
3250 *
3251 * Pads up a buffer to ensure the trailing bytes exist and are
3252 * blanked. If the buffer already contains sufficient data it
3253 * is untouched. Otherwise it is extended. Returns zero on
3254 * success. The skb is freed on error.
3255 */
skb_put_padto(struct sk_buff * skb,unsigned int len)3256 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3257 {
3258 return __skb_put_padto(skb, len, true);
3259 }
3260
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3261 static inline int skb_add_data(struct sk_buff *skb,
3262 struct iov_iter *from, int copy)
3263 {
3264 const int off = skb->len;
3265
3266 if (skb->ip_summed == CHECKSUM_NONE) {
3267 __wsum csum = 0;
3268 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3269 &csum, from)) {
3270 skb->csum = csum_block_add(skb->csum, csum, off);
3271 return 0;
3272 }
3273 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3274 return 0;
3275
3276 __skb_trim(skb, off);
3277 return -EFAULT;
3278 }
3279
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3280 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3281 const struct page *page, int off)
3282 {
3283 if (skb_zcopy(skb))
3284 return false;
3285 if (i) {
3286 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3287
3288 return page == skb_frag_page(frag) &&
3289 off == skb_frag_off(frag) + skb_frag_size(frag);
3290 }
3291 return false;
3292 }
3293
__skb_linearize(struct sk_buff * skb)3294 static inline int __skb_linearize(struct sk_buff *skb)
3295 {
3296 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3297 }
3298
3299 /**
3300 * skb_linearize - convert paged skb to linear one
3301 * @skb: buffer to linarize
3302 *
3303 * If there is no free memory -ENOMEM is returned, otherwise zero
3304 * is returned and the old skb data released.
3305 */
skb_linearize(struct sk_buff * skb)3306 static inline int skb_linearize(struct sk_buff *skb)
3307 {
3308 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3309 }
3310
3311 /**
3312 * skb_has_shared_frag - can any frag be overwritten
3313 * @skb: buffer to test
3314 *
3315 * Return true if the skb has at least one frag that might be modified
3316 * by an external entity (as in vmsplice()/sendfile())
3317 */
skb_has_shared_frag(const struct sk_buff * skb)3318 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3319 {
3320 return skb_is_nonlinear(skb) &&
3321 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3322 }
3323
3324 /**
3325 * skb_linearize_cow - make sure skb is linear and writable
3326 * @skb: buffer to process
3327 *
3328 * If there is no free memory -ENOMEM is returned, otherwise zero
3329 * is returned and the old skb data released.
3330 */
skb_linearize_cow(struct sk_buff * skb)3331 static inline int skb_linearize_cow(struct sk_buff *skb)
3332 {
3333 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3334 __skb_linearize(skb) : 0;
3335 }
3336
3337 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3338 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3339 unsigned int off)
3340 {
3341 if (skb->ip_summed == CHECKSUM_COMPLETE)
3342 skb->csum = csum_block_sub(skb->csum,
3343 csum_partial(start, len, 0), off);
3344 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3345 skb_checksum_start_offset(skb) < 0)
3346 skb->ip_summed = CHECKSUM_NONE;
3347 }
3348
3349 /**
3350 * skb_postpull_rcsum - update checksum for received skb after pull
3351 * @skb: buffer to update
3352 * @start: start of data before pull
3353 * @len: length of data pulled
3354 *
3355 * After doing a pull on a received packet, you need to call this to
3356 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3357 * CHECKSUM_NONE so that it can be recomputed from scratch.
3358 */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3359 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3360 const void *start, unsigned int len)
3361 {
3362 __skb_postpull_rcsum(skb, start, len, 0);
3363 }
3364
3365 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3366 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3367 unsigned int off)
3368 {
3369 if (skb->ip_summed == CHECKSUM_COMPLETE)
3370 skb->csum = csum_block_add(skb->csum,
3371 csum_partial(start, len, 0), off);
3372 }
3373
3374 /**
3375 * skb_postpush_rcsum - update checksum for received skb after push
3376 * @skb: buffer to update
3377 * @start: start of data after push
3378 * @len: length of data pushed
3379 *
3380 * After doing a push on a received packet, you need to call this to
3381 * update the CHECKSUM_COMPLETE checksum.
3382 */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3383 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3384 const void *start, unsigned int len)
3385 {
3386 __skb_postpush_rcsum(skb, start, len, 0);
3387 }
3388
3389 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3390
3391 /**
3392 * skb_push_rcsum - push skb and update receive checksum
3393 * @skb: buffer to update
3394 * @len: length of data pulled
3395 *
3396 * This function performs an skb_push on the packet and updates
3397 * the CHECKSUM_COMPLETE checksum. It should be used on
3398 * receive path processing instead of skb_push unless you know
3399 * that the checksum difference is zero (e.g., a valid IP header)
3400 * or you are setting ip_summed to CHECKSUM_NONE.
3401 */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3402 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3403 {
3404 skb_push(skb, len);
3405 skb_postpush_rcsum(skb, skb->data, len);
3406 return skb->data;
3407 }
3408
3409 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3410 /**
3411 * pskb_trim_rcsum - trim received skb and update checksum
3412 * @skb: buffer to trim
3413 * @len: new length
3414 *
3415 * This is exactly the same as pskb_trim except that it ensures the
3416 * checksum of received packets are still valid after the operation.
3417 * It can change skb pointers.
3418 */
3419
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)3420 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3421 {
3422 if (likely(len >= skb->len))
3423 return 0;
3424 return pskb_trim_rcsum_slow(skb, len);
3425 }
3426
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)3427 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3428 {
3429 if (skb->ip_summed == CHECKSUM_COMPLETE)
3430 skb->ip_summed = CHECKSUM_NONE;
3431 __skb_trim(skb, len);
3432 return 0;
3433 }
3434
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)3435 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3436 {
3437 if (skb->ip_summed == CHECKSUM_COMPLETE)
3438 skb->ip_summed = CHECKSUM_NONE;
3439 return __skb_grow(skb, len);
3440 }
3441
3442 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3443 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3444 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3445 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3446 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3447
3448 #define skb_queue_walk(queue, skb) \
3449 for (skb = (queue)->next; \
3450 skb != (struct sk_buff *)(queue); \
3451 skb = skb->next)
3452
3453 #define skb_queue_walk_safe(queue, skb, tmp) \
3454 for (skb = (queue)->next, tmp = skb->next; \
3455 skb != (struct sk_buff *)(queue); \
3456 skb = tmp, tmp = skb->next)
3457
3458 #define skb_queue_walk_from(queue, skb) \
3459 for (; skb != (struct sk_buff *)(queue); \
3460 skb = skb->next)
3461
3462 #define skb_rbtree_walk(skb, root) \
3463 for (skb = skb_rb_first(root); skb != NULL; \
3464 skb = skb_rb_next(skb))
3465
3466 #define skb_rbtree_walk_from(skb) \
3467 for (; skb != NULL; \
3468 skb = skb_rb_next(skb))
3469
3470 #define skb_rbtree_walk_from_safe(skb, tmp) \
3471 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3472 skb = tmp)
3473
3474 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3475 for (tmp = skb->next; \
3476 skb != (struct sk_buff *)(queue); \
3477 skb = tmp, tmp = skb->next)
3478
3479 #define skb_queue_reverse_walk(queue, skb) \
3480 for (skb = (queue)->prev; \
3481 skb != (struct sk_buff *)(queue); \
3482 skb = skb->prev)
3483
3484 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3485 for (skb = (queue)->prev, tmp = skb->prev; \
3486 skb != (struct sk_buff *)(queue); \
3487 skb = tmp, tmp = skb->prev)
3488
3489 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3490 for (tmp = skb->prev; \
3491 skb != (struct sk_buff *)(queue); \
3492 skb = tmp, tmp = skb->prev)
3493
skb_has_frag_list(const struct sk_buff * skb)3494 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3495 {
3496 return skb_shinfo(skb)->frag_list != NULL;
3497 }
3498
skb_frag_list_init(struct sk_buff * skb)3499 static inline void skb_frag_list_init(struct sk_buff *skb)
3500 {
3501 skb_shinfo(skb)->frag_list = NULL;
3502 }
3503
3504 #define skb_walk_frags(skb, iter) \
3505 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3506
3507
3508 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3509 int *err, long *timeo_p,
3510 const struct sk_buff *skb);
3511 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3512 struct sk_buff_head *queue,
3513 unsigned int flags,
3514 int *off, int *err,
3515 struct sk_buff **last);
3516 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3517 struct sk_buff_head *queue,
3518 unsigned int flags, int *off, int *err,
3519 struct sk_buff **last);
3520 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3521 struct sk_buff_head *sk_queue,
3522 unsigned int flags, int *off, int *err);
3523 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3524 int *err);
3525 __poll_t datagram_poll(struct file *file, struct socket *sock,
3526 struct poll_table_struct *wait);
3527 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3528 struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)3529 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3530 struct msghdr *msg, int size)
3531 {
3532 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3533 }
3534 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3535 struct msghdr *msg);
3536 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3537 struct iov_iter *to, int len,
3538 struct ahash_request *hash);
3539 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3540 struct iov_iter *from, int len);
3541 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3542 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3543 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)3544 static inline void skb_free_datagram_locked(struct sock *sk,
3545 struct sk_buff *skb)
3546 {
3547 __skb_free_datagram_locked(sk, skb, 0);
3548 }
3549 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3550 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3551 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3552 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3553 int len);
3554 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3555 struct pipe_inode_info *pipe, unsigned int len,
3556 unsigned int flags);
3557 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3558 int len);
3559 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3560 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3561 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3562 int len, int hlen);
3563 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3564 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3565 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3566 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3567 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3568 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3569 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3570 unsigned int offset);
3571 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3572 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3573 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3574 int skb_vlan_pop(struct sk_buff *skb);
3575 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3576 int skb_eth_pop(struct sk_buff *skb);
3577 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3578 const unsigned char *src);
3579 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3580 int mac_len, bool ethernet);
3581 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3582 bool ethernet);
3583 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3584 int skb_mpls_dec_ttl(struct sk_buff *skb);
3585 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3586 gfp_t gfp);
3587
memcpy_from_msg(void * data,struct msghdr * msg,int len)3588 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3589 {
3590 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3591 }
3592
memcpy_to_msg(struct msghdr * msg,void * data,int len)3593 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3594 {
3595 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3596 }
3597
3598 struct skb_checksum_ops {
3599 __wsum (*update)(const void *mem, int len, __wsum wsum);
3600 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3601 };
3602
3603 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3604
3605 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3606 __wsum csum, const struct skb_checksum_ops *ops);
3607 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3608 __wsum csum);
3609
3610 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * data,int hlen,void * buffer)3611 __skb_header_pointer(const struct sk_buff *skb, int offset,
3612 int len, void *data, int hlen, void *buffer)
3613 {
3614 if (hlen - offset >= len)
3615 return data + offset;
3616
3617 if (!skb ||
3618 skb_copy_bits(skb, offset, buffer, len) < 0)
3619 return NULL;
3620
3621 return buffer;
3622 }
3623
3624 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)3625 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3626 {
3627 return __skb_header_pointer(skb, offset, len, skb->data,
3628 skb_headlen(skb), buffer);
3629 }
3630
3631 /**
3632 * skb_needs_linearize - check if we need to linearize a given skb
3633 * depending on the given device features.
3634 * @skb: socket buffer to check
3635 * @features: net device features
3636 *
3637 * Returns true if either:
3638 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3639 * 2. skb is fragmented and the device does not support SG.
3640 */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)3641 static inline bool skb_needs_linearize(struct sk_buff *skb,
3642 netdev_features_t features)
3643 {
3644 return skb_is_nonlinear(skb) &&
3645 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3646 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3647 }
3648
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)3649 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3650 void *to,
3651 const unsigned int len)
3652 {
3653 memcpy(to, skb->data, len);
3654 }
3655
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)3656 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3657 const int offset, void *to,
3658 const unsigned int len)
3659 {
3660 memcpy(to, skb->data + offset, len);
3661 }
3662
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)3663 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3664 const void *from,
3665 const unsigned int len)
3666 {
3667 memcpy(skb->data, from, len);
3668 }
3669
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)3670 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3671 const int offset,
3672 const void *from,
3673 const unsigned int len)
3674 {
3675 memcpy(skb->data + offset, from, len);
3676 }
3677
3678 void skb_init(void);
3679
skb_get_ktime(const struct sk_buff * skb)3680 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3681 {
3682 return skb->tstamp;
3683 }
3684
3685 /**
3686 * skb_get_timestamp - get timestamp from a skb
3687 * @skb: skb to get stamp from
3688 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3689 *
3690 * Timestamps are stored in the skb as offsets to a base timestamp.
3691 * This function converts the offset back to a struct timeval and stores
3692 * it in stamp.
3693 */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)3694 static inline void skb_get_timestamp(const struct sk_buff *skb,
3695 struct __kernel_old_timeval *stamp)
3696 {
3697 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3698 }
3699
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)3700 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3701 struct __kernel_sock_timeval *stamp)
3702 {
3703 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3704
3705 stamp->tv_sec = ts.tv_sec;
3706 stamp->tv_usec = ts.tv_nsec / 1000;
3707 }
3708
skb_get_timestampns(const struct sk_buff * skb,struct __kernel_old_timespec * stamp)3709 static inline void skb_get_timestampns(const struct sk_buff *skb,
3710 struct __kernel_old_timespec *stamp)
3711 {
3712 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3713
3714 stamp->tv_sec = ts.tv_sec;
3715 stamp->tv_nsec = ts.tv_nsec;
3716 }
3717
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)3718 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3719 struct __kernel_timespec *stamp)
3720 {
3721 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3722
3723 stamp->tv_sec = ts.tv_sec;
3724 stamp->tv_nsec = ts.tv_nsec;
3725 }
3726
__net_timestamp(struct sk_buff * skb)3727 static inline void __net_timestamp(struct sk_buff *skb)
3728 {
3729 skb->tstamp = ktime_get_real();
3730 }
3731
net_timedelta(ktime_t t)3732 static inline ktime_t net_timedelta(ktime_t t)
3733 {
3734 return ktime_sub(ktime_get_real(), t);
3735 }
3736
net_invalid_timestamp(void)3737 static inline ktime_t net_invalid_timestamp(void)
3738 {
3739 return 0;
3740 }
3741
skb_metadata_len(const struct sk_buff * skb)3742 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3743 {
3744 return skb_shinfo(skb)->meta_len;
3745 }
3746
skb_metadata_end(const struct sk_buff * skb)3747 static inline void *skb_metadata_end(const struct sk_buff *skb)
3748 {
3749 return skb_mac_header(skb);
3750 }
3751
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)3752 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3753 const struct sk_buff *skb_b,
3754 u8 meta_len)
3755 {
3756 const void *a = skb_metadata_end(skb_a);
3757 const void *b = skb_metadata_end(skb_b);
3758 /* Using more efficient varaiant than plain call to memcmp(). */
3759 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3760 u64 diffs = 0;
3761
3762 switch (meta_len) {
3763 #define __it(x, op) (x -= sizeof(u##op))
3764 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3765 case 32: diffs |= __it_diff(a, b, 64);
3766 fallthrough;
3767 case 24: diffs |= __it_diff(a, b, 64);
3768 fallthrough;
3769 case 16: diffs |= __it_diff(a, b, 64);
3770 fallthrough;
3771 case 8: diffs |= __it_diff(a, b, 64);
3772 break;
3773 case 28: diffs |= __it_diff(a, b, 64);
3774 fallthrough;
3775 case 20: diffs |= __it_diff(a, b, 64);
3776 fallthrough;
3777 case 12: diffs |= __it_diff(a, b, 64);
3778 fallthrough;
3779 case 4: diffs |= __it_diff(a, b, 32);
3780 break;
3781 }
3782 return diffs;
3783 #else
3784 return memcmp(a - meta_len, b - meta_len, meta_len);
3785 #endif
3786 }
3787
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)3788 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3789 const struct sk_buff *skb_b)
3790 {
3791 u8 len_a = skb_metadata_len(skb_a);
3792 u8 len_b = skb_metadata_len(skb_b);
3793
3794 if (!(len_a | len_b))
3795 return false;
3796
3797 return len_a != len_b ?
3798 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3799 }
3800
skb_metadata_set(struct sk_buff * skb,u8 meta_len)3801 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3802 {
3803 skb_shinfo(skb)->meta_len = meta_len;
3804 }
3805
skb_metadata_clear(struct sk_buff * skb)3806 static inline void skb_metadata_clear(struct sk_buff *skb)
3807 {
3808 skb_metadata_set(skb, 0);
3809 }
3810
3811 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3812
3813 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3814
3815 void skb_clone_tx_timestamp(struct sk_buff *skb);
3816 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3817
3818 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3819
skb_clone_tx_timestamp(struct sk_buff * skb)3820 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3821 {
3822 }
3823
skb_defer_rx_timestamp(struct sk_buff * skb)3824 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3825 {
3826 return false;
3827 }
3828
3829 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3830
3831 /**
3832 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3833 *
3834 * PHY drivers may accept clones of transmitted packets for
3835 * timestamping via their phy_driver.txtstamp method. These drivers
3836 * must call this function to return the skb back to the stack with a
3837 * timestamp.
3838 *
3839 * @skb: clone of the original outgoing packet
3840 * @hwtstamps: hardware time stamps
3841 *
3842 */
3843 void skb_complete_tx_timestamp(struct sk_buff *skb,
3844 struct skb_shared_hwtstamps *hwtstamps);
3845
3846 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3847 struct skb_shared_hwtstamps *hwtstamps,
3848 struct sock *sk, int tstype);
3849
3850 /**
3851 * skb_tstamp_tx - queue clone of skb with send time stamps
3852 * @orig_skb: the original outgoing packet
3853 * @hwtstamps: hardware time stamps, may be NULL if not available
3854 *
3855 * If the skb has a socket associated, then this function clones the
3856 * skb (thus sharing the actual data and optional structures), stores
3857 * the optional hardware time stamping information (if non NULL) or
3858 * generates a software time stamp (otherwise), then queues the clone
3859 * to the error queue of the socket. Errors are silently ignored.
3860 */
3861 void skb_tstamp_tx(struct sk_buff *orig_skb,
3862 struct skb_shared_hwtstamps *hwtstamps);
3863
3864 /**
3865 * skb_tx_timestamp() - Driver hook for transmit timestamping
3866 *
3867 * Ethernet MAC Drivers should call this function in their hard_xmit()
3868 * function immediately before giving the sk_buff to the MAC hardware.
3869 *
3870 * Specifically, one should make absolutely sure that this function is
3871 * called before TX completion of this packet can trigger. Otherwise
3872 * the packet could potentially already be freed.
3873 *
3874 * @skb: A socket buffer.
3875 */
skb_tx_timestamp(struct sk_buff * skb)3876 static inline void skb_tx_timestamp(struct sk_buff *skb)
3877 {
3878 skb_clone_tx_timestamp(skb);
3879 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3880 skb_tstamp_tx(skb, NULL);
3881 }
3882
3883 /**
3884 * skb_complete_wifi_ack - deliver skb with wifi status
3885 *
3886 * @skb: the original outgoing packet
3887 * @acked: ack status
3888 *
3889 */
3890 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3891
3892 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3893 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3894
skb_csum_unnecessary(const struct sk_buff * skb)3895 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3896 {
3897 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3898 skb->csum_valid ||
3899 (skb->ip_summed == CHECKSUM_PARTIAL &&
3900 skb_checksum_start_offset(skb) >= 0));
3901 }
3902
3903 /**
3904 * skb_checksum_complete - Calculate checksum of an entire packet
3905 * @skb: packet to process
3906 *
3907 * This function calculates the checksum over the entire packet plus
3908 * the value of skb->csum. The latter can be used to supply the
3909 * checksum of a pseudo header as used by TCP/UDP. It returns the
3910 * checksum.
3911 *
3912 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3913 * this function can be used to verify that checksum on received
3914 * packets. In that case the function should return zero if the
3915 * checksum is correct. In particular, this function will return zero
3916 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3917 * hardware has already verified the correctness of the checksum.
3918 */
skb_checksum_complete(struct sk_buff * skb)3919 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3920 {
3921 return skb_csum_unnecessary(skb) ?
3922 0 : __skb_checksum_complete(skb);
3923 }
3924
__skb_decr_checksum_unnecessary(struct sk_buff * skb)3925 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3926 {
3927 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3928 if (skb->csum_level == 0)
3929 skb->ip_summed = CHECKSUM_NONE;
3930 else
3931 skb->csum_level--;
3932 }
3933 }
3934
__skb_incr_checksum_unnecessary(struct sk_buff * skb)3935 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3936 {
3937 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3938 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3939 skb->csum_level++;
3940 } else if (skb->ip_summed == CHECKSUM_NONE) {
3941 skb->ip_summed = CHECKSUM_UNNECESSARY;
3942 skb->csum_level = 0;
3943 }
3944 }
3945
__skb_reset_checksum_unnecessary(struct sk_buff * skb)3946 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
3947 {
3948 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3949 skb->ip_summed = CHECKSUM_NONE;
3950 skb->csum_level = 0;
3951 }
3952 }
3953
3954 /* Check if we need to perform checksum complete validation.
3955 *
3956 * Returns true if checksum complete is needed, false otherwise
3957 * (either checksum is unnecessary or zero checksum is allowed).
3958 */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)3959 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3960 bool zero_okay,
3961 __sum16 check)
3962 {
3963 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3964 skb->csum_valid = 1;
3965 __skb_decr_checksum_unnecessary(skb);
3966 return false;
3967 }
3968
3969 return true;
3970 }
3971
3972 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
3973 * in checksum_init.
3974 */
3975 #define CHECKSUM_BREAK 76
3976
3977 /* Unset checksum-complete
3978 *
3979 * Unset checksum complete can be done when packet is being modified
3980 * (uncompressed for instance) and checksum-complete value is
3981 * invalidated.
3982 */
skb_checksum_complete_unset(struct sk_buff * skb)3983 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3984 {
3985 if (skb->ip_summed == CHECKSUM_COMPLETE)
3986 skb->ip_summed = CHECKSUM_NONE;
3987 }
3988
3989 /* Validate (init) checksum based on checksum complete.
3990 *
3991 * Return values:
3992 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3993 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3994 * checksum is stored in skb->csum for use in __skb_checksum_complete
3995 * non-zero: value of invalid checksum
3996 *
3997 */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)3998 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3999 bool complete,
4000 __wsum psum)
4001 {
4002 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4003 if (!csum_fold(csum_add(psum, skb->csum))) {
4004 skb->csum_valid = 1;
4005 return 0;
4006 }
4007 }
4008
4009 skb->csum = psum;
4010
4011 if (complete || skb->len <= CHECKSUM_BREAK) {
4012 __sum16 csum;
4013
4014 csum = __skb_checksum_complete(skb);
4015 skb->csum_valid = !csum;
4016 return csum;
4017 }
4018
4019 return 0;
4020 }
4021
null_compute_pseudo(struct sk_buff * skb,int proto)4022 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4023 {
4024 return 0;
4025 }
4026
4027 /* Perform checksum validate (init). Note that this is a macro since we only
4028 * want to calculate the pseudo header which is an input function if necessary.
4029 * First we try to validate without any computation (checksum unnecessary) and
4030 * then calculate based on checksum complete calling the function to compute
4031 * pseudo header.
4032 *
4033 * Return values:
4034 * 0: checksum is validated or try to in skb_checksum_complete
4035 * non-zero: value of invalid checksum
4036 */
4037 #define __skb_checksum_validate(skb, proto, complete, \
4038 zero_okay, check, compute_pseudo) \
4039 ({ \
4040 __sum16 __ret = 0; \
4041 skb->csum_valid = 0; \
4042 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4043 __ret = __skb_checksum_validate_complete(skb, \
4044 complete, compute_pseudo(skb, proto)); \
4045 __ret; \
4046 })
4047
4048 #define skb_checksum_init(skb, proto, compute_pseudo) \
4049 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4050
4051 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4052 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4053
4054 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4055 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4056
4057 #define skb_checksum_validate_zero_check(skb, proto, check, \
4058 compute_pseudo) \
4059 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4060
4061 #define skb_checksum_simple_validate(skb) \
4062 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4063
__skb_checksum_convert_check(struct sk_buff * skb)4064 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4065 {
4066 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4067 }
4068
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4069 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4070 {
4071 skb->csum = ~pseudo;
4072 skb->ip_summed = CHECKSUM_COMPLETE;
4073 }
4074
4075 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4076 do { \
4077 if (__skb_checksum_convert_check(skb)) \
4078 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4079 } while (0)
4080
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4081 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4082 u16 start, u16 offset)
4083 {
4084 skb->ip_summed = CHECKSUM_PARTIAL;
4085 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4086 skb->csum_offset = offset - start;
4087 }
4088
4089 /* Update skbuf and packet to reflect the remote checksum offload operation.
4090 * When called, ptr indicates the starting point for skb->csum when
4091 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4092 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4093 */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4094 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4095 int start, int offset, bool nopartial)
4096 {
4097 __wsum delta;
4098
4099 if (!nopartial) {
4100 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4101 return;
4102 }
4103
4104 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4105 __skb_checksum_complete(skb);
4106 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4107 }
4108
4109 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4110
4111 /* Adjust skb->csum since we changed the packet */
4112 skb->csum = csum_add(skb->csum, delta);
4113 }
4114
skb_nfct(const struct sk_buff * skb)4115 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4116 {
4117 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4118 return (void *)(skb->_nfct & NFCT_PTRMASK);
4119 #else
4120 return NULL;
4121 #endif
4122 }
4123
skb_get_nfct(const struct sk_buff * skb)4124 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4125 {
4126 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4127 return skb->_nfct;
4128 #else
4129 return 0UL;
4130 #endif
4131 }
4132
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4133 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4134 {
4135 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4136 skb->_nfct = nfct;
4137 #endif
4138 }
4139
4140 #ifdef CONFIG_SKB_EXTENSIONS
4141 enum skb_ext_id {
4142 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4143 SKB_EXT_BRIDGE_NF,
4144 #endif
4145 #ifdef CONFIG_XFRM
4146 SKB_EXT_SEC_PATH,
4147 #endif
4148 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4149 TC_SKB_EXT,
4150 #endif
4151 #if IS_ENABLED(CONFIG_MPTCP)
4152 SKB_EXT_MPTCP,
4153 #endif
4154 SKB_EXT_NUM, /* must be last */
4155 };
4156
4157 /**
4158 * struct skb_ext - sk_buff extensions
4159 * @refcnt: 1 on allocation, deallocated on 0
4160 * @offset: offset to add to @data to obtain extension address
4161 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4162 * @data: start of extension data, variable sized
4163 *
4164 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4165 * to use 'u8' types while allowing up to 2kb worth of extension data.
4166 */
4167 struct skb_ext {
4168 refcount_t refcnt;
4169 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4170 u8 chunks; /* same */
4171 char data[] __aligned(8);
4172 };
4173
4174 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4175 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4176 struct skb_ext *ext);
4177 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4178 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4179 void __skb_ext_put(struct skb_ext *ext);
4180
skb_ext_put(struct sk_buff * skb)4181 static inline void skb_ext_put(struct sk_buff *skb)
4182 {
4183 if (skb->active_extensions)
4184 __skb_ext_put(skb->extensions);
4185 }
4186
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4187 static inline void __skb_ext_copy(struct sk_buff *dst,
4188 const struct sk_buff *src)
4189 {
4190 dst->active_extensions = src->active_extensions;
4191
4192 if (src->active_extensions) {
4193 struct skb_ext *ext = src->extensions;
4194
4195 refcount_inc(&ext->refcnt);
4196 dst->extensions = ext;
4197 }
4198 }
4199
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4200 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4201 {
4202 skb_ext_put(dst);
4203 __skb_ext_copy(dst, src);
4204 }
4205
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4206 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4207 {
4208 return !!ext->offset[i];
4209 }
4210
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4211 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4212 {
4213 return skb->active_extensions & (1 << id);
4214 }
4215
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4216 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4217 {
4218 if (skb_ext_exist(skb, id))
4219 __skb_ext_del(skb, id);
4220 }
4221
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4222 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4223 {
4224 if (skb_ext_exist(skb, id)) {
4225 struct skb_ext *ext = skb->extensions;
4226
4227 return (void *)ext + (ext->offset[id] << 3);
4228 }
4229
4230 return NULL;
4231 }
4232
skb_ext_reset(struct sk_buff * skb)4233 static inline void skb_ext_reset(struct sk_buff *skb)
4234 {
4235 if (unlikely(skb->active_extensions)) {
4236 __skb_ext_put(skb->extensions);
4237 skb->active_extensions = 0;
4238 }
4239 }
4240
skb_has_extensions(struct sk_buff * skb)4241 static inline bool skb_has_extensions(struct sk_buff *skb)
4242 {
4243 return unlikely(skb->active_extensions);
4244 }
4245 #else
skb_ext_put(struct sk_buff * skb)4246 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4247 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4248 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4249 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)4250 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4251 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4252 #endif /* CONFIG_SKB_EXTENSIONS */
4253
nf_reset_ct(struct sk_buff * skb)4254 static inline void nf_reset_ct(struct sk_buff *skb)
4255 {
4256 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4257 nf_conntrack_put(skb_nfct(skb));
4258 skb->_nfct = 0;
4259 #endif
4260 }
4261
nf_reset_trace(struct sk_buff * skb)4262 static inline void nf_reset_trace(struct sk_buff *skb)
4263 {
4264 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4265 skb->nf_trace = 0;
4266 #endif
4267 }
4268
ipvs_reset(struct sk_buff * skb)4269 static inline void ipvs_reset(struct sk_buff *skb)
4270 {
4271 #if IS_ENABLED(CONFIG_IP_VS)
4272 skb->ipvs_property = 0;
4273 #endif
4274 }
4275
4276 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4277 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4278 bool copy)
4279 {
4280 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4281 dst->_nfct = src->_nfct;
4282 nf_conntrack_get(skb_nfct(src));
4283 #endif
4284 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4285 if (copy)
4286 dst->nf_trace = src->nf_trace;
4287 #endif
4288 }
4289
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4290 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4291 {
4292 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4293 nf_conntrack_put(skb_nfct(dst));
4294 #endif
4295 __nf_copy(dst, src, true);
4296 }
4297
4298 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4299 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4300 {
4301 to->secmark = from->secmark;
4302 }
4303
skb_init_secmark(struct sk_buff * skb)4304 static inline void skb_init_secmark(struct sk_buff *skb)
4305 {
4306 skb->secmark = 0;
4307 }
4308 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4309 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4310 { }
4311
skb_init_secmark(struct sk_buff * skb)4312 static inline void skb_init_secmark(struct sk_buff *skb)
4313 { }
4314 #endif
4315
secpath_exists(const struct sk_buff * skb)4316 static inline int secpath_exists(const struct sk_buff *skb)
4317 {
4318 #ifdef CONFIG_XFRM
4319 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4320 #else
4321 return 0;
4322 #endif
4323 }
4324
skb_irq_freeable(const struct sk_buff * skb)4325 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4326 {
4327 return !skb->destructor &&
4328 !secpath_exists(skb) &&
4329 !skb_nfct(skb) &&
4330 !skb->_skb_refdst &&
4331 !skb_has_frag_list(skb);
4332 }
4333
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)4334 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4335 {
4336 skb->queue_mapping = queue_mapping;
4337 }
4338
skb_get_queue_mapping(const struct sk_buff * skb)4339 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4340 {
4341 return skb->queue_mapping;
4342 }
4343
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)4344 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4345 {
4346 to->queue_mapping = from->queue_mapping;
4347 }
4348
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)4349 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4350 {
4351 skb->queue_mapping = rx_queue + 1;
4352 }
4353
skb_get_rx_queue(const struct sk_buff * skb)4354 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4355 {
4356 return skb->queue_mapping - 1;
4357 }
4358
skb_rx_queue_recorded(const struct sk_buff * skb)4359 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4360 {
4361 return skb->queue_mapping != 0;
4362 }
4363
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)4364 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4365 {
4366 skb->dst_pending_confirm = val;
4367 }
4368
skb_get_dst_pending_confirm(const struct sk_buff * skb)4369 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4370 {
4371 return skb->dst_pending_confirm != 0;
4372 }
4373
skb_sec_path(const struct sk_buff * skb)4374 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4375 {
4376 #ifdef CONFIG_XFRM
4377 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4378 #else
4379 return NULL;
4380 #endif
4381 }
4382
4383 /* Keeps track of mac header offset relative to skb->head.
4384 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4385 * For non-tunnel skb it points to skb_mac_header() and for
4386 * tunnel skb it points to outer mac header.
4387 * Keeps track of level of encapsulation of network headers.
4388 */
4389 struct skb_gso_cb {
4390 union {
4391 int mac_offset;
4392 int data_offset;
4393 };
4394 int encap_level;
4395 __wsum csum;
4396 __u16 csum_start;
4397 };
4398 #define SKB_GSO_CB_OFFSET 32
4399 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4400
skb_tnl_header_len(const struct sk_buff * inner_skb)4401 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4402 {
4403 return (skb_mac_header(inner_skb) - inner_skb->head) -
4404 SKB_GSO_CB(inner_skb)->mac_offset;
4405 }
4406
gso_pskb_expand_head(struct sk_buff * skb,int extra)4407 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4408 {
4409 int new_headroom, headroom;
4410 int ret;
4411
4412 headroom = skb_headroom(skb);
4413 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4414 if (ret)
4415 return ret;
4416
4417 new_headroom = skb_headroom(skb);
4418 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4419 return 0;
4420 }
4421
gso_reset_checksum(struct sk_buff * skb,__wsum res)4422 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4423 {
4424 /* Do not update partial checksums if remote checksum is enabled. */
4425 if (skb->remcsum_offload)
4426 return;
4427
4428 SKB_GSO_CB(skb)->csum = res;
4429 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4430 }
4431
4432 /* Compute the checksum for a gso segment. First compute the checksum value
4433 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4434 * then add in skb->csum (checksum from csum_start to end of packet).
4435 * skb->csum and csum_start are then updated to reflect the checksum of the
4436 * resultant packet starting from the transport header-- the resultant checksum
4437 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4438 * header.
4439 */
gso_make_checksum(struct sk_buff * skb,__wsum res)4440 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4441 {
4442 unsigned char *csum_start = skb_transport_header(skb);
4443 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4444 __wsum partial = SKB_GSO_CB(skb)->csum;
4445
4446 SKB_GSO_CB(skb)->csum = res;
4447 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4448
4449 return csum_fold(csum_partial(csum_start, plen, partial));
4450 }
4451
skb_is_gso(const struct sk_buff * skb)4452 static inline bool skb_is_gso(const struct sk_buff *skb)
4453 {
4454 return skb_shinfo(skb)->gso_size;
4455 }
4456
4457 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)4458 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4459 {
4460 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4461 }
4462
4463 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)4464 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4465 {
4466 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4467 }
4468
4469 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)4470 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4471 {
4472 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4473 }
4474
skb_gso_reset(struct sk_buff * skb)4475 static inline void skb_gso_reset(struct sk_buff *skb)
4476 {
4477 skb_shinfo(skb)->gso_size = 0;
4478 skb_shinfo(skb)->gso_segs = 0;
4479 skb_shinfo(skb)->gso_type = 0;
4480 }
4481
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)4482 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4483 u16 increment)
4484 {
4485 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4486 return;
4487 shinfo->gso_size += increment;
4488 }
4489
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)4490 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4491 u16 decrement)
4492 {
4493 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4494 return;
4495 shinfo->gso_size -= decrement;
4496 }
4497
4498 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4499
skb_warn_if_lro(const struct sk_buff * skb)4500 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4501 {
4502 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4503 * wanted then gso_type will be set. */
4504 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4505
4506 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4507 unlikely(shinfo->gso_type == 0)) {
4508 __skb_warn_lro_forwarding(skb);
4509 return true;
4510 }
4511 return false;
4512 }
4513
skb_forward_csum(struct sk_buff * skb)4514 static inline void skb_forward_csum(struct sk_buff *skb)
4515 {
4516 /* Unfortunately we don't support this one. Any brave souls? */
4517 if (skb->ip_summed == CHECKSUM_COMPLETE)
4518 skb->ip_summed = CHECKSUM_NONE;
4519 }
4520
4521 /**
4522 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4523 * @skb: skb to check
4524 *
4525 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4526 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4527 * use this helper, to document places where we make this assertion.
4528 */
skb_checksum_none_assert(const struct sk_buff * skb)4529 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4530 {
4531 #ifdef DEBUG
4532 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4533 #endif
4534 }
4535
4536 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4537
4538 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4539 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4540 unsigned int transport_len,
4541 __sum16(*skb_chkf)(struct sk_buff *skb));
4542
4543 /**
4544 * skb_head_is_locked - Determine if the skb->head is locked down
4545 * @skb: skb to check
4546 *
4547 * The head on skbs build around a head frag can be removed if they are
4548 * not cloned. This function returns true if the skb head is locked down
4549 * due to either being allocated via kmalloc, or by being a clone with
4550 * multiple references to the head.
4551 */
skb_head_is_locked(const struct sk_buff * skb)4552 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4553 {
4554 return !skb->head_frag || skb_cloned(skb);
4555 }
4556
4557 /* Local Checksum Offload.
4558 * Compute outer checksum based on the assumption that the
4559 * inner checksum will be offloaded later.
4560 * See Documentation/networking/checksum-offloads.rst for
4561 * explanation of how this works.
4562 * Fill in outer checksum adjustment (e.g. with sum of outer
4563 * pseudo-header) before calling.
4564 * Also ensure that inner checksum is in linear data area.
4565 */
lco_csum(struct sk_buff * skb)4566 static inline __wsum lco_csum(struct sk_buff *skb)
4567 {
4568 unsigned char *csum_start = skb_checksum_start(skb);
4569 unsigned char *l4_hdr = skb_transport_header(skb);
4570 __wsum partial;
4571
4572 /* Start with complement of inner checksum adjustment */
4573 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4574 skb->csum_offset));
4575
4576 /* Add in checksum of our headers (incl. outer checksum
4577 * adjustment filled in by caller) and return result.
4578 */
4579 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4580 }
4581
skb_is_redirected(const struct sk_buff * skb)4582 static inline bool skb_is_redirected(const struct sk_buff *skb)
4583 {
4584 #ifdef CONFIG_NET_REDIRECT
4585 return skb->redirected;
4586 #else
4587 return false;
4588 #endif
4589 }
4590
skb_set_redirected(struct sk_buff * skb,bool from_ingress)4591 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4592 {
4593 #ifdef CONFIG_NET_REDIRECT
4594 skb->redirected = 1;
4595 skb->from_ingress = from_ingress;
4596 if (skb->from_ingress)
4597 skb->tstamp = 0;
4598 #endif
4599 }
4600
skb_reset_redirect(struct sk_buff * skb)4601 static inline void skb_reset_redirect(struct sk_buff *skb)
4602 {
4603 #ifdef CONFIG_NET_REDIRECT
4604 skb->redirected = 0;
4605 #endif
4606 }
4607
4608 #endif /* __KERNEL__ */
4609 #endif /* _LINUX_SKBUFF_H */
4610