]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - include/linux/skbuff.h
Merge tag 'fs.move_mount.move_mount_set_group.v5.15' of git://git.kernel.org/pub...
[mirror_ubuntu-jammy-kernel.git] / include / linux / skbuff.h
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 #include <net/page_pool.h>
41 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
42 #include <linux/netfilter/nf_conntrack_common.h>
43 #endif
44
45 /* The interface for checksum offload between the stack and networking drivers
46 * is as follows...
47 *
48 * A. IP checksum related features
49 *
50 * Drivers advertise checksum offload capabilities in the features of a device.
51 * From the stack's point of view these are capabilities offered by the driver.
52 * A driver typically only advertises features that it is capable of offloading
53 * to its device.
54 *
55 * The checksum related features are:
56 *
57 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
58 * IP (one's complement) checksum for any combination
59 * of protocols or protocol layering. The checksum is
60 * computed and set in a packet per the CHECKSUM_PARTIAL
61 * interface (see below).
62 *
63 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
64 * TCP or UDP packets over IPv4. These are specifically
65 * unencapsulated packets of the form IPv4|TCP or
66 * IPv4|UDP where the Protocol field in the IPv4 header
67 * is TCP or UDP. The IPv4 header may contain IP options.
68 * This feature cannot be set in features for a device
69 * with NETIF_F_HW_CSUM also set. This feature is being
70 * DEPRECATED (see below).
71 *
72 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
73 * TCP or UDP packets over IPv6. These are specifically
74 * unencapsulated packets of the form IPv6|TCP or
75 * IPv6|UDP where the Next Header field in the IPv6
76 * header is either TCP or UDP. IPv6 extension headers
77 * are not supported with this feature. This feature
78 * cannot be set in features for a device with
79 * NETIF_F_HW_CSUM also set. This feature is being
80 * DEPRECATED (see below).
81 *
82 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
83 * This flag is only used to disable the RX checksum
84 * feature for a device. The stack will accept receive
85 * checksum indication in packets received on a device
86 * regardless of whether NETIF_F_RXCSUM is set.
87 *
88 * B. Checksumming of received packets by device. Indication of checksum
89 * verification is set in skb->ip_summed. Possible values are:
90 *
91 * CHECKSUM_NONE:
92 *
93 * Device did not checksum this packet e.g. due to lack of capabilities.
94 * The packet contains full (though not verified) checksum in packet but
95 * not in skb->csum. Thus, skb->csum is undefined in this case.
96 *
97 * CHECKSUM_UNNECESSARY:
98 *
99 * The hardware you're dealing with doesn't calculate the full checksum
100 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
101 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
102 * if their checksums are okay. skb->csum is still undefined in this case
103 * though. A driver or device must never modify the checksum field in the
104 * packet even if checksum is verified.
105 *
106 * CHECKSUM_UNNECESSARY is applicable to following protocols:
107 * TCP: IPv6 and IPv4.
108 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
109 * zero UDP checksum for either IPv4 or IPv6, the networking stack
110 * may perform further validation in this case.
111 * GRE: only if the checksum is present in the header.
112 * SCTP: indicates the CRC in SCTP header has been validated.
113 * FCOE: indicates the CRC in FC frame has been validated.
114 *
115 * skb->csum_level indicates the number of consecutive checksums found in
116 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
117 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
118 * and a device is able to verify the checksums for UDP (possibly zero),
119 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
120 * two. If the device were only able to verify the UDP checksum and not
121 * GRE, either because it doesn't support GRE checksum or because GRE
122 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
123 * not considered in this case).
124 *
125 * CHECKSUM_COMPLETE:
126 *
127 * This is the most generic way. The device supplied checksum of the _whole_
128 * packet as seen by netif_rx() and fills in skb->csum. This means the
129 * hardware doesn't need to parse L3/L4 headers to implement this.
130 *
131 * Notes:
132 * - Even if device supports only some protocols, but is able to produce
133 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
134 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
135 *
136 * CHECKSUM_PARTIAL:
137 *
138 * A checksum is set up to be offloaded to a device as described in the
139 * output description for CHECKSUM_PARTIAL. This may occur on a packet
140 * received directly from another Linux OS, e.g., a virtualized Linux kernel
141 * on the same host, or it may be set in the input path in GRO or remote
142 * checksum offload. For the purposes of checksum verification, the checksum
143 * referred to by skb->csum_start + skb->csum_offset and any preceding
144 * checksums in the packet are considered verified. Any checksums in the
145 * packet that are after the checksum being offloaded are not considered to
146 * be verified.
147 *
148 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
149 * in the skb->ip_summed for a packet. Values are:
150 *
151 * CHECKSUM_PARTIAL:
152 *
153 * The driver is required to checksum the packet as seen by hard_start_xmit()
154 * from skb->csum_start up to the end, and to record/write the checksum at
155 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
156 * csum_start and csum_offset values are valid values given the length and
157 * offset of the packet, but it should not attempt to validate that the
158 * checksum refers to a legitimate transport layer checksum -- it is the
159 * purview of the stack to validate that csum_start and csum_offset are set
160 * correctly.
161 *
162 * When the stack requests checksum offload for a packet, the driver MUST
163 * ensure that the checksum is set correctly. A driver can either offload the
164 * checksum calculation to the device, or call skb_checksum_help (in the case
165 * that the device does not support offload for a particular checksum).
166 *
167 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
168 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
169 * checksum offload capability.
170 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
171 * on network device checksumming capabilities: if a packet does not match
172 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
173 * csum_not_inet, see item D.) is called to resolve the checksum.
174 *
175 * CHECKSUM_NONE:
176 *
177 * The skb was already checksummed by the protocol, or a checksum is not
178 * required.
179 *
180 * CHECKSUM_UNNECESSARY:
181 *
182 * This has the same meaning as CHECKSUM_NONE for checksum offload on
183 * output.
184 *
185 * CHECKSUM_COMPLETE:
186 * Not used in checksum output. If a driver observes a packet with this value
187 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
188 *
189 * D. Non-IP checksum (CRC) offloads
190 *
191 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
192 * offloading the SCTP CRC in a packet. To perform this offload the stack
193 * will set csum_start and csum_offset accordingly, set ip_summed to
194 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
195 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
196 * A driver that supports both IP checksum offload and SCTP CRC32c offload
197 * must verify which offload is configured for a packet by testing the
198 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
199 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
200 *
201 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
202 * offloading the FCOE CRC in a packet. To perform this offload the stack
203 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
204 * accordingly. Note that there is no indication in the skbuff that the
205 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
206 * both IP checksum offload and FCOE CRC offload must verify which offload
207 * is configured for a packet, presumably by inspecting packet headers.
208 *
209 * E. Checksumming on output with GSO.
210 *
211 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
212 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
213 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
214 * part of the GSO operation is implied. If a checksum is being offloaded
215 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
216 * csum_offset are set to refer to the outermost checksum being offloaded
217 * (two offloaded checksums are possible with UDP encapsulation).
218 */
219
220 /* Don't change this without changing skb_csum_unnecessary! */
221 #define CHECKSUM_NONE 0
222 #define CHECKSUM_UNNECESSARY 1
223 #define CHECKSUM_COMPLETE 2
224 #define CHECKSUM_PARTIAL 3
225
226 /* Maximum value in skb->csum_level */
227 #define SKB_MAX_CSUM_LEVEL 3
228
229 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
230 #define SKB_WITH_OVERHEAD(X) \
231 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
232 #define SKB_MAX_ORDER(X, ORDER) \
233 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
234 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
235 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
236
237 /* return minimum truesize of one skb containing X bytes of data */
238 #define SKB_TRUESIZE(X) ((X) + \
239 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
240 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
241
242 struct ahash_request;
243 struct net_device;
244 struct scatterlist;
245 struct pipe_inode_info;
246 struct iov_iter;
247 struct napi_struct;
248 struct bpf_prog;
249 union bpf_attr;
250 struct skb_ext;
251
252 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
253 struct nf_bridge_info {
254 enum {
255 BRNF_PROTO_UNCHANGED,
256 BRNF_PROTO_8021Q,
257 BRNF_PROTO_PPPOE
258 } orig_proto:8;
259 u8 pkt_otherhost:1;
260 u8 in_prerouting:1;
261 u8 bridged_dnat:1;
262 __u16 frag_max_size;
263 struct net_device *physindev;
264
265 /* always valid & non-NULL from FORWARD on, for physdev match */
266 struct net_device *physoutdev;
267 union {
268 /* prerouting: detect dnat in orig/reply direction */
269 __be32 ipv4_daddr;
270 struct in6_addr ipv6_daddr;
271
272 /* after prerouting + nat detected: store original source
273 * mac since neigh resolution overwrites it, only used while
274 * skb is out in neigh layer.
275 */
276 char neigh_header[8];
277 };
278 };
279 #endif
280
281 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
282 /* Chain in tc_skb_ext will be used to share the tc chain with
283 * ovs recirc_id. It will be set to the current chain by tc
284 * and read by ovs to recirc_id.
285 */
286 struct tc_skb_ext {
287 __u32 chain;
288 __u16 mru;
289 bool post_ct;
290 };
291 #endif
292
293 struct sk_buff_head {
294 /* These two members must be first. */
295 struct sk_buff *next;
296 struct sk_buff *prev;
297
298 __u32 qlen;
299 spinlock_t lock;
300 };
301
302 struct sk_buff;
303
304 /* To allow 64K frame to be packed as single skb without frag_list we
305 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
306 * buffers which do not start on a page boundary.
307 *
308 * Since GRO uses frags we allocate at least 16 regardless of page
309 * size.
310 */
311 #if (65536/PAGE_SIZE + 1) < 16
312 #define MAX_SKB_FRAGS 16UL
313 #else
314 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
315 #endif
316 extern int sysctl_max_skb_frags;
317
318 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
319 * segment using its current segmentation instead.
320 */
321 #define GSO_BY_FRAGS 0xFFFF
322
323 typedef struct bio_vec skb_frag_t;
324
325 /**
326 * skb_frag_size() - Returns the size of a skb fragment
327 * @frag: skb fragment
328 */
329 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
330 {
331 return frag->bv_len;
332 }
333
334 /**
335 * skb_frag_size_set() - Sets the size of a skb fragment
336 * @frag: skb fragment
337 * @size: size of fragment
338 */
339 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
340 {
341 frag->bv_len = size;
342 }
343
344 /**
345 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
346 * @frag: skb fragment
347 * @delta: value to add
348 */
349 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
350 {
351 frag->bv_len += delta;
352 }
353
354 /**
355 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
356 * @frag: skb fragment
357 * @delta: value to subtract
358 */
359 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
360 {
361 frag->bv_len -= delta;
362 }
363
364 /**
365 * skb_frag_must_loop - Test if %p is a high memory page
366 * @p: fragment's page
367 */
368 static inline bool skb_frag_must_loop(struct page *p)
369 {
370 #if defined(CONFIG_HIGHMEM)
371 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
372 return true;
373 #endif
374 return false;
375 }
376
377 /**
378 * skb_frag_foreach_page - loop over pages in a fragment
379 *
380 * @f: skb frag to operate on
381 * @f_off: offset from start of f->bv_page
382 * @f_len: length from f_off to loop over
383 * @p: (temp var) current page
384 * @p_off: (temp var) offset from start of current page,
385 * non-zero only on first page.
386 * @p_len: (temp var) length in current page,
387 * < PAGE_SIZE only on first and last page.
388 * @copied: (temp var) length so far, excluding current p_len.
389 *
390 * A fragment can hold a compound page, in which case per-page
391 * operations, notably kmap_atomic, must be called for each
392 * regular page.
393 */
394 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
395 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
396 p_off = (f_off) & (PAGE_SIZE - 1), \
397 p_len = skb_frag_must_loop(p) ? \
398 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
399 copied = 0; \
400 copied < f_len; \
401 copied += p_len, p++, p_off = 0, \
402 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
403
404 #define HAVE_HW_TIME_STAMP
405
406 /**
407 * struct skb_shared_hwtstamps - hardware time stamps
408 * @hwtstamp: hardware time stamp transformed into duration
409 * since arbitrary point in time
410 *
411 * Software time stamps generated by ktime_get_real() are stored in
412 * skb->tstamp.
413 *
414 * hwtstamps can only be compared against other hwtstamps from
415 * the same device.
416 *
417 * This structure is attached to packets as part of the
418 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
419 */
420 struct skb_shared_hwtstamps {
421 ktime_t hwtstamp;
422 };
423
424 /* Definitions for tx_flags in struct skb_shared_info */
425 enum {
426 /* generate hardware time stamp */
427 SKBTX_HW_TSTAMP = 1 << 0,
428
429 /* generate software time stamp when queueing packet to NIC */
430 SKBTX_SW_TSTAMP = 1 << 1,
431
432 /* device driver is going to provide hardware time stamp */
433 SKBTX_IN_PROGRESS = 1 << 2,
434
435 /* generate wifi status information (where possible) */
436 SKBTX_WIFI_STATUS = 1 << 4,
437
438 /* generate software time stamp when entering packet scheduling */
439 SKBTX_SCHED_TSTAMP = 1 << 6,
440 };
441
442 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
443 SKBTX_SCHED_TSTAMP)
444 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
445
446 /* Definitions for flags in struct skb_shared_info */
447 enum {
448 /* use zcopy routines */
449 SKBFL_ZEROCOPY_ENABLE = BIT(0),
450
451 /* This indicates at least one fragment might be overwritten
452 * (as in vmsplice(), sendfile() ...)
453 * If we need to compute a TX checksum, we'll need to copy
454 * all frags to avoid possible bad checksum
455 */
456 SKBFL_SHARED_FRAG = BIT(1),
457 };
458
459 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
460
461 /*
462 * The callback notifies userspace to release buffers when skb DMA is done in
463 * lower device, the skb last reference should be 0 when calling this.
464 * The zerocopy_success argument is true if zero copy transmit occurred,
465 * false on data copy or out of memory error caused by data copy attempt.
466 * The ctx field is used to track device context.
467 * The desc field is used to track userspace buffer index.
468 */
469 struct ubuf_info {
470 void (*callback)(struct sk_buff *, struct ubuf_info *,
471 bool zerocopy_success);
472 union {
473 struct {
474 unsigned long desc;
475 void *ctx;
476 };
477 struct {
478 u32 id;
479 u16 len;
480 u16 zerocopy:1;
481 u32 bytelen;
482 };
483 };
484 refcount_t refcnt;
485 u8 flags;
486
487 struct mmpin {
488 struct user_struct *user;
489 unsigned int num_pg;
490 } mmp;
491 };
492
493 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
494
495 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
496 void mm_unaccount_pinned_pages(struct mmpin *mmp);
497
498 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
499 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
500 struct ubuf_info *uarg);
501
502 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
503
504 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
505 bool success);
506
507 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
508 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
509 struct msghdr *msg, int len,
510 struct ubuf_info *uarg);
511
512 /* This data is invariant across clones and lives at
513 * the end of the header data, ie. at skb->end.
514 */
515 struct skb_shared_info {
516 __u8 flags;
517 __u8 meta_len;
518 __u8 nr_frags;
519 __u8 tx_flags;
520 unsigned short gso_size;
521 /* Warning: this field is not always filled in (UFO)! */
522 unsigned short gso_segs;
523 struct sk_buff *frag_list;
524 struct skb_shared_hwtstamps hwtstamps;
525 unsigned int gso_type;
526 u32 tskey;
527
528 /*
529 * Warning : all fields before dataref are cleared in __alloc_skb()
530 */
531 atomic_t dataref;
532
533 /* Intermediate layers must ensure that destructor_arg
534 * remains valid until skb destructor */
535 void * destructor_arg;
536
537 /* must be last field, see pskb_expand_head() */
538 skb_frag_t frags[MAX_SKB_FRAGS];
539 };
540
541 /* We divide dataref into two halves. The higher 16 bits hold references
542 * to the payload part of skb->data. The lower 16 bits hold references to
543 * the entire skb->data. A clone of a headerless skb holds the length of
544 * the header in skb->hdr_len.
545 *
546 * All users must obey the rule that the skb->data reference count must be
547 * greater than or equal to the payload reference count.
548 *
549 * Holding a reference to the payload part means that the user does not
550 * care about modifications to the header part of skb->data.
551 */
552 #define SKB_DATAREF_SHIFT 16
553 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
554
555
556 enum {
557 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
558 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
559 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
560 };
561
562 enum {
563 SKB_GSO_TCPV4 = 1 << 0,
564
565 /* This indicates the skb is from an untrusted source. */
566 SKB_GSO_DODGY = 1 << 1,
567
568 /* This indicates the tcp segment has CWR set. */
569 SKB_GSO_TCP_ECN = 1 << 2,
570
571 SKB_GSO_TCP_FIXEDID = 1 << 3,
572
573 SKB_GSO_TCPV6 = 1 << 4,
574
575 SKB_GSO_FCOE = 1 << 5,
576
577 SKB_GSO_GRE = 1 << 6,
578
579 SKB_GSO_GRE_CSUM = 1 << 7,
580
581 SKB_GSO_IPXIP4 = 1 << 8,
582
583 SKB_GSO_IPXIP6 = 1 << 9,
584
585 SKB_GSO_UDP_TUNNEL = 1 << 10,
586
587 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
588
589 SKB_GSO_PARTIAL = 1 << 12,
590
591 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
592
593 SKB_GSO_SCTP = 1 << 14,
594
595 SKB_GSO_ESP = 1 << 15,
596
597 SKB_GSO_UDP = 1 << 16,
598
599 SKB_GSO_UDP_L4 = 1 << 17,
600
601 SKB_GSO_FRAGLIST = 1 << 18,
602 };
603
604 #if BITS_PER_LONG > 32
605 #define NET_SKBUFF_DATA_USES_OFFSET 1
606 #endif
607
608 #ifdef NET_SKBUFF_DATA_USES_OFFSET
609 typedef unsigned int sk_buff_data_t;
610 #else
611 typedef unsigned char *sk_buff_data_t;
612 #endif
613
614 /**
615 * struct sk_buff - socket buffer
616 * @next: Next buffer in list
617 * @prev: Previous buffer in list
618 * @tstamp: Time we arrived/left
619 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
620 * for retransmit timer
621 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
622 * @list: queue head
623 * @sk: Socket we are owned by
624 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
625 * fragmentation management
626 * @dev: Device we arrived on/are leaving by
627 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
628 * @cb: Control buffer. Free for use by every layer. Put private vars here
629 * @_skb_refdst: destination entry (with norefcount bit)
630 * @sp: the security path, used for xfrm
631 * @len: Length of actual data
632 * @data_len: Data length
633 * @mac_len: Length of link layer header
634 * @hdr_len: writable header length of cloned skb
635 * @csum: Checksum (must include start/offset pair)
636 * @csum_start: Offset from skb->head where checksumming should start
637 * @csum_offset: Offset from csum_start where checksum should be stored
638 * @priority: Packet queueing priority
639 * @ignore_df: allow local fragmentation
640 * @cloned: Head may be cloned (check refcnt to be sure)
641 * @ip_summed: Driver fed us an IP checksum
642 * @nohdr: Payload reference only, must not modify header
643 * @pkt_type: Packet class
644 * @fclone: skbuff clone status
645 * @ipvs_property: skbuff is owned by ipvs
646 * @inner_protocol_type: whether the inner protocol is
647 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
648 * @remcsum_offload: remote checksum offload is enabled
649 * @offload_fwd_mark: Packet was L2-forwarded in hardware
650 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
651 * @tc_skip_classify: do not classify packet. set by IFB device
652 * @tc_at_ingress: used within tc_classify to distinguish in/egress
653 * @redirected: packet was redirected by packet classifier
654 * @from_ingress: packet was redirected from the ingress path
655 * @peeked: this packet has been seen already, so stats have been
656 * done for it, don't do them again
657 * @nf_trace: netfilter packet trace flag
658 * @protocol: Packet protocol from driver
659 * @destructor: Destruct function
660 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
661 * @_sk_redir: socket redirection information for skmsg
662 * @_nfct: Associated connection, if any (with nfctinfo bits)
663 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
664 * @skb_iif: ifindex of device we arrived on
665 * @tc_index: Traffic control index
666 * @hash: the packet hash
667 * @queue_mapping: Queue mapping for multiqueue devices
668 * @head_frag: skb was allocated from page fragments,
669 * not allocated by kmalloc() or vmalloc().
670 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
671 * @pp_recycle: mark the packet for recycling instead of freeing (implies
672 * page_pool support on driver)
673 * @active_extensions: active extensions (skb_ext_id types)
674 * @ndisc_nodetype: router type (from link layer)
675 * @ooo_okay: allow the mapping of a socket to a queue to be changed
676 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
677 * ports.
678 * @sw_hash: indicates hash was computed in software stack
679 * @wifi_acked_valid: wifi_acked was set
680 * @wifi_acked: whether frame was acked on wifi or not
681 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
682 * @encapsulation: indicates the inner headers in the skbuff are valid
683 * @encap_hdr_csum: software checksum is needed
684 * @csum_valid: checksum is already valid
685 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
686 * @csum_complete_sw: checksum was completed by software
687 * @csum_level: indicates the number of consecutive checksums found in
688 * the packet minus one that have been verified as
689 * CHECKSUM_UNNECESSARY (max 3)
690 * @dst_pending_confirm: need to confirm neighbour
691 * @decrypted: Decrypted SKB
692 * @napi_id: id of the NAPI struct this skb came from
693 * @sender_cpu: (aka @napi_id) source CPU in XPS
694 * @secmark: security marking
695 * @mark: Generic packet mark
696 * @reserved_tailroom: (aka @mark) number of bytes of free space available
697 * at the tail of an sk_buff
698 * @vlan_present: VLAN tag is present
699 * @vlan_proto: vlan encapsulation protocol
700 * @vlan_tci: vlan tag control information
701 * @inner_protocol: Protocol (encapsulation)
702 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
703 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
704 * @inner_transport_header: Inner transport layer header (encapsulation)
705 * @inner_network_header: Network layer header (encapsulation)
706 * @inner_mac_header: Link layer header (encapsulation)
707 * @transport_header: Transport layer header
708 * @network_header: Network layer header
709 * @mac_header: Link layer header
710 * @kcov_handle: KCOV remote handle for remote coverage collection
711 * @tail: Tail pointer
712 * @end: End pointer
713 * @head: Head of buffer
714 * @data: Data head pointer
715 * @truesize: Buffer size
716 * @users: User count - see {datagram,tcp}.c
717 * @extensions: allocated extensions, valid if active_extensions is nonzero
718 */
719
720 struct sk_buff {
721 union {
722 struct {
723 /* These two members must be first. */
724 struct sk_buff *next;
725 struct sk_buff *prev;
726
727 union {
728 struct net_device *dev;
729 /* Some protocols might use this space to store information,
730 * while device pointer would be NULL.
731 * UDP receive path is one user.
732 */
733 unsigned long dev_scratch;
734 };
735 };
736 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
737 struct list_head list;
738 };
739
740 union {
741 struct sock *sk;
742 int ip_defrag_offset;
743 };
744
745 union {
746 ktime_t tstamp;
747 u64 skb_mstamp_ns; /* earliest departure time */
748 };
749 /*
750 * This is the control buffer. It is free to use for every
751 * layer. Please put your private variables there. If you
752 * want to keep them across layers you have to do a skb_clone()
753 * first. This is owned by whoever has the skb queued ATM.
754 */
755 char cb[48] __aligned(8);
756
757 union {
758 struct {
759 unsigned long _skb_refdst;
760 void (*destructor)(struct sk_buff *skb);
761 };
762 struct list_head tcp_tsorted_anchor;
763 #ifdef CONFIG_NET_SOCK_MSG
764 unsigned long _sk_redir;
765 #endif
766 };
767
768 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
769 unsigned long _nfct;
770 #endif
771 unsigned int len,
772 data_len;
773 __u16 mac_len,
774 hdr_len;
775
776 /* Following fields are _not_ copied in __copy_skb_header()
777 * Note that queue_mapping is here mostly to fill a hole.
778 */
779 __u16 queue_mapping;
780
781 /* if you move cloned around you also must adapt those constants */
782 #ifdef __BIG_ENDIAN_BITFIELD
783 #define CLONED_MASK (1 << 7)
784 #else
785 #define CLONED_MASK 1
786 #endif
787 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
788
789 /* private: */
790 __u8 __cloned_offset[0];
791 /* public: */
792 __u8 cloned:1,
793 nohdr:1,
794 fclone:2,
795 peeked:1,
796 head_frag:1,
797 pfmemalloc:1,
798 pp_recycle:1; /* page_pool recycle indicator */
799 #ifdef CONFIG_SKB_EXTENSIONS
800 __u8 active_extensions;
801 #endif
802
803 /* fields enclosed in headers_start/headers_end are copied
804 * using a single memcpy() in __copy_skb_header()
805 */
806 /* private: */
807 __u32 headers_start[0];
808 /* public: */
809
810 /* if you move pkt_type around you also must adapt those constants */
811 #ifdef __BIG_ENDIAN_BITFIELD
812 #define PKT_TYPE_MAX (7 << 5)
813 #else
814 #define PKT_TYPE_MAX 7
815 #endif
816 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
817
818 /* private: */
819 __u8 __pkt_type_offset[0];
820 /* public: */
821 __u8 pkt_type:3;
822 __u8 ignore_df:1;
823 __u8 nf_trace:1;
824 __u8 ip_summed:2;
825 __u8 ooo_okay:1;
826
827 __u8 l4_hash:1;
828 __u8 sw_hash:1;
829 __u8 wifi_acked_valid:1;
830 __u8 wifi_acked:1;
831 __u8 no_fcs:1;
832 /* Indicates the inner headers are valid in the skbuff. */
833 __u8 encapsulation:1;
834 __u8 encap_hdr_csum:1;
835 __u8 csum_valid:1;
836
837 #ifdef __BIG_ENDIAN_BITFIELD
838 #define PKT_VLAN_PRESENT_BIT 7
839 #else
840 #define PKT_VLAN_PRESENT_BIT 0
841 #endif
842 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
843 /* private: */
844 __u8 __pkt_vlan_present_offset[0];
845 /* public: */
846 __u8 vlan_present:1;
847 __u8 csum_complete_sw:1;
848 __u8 csum_level:2;
849 __u8 csum_not_inet:1;
850 __u8 dst_pending_confirm:1;
851 #ifdef CONFIG_IPV6_NDISC_NODETYPE
852 __u8 ndisc_nodetype:2;
853 #endif
854
855 __u8 ipvs_property:1;
856 __u8 inner_protocol_type:1;
857 __u8 remcsum_offload:1;
858 #ifdef CONFIG_NET_SWITCHDEV
859 __u8 offload_fwd_mark:1;
860 __u8 offload_l3_fwd_mark:1;
861 #endif
862 #ifdef CONFIG_NET_CLS_ACT
863 __u8 tc_skip_classify:1;
864 __u8 tc_at_ingress:1;
865 #endif
866 #ifdef CONFIG_NET_REDIRECT
867 __u8 redirected:1;
868 __u8 from_ingress:1;
869 #endif
870 #ifdef CONFIG_TLS_DEVICE
871 __u8 decrypted:1;
872 #endif
873
874 #ifdef CONFIG_NET_SCHED
875 __u16 tc_index; /* traffic control index */
876 #endif
877
878 union {
879 __wsum csum;
880 struct {
881 __u16 csum_start;
882 __u16 csum_offset;
883 };
884 };
885 __u32 priority;
886 int skb_iif;
887 __u32 hash;
888 __be16 vlan_proto;
889 __u16 vlan_tci;
890 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
891 union {
892 unsigned int napi_id;
893 unsigned int sender_cpu;
894 };
895 #endif
896 #ifdef CONFIG_NETWORK_SECMARK
897 __u32 secmark;
898 #endif
899
900 union {
901 __u32 mark;
902 __u32 reserved_tailroom;
903 };
904
905 union {
906 __be16 inner_protocol;
907 __u8 inner_ipproto;
908 };
909
910 __u16 inner_transport_header;
911 __u16 inner_network_header;
912 __u16 inner_mac_header;
913
914 __be16 protocol;
915 __u16 transport_header;
916 __u16 network_header;
917 __u16 mac_header;
918
919 #ifdef CONFIG_KCOV
920 u64 kcov_handle;
921 #endif
922
923 /* private: */
924 __u32 headers_end[0];
925 /* public: */
926
927 /* These elements must be at the end, see alloc_skb() for details. */
928 sk_buff_data_t tail;
929 sk_buff_data_t end;
930 unsigned char *head,
931 *data;
932 unsigned int truesize;
933 refcount_t users;
934
935 #ifdef CONFIG_SKB_EXTENSIONS
936 /* only useable after checking ->active_extensions != 0 */
937 struct skb_ext *extensions;
938 #endif
939 };
940
941 #ifdef __KERNEL__
942 /*
943 * Handling routines are only of interest to the kernel
944 */
945
946 #define SKB_ALLOC_FCLONE 0x01
947 #define SKB_ALLOC_RX 0x02
948 #define SKB_ALLOC_NAPI 0x04
949
950 /**
951 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
952 * @skb: buffer
953 */
954 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
955 {
956 return unlikely(skb->pfmemalloc);
957 }
958
959 /*
960 * skb might have a dst pointer attached, refcounted or not.
961 * _skb_refdst low order bit is set if refcount was _not_ taken
962 */
963 #define SKB_DST_NOREF 1UL
964 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
965
966 /**
967 * skb_dst - returns skb dst_entry
968 * @skb: buffer
969 *
970 * Returns skb dst_entry, regardless of reference taken or not.
971 */
972 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
973 {
974 /* If refdst was not refcounted, check we still are in a
975 * rcu_read_lock section
976 */
977 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
978 !rcu_read_lock_held() &&
979 !rcu_read_lock_bh_held());
980 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
981 }
982
983 /**
984 * skb_dst_set - sets skb dst
985 * @skb: buffer
986 * @dst: dst entry
987 *
988 * Sets skb dst, assuming a reference was taken on dst and should
989 * be released by skb_dst_drop()
990 */
991 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
992 {
993 skb->_skb_refdst = (unsigned long)dst;
994 }
995
996 /**
997 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
998 * @skb: buffer
999 * @dst: dst entry
1000 *
1001 * Sets skb dst, assuming a reference was not taken on dst.
1002 * If dst entry is cached, we do not take reference and dst_release
1003 * will be avoided by refdst_drop. If dst entry is not cached, we take
1004 * reference, so that last dst_release can destroy the dst immediately.
1005 */
1006 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1007 {
1008 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1009 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1010 }
1011
1012 /**
1013 * skb_dst_is_noref - Test if skb dst isn't refcounted
1014 * @skb: buffer
1015 */
1016 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1017 {
1018 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1019 }
1020
1021 /**
1022 * skb_rtable - Returns the skb &rtable
1023 * @skb: buffer
1024 */
1025 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1026 {
1027 return (struct rtable *)skb_dst(skb);
1028 }
1029
1030 /* For mangling skb->pkt_type from user space side from applications
1031 * such as nft, tc, etc, we only allow a conservative subset of
1032 * possible pkt_types to be set.
1033 */
1034 static inline bool skb_pkt_type_ok(u32 ptype)
1035 {
1036 return ptype <= PACKET_OTHERHOST;
1037 }
1038
1039 /**
1040 * skb_napi_id - Returns the skb's NAPI id
1041 * @skb: buffer
1042 */
1043 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1044 {
1045 #ifdef CONFIG_NET_RX_BUSY_POLL
1046 return skb->napi_id;
1047 #else
1048 return 0;
1049 #endif
1050 }
1051
1052 /**
1053 * skb_unref - decrement the skb's reference count
1054 * @skb: buffer
1055 *
1056 * Returns true if we can free the skb.
1057 */
1058 static inline bool skb_unref(struct sk_buff *skb)
1059 {
1060 if (unlikely(!skb))
1061 return false;
1062 if (likely(refcount_read(&skb->users) == 1))
1063 smp_rmb();
1064 else if (likely(!refcount_dec_and_test(&skb->users)))
1065 return false;
1066
1067 return true;
1068 }
1069
1070 void skb_release_head_state(struct sk_buff *skb);
1071 void kfree_skb(struct sk_buff *skb);
1072 void kfree_skb_list(struct sk_buff *segs);
1073 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1074 void skb_tx_error(struct sk_buff *skb);
1075
1076 #ifdef CONFIG_TRACEPOINTS
1077 void consume_skb(struct sk_buff *skb);
1078 #else
1079 static inline void consume_skb(struct sk_buff *skb)
1080 {
1081 return kfree_skb(skb);
1082 }
1083 #endif
1084
1085 void __consume_stateless_skb(struct sk_buff *skb);
1086 void __kfree_skb(struct sk_buff *skb);
1087 extern struct kmem_cache *skbuff_head_cache;
1088
1089 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1090 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1091 bool *fragstolen, int *delta_truesize);
1092
1093 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1094 int node);
1095 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1096 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1097 struct sk_buff *build_skb_around(struct sk_buff *skb,
1098 void *data, unsigned int frag_size);
1099
1100 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1101
1102 /**
1103 * alloc_skb - allocate a network buffer
1104 * @size: size to allocate
1105 * @priority: allocation mask
1106 *
1107 * This function is a convenient wrapper around __alloc_skb().
1108 */
1109 static inline struct sk_buff *alloc_skb(unsigned int size,
1110 gfp_t priority)
1111 {
1112 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1113 }
1114
1115 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1116 unsigned long data_len,
1117 int max_page_order,
1118 int *errcode,
1119 gfp_t gfp_mask);
1120 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1121
1122 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1123 struct sk_buff_fclones {
1124 struct sk_buff skb1;
1125
1126 struct sk_buff skb2;
1127
1128 refcount_t fclone_ref;
1129 };
1130
1131 /**
1132 * skb_fclone_busy - check if fclone is busy
1133 * @sk: socket
1134 * @skb: buffer
1135 *
1136 * Returns true if skb is a fast clone, and its clone is not freed.
1137 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1138 * so we also check that this didnt happen.
1139 */
1140 static inline bool skb_fclone_busy(const struct sock *sk,
1141 const struct sk_buff *skb)
1142 {
1143 const struct sk_buff_fclones *fclones;
1144
1145 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1146
1147 return skb->fclone == SKB_FCLONE_ORIG &&
1148 refcount_read(&fclones->fclone_ref) > 1 &&
1149 READ_ONCE(fclones->skb2.sk) == sk;
1150 }
1151
1152 /**
1153 * alloc_skb_fclone - allocate a network buffer from fclone cache
1154 * @size: size to allocate
1155 * @priority: allocation mask
1156 *
1157 * This function is a convenient wrapper around __alloc_skb().
1158 */
1159 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1160 gfp_t priority)
1161 {
1162 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1163 }
1164
1165 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1166 void skb_headers_offset_update(struct sk_buff *skb, int off);
1167 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1168 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1169 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1170 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1171 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1172 gfp_t gfp_mask, bool fclone);
1173 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1174 gfp_t gfp_mask)
1175 {
1176 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1177 }
1178
1179 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1180 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1181 unsigned int headroom);
1182 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1183 int newtailroom, gfp_t priority);
1184 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1185 int offset, int len);
1186 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1187 int offset, int len);
1188 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1189 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1190
1191 /**
1192 * skb_pad - zero pad the tail of an skb
1193 * @skb: buffer to pad
1194 * @pad: space to pad
1195 *
1196 * Ensure that a buffer is followed by a padding area that is zero
1197 * filled. Used by network drivers which may DMA or transfer data
1198 * beyond the buffer end onto the wire.
1199 *
1200 * May return error in out of memory cases. The skb is freed on error.
1201 */
1202 static inline int skb_pad(struct sk_buff *skb, int pad)
1203 {
1204 return __skb_pad(skb, pad, true);
1205 }
1206 #define dev_kfree_skb(a) consume_skb(a)
1207
1208 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1209 int offset, size_t size);
1210
1211 struct skb_seq_state {
1212 __u32 lower_offset;
1213 __u32 upper_offset;
1214 __u32 frag_idx;
1215 __u32 stepped_offset;
1216 struct sk_buff *root_skb;
1217 struct sk_buff *cur_skb;
1218 __u8 *frag_data;
1219 __u32 frag_off;
1220 };
1221
1222 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1223 unsigned int to, struct skb_seq_state *st);
1224 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1225 struct skb_seq_state *st);
1226 void skb_abort_seq_read(struct skb_seq_state *st);
1227
1228 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1229 unsigned int to, struct ts_config *config);
1230
1231 /*
1232 * Packet hash types specify the type of hash in skb_set_hash.
1233 *
1234 * Hash types refer to the protocol layer addresses which are used to
1235 * construct a packet's hash. The hashes are used to differentiate or identify
1236 * flows of the protocol layer for the hash type. Hash types are either
1237 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1238 *
1239 * Properties of hashes:
1240 *
1241 * 1) Two packets in different flows have different hash values
1242 * 2) Two packets in the same flow should have the same hash value
1243 *
1244 * A hash at a higher layer is considered to be more specific. A driver should
1245 * set the most specific hash possible.
1246 *
1247 * A driver cannot indicate a more specific hash than the layer at which a hash
1248 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1249 *
1250 * A driver may indicate a hash level which is less specific than the
1251 * actual layer the hash was computed on. For instance, a hash computed
1252 * at L4 may be considered an L3 hash. This should only be done if the
1253 * driver can't unambiguously determine that the HW computed the hash at
1254 * the higher layer. Note that the "should" in the second property above
1255 * permits this.
1256 */
1257 enum pkt_hash_types {
1258 PKT_HASH_TYPE_NONE, /* Undefined type */
1259 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1260 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1261 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1262 };
1263
1264 static inline void skb_clear_hash(struct sk_buff *skb)
1265 {
1266 skb->hash = 0;
1267 skb->sw_hash = 0;
1268 skb->l4_hash = 0;
1269 }
1270
1271 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1272 {
1273 if (!skb->l4_hash)
1274 skb_clear_hash(skb);
1275 }
1276
1277 static inline void
1278 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1279 {
1280 skb->l4_hash = is_l4;
1281 skb->sw_hash = is_sw;
1282 skb->hash = hash;
1283 }
1284
1285 static inline void
1286 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1287 {
1288 /* Used by drivers to set hash from HW */
1289 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1290 }
1291
1292 static inline void
1293 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1294 {
1295 __skb_set_hash(skb, hash, true, is_l4);
1296 }
1297
1298 void __skb_get_hash(struct sk_buff *skb);
1299 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1300 u32 skb_get_poff(const struct sk_buff *skb);
1301 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1302 const struct flow_keys_basic *keys, int hlen);
1303 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1304 const void *data, int hlen_proto);
1305
1306 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1307 int thoff, u8 ip_proto)
1308 {
1309 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1310 }
1311
1312 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1313 const struct flow_dissector_key *key,
1314 unsigned int key_count);
1315
1316 struct bpf_flow_dissector;
1317 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1318 __be16 proto, int nhoff, int hlen, unsigned int flags);
1319
1320 bool __skb_flow_dissect(const struct net *net,
1321 const struct sk_buff *skb,
1322 struct flow_dissector *flow_dissector,
1323 void *target_container, const void *data,
1324 __be16 proto, int nhoff, int hlen, unsigned int flags);
1325
1326 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1327 struct flow_dissector *flow_dissector,
1328 void *target_container, unsigned int flags)
1329 {
1330 return __skb_flow_dissect(NULL, skb, flow_dissector,
1331 target_container, NULL, 0, 0, 0, flags);
1332 }
1333
1334 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1335 struct flow_keys *flow,
1336 unsigned int flags)
1337 {
1338 memset(flow, 0, sizeof(*flow));
1339 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1340 flow, NULL, 0, 0, 0, flags);
1341 }
1342
1343 static inline bool
1344 skb_flow_dissect_flow_keys_basic(const struct net *net,
1345 const struct sk_buff *skb,
1346 struct flow_keys_basic *flow,
1347 const void *data, __be16 proto,
1348 int nhoff, int hlen, unsigned int flags)
1349 {
1350 memset(flow, 0, sizeof(*flow));
1351 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1352 data, proto, nhoff, hlen, flags);
1353 }
1354
1355 void skb_flow_dissect_meta(const struct sk_buff *skb,
1356 struct flow_dissector *flow_dissector,
1357 void *target_container);
1358
1359 /* Gets a skb connection tracking info, ctinfo map should be a
1360 * map of mapsize to translate enum ip_conntrack_info states
1361 * to user states.
1362 */
1363 void
1364 skb_flow_dissect_ct(const struct sk_buff *skb,
1365 struct flow_dissector *flow_dissector,
1366 void *target_container,
1367 u16 *ctinfo_map, size_t mapsize,
1368 bool post_ct);
1369 void
1370 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1371 struct flow_dissector *flow_dissector,
1372 void *target_container);
1373
1374 void skb_flow_dissect_hash(const struct sk_buff *skb,
1375 struct flow_dissector *flow_dissector,
1376 void *target_container);
1377
1378 static inline __u32 skb_get_hash(struct sk_buff *skb)
1379 {
1380 if (!skb->l4_hash && !skb->sw_hash)
1381 __skb_get_hash(skb);
1382
1383 return skb->hash;
1384 }
1385
1386 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1387 {
1388 if (!skb->l4_hash && !skb->sw_hash) {
1389 struct flow_keys keys;
1390 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1391
1392 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1393 }
1394
1395 return skb->hash;
1396 }
1397
1398 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1399 const siphash_key_t *perturb);
1400
1401 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1402 {
1403 return skb->hash;
1404 }
1405
1406 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1407 {
1408 to->hash = from->hash;
1409 to->sw_hash = from->sw_hash;
1410 to->l4_hash = from->l4_hash;
1411 };
1412
1413 static inline void skb_copy_decrypted(struct sk_buff *to,
1414 const struct sk_buff *from)
1415 {
1416 #ifdef CONFIG_TLS_DEVICE
1417 to->decrypted = from->decrypted;
1418 #endif
1419 }
1420
1421 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1422 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1423 {
1424 return skb->head + skb->end;
1425 }
1426
1427 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1428 {
1429 return skb->end;
1430 }
1431 #else
1432 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1433 {
1434 return skb->end;
1435 }
1436
1437 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1438 {
1439 return skb->end - skb->head;
1440 }
1441 #endif
1442
1443 /* Internal */
1444 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1445
1446 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1447 {
1448 return &skb_shinfo(skb)->hwtstamps;
1449 }
1450
1451 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1452 {
1453 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1454
1455 return is_zcopy ? skb_uarg(skb) : NULL;
1456 }
1457
1458 static inline void net_zcopy_get(struct ubuf_info *uarg)
1459 {
1460 refcount_inc(&uarg->refcnt);
1461 }
1462
1463 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1464 {
1465 skb_shinfo(skb)->destructor_arg = uarg;
1466 skb_shinfo(skb)->flags |= uarg->flags;
1467 }
1468
1469 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1470 bool *have_ref)
1471 {
1472 if (skb && uarg && !skb_zcopy(skb)) {
1473 if (unlikely(have_ref && *have_ref))
1474 *have_ref = false;
1475 else
1476 net_zcopy_get(uarg);
1477 skb_zcopy_init(skb, uarg);
1478 }
1479 }
1480
1481 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1482 {
1483 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1484 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1485 }
1486
1487 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1488 {
1489 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1490 }
1491
1492 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1493 {
1494 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1495 }
1496
1497 static inline void net_zcopy_put(struct ubuf_info *uarg)
1498 {
1499 if (uarg)
1500 uarg->callback(NULL, uarg, true);
1501 }
1502
1503 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1504 {
1505 if (uarg) {
1506 if (uarg->callback == msg_zerocopy_callback)
1507 msg_zerocopy_put_abort(uarg, have_uref);
1508 else if (have_uref)
1509 net_zcopy_put(uarg);
1510 }
1511 }
1512
1513 /* Release a reference on a zerocopy structure */
1514 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1515 {
1516 struct ubuf_info *uarg = skb_zcopy(skb);
1517
1518 if (uarg) {
1519 if (!skb_zcopy_is_nouarg(skb))
1520 uarg->callback(skb, uarg, zerocopy_success);
1521
1522 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
1523 }
1524 }
1525
1526 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1527 {
1528 skb->next = NULL;
1529 }
1530
1531 /* Iterate through singly-linked GSO fragments of an skb. */
1532 #define skb_list_walk_safe(first, skb, next_skb) \
1533 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1534 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1535
1536 static inline void skb_list_del_init(struct sk_buff *skb)
1537 {
1538 __list_del_entry(&skb->list);
1539 skb_mark_not_on_list(skb);
1540 }
1541
1542 /**
1543 * skb_queue_empty - check if a queue is empty
1544 * @list: queue head
1545 *
1546 * Returns true if the queue is empty, false otherwise.
1547 */
1548 static inline int skb_queue_empty(const struct sk_buff_head *list)
1549 {
1550 return list->next == (const struct sk_buff *) list;
1551 }
1552
1553 /**
1554 * skb_queue_empty_lockless - check if a queue is empty
1555 * @list: queue head
1556 *
1557 * Returns true if the queue is empty, false otherwise.
1558 * This variant can be used in lockless contexts.
1559 */
1560 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1561 {
1562 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1563 }
1564
1565
1566 /**
1567 * skb_queue_is_last - check if skb is the last entry in the queue
1568 * @list: queue head
1569 * @skb: buffer
1570 *
1571 * Returns true if @skb is the last buffer on the list.
1572 */
1573 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1574 const struct sk_buff *skb)
1575 {
1576 return skb->next == (const struct sk_buff *) list;
1577 }
1578
1579 /**
1580 * skb_queue_is_first - check if skb is the first entry in the queue
1581 * @list: queue head
1582 * @skb: buffer
1583 *
1584 * Returns true if @skb is the first buffer on the list.
1585 */
1586 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1587 const struct sk_buff *skb)
1588 {
1589 return skb->prev == (const struct sk_buff *) list;
1590 }
1591
1592 /**
1593 * skb_queue_next - return the next packet in the queue
1594 * @list: queue head
1595 * @skb: current buffer
1596 *
1597 * Return the next packet in @list after @skb. It is only valid to
1598 * call this if skb_queue_is_last() evaluates to false.
1599 */
1600 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1601 const struct sk_buff *skb)
1602 {
1603 /* This BUG_ON may seem severe, but if we just return then we
1604 * are going to dereference garbage.
1605 */
1606 BUG_ON(skb_queue_is_last(list, skb));
1607 return skb->next;
1608 }
1609
1610 /**
1611 * skb_queue_prev - return the prev packet in the queue
1612 * @list: queue head
1613 * @skb: current buffer
1614 *
1615 * Return the prev packet in @list before @skb. It is only valid to
1616 * call this if skb_queue_is_first() evaluates to false.
1617 */
1618 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1619 const struct sk_buff *skb)
1620 {
1621 /* This BUG_ON may seem severe, but if we just return then we
1622 * are going to dereference garbage.
1623 */
1624 BUG_ON(skb_queue_is_first(list, skb));
1625 return skb->prev;
1626 }
1627
1628 /**
1629 * skb_get - reference buffer
1630 * @skb: buffer to reference
1631 *
1632 * Makes another reference to a socket buffer and returns a pointer
1633 * to the buffer.
1634 */
1635 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1636 {
1637 refcount_inc(&skb->users);
1638 return skb;
1639 }
1640
1641 /*
1642 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1643 */
1644
1645 /**
1646 * skb_cloned - is the buffer a clone
1647 * @skb: buffer to check
1648 *
1649 * Returns true if the buffer was generated with skb_clone() and is
1650 * one of multiple shared copies of the buffer. Cloned buffers are
1651 * shared data so must not be written to under normal circumstances.
1652 */
1653 static inline int skb_cloned(const struct sk_buff *skb)
1654 {
1655 return skb->cloned &&
1656 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1657 }
1658
1659 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1660 {
1661 might_sleep_if(gfpflags_allow_blocking(pri));
1662
1663 if (skb_cloned(skb))
1664 return pskb_expand_head(skb, 0, 0, pri);
1665
1666 return 0;
1667 }
1668
1669 /**
1670 * skb_header_cloned - is the header a clone
1671 * @skb: buffer to check
1672 *
1673 * Returns true if modifying the header part of the buffer requires
1674 * the data to be copied.
1675 */
1676 static inline int skb_header_cloned(const struct sk_buff *skb)
1677 {
1678 int dataref;
1679
1680 if (!skb->cloned)
1681 return 0;
1682
1683 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1684 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1685 return dataref != 1;
1686 }
1687
1688 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1689 {
1690 might_sleep_if(gfpflags_allow_blocking(pri));
1691
1692 if (skb_header_cloned(skb))
1693 return pskb_expand_head(skb, 0, 0, pri);
1694
1695 return 0;
1696 }
1697
1698 /**
1699 * __skb_header_release - release reference to header
1700 * @skb: buffer to operate on
1701 */
1702 static inline void __skb_header_release(struct sk_buff *skb)
1703 {
1704 skb->nohdr = 1;
1705 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1706 }
1707
1708
1709 /**
1710 * skb_shared - is the buffer shared
1711 * @skb: buffer to check
1712 *
1713 * Returns true if more than one person has a reference to this
1714 * buffer.
1715 */
1716 static inline int skb_shared(const struct sk_buff *skb)
1717 {
1718 return refcount_read(&skb->users) != 1;
1719 }
1720
1721 /**
1722 * skb_share_check - check if buffer is shared and if so clone it
1723 * @skb: buffer to check
1724 * @pri: priority for memory allocation
1725 *
1726 * If the buffer is shared the buffer is cloned and the old copy
1727 * drops a reference. A new clone with a single reference is returned.
1728 * If the buffer is not shared the original buffer is returned. When
1729 * being called from interrupt status or with spinlocks held pri must
1730 * be GFP_ATOMIC.
1731 *
1732 * NULL is returned on a memory allocation failure.
1733 */
1734 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1735 {
1736 might_sleep_if(gfpflags_allow_blocking(pri));
1737 if (skb_shared(skb)) {
1738 struct sk_buff *nskb = skb_clone(skb, pri);
1739
1740 if (likely(nskb))
1741 consume_skb(skb);
1742 else
1743 kfree_skb(skb);
1744 skb = nskb;
1745 }
1746 return skb;
1747 }
1748
1749 /*
1750 * Copy shared buffers into a new sk_buff. We effectively do COW on
1751 * packets to handle cases where we have a local reader and forward
1752 * and a couple of other messy ones. The normal one is tcpdumping
1753 * a packet thats being forwarded.
1754 */
1755
1756 /**
1757 * skb_unshare - make a copy of a shared buffer
1758 * @skb: buffer to check
1759 * @pri: priority for memory allocation
1760 *
1761 * If the socket buffer is a clone then this function creates a new
1762 * copy of the data, drops a reference count on the old copy and returns
1763 * the new copy with the reference count at 1. If the buffer is not a clone
1764 * the original buffer is returned. When called with a spinlock held or
1765 * from interrupt state @pri must be %GFP_ATOMIC
1766 *
1767 * %NULL is returned on a memory allocation failure.
1768 */
1769 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1770 gfp_t pri)
1771 {
1772 might_sleep_if(gfpflags_allow_blocking(pri));
1773 if (skb_cloned(skb)) {
1774 struct sk_buff *nskb = skb_copy(skb, pri);
1775
1776 /* Free our shared copy */
1777 if (likely(nskb))
1778 consume_skb(skb);
1779 else
1780 kfree_skb(skb);
1781 skb = nskb;
1782 }
1783 return skb;
1784 }
1785
1786 /**
1787 * skb_peek - peek at the head of an &sk_buff_head
1788 * @list_: list to peek at
1789 *
1790 * Peek an &sk_buff. Unlike most other operations you _MUST_
1791 * be careful with this one. A peek leaves the buffer on the
1792 * list and someone else may run off with it. You must hold
1793 * the appropriate locks or have a private queue to do this.
1794 *
1795 * Returns %NULL for an empty list or a pointer to the head element.
1796 * The reference count is not incremented and the reference is therefore
1797 * volatile. Use with caution.
1798 */
1799 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1800 {
1801 struct sk_buff *skb = list_->next;
1802
1803 if (skb == (struct sk_buff *)list_)
1804 skb = NULL;
1805 return skb;
1806 }
1807
1808 /**
1809 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1810 * @list_: list to peek at
1811 *
1812 * Like skb_peek(), but the caller knows that the list is not empty.
1813 */
1814 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1815 {
1816 return list_->next;
1817 }
1818
1819 /**
1820 * skb_peek_next - peek skb following the given one from a queue
1821 * @skb: skb to start from
1822 * @list_: list to peek at
1823 *
1824 * Returns %NULL when the end of the list is met or a pointer to the
1825 * next element. The reference count is not incremented and the
1826 * reference is therefore volatile. Use with caution.
1827 */
1828 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1829 const struct sk_buff_head *list_)
1830 {
1831 struct sk_buff *next = skb->next;
1832
1833 if (next == (struct sk_buff *)list_)
1834 next = NULL;
1835 return next;
1836 }
1837
1838 /**
1839 * skb_peek_tail - peek at the tail of an &sk_buff_head
1840 * @list_: list to peek at
1841 *
1842 * Peek an &sk_buff. Unlike most other operations you _MUST_
1843 * be careful with this one. A peek leaves the buffer on the
1844 * list and someone else may run off with it. You must hold
1845 * the appropriate locks or have a private queue to do this.
1846 *
1847 * Returns %NULL for an empty list or a pointer to the tail element.
1848 * The reference count is not incremented and the reference is therefore
1849 * volatile. Use with caution.
1850 */
1851 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1852 {
1853 struct sk_buff *skb = READ_ONCE(list_->prev);
1854
1855 if (skb == (struct sk_buff *)list_)
1856 skb = NULL;
1857 return skb;
1858
1859 }
1860
1861 /**
1862 * skb_queue_len - get queue length
1863 * @list_: list to measure
1864 *
1865 * Return the length of an &sk_buff queue.
1866 */
1867 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1868 {
1869 return list_->qlen;
1870 }
1871
1872 /**
1873 * skb_queue_len_lockless - get queue length
1874 * @list_: list to measure
1875 *
1876 * Return the length of an &sk_buff queue.
1877 * This variant can be used in lockless contexts.
1878 */
1879 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1880 {
1881 return READ_ONCE(list_->qlen);
1882 }
1883
1884 /**
1885 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1886 * @list: queue to initialize
1887 *
1888 * This initializes only the list and queue length aspects of
1889 * an sk_buff_head object. This allows to initialize the list
1890 * aspects of an sk_buff_head without reinitializing things like
1891 * the spinlock. It can also be used for on-stack sk_buff_head
1892 * objects where the spinlock is known to not be used.
1893 */
1894 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1895 {
1896 list->prev = list->next = (struct sk_buff *)list;
1897 list->qlen = 0;
1898 }
1899
1900 /*
1901 * This function creates a split out lock class for each invocation;
1902 * this is needed for now since a whole lot of users of the skb-queue
1903 * infrastructure in drivers have different locking usage (in hardirq)
1904 * than the networking core (in softirq only). In the long run either the
1905 * network layer or drivers should need annotation to consolidate the
1906 * main types of usage into 3 classes.
1907 */
1908 static inline void skb_queue_head_init(struct sk_buff_head *list)
1909 {
1910 spin_lock_init(&list->lock);
1911 __skb_queue_head_init(list);
1912 }
1913
1914 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1915 struct lock_class_key *class)
1916 {
1917 skb_queue_head_init(list);
1918 lockdep_set_class(&list->lock, class);
1919 }
1920
1921 /*
1922 * Insert an sk_buff on a list.
1923 *
1924 * The "__skb_xxxx()" functions are the non-atomic ones that
1925 * can only be called with interrupts disabled.
1926 */
1927 static inline void __skb_insert(struct sk_buff *newsk,
1928 struct sk_buff *prev, struct sk_buff *next,
1929 struct sk_buff_head *list)
1930 {
1931 /* See skb_queue_empty_lockless() and skb_peek_tail()
1932 * for the opposite READ_ONCE()
1933 */
1934 WRITE_ONCE(newsk->next, next);
1935 WRITE_ONCE(newsk->prev, prev);
1936 WRITE_ONCE(next->prev, newsk);
1937 WRITE_ONCE(prev->next, newsk);
1938 list->qlen++;
1939 }
1940
1941 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1942 struct sk_buff *prev,
1943 struct sk_buff *next)
1944 {
1945 struct sk_buff *first = list->next;
1946 struct sk_buff *last = list->prev;
1947
1948 WRITE_ONCE(first->prev, prev);
1949 WRITE_ONCE(prev->next, first);
1950
1951 WRITE_ONCE(last->next, next);
1952 WRITE_ONCE(next->prev, last);
1953 }
1954
1955 /**
1956 * skb_queue_splice - join two skb lists, this is designed for stacks
1957 * @list: the new list to add
1958 * @head: the place to add it in the first list
1959 */
1960 static inline void skb_queue_splice(const struct sk_buff_head *list,
1961 struct sk_buff_head *head)
1962 {
1963 if (!skb_queue_empty(list)) {
1964 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1965 head->qlen += list->qlen;
1966 }
1967 }
1968
1969 /**
1970 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1971 * @list: the new list to add
1972 * @head: the place to add it in the first list
1973 *
1974 * The list at @list is reinitialised
1975 */
1976 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1977 struct sk_buff_head *head)
1978 {
1979 if (!skb_queue_empty(list)) {
1980 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1981 head->qlen += list->qlen;
1982 __skb_queue_head_init(list);
1983 }
1984 }
1985
1986 /**
1987 * skb_queue_splice_tail - join two skb lists, each list being a queue
1988 * @list: the new list to add
1989 * @head: the place to add it in the first list
1990 */
1991 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1992 struct sk_buff_head *head)
1993 {
1994 if (!skb_queue_empty(list)) {
1995 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1996 head->qlen += list->qlen;
1997 }
1998 }
1999
2000 /**
2001 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2002 * @list: the new list to add
2003 * @head: the place to add it in the first list
2004 *
2005 * Each of the lists is a queue.
2006 * The list at @list is reinitialised
2007 */
2008 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2009 struct sk_buff_head *head)
2010 {
2011 if (!skb_queue_empty(list)) {
2012 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2013 head->qlen += list->qlen;
2014 __skb_queue_head_init(list);
2015 }
2016 }
2017
2018 /**
2019 * __skb_queue_after - queue a buffer at the list head
2020 * @list: list to use
2021 * @prev: place after this buffer
2022 * @newsk: buffer to queue
2023 *
2024 * Queue a buffer int the middle 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 */
2029 static inline void __skb_queue_after(struct sk_buff_head *list,
2030 struct sk_buff *prev,
2031 struct sk_buff *newsk)
2032 {
2033 __skb_insert(newsk, prev, prev->next, list);
2034 }
2035
2036 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2037 struct sk_buff_head *list);
2038
2039 static inline void __skb_queue_before(struct sk_buff_head *list,
2040 struct sk_buff *next,
2041 struct sk_buff *newsk)
2042 {
2043 __skb_insert(newsk, next->prev, next, list);
2044 }
2045
2046 /**
2047 * __skb_queue_head - queue a buffer at the list head
2048 * @list: list to use
2049 * @newsk: buffer to queue
2050 *
2051 * Queue a buffer at the start of a list. This function takes no locks
2052 * and you must therefore hold required locks before calling it.
2053 *
2054 * A buffer cannot be placed on two lists at the same time.
2055 */
2056 static inline void __skb_queue_head(struct sk_buff_head *list,
2057 struct sk_buff *newsk)
2058 {
2059 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2060 }
2061 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2062
2063 /**
2064 * __skb_queue_tail - queue a buffer at the list tail
2065 * @list: list to use
2066 * @newsk: buffer to queue
2067 *
2068 * Queue a buffer at the end of a list. This function takes no locks
2069 * and you must therefore hold required locks before calling it.
2070 *
2071 * A buffer cannot be placed on two lists at the same time.
2072 */
2073 static inline void __skb_queue_tail(struct sk_buff_head *list,
2074 struct sk_buff *newsk)
2075 {
2076 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2077 }
2078 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2079
2080 /*
2081 * remove sk_buff from list. _Must_ be called atomically, and with
2082 * the list known..
2083 */
2084 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2085 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2086 {
2087 struct sk_buff *next, *prev;
2088
2089 WRITE_ONCE(list->qlen, list->qlen - 1);
2090 next = skb->next;
2091 prev = skb->prev;
2092 skb->next = skb->prev = NULL;
2093 WRITE_ONCE(next->prev, prev);
2094 WRITE_ONCE(prev->next, next);
2095 }
2096
2097 /**
2098 * __skb_dequeue - remove from the head of the queue
2099 * @list: list to dequeue from
2100 *
2101 * Remove the head of the list. This function does not take any locks
2102 * so must be used with appropriate locks held only. The head item is
2103 * returned or %NULL if the list is empty.
2104 */
2105 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2106 {
2107 struct sk_buff *skb = skb_peek(list);
2108 if (skb)
2109 __skb_unlink(skb, list);
2110 return skb;
2111 }
2112 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2113
2114 /**
2115 * __skb_dequeue_tail - remove from the tail of the queue
2116 * @list: list to dequeue from
2117 *
2118 * Remove the tail of the list. This function does not take any locks
2119 * so must be used with appropriate locks held only. The tail item is
2120 * returned or %NULL if the list is empty.
2121 */
2122 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2123 {
2124 struct sk_buff *skb = skb_peek_tail(list);
2125 if (skb)
2126 __skb_unlink(skb, list);
2127 return skb;
2128 }
2129 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2130
2131
2132 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2133 {
2134 return skb->data_len;
2135 }
2136
2137 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2138 {
2139 return skb->len - skb->data_len;
2140 }
2141
2142 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2143 {
2144 unsigned int i, len = 0;
2145
2146 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2147 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2148 return len;
2149 }
2150
2151 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2152 {
2153 return skb_headlen(skb) + __skb_pagelen(skb);
2154 }
2155
2156 /**
2157 * __skb_fill_page_desc - initialise a paged fragment in an skb
2158 * @skb: buffer containing fragment to be initialised
2159 * @i: paged fragment index to initialise
2160 * @page: the page to use for this fragment
2161 * @off: the offset to the data with @page
2162 * @size: the length of the data
2163 *
2164 * Initialises the @i'th fragment of @skb to point to &size bytes at
2165 * offset @off within @page.
2166 *
2167 * Does not take any additional reference on the fragment.
2168 */
2169 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2170 struct page *page, int off, int size)
2171 {
2172 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2173
2174 /*
2175 * Propagate page pfmemalloc to the skb if we can. The problem is
2176 * that not all callers have unique ownership of the page but rely
2177 * on page_is_pfmemalloc doing the right thing(tm).
2178 */
2179 frag->bv_page = page;
2180 frag->bv_offset = off;
2181 skb_frag_size_set(frag, size);
2182
2183 page = compound_head(page);
2184 if (page_is_pfmemalloc(page))
2185 skb->pfmemalloc = true;
2186 }
2187
2188 /**
2189 * skb_fill_page_desc - initialise a paged fragment in an skb
2190 * @skb: buffer containing fragment to be initialised
2191 * @i: paged fragment index to initialise
2192 * @page: the page to use for this fragment
2193 * @off: the offset to the data with @page
2194 * @size: the length of the data
2195 *
2196 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2197 * @skb to point to @size bytes at offset @off within @page. In
2198 * addition updates @skb such that @i is the last fragment.
2199 *
2200 * Does not take any additional reference on the fragment.
2201 */
2202 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2203 struct page *page, int off, int size)
2204 {
2205 __skb_fill_page_desc(skb, i, page, off, size);
2206 skb_shinfo(skb)->nr_frags = i + 1;
2207 }
2208
2209 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2210 int size, unsigned int truesize);
2211
2212 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2213 unsigned int truesize);
2214
2215 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2216
2217 #ifdef NET_SKBUFF_DATA_USES_OFFSET
2218 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2219 {
2220 return skb->head + skb->tail;
2221 }
2222
2223 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2224 {
2225 skb->tail = skb->data - skb->head;
2226 }
2227
2228 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2229 {
2230 skb_reset_tail_pointer(skb);
2231 skb->tail += offset;
2232 }
2233
2234 #else /* NET_SKBUFF_DATA_USES_OFFSET */
2235 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2236 {
2237 return skb->tail;
2238 }
2239
2240 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2241 {
2242 skb->tail = skb->data;
2243 }
2244
2245 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2246 {
2247 skb->tail = skb->data + offset;
2248 }
2249
2250 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2251
2252 /*
2253 * Add data to an sk_buff
2254 */
2255 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2256 void *skb_put(struct sk_buff *skb, unsigned int len);
2257 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2258 {
2259 void *tmp = skb_tail_pointer(skb);
2260 SKB_LINEAR_ASSERT(skb);
2261 skb->tail += len;
2262 skb->len += len;
2263 return tmp;
2264 }
2265
2266 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2267 {
2268 void *tmp = __skb_put(skb, len);
2269
2270 memset(tmp, 0, len);
2271 return tmp;
2272 }
2273
2274 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2275 unsigned int len)
2276 {
2277 void *tmp = __skb_put(skb, len);
2278
2279 memcpy(tmp, data, len);
2280 return tmp;
2281 }
2282
2283 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2284 {
2285 *(u8 *)__skb_put(skb, 1) = val;
2286 }
2287
2288 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2289 {
2290 void *tmp = skb_put(skb, len);
2291
2292 memset(tmp, 0, len);
2293
2294 return tmp;
2295 }
2296
2297 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2298 unsigned int len)
2299 {
2300 void *tmp = skb_put(skb, len);
2301
2302 memcpy(tmp, data, len);
2303
2304 return tmp;
2305 }
2306
2307 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2308 {
2309 *(u8 *)skb_put(skb, 1) = val;
2310 }
2311
2312 void *skb_push(struct sk_buff *skb, unsigned int len);
2313 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2314 {
2315 skb->data -= len;
2316 skb->len += len;
2317 return skb->data;
2318 }
2319
2320 void *skb_pull(struct sk_buff *skb, unsigned int len);
2321 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2322 {
2323 skb->len -= len;
2324 BUG_ON(skb->len < skb->data_len);
2325 return skb->data += len;
2326 }
2327
2328 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2329 {
2330 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2331 }
2332
2333 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2334
2335 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2336 {
2337 if (len > skb_headlen(skb) &&
2338 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2339 return NULL;
2340 skb->len -= len;
2341 return skb->data += len;
2342 }
2343
2344 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2345 {
2346 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2347 }
2348
2349 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2350 {
2351 if (likely(len <= skb_headlen(skb)))
2352 return true;
2353 if (unlikely(len > skb->len))
2354 return false;
2355 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2356 }
2357
2358 void skb_condense(struct sk_buff *skb);
2359
2360 /**
2361 * skb_headroom - bytes at buffer head
2362 * @skb: buffer to check
2363 *
2364 * Return the number of bytes of free space at the head of an &sk_buff.
2365 */
2366 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2367 {
2368 return skb->data - skb->head;
2369 }
2370
2371 /**
2372 * skb_tailroom - bytes at buffer end
2373 * @skb: buffer to check
2374 *
2375 * Return the number of bytes of free space at the tail of an sk_buff
2376 */
2377 static inline int skb_tailroom(const struct sk_buff *skb)
2378 {
2379 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2380 }
2381
2382 /**
2383 * skb_availroom - bytes at buffer end
2384 * @skb: buffer to check
2385 *
2386 * Return the number of bytes of free space at the tail of an sk_buff
2387 * allocated by sk_stream_alloc()
2388 */
2389 static inline int skb_availroom(const struct sk_buff *skb)
2390 {
2391 if (skb_is_nonlinear(skb))
2392 return 0;
2393
2394 return skb->end - skb->tail - skb->reserved_tailroom;
2395 }
2396
2397 /**
2398 * skb_reserve - adjust headroom
2399 * @skb: buffer to alter
2400 * @len: bytes to move
2401 *
2402 * Increase the headroom of an empty &sk_buff by reducing the tail
2403 * room. This is only allowed for an empty buffer.
2404 */
2405 static inline void skb_reserve(struct sk_buff *skb, int len)
2406 {
2407 skb->data += len;
2408 skb->tail += len;
2409 }
2410
2411 /**
2412 * skb_tailroom_reserve - adjust reserved_tailroom
2413 * @skb: buffer to alter
2414 * @mtu: maximum amount of headlen permitted
2415 * @needed_tailroom: minimum amount of reserved_tailroom
2416 *
2417 * Set reserved_tailroom so that headlen can be as large as possible but
2418 * not larger than mtu and tailroom cannot be smaller than
2419 * needed_tailroom.
2420 * The required headroom should already have been reserved before using
2421 * this function.
2422 */
2423 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2424 unsigned int needed_tailroom)
2425 {
2426 SKB_LINEAR_ASSERT(skb);
2427 if (mtu < skb_tailroom(skb) - needed_tailroom)
2428 /* use at most mtu */
2429 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2430 else
2431 /* use up to all available space */
2432 skb->reserved_tailroom = needed_tailroom;
2433 }
2434
2435 #define ENCAP_TYPE_ETHER 0
2436 #define ENCAP_TYPE_IPPROTO 1
2437
2438 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2439 __be16 protocol)
2440 {
2441 skb->inner_protocol = protocol;
2442 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2443 }
2444
2445 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2446 __u8 ipproto)
2447 {
2448 skb->inner_ipproto = ipproto;
2449 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2450 }
2451
2452 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2453 {
2454 skb->inner_mac_header = skb->mac_header;
2455 skb->inner_network_header = skb->network_header;
2456 skb->inner_transport_header = skb->transport_header;
2457 }
2458
2459 static inline void skb_reset_mac_len(struct sk_buff *skb)
2460 {
2461 skb->mac_len = skb->network_header - skb->mac_header;
2462 }
2463
2464 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2465 *skb)
2466 {
2467 return skb->head + skb->inner_transport_header;
2468 }
2469
2470 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2471 {
2472 return skb_inner_transport_header(skb) - skb->data;
2473 }
2474
2475 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2476 {
2477 skb->inner_transport_header = skb->data - skb->head;
2478 }
2479
2480 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2481 const int offset)
2482 {
2483 skb_reset_inner_transport_header(skb);
2484 skb->inner_transport_header += offset;
2485 }
2486
2487 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2488 {
2489 return skb->head + skb->inner_network_header;
2490 }
2491
2492 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2493 {
2494 skb->inner_network_header = skb->data - skb->head;
2495 }
2496
2497 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2498 const int offset)
2499 {
2500 skb_reset_inner_network_header(skb);
2501 skb->inner_network_header += offset;
2502 }
2503
2504 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2505 {
2506 return skb->head + skb->inner_mac_header;
2507 }
2508
2509 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2510 {
2511 skb->inner_mac_header = skb->data - skb->head;
2512 }
2513
2514 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2515 const int offset)
2516 {
2517 skb_reset_inner_mac_header(skb);
2518 skb->inner_mac_header += offset;
2519 }
2520 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2521 {
2522 return skb->transport_header != (typeof(skb->transport_header))~0U;
2523 }
2524
2525 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2526 {
2527 return skb->head + skb->transport_header;
2528 }
2529
2530 static inline void skb_reset_transport_header(struct sk_buff *skb)
2531 {
2532 skb->transport_header = skb->data - skb->head;
2533 }
2534
2535 static inline void skb_set_transport_header(struct sk_buff *skb,
2536 const int offset)
2537 {
2538 skb_reset_transport_header(skb);
2539 skb->transport_header += offset;
2540 }
2541
2542 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2543 {
2544 return skb->head + skb->network_header;
2545 }
2546
2547 static inline void skb_reset_network_header(struct sk_buff *skb)
2548 {
2549 skb->network_header = skb->data - skb->head;
2550 }
2551
2552 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2553 {
2554 skb_reset_network_header(skb);
2555 skb->network_header += offset;
2556 }
2557
2558 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2559 {
2560 return skb->head + skb->mac_header;
2561 }
2562
2563 static inline int skb_mac_offset(const struct sk_buff *skb)
2564 {
2565 return skb_mac_header(skb) - skb->data;
2566 }
2567
2568 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2569 {
2570 return skb->network_header - skb->mac_header;
2571 }
2572
2573 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2574 {
2575 return skb->mac_header != (typeof(skb->mac_header))~0U;
2576 }
2577
2578 static inline void skb_unset_mac_header(struct sk_buff *skb)
2579 {
2580 skb->mac_header = (typeof(skb->mac_header))~0U;
2581 }
2582
2583 static inline void skb_reset_mac_header(struct sk_buff *skb)
2584 {
2585 skb->mac_header = skb->data - skb->head;
2586 }
2587
2588 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2589 {
2590 skb_reset_mac_header(skb);
2591 skb->mac_header += offset;
2592 }
2593
2594 static inline void skb_pop_mac_header(struct sk_buff *skb)
2595 {
2596 skb->mac_header = skb->network_header;
2597 }
2598
2599 static inline void skb_probe_transport_header(struct sk_buff *skb)
2600 {
2601 struct flow_keys_basic keys;
2602
2603 if (skb_transport_header_was_set(skb))
2604 return;
2605
2606 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2607 NULL, 0, 0, 0, 0))
2608 skb_set_transport_header(skb, keys.control.thoff);
2609 }
2610
2611 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2612 {
2613 if (skb_mac_header_was_set(skb)) {
2614 const unsigned char *old_mac = skb_mac_header(skb);
2615
2616 skb_set_mac_header(skb, -skb->mac_len);
2617 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2618 }
2619 }
2620
2621 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2622 {
2623 return skb->csum_start - skb_headroom(skb);
2624 }
2625
2626 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2627 {
2628 return skb->head + skb->csum_start;
2629 }
2630
2631 static inline int skb_transport_offset(const struct sk_buff *skb)
2632 {
2633 return skb_transport_header(skb) - skb->data;
2634 }
2635
2636 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2637 {
2638 return skb->transport_header - skb->network_header;
2639 }
2640
2641 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2642 {
2643 return skb->inner_transport_header - skb->inner_network_header;
2644 }
2645
2646 static inline int skb_network_offset(const struct sk_buff *skb)
2647 {
2648 return skb_network_header(skb) - skb->data;
2649 }
2650
2651 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2652 {
2653 return skb_inner_network_header(skb) - skb->data;
2654 }
2655
2656 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2657 {
2658 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2659 }
2660
2661 /*
2662 * CPUs often take a performance hit when accessing unaligned memory
2663 * locations. The actual performance hit varies, it can be small if the
2664 * hardware handles it or large if we have to take an exception and fix it
2665 * in software.
2666 *
2667 * Since an ethernet header is 14 bytes network drivers often end up with
2668 * the IP header at an unaligned offset. The IP header can be aligned by
2669 * shifting the start of the packet by 2 bytes. Drivers should do this
2670 * with:
2671 *
2672 * skb_reserve(skb, NET_IP_ALIGN);
2673 *
2674 * The downside to this alignment of the IP header is that the DMA is now
2675 * unaligned. On some architectures the cost of an unaligned DMA is high
2676 * and this cost outweighs the gains made by aligning the IP header.
2677 *
2678 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2679 * to be overridden.
2680 */
2681 #ifndef NET_IP_ALIGN
2682 #define NET_IP_ALIGN 2
2683 #endif
2684
2685 /*
2686 * The networking layer reserves some headroom in skb data (via
2687 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2688 * the header has to grow. In the default case, if the header has to grow
2689 * 32 bytes or less we avoid the reallocation.
2690 *
2691 * Unfortunately this headroom changes the DMA alignment of the resulting
2692 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2693 * on some architectures. An architecture can override this value,
2694 * perhaps setting it to a cacheline in size (since that will maintain
2695 * cacheline alignment of the DMA). It must be a power of 2.
2696 *
2697 * Various parts of the networking layer expect at least 32 bytes of
2698 * headroom, you should not reduce this.
2699 *
2700 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2701 * to reduce average number of cache lines per packet.
2702 * get_rps_cpu() for example only access one 64 bytes aligned block :
2703 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2704 */
2705 #ifndef NET_SKB_PAD
2706 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2707 #endif
2708
2709 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2710
2711 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2712 {
2713 if (WARN_ON(skb_is_nonlinear(skb)))
2714 return;
2715 skb->len = len;
2716 skb_set_tail_pointer(skb, len);
2717 }
2718
2719 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2720 {
2721 __skb_set_length(skb, len);
2722 }
2723
2724 void skb_trim(struct sk_buff *skb, unsigned int len);
2725
2726 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2727 {
2728 if (skb->data_len)
2729 return ___pskb_trim(skb, len);
2730 __skb_trim(skb, len);
2731 return 0;
2732 }
2733
2734 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2735 {
2736 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2737 }
2738
2739 /**
2740 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2741 * @skb: buffer to alter
2742 * @len: new length
2743 *
2744 * This is identical to pskb_trim except that the caller knows that
2745 * the skb is not cloned so we should never get an error due to out-
2746 * of-memory.
2747 */
2748 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2749 {
2750 int err = pskb_trim(skb, len);
2751 BUG_ON(err);
2752 }
2753
2754 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2755 {
2756 unsigned int diff = len - skb->len;
2757
2758 if (skb_tailroom(skb) < diff) {
2759 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2760 GFP_ATOMIC);
2761 if (ret)
2762 return ret;
2763 }
2764 __skb_set_length(skb, len);
2765 return 0;
2766 }
2767
2768 /**
2769 * skb_orphan - orphan a buffer
2770 * @skb: buffer to orphan
2771 *
2772 * If a buffer currently has an owner then we call the owner's
2773 * destructor function and make the @skb unowned. The buffer continues
2774 * to exist but is no longer charged to its former owner.
2775 */
2776 static inline void skb_orphan(struct sk_buff *skb)
2777 {
2778 if (skb->destructor) {
2779 skb->destructor(skb);
2780 skb->destructor = NULL;
2781 skb->sk = NULL;
2782 } else {
2783 BUG_ON(skb->sk);
2784 }
2785 }
2786
2787 /**
2788 * skb_orphan_frags - orphan the frags contained in a buffer
2789 * @skb: buffer to orphan frags from
2790 * @gfp_mask: allocation mask for replacement pages
2791 *
2792 * For each frag in the SKB which needs a destructor (i.e. has an
2793 * owner) create a copy of that frag and release the original
2794 * page by calling the destructor.
2795 */
2796 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2797 {
2798 if (likely(!skb_zcopy(skb)))
2799 return 0;
2800 if (!skb_zcopy_is_nouarg(skb) &&
2801 skb_uarg(skb)->callback == msg_zerocopy_callback)
2802 return 0;
2803 return skb_copy_ubufs(skb, gfp_mask);
2804 }
2805
2806 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2807 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2808 {
2809 if (likely(!skb_zcopy(skb)))
2810 return 0;
2811 return skb_copy_ubufs(skb, gfp_mask);
2812 }
2813
2814 /**
2815 * __skb_queue_purge - empty a list
2816 * @list: list to empty
2817 *
2818 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2819 * the list and one reference dropped. This function does not take the
2820 * list lock and the caller must hold the relevant locks to use it.
2821 */
2822 static inline void __skb_queue_purge(struct sk_buff_head *list)
2823 {
2824 struct sk_buff *skb;
2825 while ((skb = __skb_dequeue(list)) != NULL)
2826 kfree_skb(skb);
2827 }
2828 void skb_queue_purge(struct sk_buff_head *list);
2829
2830 unsigned int skb_rbtree_purge(struct rb_root *root);
2831
2832 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2833
2834 /**
2835 * netdev_alloc_frag - allocate a page fragment
2836 * @fragsz: fragment size
2837 *
2838 * Allocates a frag from a page for receive buffer.
2839 * Uses GFP_ATOMIC allocations.
2840 */
2841 static inline void *netdev_alloc_frag(unsigned int fragsz)
2842 {
2843 return __netdev_alloc_frag_align(fragsz, ~0u);
2844 }
2845
2846 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2847 unsigned int align)
2848 {
2849 WARN_ON_ONCE(!is_power_of_2(align));
2850 return __netdev_alloc_frag_align(fragsz, -align);
2851 }
2852
2853 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2854 gfp_t gfp_mask);
2855
2856 /**
2857 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2858 * @dev: network device to receive on
2859 * @length: length to allocate
2860 *
2861 * Allocate a new &sk_buff and assign it a usage count of one. The
2862 * buffer has unspecified headroom built in. Users should allocate
2863 * the headroom they think they need without accounting for the
2864 * built in space. The built in space is used for optimisations.
2865 *
2866 * %NULL is returned if there is no free memory. Although this function
2867 * allocates memory it can be called from an interrupt.
2868 */
2869 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2870 unsigned int length)
2871 {
2872 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2873 }
2874
2875 /* legacy helper around __netdev_alloc_skb() */
2876 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2877 gfp_t gfp_mask)
2878 {
2879 return __netdev_alloc_skb(NULL, length, gfp_mask);
2880 }
2881
2882 /* legacy helper around netdev_alloc_skb() */
2883 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2884 {
2885 return netdev_alloc_skb(NULL, length);
2886 }
2887
2888
2889 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2890 unsigned int length, gfp_t gfp)
2891 {
2892 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2893
2894 if (NET_IP_ALIGN && skb)
2895 skb_reserve(skb, NET_IP_ALIGN);
2896 return skb;
2897 }
2898
2899 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2900 unsigned int length)
2901 {
2902 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2903 }
2904
2905 static inline void skb_free_frag(void *addr)
2906 {
2907 page_frag_free(addr);
2908 }
2909
2910 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2911
2912 static inline void *napi_alloc_frag(unsigned int fragsz)
2913 {
2914 return __napi_alloc_frag_align(fragsz, ~0u);
2915 }
2916
2917 static inline void *napi_alloc_frag_align(unsigned int fragsz,
2918 unsigned int align)
2919 {
2920 WARN_ON_ONCE(!is_power_of_2(align));
2921 return __napi_alloc_frag_align(fragsz, -align);
2922 }
2923
2924 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2925 unsigned int length, gfp_t gfp_mask);
2926 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2927 unsigned int length)
2928 {
2929 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2930 }
2931 void napi_consume_skb(struct sk_buff *skb, int budget);
2932
2933 void napi_skb_free_stolen_head(struct sk_buff *skb);
2934 void __kfree_skb_defer(struct sk_buff *skb);
2935
2936 /**
2937 * __dev_alloc_pages - allocate page for network Rx
2938 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2939 * @order: size of the allocation
2940 *
2941 * Allocate a new page.
2942 *
2943 * %NULL is returned if there is no free memory.
2944 */
2945 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2946 unsigned int order)
2947 {
2948 /* This piece of code contains several assumptions.
2949 * 1. This is for device Rx, therefor a cold page is preferred.
2950 * 2. The expectation is the user wants a compound page.
2951 * 3. If requesting a order 0 page it will not be compound
2952 * due to the check to see if order has a value in prep_new_page
2953 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2954 * code in gfp_to_alloc_flags that should be enforcing this.
2955 */
2956 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2957
2958 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2959 }
2960
2961 static inline struct page *dev_alloc_pages(unsigned int order)
2962 {
2963 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2964 }
2965
2966 /**
2967 * __dev_alloc_page - allocate a page for network Rx
2968 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2969 *
2970 * Allocate a new page.
2971 *
2972 * %NULL is returned if there is no free memory.
2973 */
2974 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2975 {
2976 return __dev_alloc_pages(gfp_mask, 0);
2977 }
2978
2979 static inline struct page *dev_alloc_page(void)
2980 {
2981 return dev_alloc_pages(0);
2982 }
2983
2984 /**
2985 * dev_page_is_reusable - check whether a page can be reused for network Rx
2986 * @page: the page to test
2987 *
2988 * A page shouldn't be considered for reusing/recycling if it was allocated
2989 * under memory pressure or at a distant memory node.
2990 *
2991 * Returns false if this page should be returned to page allocator, true
2992 * otherwise.
2993 */
2994 static inline bool dev_page_is_reusable(const struct page *page)
2995 {
2996 return likely(page_to_nid(page) == numa_mem_id() &&
2997 !page_is_pfmemalloc(page));
2998 }
2999
3000 /**
3001 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3002 * @page: The page that was allocated from skb_alloc_page
3003 * @skb: The skb that may need pfmemalloc set
3004 */
3005 static inline void skb_propagate_pfmemalloc(const struct page *page,
3006 struct sk_buff *skb)
3007 {
3008 if (page_is_pfmemalloc(page))
3009 skb->pfmemalloc = true;
3010 }
3011
3012 /**
3013 * skb_frag_off() - Returns the offset of a skb fragment
3014 * @frag: the paged fragment
3015 */
3016 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3017 {
3018 return frag->bv_offset;
3019 }
3020
3021 /**
3022 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3023 * @frag: skb fragment
3024 * @delta: value to add
3025 */
3026 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3027 {
3028 frag->bv_offset += delta;
3029 }
3030
3031 /**
3032 * skb_frag_off_set() - Sets the offset of a skb fragment
3033 * @frag: skb fragment
3034 * @offset: offset of fragment
3035 */
3036 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3037 {
3038 frag->bv_offset = offset;
3039 }
3040
3041 /**
3042 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3043 * @fragto: skb fragment where offset is set
3044 * @fragfrom: skb fragment offset is copied from
3045 */
3046 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3047 const skb_frag_t *fragfrom)
3048 {
3049 fragto->bv_offset = fragfrom->bv_offset;
3050 }
3051
3052 /**
3053 * skb_frag_page - retrieve the page referred to by a paged fragment
3054 * @frag: the paged fragment
3055 *
3056 * Returns the &struct page associated with @frag.
3057 */
3058 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3059 {
3060 return frag->bv_page;
3061 }
3062
3063 /**
3064 * __skb_frag_ref - take an addition reference on a paged fragment.
3065 * @frag: the paged fragment
3066 *
3067 * Takes an additional reference on the paged fragment @frag.
3068 */
3069 static inline void __skb_frag_ref(skb_frag_t *frag)
3070 {
3071 get_page(skb_frag_page(frag));
3072 }
3073
3074 /**
3075 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3076 * @skb: the buffer
3077 * @f: the fragment offset.
3078 *
3079 * Takes an additional reference on the @f'th paged fragment of @skb.
3080 */
3081 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3082 {
3083 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3084 }
3085
3086 /**
3087 * __skb_frag_unref - release a reference on a paged fragment.
3088 * @frag: the paged fragment
3089 * @recycle: recycle the page if allocated via page_pool
3090 *
3091 * Releases a reference on the paged fragment @frag
3092 * or recycles the page via the page_pool API.
3093 */
3094 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3095 {
3096 struct page *page = skb_frag_page(frag);
3097
3098 #ifdef CONFIG_PAGE_POOL
3099 if (recycle && page_pool_return_skb_page(page))
3100 return;
3101 #endif
3102 put_page(page);
3103 }
3104
3105 /**
3106 * skb_frag_unref - release a reference on a paged fragment of an skb.
3107 * @skb: the buffer
3108 * @f: the fragment offset
3109 *
3110 * Releases a reference on the @f'th paged fragment of @skb.
3111 */
3112 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3113 {
3114 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
3115 }
3116
3117 /**
3118 * skb_frag_address - gets the address of the data contained in a paged fragment
3119 * @frag: the paged fragment buffer
3120 *
3121 * Returns the address of the data within @frag. The page must already
3122 * be mapped.
3123 */
3124 static inline void *skb_frag_address(const skb_frag_t *frag)
3125 {
3126 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3127 }
3128
3129 /**
3130 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3131 * @frag: the paged fragment buffer
3132 *
3133 * Returns the address of the data within @frag. Checks that the page
3134 * is mapped and returns %NULL otherwise.
3135 */
3136 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3137 {
3138 void *ptr = page_address(skb_frag_page(frag));
3139 if (unlikely(!ptr))
3140 return NULL;
3141
3142 return ptr + skb_frag_off(frag);
3143 }
3144
3145 /**
3146 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3147 * @fragto: skb fragment where page is set
3148 * @fragfrom: skb fragment page is copied from
3149 */
3150 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3151 const skb_frag_t *fragfrom)
3152 {
3153 fragto->bv_page = fragfrom->bv_page;
3154 }
3155
3156 /**
3157 * __skb_frag_set_page - sets the page contained in a paged fragment
3158 * @frag: the paged fragment
3159 * @page: the page to set
3160 *
3161 * Sets the fragment @frag to contain @page.
3162 */
3163 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3164 {
3165 frag->bv_page = page;
3166 }
3167
3168 /**
3169 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3170 * @skb: the buffer
3171 * @f: the fragment offset
3172 * @page: the page to set
3173 *
3174 * Sets the @f'th fragment of @skb to contain @page.
3175 */
3176 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3177 struct page *page)
3178 {
3179 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3180 }
3181
3182 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3183
3184 /**
3185 * skb_frag_dma_map - maps a paged fragment via the DMA API
3186 * @dev: the device to map the fragment to
3187 * @frag: the paged fragment to map
3188 * @offset: the offset within the fragment (starting at the
3189 * fragment's own offset)
3190 * @size: the number of bytes to map
3191 * @dir: the direction of the mapping (``PCI_DMA_*``)
3192 *
3193 * Maps the page associated with @frag to @device.
3194 */
3195 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3196 const skb_frag_t *frag,
3197 size_t offset, size_t size,
3198 enum dma_data_direction dir)
3199 {
3200 return dma_map_page(dev, skb_frag_page(frag),
3201 skb_frag_off(frag) + offset, size, dir);
3202 }
3203
3204 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3205 gfp_t gfp_mask)
3206 {
3207 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3208 }
3209
3210
3211 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3212 gfp_t gfp_mask)
3213 {
3214 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3215 }
3216
3217
3218 /**
3219 * skb_clone_writable - is the header of a clone writable
3220 * @skb: buffer to check
3221 * @len: length up to which to write
3222 *
3223 * Returns true if modifying the header part of the cloned buffer
3224 * does not requires the data to be copied.
3225 */
3226 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3227 {
3228 return !skb_header_cloned(skb) &&
3229 skb_headroom(skb) + len <= skb->hdr_len;
3230 }
3231
3232 static inline int skb_try_make_writable(struct sk_buff *skb,
3233 unsigned int write_len)
3234 {
3235 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3236 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3237 }
3238
3239 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3240 int cloned)
3241 {
3242 int delta = 0;
3243
3244 if (headroom > skb_headroom(skb))
3245 delta = headroom - skb_headroom(skb);
3246
3247 if (delta || cloned)
3248 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3249 GFP_ATOMIC);
3250 return 0;
3251 }
3252
3253 /**
3254 * skb_cow - copy header of skb when it is required
3255 * @skb: buffer to cow
3256 * @headroom: needed headroom
3257 *
3258 * If the skb passed lacks sufficient headroom or its data part
3259 * is shared, data is reallocated. If reallocation fails, an error
3260 * is returned and original skb is not changed.
3261 *
3262 * The result is skb with writable area skb->head...skb->tail
3263 * and at least @headroom of space at head.
3264 */
3265 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3266 {
3267 return __skb_cow(skb, headroom, skb_cloned(skb));
3268 }
3269
3270 /**
3271 * skb_cow_head - skb_cow but only making the head writable
3272 * @skb: buffer to cow
3273 * @headroom: needed headroom
3274 *
3275 * This function is identical to skb_cow except that we replace the
3276 * skb_cloned check by skb_header_cloned. It should be used when
3277 * you only need to push on some header and do not need to modify
3278 * the data.
3279 */
3280 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3281 {
3282 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3283 }
3284
3285 /**
3286 * skb_padto - pad an skbuff up to a minimal size
3287 * @skb: buffer to pad
3288 * @len: minimal length
3289 *
3290 * Pads up a buffer to ensure the trailing bytes exist and are
3291 * blanked. If the buffer already contains sufficient data it
3292 * is untouched. Otherwise it is extended. Returns zero on
3293 * success. The skb is freed on error.
3294 */
3295 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3296 {
3297 unsigned int size = skb->len;
3298 if (likely(size >= len))
3299 return 0;
3300 return skb_pad(skb, len - size);
3301 }
3302
3303 /**
3304 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3305 * @skb: buffer to pad
3306 * @len: minimal length
3307 * @free_on_error: free buffer on error
3308 *
3309 * Pads up a buffer to ensure the trailing bytes exist and are
3310 * blanked. If the buffer already contains sufficient data it
3311 * is untouched. Otherwise it is extended. Returns zero on
3312 * success. The skb is freed on error if @free_on_error is true.
3313 */
3314 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3315 unsigned int len,
3316 bool free_on_error)
3317 {
3318 unsigned int size = skb->len;
3319
3320 if (unlikely(size < len)) {
3321 len -= size;
3322 if (__skb_pad(skb, len, free_on_error))
3323 return -ENOMEM;
3324 __skb_put(skb, len);
3325 }
3326 return 0;
3327 }
3328
3329 /**
3330 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3331 * @skb: buffer to pad
3332 * @len: minimal length
3333 *
3334 * Pads up a buffer to ensure the trailing bytes exist and are
3335 * blanked. If the buffer already contains sufficient data it
3336 * is untouched. Otherwise it is extended. Returns zero on
3337 * success. The skb is freed on error.
3338 */
3339 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3340 {
3341 return __skb_put_padto(skb, len, true);
3342 }
3343
3344 static inline int skb_add_data(struct sk_buff *skb,
3345 struct iov_iter *from, int copy)
3346 {
3347 const int off = skb->len;
3348
3349 if (skb->ip_summed == CHECKSUM_NONE) {
3350 __wsum csum = 0;
3351 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3352 &csum, from)) {
3353 skb->csum = csum_block_add(skb->csum, csum, off);
3354 return 0;
3355 }
3356 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3357 return 0;
3358
3359 __skb_trim(skb, off);
3360 return -EFAULT;
3361 }
3362
3363 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3364 const struct page *page, int off)
3365 {
3366 if (skb_zcopy(skb))
3367 return false;
3368 if (i) {
3369 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3370
3371 return page == skb_frag_page(frag) &&
3372 off == skb_frag_off(frag) + skb_frag_size(frag);
3373 }
3374 return false;
3375 }
3376
3377 static inline int __skb_linearize(struct sk_buff *skb)
3378 {
3379 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3380 }
3381
3382 /**
3383 * skb_linearize - convert paged skb to linear one
3384 * @skb: buffer to linarize
3385 *
3386 * If there is no free memory -ENOMEM is returned, otherwise zero
3387 * is returned and the old skb data released.
3388 */
3389 static inline int skb_linearize(struct sk_buff *skb)
3390 {
3391 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3392 }
3393
3394 /**
3395 * skb_has_shared_frag - can any frag be overwritten
3396 * @skb: buffer to test
3397 *
3398 * Return true if the skb has at least one frag that might be modified
3399 * by an external entity (as in vmsplice()/sendfile())
3400 */
3401 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3402 {
3403 return skb_is_nonlinear(skb) &&
3404 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3405 }
3406
3407 /**
3408 * skb_linearize_cow - make sure skb is linear and writable
3409 * @skb: buffer to process
3410 *
3411 * If there is no free memory -ENOMEM is returned, otherwise zero
3412 * is returned and the old skb data released.
3413 */
3414 static inline int skb_linearize_cow(struct sk_buff *skb)
3415 {
3416 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3417 __skb_linearize(skb) : 0;
3418 }
3419
3420 static __always_inline void
3421 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3422 unsigned int off)
3423 {
3424 if (skb->ip_summed == CHECKSUM_COMPLETE)
3425 skb->csum = csum_block_sub(skb->csum,
3426 csum_partial(start, len, 0), off);
3427 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3428 skb_checksum_start_offset(skb) < 0)
3429 skb->ip_summed = CHECKSUM_NONE;
3430 }
3431
3432 /**
3433 * skb_postpull_rcsum - update checksum for received skb after pull
3434 * @skb: buffer to update
3435 * @start: start of data before pull
3436 * @len: length of data pulled
3437 *
3438 * After doing a pull on a received packet, you need to call this to
3439 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3440 * CHECKSUM_NONE so that it can be recomputed from scratch.
3441 */
3442 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3443 const void *start, unsigned int len)
3444 {
3445 __skb_postpull_rcsum(skb, start, len, 0);
3446 }
3447
3448 static __always_inline void
3449 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3450 unsigned int off)
3451 {
3452 if (skb->ip_summed == CHECKSUM_COMPLETE)
3453 skb->csum = csum_block_add(skb->csum,
3454 csum_partial(start, len, 0), off);
3455 }
3456
3457 /**
3458 * skb_postpush_rcsum - update checksum for received skb after push
3459 * @skb: buffer to update
3460 * @start: start of data after push
3461 * @len: length of data pushed
3462 *
3463 * After doing a push on a received packet, you need to call this to
3464 * update the CHECKSUM_COMPLETE checksum.
3465 */
3466 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3467 const void *start, unsigned int len)
3468 {
3469 __skb_postpush_rcsum(skb, start, len, 0);
3470 }
3471
3472 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3473
3474 /**
3475 * skb_push_rcsum - push skb and update receive checksum
3476 * @skb: buffer to update
3477 * @len: length of data pulled
3478 *
3479 * This function performs an skb_push on the packet and updates
3480 * the CHECKSUM_COMPLETE checksum. It should be used on
3481 * receive path processing instead of skb_push unless you know
3482 * that the checksum difference is zero (e.g., a valid IP header)
3483 * or you are setting ip_summed to CHECKSUM_NONE.
3484 */
3485 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3486 {
3487 skb_push(skb, len);
3488 skb_postpush_rcsum(skb, skb->data, len);
3489 return skb->data;
3490 }
3491
3492 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3493 /**
3494 * pskb_trim_rcsum - trim received skb and update checksum
3495 * @skb: buffer to trim
3496 * @len: new length
3497 *
3498 * This is exactly the same as pskb_trim except that it ensures the
3499 * checksum of received packets are still valid after the operation.
3500 * It can change skb pointers.
3501 */
3502
3503 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3504 {
3505 if (likely(len >= skb->len))
3506 return 0;
3507 return pskb_trim_rcsum_slow(skb, len);
3508 }
3509
3510 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3511 {
3512 if (skb->ip_summed == CHECKSUM_COMPLETE)
3513 skb->ip_summed = CHECKSUM_NONE;
3514 __skb_trim(skb, len);
3515 return 0;
3516 }
3517
3518 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3519 {
3520 if (skb->ip_summed == CHECKSUM_COMPLETE)
3521 skb->ip_summed = CHECKSUM_NONE;
3522 return __skb_grow(skb, len);
3523 }
3524
3525 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3526 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3527 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3528 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3529 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3530
3531 #define skb_queue_walk(queue, skb) \
3532 for (skb = (queue)->next; \
3533 skb != (struct sk_buff *)(queue); \
3534 skb = skb->next)
3535
3536 #define skb_queue_walk_safe(queue, skb, tmp) \
3537 for (skb = (queue)->next, tmp = skb->next; \
3538 skb != (struct sk_buff *)(queue); \
3539 skb = tmp, tmp = skb->next)
3540
3541 #define skb_queue_walk_from(queue, skb) \
3542 for (; skb != (struct sk_buff *)(queue); \
3543 skb = skb->next)
3544
3545 #define skb_rbtree_walk(skb, root) \
3546 for (skb = skb_rb_first(root); skb != NULL; \
3547 skb = skb_rb_next(skb))
3548
3549 #define skb_rbtree_walk_from(skb) \
3550 for (; skb != NULL; \
3551 skb = skb_rb_next(skb))
3552
3553 #define skb_rbtree_walk_from_safe(skb, tmp) \
3554 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3555 skb = tmp)
3556
3557 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3558 for (tmp = skb->next; \
3559 skb != (struct sk_buff *)(queue); \
3560 skb = tmp, tmp = skb->next)
3561
3562 #define skb_queue_reverse_walk(queue, skb) \
3563 for (skb = (queue)->prev; \
3564 skb != (struct sk_buff *)(queue); \
3565 skb = skb->prev)
3566
3567 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3568 for (skb = (queue)->prev, tmp = skb->prev; \
3569 skb != (struct sk_buff *)(queue); \
3570 skb = tmp, tmp = skb->prev)
3571
3572 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3573 for (tmp = skb->prev; \
3574 skb != (struct sk_buff *)(queue); \
3575 skb = tmp, tmp = skb->prev)
3576
3577 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3578 {
3579 return skb_shinfo(skb)->frag_list != NULL;
3580 }
3581
3582 static inline void skb_frag_list_init(struct sk_buff *skb)
3583 {
3584 skb_shinfo(skb)->frag_list = NULL;
3585 }
3586
3587 #define skb_walk_frags(skb, iter) \
3588 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3589
3590
3591 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3592 int *err, long *timeo_p,
3593 const struct sk_buff *skb);
3594 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3595 struct sk_buff_head *queue,
3596 unsigned int flags,
3597 int *off, int *err,
3598 struct sk_buff **last);
3599 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3600 struct sk_buff_head *queue,
3601 unsigned int flags, int *off, int *err,
3602 struct sk_buff **last);
3603 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3604 struct sk_buff_head *sk_queue,
3605 unsigned int flags, int *off, int *err);
3606 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3607 int *err);
3608 __poll_t datagram_poll(struct file *file, struct socket *sock,
3609 struct poll_table_struct *wait);
3610 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3611 struct iov_iter *to, int size);
3612 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3613 struct msghdr *msg, int size)
3614 {
3615 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3616 }
3617 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3618 struct msghdr *msg);
3619 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3620 struct iov_iter *to, int len,
3621 struct ahash_request *hash);
3622 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3623 struct iov_iter *from, int len);
3624 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3625 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3626 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3627 static inline void skb_free_datagram_locked(struct sock *sk,
3628 struct sk_buff *skb)
3629 {
3630 __skb_free_datagram_locked(sk, skb, 0);
3631 }
3632 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3633 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3634 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3635 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3636 int len);
3637 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3638 struct pipe_inode_info *pipe, unsigned int len,
3639 unsigned int flags);
3640 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3641 int len);
3642 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3643 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3644 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3645 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3646 int len, int hlen);
3647 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3648 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3649 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3650 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3651 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3652 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3653 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3654 unsigned int offset);
3655 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3656 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3657 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3658 int skb_vlan_pop(struct sk_buff *skb);
3659 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3660 int skb_eth_pop(struct sk_buff *skb);
3661 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3662 const unsigned char *src);
3663 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3664 int mac_len, bool ethernet);
3665 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3666 bool ethernet);
3667 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3668 int skb_mpls_dec_ttl(struct sk_buff *skb);
3669 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3670 gfp_t gfp);
3671
3672 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3673 {
3674 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3675 }
3676
3677 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3678 {
3679 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3680 }
3681
3682 struct skb_checksum_ops {
3683 __wsum (*update)(const void *mem, int len, __wsum wsum);
3684 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3685 };
3686
3687 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3688
3689 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3690 __wsum csum, const struct skb_checksum_ops *ops);
3691 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3692 __wsum csum);
3693
3694 static inline void * __must_check
3695 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3696 const void *data, int hlen, void *buffer)
3697 {
3698 if (likely(hlen - offset >= len))
3699 return (void *)data + offset;
3700
3701 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3702 return NULL;
3703
3704 return buffer;
3705 }
3706
3707 static inline void * __must_check
3708 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3709 {
3710 return __skb_header_pointer(skb, offset, len, skb->data,
3711 skb_headlen(skb), buffer);
3712 }
3713
3714 /**
3715 * skb_needs_linearize - check if we need to linearize a given skb
3716 * depending on the given device features.
3717 * @skb: socket buffer to check
3718 * @features: net device features
3719 *
3720 * Returns true if either:
3721 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3722 * 2. skb is fragmented and the device does not support SG.
3723 */
3724 static inline bool skb_needs_linearize(struct sk_buff *skb,
3725 netdev_features_t features)
3726 {
3727 return skb_is_nonlinear(skb) &&
3728 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3729 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3730 }
3731
3732 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3733 void *to,
3734 const unsigned int len)
3735 {
3736 memcpy(to, skb->data, len);
3737 }
3738
3739 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3740 const int offset, void *to,
3741 const unsigned int len)
3742 {
3743 memcpy(to, skb->data + offset, len);
3744 }
3745
3746 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3747 const void *from,
3748 const unsigned int len)
3749 {
3750 memcpy(skb->data, from, len);
3751 }
3752
3753 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3754 const int offset,
3755 const void *from,
3756 const unsigned int len)
3757 {
3758 memcpy(skb->data + offset, from, len);
3759 }
3760
3761 void skb_init(void);
3762
3763 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3764 {
3765 return skb->tstamp;
3766 }
3767
3768 /**
3769 * skb_get_timestamp - get timestamp from a skb
3770 * @skb: skb to get stamp from
3771 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3772 *
3773 * Timestamps are stored in the skb as offsets to a base timestamp.
3774 * This function converts the offset back to a struct timeval and stores
3775 * it in stamp.
3776 */
3777 static inline void skb_get_timestamp(const struct sk_buff *skb,
3778 struct __kernel_old_timeval *stamp)
3779 {
3780 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3781 }
3782
3783 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3784 struct __kernel_sock_timeval *stamp)
3785 {
3786 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3787
3788 stamp->tv_sec = ts.tv_sec;
3789 stamp->tv_usec = ts.tv_nsec / 1000;
3790 }
3791
3792 static inline void skb_get_timestampns(const struct sk_buff *skb,
3793 struct __kernel_old_timespec *stamp)
3794 {
3795 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3796
3797 stamp->tv_sec = ts.tv_sec;
3798 stamp->tv_nsec = ts.tv_nsec;
3799 }
3800
3801 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3802 struct __kernel_timespec *stamp)
3803 {
3804 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3805
3806 stamp->tv_sec = ts.tv_sec;
3807 stamp->tv_nsec = ts.tv_nsec;
3808 }
3809
3810 static inline void __net_timestamp(struct sk_buff *skb)
3811 {
3812 skb->tstamp = ktime_get_real();
3813 }
3814
3815 static inline ktime_t net_timedelta(ktime_t t)
3816 {
3817 return ktime_sub(ktime_get_real(), t);
3818 }
3819
3820 static inline ktime_t net_invalid_timestamp(void)
3821 {
3822 return 0;
3823 }
3824
3825 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3826 {
3827 return skb_shinfo(skb)->meta_len;
3828 }
3829
3830 static inline void *skb_metadata_end(const struct sk_buff *skb)
3831 {
3832 return skb_mac_header(skb);
3833 }
3834
3835 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3836 const struct sk_buff *skb_b,
3837 u8 meta_len)
3838 {
3839 const void *a = skb_metadata_end(skb_a);
3840 const void *b = skb_metadata_end(skb_b);
3841 /* Using more efficient varaiant than plain call to memcmp(). */
3842 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3843 u64 diffs = 0;
3844
3845 switch (meta_len) {
3846 #define __it(x, op) (x -= sizeof(u##op))
3847 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3848 case 32: diffs |= __it_diff(a, b, 64);
3849 fallthrough;
3850 case 24: diffs |= __it_diff(a, b, 64);
3851 fallthrough;
3852 case 16: diffs |= __it_diff(a, b, 64);
3853 fallthrough;
3854 case 8: diffs |= __it_diff(a, b, 64);
3855 break;
3856 case 28: diffs |= __it_diff(a, b, 64);
3857 fallthrough;
3858 case 20: diffs |= __it_diff(a, b, 64);
3859 fallthrough;
3860 case 12: diffs |= __it_diff(a, b, 64);
3861 fallthrough;
3862 case 4: diffs |= __it_diff(a, b, 32);
3863 break;
3864 }
3865 return diffs;
3866 #else
3867 return memcmp(a - meta_len, b - meta_len, meta_len);
3868 #endif
3869 }
3870
3871 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3872 const struct sk_buff *skb_b)
3873 {
3874 u8 len_a = skb_metadata_len(skb_a);
3875 u8 len_b = skb_metadata_len(skb_b);
3876
3877 if (!(len_a | len_b))
3878 return false;
3879
3880 return len_a != len_b ?
3881 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3882 }
3883
3884 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3885 {
3886 skb_shinfo(skb)->meta_len = meta_len;
3887 }
3888
3889 static inline void skb_metadata_clear(struct sk_buff *skb)
3890 {
3891 skb_metadata_set(skb, 0);
3892 }
3893
3894 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3895
3896 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3897
3898 void skb_clone_tx_timestamp(struct sk_buff *skb);
3899 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3900
3901 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3902
3903 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3904 {
3905 }
3906
3907 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3908 {
3909 return false;
3910 }
3911
3912 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3913
3914 /**
3915 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3916 *
3917 * PHY drivers may accept clones of transmitted packets for
3918 * timestamping via their phy_driver.txtstamp method. These drivers
3919 * must call this function to return the skb back to the stack with a
3920 * timestamp.
3921 *
3922 * @skb: clone of the original outgoing packet
3923 * @hwtstamps: hardware time stamps
3924 *
3925 */
3926 void skb_complete_tx_timestamp(struct sk_buff *skb,
3927 struct skb_shared_hwtstamps *hwtstamps);
3928
3929 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
3930 struct skb_shared_hwtstamps *hwtstamps,
3931 struct sock *sk, int tstype);
3932
3933 /**
3934 * skb_tstamp_tx - queue clone of skb with send time stamps
3935 * @orig_skb: the original outgoing packet
3936 * @hwtstamps: hardware time stamps, may be NULL if not available
3937 *
3938 * If the skb has a socket associated, then this function clones the
3939 * skb (thus sharing the actual data and optional structures), stores
3940 * the optional hardware time stamping information (if non NULL) or
3941 * generates a software time stamp (otherwise), then queues the clone
3942 * to the error queue of the socket. Errors are silently ignored.
3943 */
3944 void skb_tstamp_tx(struct sk_buff *orig_skb,
3945 struct skb_shared_hwtstamps *hwtstamps);
3946
3947 /**
3948 * skb_tx_timestamp() - Driver hook for transmit timestamping
3949 *
3950 * Ethernet MAC Drivers should call this function in their hard_xmit()
3951 * function immediately before giving the sk_buff to the MAC hardware.
3952 *
3953 * Specifically, one should make absolutely sure that this function is
3954 * called before TX completion of this packet can trigger. Otherwise
3955 * the packet could potentially already be freed.
3956 *
3957 * @skb: A socket buffer.
3958 */
3959 static inline void skb_tx_timestamp(struct sk_buff *skb)
3960 {
3961 skb_clone_tx_timestamp(skb);
3962 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3963 skb_tstamp_tx(skb, NULL);
3964 }
3965
3966 /**
3967 * skb_complete_wifi_ack - deliver skb with wifi status
3968 *
3969 * @skb: the original outgoing packet
3970 * @acked: ack status
3971 *
3972 */
3973 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3974
3975 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3976 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3977
3978 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3979 {
3980 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3981 skb->csum_valid ||
3982 (skb->ip_summed == CHECKSUM_PARTIAL &&
3983 skb_checksum_start_offset(skb) >= 0));
3984 }
3985
3986 /**
3987 * skb_checksum_complete - Calculate checksum of an entire packet
3988 * @skb: packet to process
3989 *
3990 * This function calculates the checksum over the entire packet plus
3991 * the value of skb->csum. The latter can be used to supply the
3992 * checksum of a pseudo header as used by TCP/UDP. It returns the
3993 * checksum.
3994 *
3995 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3996 * this function can be used to verify that checksum on received
3997 * packets. In that case the function should return zero if the
3998 * checksum is correct. In particular, this function will return zero
3999 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4000 * hardware has already verified the correctness of the checksum.
4001 */
4002 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4003 {
4004 return skb_csum_unnecessary(skb) ?
4005 0 : __skb_checksum_complete(skb);
4006 }
4007
4008 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4009 {
4010 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4011 if (skb->csum_level == 0)
4012 skb->ip_summed = CHECKSUM_NONE;
4013 else
4014 skb->csum_level--;
4015 }
4016 }
4017
4018 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4019 {
4020 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4021 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4022 skb->csum_level++;
4023 } else if (skb->ip_summed == CHECKSUM_NONE) {
4024 skb->ip_summed = CHECKSUM_UNNECESSARY;
4025 skb->csum_level = 0;
4026 }
4027 }
4028
4029 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4030 {
4031 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4032 skb->ip_summed = CHECKSUM_NONE;
4033 skb->csum_level = 0;
4034 }
4035 }
4036
4037 /* Check if we need to perform checksum complete validation.
4038 *
4039 * Returns true if checksum complete is needed, false otherwise
4040 * (either checksum is unnecessary or zero checksum is allowed).
4041 */
4042 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4043 bool zero_okay,
4044 __sum16 check)
4045 {
4046 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4047 skb->csum_valid = 1;
4048 __skb_decr_checksum_unnecessary(skb);
4049 return false;
4050 }
4051
4052 return true;
4053 }
4054
4055 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4056 * in checksum_init.
4057 */
4058 #define CHECKSUM_BREAK 76
4059
4060 /* Unset checksum-complete
4061 *
4062 * Unset checksum complete can be done when packet is being modified
4063 * (uncompressed for instance) and checksum-complete value is
4064 * invalidated.
4065 */
4066 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4067 {
4068 if (skb->ip_summed == CHECKSUM_COMPLETE)
4069 skb->ip_summed = CHECKSUM_NONE;
4070 }
4071
4072 /* Validate (init) checksum based on checksum complete.
4073 *
4074 * Return values:
4075 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4076 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4077 * checksum is stored in skb->csum for use in __skb_checksum_complete
4078 * non-zero: value of invalid checksum
4079 *
4080 */
4081 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4082 bool complete,
4083 __wsum psum)
4084 {
4085 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4086 if (!csum_fold(csum_add(psum, skb->csum))) {
4087 skb->csum_valid = 1;
4088 return 0;
4089 }
4090 }
4091
4092 skb->csum = psum;
4093
4094 if (complete || skb->len <= CHECKSUM_BREAK) {
4095 __sum16 csum;
4096
4097 csum = __skb_checksum_complete(skb);
4098 skb->csum_valid = !csum;
4099 return csum;
4100 }
4101
4102 return 0;
4103 }
4104
4105 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4106 {
4107 return 0;
4108 }
4109
4110 /* Perform checksum validate (init). Note that this is a macro since we only
4111 * want to calculate the pseudo header which is an input function if necessary.
4112 * First we try to validate without any computation (checksum unnecessary) and
4113 * then calculate based on checksum complete calling the function to compute
4114 * pseudo header.
4115 *
4116 * Return values:
4117 * 0: checksum is validated or try to in skb_checksum_complete
4118 * non-zero: value of invalid checksum
4119 */
4120 #define __skb_checksum_validate(skb, proto, complete, \
4121 zero_okay, check, compute_pseudo) \
4122 ({ \
4123 __sum16 __ret = 0; \
4124 skb->csum_valid = 0; \
4125 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4126 __ret = __skb_checksum_validate_complete(skb, \
4127 complete, compute_pseudo(skb, proto)); \
4128 __ret; \
4129 })
4130
4131 #define skb_checksum_init(skb, proto, compute_pseudo) \
4132 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4133
4134 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4135 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4136
4137 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4138 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4139
4140 #define skb_checksum_validate_zero_check(skb, proto, check, \
4141 compute_pseudo) \
4142 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4143
4144 #define skb_checksum_simple_validate(skb) \
4145 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4146
4147 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4148 {
4149 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4150 }
4151
4152 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4153 {
4154 skb->csum = ~pseudo;
4155 skb->ip_summed = CHECKSUM_COMPLETE;
4156 }
4157
4158 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4159 do { \
4160 if (__skb_checksum_convert_check(skb)) \
4161 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4162 } while (0)
4163
4164 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4165 u16 start, u16 offset)
4166 {
4167 skb->ip_summed = CHECKSUM_PARTIAL;
4168 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4169 skb->csum_offset = offset - start;
4170 }
4171
4172 /* Update skbuf and packet to reflect the remote checksum offload operation.
4173 * When called, ptr indicates the starting point for skb->csum when
4174 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4175 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4176 */
4177 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4178 int start, int offset, bool nopartial)
4179 {
4180 __wsum delta;
4181
4182 if (!nopartial) {
4183 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4184 return;
4185 }
4186
4187 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4188 __skb_checksum_complete(skb);
4189 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4190 }
4191
4192 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4193
4194 /* Adjust skb->csum since we changed the packet */
4195 skb->csum = csum_add(skb->csum, delta);
4196 }
4197
4198 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4199 {
4200 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4201 return (void *)(skb->_nfct & NFCT_PTRMASK);
4202 #else
4203 return NULL;
4204 #endif
4205 }
4206
4207 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4208 {
4209 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4210 return skb->_nfct;
4211 #else
4212 return 0UL;
4213 #endif
4214 }
4215
4216 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4217 {
4218 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4219 skb->_nfct = nfct;
4220 #endif
4221 }
4222
4223 #ifdef CONFIG_SKB_EXTENSIONS
4224 enum skb_ext_id {
4225 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4226 SKB_EXT_BRIDGE_NF,
4227 #endif
4228 #ifdef CONFIG_XFRM
4229 SKB_EXT_SEC_PATH,
4230 #endif
4231 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4232 TC_SKB_EXT,
4233 #endif
4234 #if IS_ENABLED(CONFIG_MPTCP)
4235 SKB_EXT_MPTCP,
4236 #endif
4237 SKB_EXT_NUM, /* must be last */
4238 };
4239
4240 /**
4241 * struct skb_ext - sk_buff extensions
4242 * @refcnt: 1 on allocation, deallocated on 0
4243 * @offset: offset to add to @data to obtain extension address
4244 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4245 * @data: start of extension data, variable sized
4246 *
4247 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4248 * to use 'u8' types while allowing up to 2kb worth of extension data.
4249 */
4250 struct skb_ext {
4251 refcount_t refcnt;
4252 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4253 u8 chunks; /* same */
4254 char data[] __aligned(8);
4255 };
4256
4257 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4258 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4259 struct skb_ext *ext);
4260 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4261 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4262 void __skb_ext_put(struct skb_ext *ext);
4263
4264 static inline void skb_ext_put(struct sk_buff *skb)
4265 {
4266 if (skb->active_extensions)
4267 __skb_ext_put(skb->extensions);
4268 }
4269
4270 static inline void __skb_ext_copy(struct sk_buff *dst,
4271 const struct sk_buff *src)
4272 {
4273 dst->active_extensions = src->active_extensions;
4274
4275 if (src->active_extensions) {
4276 struct skb_ext *ext = src->extensions;
4277
4278 refcount_inc(&ext->refcnt);
4279 dst->extensions = ext;
4280 }
4281 }
4282
4283 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4284 {
4285 skb_ext_put(dst);
4286 __skb_ext_copy(dst, src);
4287 }
4288
4289 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4290 {
4291 return !!ext->offset[i];
4292 }
4293
4294 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4295 {
4296 return skb->active_extensions & (1 << id);
4297 }
4298
4299 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4300 {
4301 if (skb_ext_exist(skb, id))
4302 __skb_ext_del(skb, id);
4303 }
4304
4305 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4306 {
4307 if (skb_ext_exist(skb, id)) {
4308 struct skb_ext *ext = skb->extensions;
4309
4310 return (void *)ext + (ext->offset[id] << 3);
4311 }
4312
4313 return NULL;
4314 }
4315
4316 static inline void skb_ext_reset(struct sk_buff *skb)
4317 {
4318 if (unlikely(skb->active_extensions)) {
4319 __skb_ext_put(skb->extensions);
4320 skb->active_extensions = 0;
4321 }
4322 }
4323
4324 static inline bool skb_has_extensions(struct sk_buff *skb)
4325 {
4326 return unlikely(skb->active_extensions);
4327 }
4328 #else
4329 static inline void skb_ext_put(struct sk_buff *skb) {}
4330 static inline void skb_ext_reset(struct sk_buff *skb) {}
4331 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4332 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4333 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4334 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4335 #endif /* CONFIG_SKB_EXTENSIONS */
4336
4337 static inline void nf_reset_ct(struct sk_buff *skb)
4338 {
4339 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4340 nf_conntrack_put(skb_nfct(skb));
4341 skb->_nfct = 0;
4342 #endif
4343 }
4344
4345 static inline void nf_reset_trace(struct sk_buff *skb)
4346 {
4347 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4348 skb->nf_trace = 0;
4349 #endif
4350 }
4351
4352 static inline void ipvs_reset(struct sk_buff *skb)
4353 {
4354 #if IS_ENABLED(CONFIG_IP_VS)
4355 skb->ipvs_property = 0;
4356 #endif
4357 }
4358
4359 /* Note: This doesn't put any conntrack info in dst. */
4360 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4361 bool copy)
4362 {
4363 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4364 dst->_nfct = src->_nfct;
4365 nf_conntrack_get(skb_nfct(src));
4366 #endif
4367 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4368 if (copy)
4369 dst->nf_trace = src->nf_trace;
4370 #endif
4371 }
4372
4373 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4374 {
4375 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4376 nf_conntrack_put(skb_nfct(dst));
4377 #endif
4378 __nf_copy(dst, src, true);
4379 }
4380
4381 #ifdef CONFIG_NETWORK_SECMARK
4382 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4383 {
4384 to->secmark = from->secmark;
4385 }
4386
4387 static inline void skb_init_secmark(struct sk_buff *skb)
4388 {
4389 skb->secmark = 0;
4390 }
4391 #else
4392 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4393 { }
4394
4395 static inline void skb_init_secmark(struct sk_buff *skb)
4396 { }
4397 #endif
4398
4399 static inline int secpath_exists(const struct sk_buff *skb)
4400 {
4401 #ifdef CONFIG_XFRM
4402 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4403 #else
4404 return 0;
4405 #endif
4406 }
4407
4408 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4409 {
4410 return !skb->destructor &&
4411 !secpath_exists(skb) &&
4412 !skb_nfct(skb) &&
4413 !skb->_skb_refdst &&
4414 !skb_has_frag_list(skb);
4415 }
4416
4417 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4418 {
4419 skb->queue_mapping = queue_mapping;
4420 }
4421
4422 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4423 {
4424 return skb->queue_mapping;
4425 }
4426
4427 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4428 {
4429 to->queue_mapping = from->queue_mapping;
4430 }
4431
4432 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4433 {
4434 skb->queue_mapping = rx_queue + 1;
4435 }
4436
4437 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4438 {
4439 return skb->queue_mapping - 1;
4440 }
4441
4442 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4443 {
4444 return skb->queue_mapping != 0;
4445 }
4446
4447 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4448 {
4449 skb->dst_pending_confirm = val;
4450 }
4451
4452 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4453 {
4454 return skb->dst_pending_confirm != 0;
4455 }
4456
4457 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4458 {
4459 #ifdef CONFIG_XFRM
4460 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4461 #else
4462 return NULL;
4463 #endif
4464 }
4465
4466 /* Keeps track of mac header offset relative to skb->head.
4467 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4468 * For non-tunnel skb it points to skb_mac_header() and for
4469 * tunnel skb it points to outer mac header.
4470 * Keeps track of level of encapsulation of network headers.
4471 */
4472 struct skb_gso_cb {
4473 union {
4474 int mac_offset;
4475 int data_offset;
4476 };
4477 int encap_level;
4478 __wsum csum;
4479 __u16 csum_start;
4480 };
4481 #define SKB_GSO_CB_OFFSET 32
4482 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4483
4484 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4485 {
4486 return (skb_mac_header(inner_skb) - inner_skb->head) -
4487 SKB_GSO_CB(inner_skb)->mac_offset;
4488 }
4489
4490 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4491 {
4492 int new_headroom, headroom;
4493 int ret;
4494
4495 headroom = skb_headroom(skb);
4496 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4497 if (ret)
4498 return ret;
4499
4500 new_headroom = skb_headroom(skb);
4501 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4502 return 0;
4503 }
4504
4505 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4506 {
4507 /* Do not update partial checksums if remote checksum is enabled. */
4508 if (skb->remcsum_offload)
4509 return;
4510
4511 SKB_GSO_CB(skb)->csum = res;
4512 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4513 }
4514
4515 /* Compute the checksum for a gso segment. First compute the checksum value
4516 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4517 * then add in skb->csum (checksum from csum_start to end of packet).
4518 * skb->csum and csum_start are then updated to reflect the checksum of the
4519 * resultant packet starting from the transport header-- the resultant checksum
4520 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4521 * header.
4522 */
4523 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4524 {
4525 unsigned char *csum_start = skb_transport_header(skb);
4526 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4527 __wsum partial = SKB_GSO_CB(skb)->csum;
4528
4529 SKB_GSO_CB(skb)->csum = res;
4530 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4531
4532 return csum_fold(csum_partial(csum_start, plen, partial));
4533 }
4534
4535 static inline bool skb_is_gso(const struct sk_buff *skb)
4536 {
4537 return skb_shinfo(skb)->gso_size;
4538 }
4539
4540 /* Note: Should be called only if skb_is_gso(skb) is true */
4541 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4542 {
4543 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4544 }
4545
4546 /* Note: Should be called only if skb_is_gso(skb) is true */
4547 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4548 {
4549 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4550 }
4551
4552 /* Note: Should be called only if skb_is_gso(skb) is true */
4553 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4554 {
4555 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4556 }
4557
4558 static inline void skb_gso_reset(struct sk_buff *skb)
4559 {
4560 skb_shinfo(skb)->gso_size = 0;
4561 skb_shinfo(skb)->gso_segs = 0;
4562 skb_shinfo(skb)->gso_type = 0;
4563 }
4564
4565 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4566 u16 increment)
4567 {
4568 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4569 return;
4570 shinfo->gso_size += increment;
4571 }
4572
4573 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4574 u16 decrement)
4575 {
4576 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4577 return;
4578 shinfo->gso_size -= decrement;
4579 }
4580
4581 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4582
4583 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4584 {
4585 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4586 * wanted then gso_type will be set. */
4587 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4588
4589 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4590 unlikely(shinfo->gso_type == 0)) {
4591 __skb_warn_lro_forwarding(skb);
4592 return true;
4593 }
4594 return false;
4595 }
4596
4597 static inline void skb_forward_csum(struct sk_buff *skb)
4598 {
4599 /* Unfortunately we don't support this one. Any brave souls? */
4600 if (skb->ip_summed == CHECKSUM_COMPLETE)
4601 skb->ip_summed = CHECKSUM_NONE;
4602 }
4603
4604 /**
4605 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4606 * @skb: skb to check
4607 *
4608 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4609 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4610 * use this helper, to document places where we make this assertion.
4611 */
4612 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4613 {
4614 #ifdef DEBUG
4615 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4616 #endif
4617 }
4618
4619 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4620
4621 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4622 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4623 unsigned int transport_len,
4624 __sum16(*skb_chkf)(struct sk_buff *skb));
4625
4626 /**
4627 * skb_head_is_locked - Determine if the skb->head is locked down
4628 * @skb: skb to check
4629 *
4630 * The head on skbs build around a head frag can be removed if they are
4631 * not cloned. This function returns true if the skb head is locked down
4632 * due to either being allocated via kmalloc, or by being a clone with
4633 * multiple references to the head.
4634 */
4635 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4636 {
4637 return !skb->head_frag || skb_cloned(skb);
4638 }
4639
4640 /* Local Checksum Offload.
4641 * Compute outer checksum based on the assumption that the
4642 * inner checksum will be offloaded later.
4643 * See Documentation/networking/checksum-offloads.rst for
4644 * explanation of how this works.
4645 * Fill in outer checksum adjustment (e.g. with sum of outer
4646 * pseudo-header) before calling.
4647 * Also ensure that inner checksum is in linear data area.
4648 */
4649 static inline __wsum lco_csum(struct sk_buff *skb)
4650 {
4651 unsigned char *csum_start = skb_checksum_start(skb);
4652 unsigned char *l4_hdr = skb_transport_header(skb);
4653 __wsum partial;
4654
4655 /* Start with complement of inner checksum adjustment */
4656 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4657 skb->csum_offset));
4658
4659 /* Add in checksum of our headers (incl. outer checksum
4660 * adjustment filled in by caller) and return result.
4661 */
4662 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4663 }
4664
4665 static inline bool skb_is_redirected(const struct sk_buff *skb)
4666 {
4667 #ifdef CONFIG_NET_REDIRECT
4668 return skb->redirected;
4669 #else
4670 return false;
4671 #endif
4672 }
4673
4674 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4675 {
4676 #ifdef CONFIG_NET_REDIRECT
4677 skb->redirected = 1;
4678 skb->from_ingress = from_ingress;
4679 if (skb->from_ingress)
4680 skb->tstamp = 0;
4681 #endif
4682 }
4683
4684 static inline void skb_reset_redirect(struct sk_buff *skb)
4685 {
4686 #ifdef CONFIG_NET_REDIRECT
4687 skb->redirected = 0;
4688 #endif
4689 }
4690
4691 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4692 {
4693 return skb->csum_not_inet;
4694 }
4695
4696 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4697 const u64 kcov_handle)
4698 {
4699 #ifdef CONFIG_KCOV
4700 skb->kcov_handle = kcov_handle;
4701 #endif
4702 }
4703
4704 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4705 {
4706 #ifdef CONFIG_KCOV
4707 return skb->kcov_handle;
4708 #else
4709 return 0;
4710 #endif
4711 }
4712
4713 #ifdef CONFIG_PAGE_POOL
4714 static inline void skb_mark_for_recycle(struct sk_buff *skb, struct page *page,
4715 struct page_pool *pp)
4716 {
4717 skb->pp_recycle = 1;
4718 page_pool_store_mem_info(page, pp);
4719 }
4720 #endif
4721
4722 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4723 {
4724 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4725 return false;
4726 return page_pool_return_skb_page(virt_to_page(data));
4727 }
4728
4729 #endif /* __KERNEL__ */
4730 #endif /* _LINUX_SKBUFF_H */