1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 */
9
10 #ifndef _LINUX_SKBUFF_H
11 #define _LINUX_SKBUFF_H
12
13 #include <linux/kernel.h>
14 #include <linux/compiler.h>
15 #include <linux/time.h>
16 #include <linux/bug.h>
17 #include <linux/bvec.h>
18 #include <linux/cache.h>
19 #include <linux/rbtree.h>
20 #include <linux/socket.h>
21 #include <linux/refcount.h>
22
23 #include <linux/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/hrtimer.h>
31 #include <linux/dma-mapping.h>
32 #include <linux/netdev_features.h>
33 #include <linux/sched.h>
34 #include <linux/sched/clock.h>
35 #include <net/flow_dissector.h>
36 #include <linux/splice.h>
37 #include <linux/in6.h>
38 #include <linux/if_packet.h>
39 #include <net/flow.h>
40 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
41 #include <linux/netfilter/nf_conntrack_common.h>
42 #endif
43 #include <linux/android_kabi.h>
44 #include <linux/android_vendor.h>
45
46 /* The interface for checksum offload between the stack and networking drivers
47 * is as follows...
48 *
49 * A. IP checksum related features
50 *
51 * Drivers advertise checksum offload capabilities in the features of a device.
52 * From the stack's point of view these are capabilities offered by the driver,
53 * a driver typically only advertises features that it is capable of offloading
54 * to its device.
55 *
56 * The checksum related features are:
57 *
58 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
59 * IP (one's complement) checksum for any combination
60 * of protocols or protocol layering. The checksum is
61 * computed and set in a packet per the CHECKSUM_PARTIAL
62 * interface (see below).
63 *
64 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
65 * TCP or UDP packets over IPv4. These are specifically
66 * unencapsulated packets of the form IPv4|TCP or
67 * IPv4|UDP where the Protocol field in the IPv4 header
68 * is TCP or UDP. The IPv4 header may contain IP options
69 * This feature cannot be set in features for a device
70 * with NETIF_F_HW_CSUM also set. This feature is being
71 * DEPRECATED (see below).
72 *
73 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
74 * TCP or UDP packets over IPv6. These are specifically
75 * unencapsulated packets of the form IPv6|TCP or
76 * IPv4|UDP where the Next Header field in the IPv6
77 * header is either TCP or UDP. IPv6 extension headers
78 * are not supported with this feature. This feature
79 * cannot be set in features for a device with
80 * NETIF_F_HW_CSUM also set. This feature is being
81 * DEPRECATED (see below).
82 *
83 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
84 * This flag is used only used to disable the RX checksum
85 * feature for a device. The stack will accept receive
86 * checksum indication in packets received on a device
87 * regardless of whether NETIF_F_RXCSUM is set.
88 *
89 * B. Checksumming of received packets by device. Indication of checksum
90 * verification is in set skb->ip_summed. Possible values are:
91 *
92 * CHECKSUM_NONE:
93 *
94 * Device did not checksum this packet e.g. due to lack of capabilities.
95 * The packet contains full (though not verified) checksum in packet but
96 * not in skb->csum. Thus, skb->csum is undefined in this case.
97 *
98 * CHECKSUM_UNNECESSARY:
99 *
100 * The hardware you're dealing with doesn't calculate the full checksum
101 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
102 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
103 * if their checksums are okay. skb->csum is still undefined in this case
104 * though. A driver or device must never modify the checksum field in the
105 * packet even if checksum is verified.
106 *
107 * CHECKSUM_UNNECESSARY is applicable to following protocols:
108 * TCP: IPv6 and IPv4.
109 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
110 * zero UDP checksum for either IPv4 or IPv6, the networking stack
111 * may perform further validation in this case.
112 * GRE: only if the checksum is present in the header.
113 * SCTP: indicates the CRC in SCTP header has been validated.
114 * FCOE: indicates the CRC in FC frame has been validated.
115 *
116 * skb->csum_level indicates the number of consecutive checksums found in
117 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
118 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
119 * and a device is able to verify the checksums for UDP (possibly zero),
120 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
121 * two. If the device were only able to verify the UDP checksum and not
122 * GRE, either because it doesn't support GRE checksum of because GRE
123 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
124 * not considered in this case).
125 *
126 * CHECKSUM_COMPLETE:
127 *
128 * This is the most generic way. The device supplied checksum of the _whole_
129 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
130 * hardware doesn't need to parse L3/L4 headers to implement this.
131 *
132 * Notes:
133 * - Even if device supports only some protocols, but is able to produce
134 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
135 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
136 *
137 * CHECKSUM_PARTIAL:
138 *
139 * A checksum is set up to be offloaded to a device as described in the
140 * output description for CHECKSUM_PARTIAL. This may occur on a packet
141 * received directly from another Linux OS, e.g., a virtualized Linux kernel
142 * on the same host, or it may be set in the input path in GRO or remote
143 * checksum offload. For the purposes of checksum verification, the checksum
144 * referred to by skb->csum_start + skb->csum_offset and any preceding
145 * checksums in the packet are considered verified. Any checksums in the
146 * packet that are after the checksum being offloaded are not considered to
147 * be verified.
148 *
149 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
150 * in the skb->ip_summed for a packet. Values are:
151 *
152 * CHECKSUM_PARTIAL:
153 *
154 * The driver is required to checksum the packet as seen by hard_start_xmit()
155 * from skb->csum_start up to the end, and to record/write the checksum at
156 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
157 * csum_start and csum_offset values are valid values given the length and
158 * offset of the packet, however they should not attempt to validate that the
159 * checksum refers to a legitimate transport layer checksum-- it is the
160 * purview of the stack to validate that csum_start and csum_offset are set
161 * correctly.
162 *
163 * When the stack requests checksum offload for a packet, the driver MUST
164 * ensure that the checksum is set correctly. A driver can either offload the
165 * checksum calculation to the device, or call skb_checksum_help (in the case
166 * that the device does not support offload for a particular checksum).
167 *
168 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
169 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
170 * checksum offload capability.
171 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
172 * on network device checksumming capabilities: if a packet does not match
173 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
174 * csum_not_inet, see item D.) is called to resolve the checksum.
175 *
176 * CHECKSUM_NONE:
177 *
178 * The skb was already checksummed by the protocol, or a checksum is not
179 * required.
180 *
181 * CHECKSUM_UNNECESSARY:
182 *
183 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
184 * output.
185 *
186 * CHECKSUM_COMPLETE:
187 * Not used in checksum output. If a driver observes a packet with this value
188 * set in skbuff, if should treat as CHECKSUM_NONE being set.
189 *
190 * D. Non-IP checksum (CRC) offloads
191 *
192 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
193 * offloading the SCTP CRC in a packet. To perform this offload the stack
194 * will set set csum_start and csum_offset accordingly, set ip_summed to
195 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
196 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
197 * A driver that supports both IP checksum offload and SCTP CRC32c offload
198 * must verify which offload is configured for a packet by testing the
199 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
200 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
201 *
202 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
203 * offloading the FCOE CRC in a packet. To perform this offload the stack
204 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
205 * accordingly. Note the there is no indication in the skbuff that the
206 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
207 * both IP checksum offload and FCOE CRC offload must verify which offload
208 * is configured for a packet presumably by inspecting packet headers.
209 *
210 * E. Checksumming on output with GSO.
211 *
212 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
213 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
214 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
215 * part of the GSO operation is implied. If a checksum is being offloaded
216 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
217 * are set to refer to the outermost checksum being offload (two offloaded
218 * checksums are possible with UDP encapsulation).
219 */
220
221 /* Don't change this without changing skb_csum_unnecessary! */
222 #define CHECKSUM_NONE 0
223 #define CHECKSUM_UNNECESSARY 1
224 #define CHECKSUM_COMPLETE 2
225 #define CHECKSUM_PARTIAL 3
226
227 /* Maximum value in skb->csum_level */
228 #define SKB_MAX_CSUM_LEVEL 3
229
230 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
231 #define SKB_WITH_OVERHEAD(X) \
232 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
233 #define SKB_MAX_ORDER(X, ORDER) \
234 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
235 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
236 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
237
238 /* return minimum truesize of one skb containing X bytes of data */
239 #define SKB_TRUESIZE(X) ((X) + \
240 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
241 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
242
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 u8 sabotage_in_done:1;
263 __u16 frag_max_size;
264 struct net_device *physindev;
265
266 /* always valid & non-NULL from FORWARD on, for physdev match */
267 struct net_device *physoutdev;
268 union {
269 /* prerouting: detect dnat in orig/reply direction */
270 __be32 ipv4_daddr;
271 struct in6_addr ipv6_daddr;
272
273 /* after prerouting + nat detected: store original source
274 * mac since neigh resolution overwrites it, only used while
275 * skb is out in neigh layer.
276 */
277 char neigh_header[8];
278 };
279 };
280 #endif
281
282 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
283 /* Chain in tc_skb_ext will be used to share the tc chain with
284 * ovs recirc_id. It will be set to the current chain by tc
285 * and read by ovs to recirc_id.
286 */
287 struct tc_skb_ext {
288 __u32 chain;
289 };
290 #endif
291
292 struct sk_buff_head {
293 /* These two members must be first. */
294 struct sk_buff *next;
295 struct sk_buff *prev;
296
297 __u32 qlen;
298 spinlock_t lock;
299 };
300
301 struct sk_buff;
302
303 /* To allow 64K frame to be packed as single skb without frag_list we
304 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
305 * buffers which do not start on a page boundary.
306 *
307 * Since GRO uses frags we allocate at least 16 regardless of page
308 * size.
309 */
310 #if (65536/PAGE_SIZE + 1) < 16
311 #define MAX_SKB_FRAGS 16UL
312 #else
313 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
314 #endif
315 extern int sysctl_max_skb_frags;
316
317 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
318 * segment using its current segmentation instead.
319 */
320 #define GSO_BY_FRAGS 0xFFFF
321
322 typedef struct bio_vec skb_frag_t;
323
324 /**
325 * skb_frag_size() - Returns the size of a skb fragment
326 * @frag: skb fragment
327 */
skb_frag_size(const skb_frag_t * frag)328 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329 {
330 return frag->bv_len;
331 }
332
333 /**
334 * skb_frag_size_set() - Sets the size of a skb fragment
335 * @frag: skb fragment
336 * @size: size of fragment
337 */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)338 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
339 {
340 frag->bv_len = size;
341 }
342
343 /**
344 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
345 * @frag: skb fragment
346 * @delta: value to add
347 */
skb_frag_size_add(skb_frag_t * frag,int delta)348 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
349 {
350 frag->bv_len += delta;
351 }
352
353 /**
354 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
355 * @frag: skb fragment
356 * @delta: value to subtract
357 */
skb_frag_size_sub(skb_frag_t * frag,int delta)358 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
359 {
360 frag->bv_len -= delta;
361 }
362
363 /**
364 * skb_frag_must_loop - Test if %p is a high memory page
365 * @p: fragment's page
366 */
skb_frag_must_loop(struct page * p)367 static inline bool skb_frag_must_loop(struct page *p)
368 {
369 #if defined(CONFIG_HIGHMEM)
370 if (PageHighMem(p))
371 return true;
372 #endif
373 return false;
374 }
375
376 /**
377 * skb_frag_foreach_page - loop over pages in a fragment
378 *
379 * @f: skb frag to operate on
380 * @f_off: offset from start of f->bv_page
381 * @f_len: length from f_off to loop over
382 * @p: (temp var) current page
383 * @p_off: (temp var) offset from start of current page,
384 * non-zero only on first page.
385 * @p_len: (temp var) length in current page,
386 * < PAGE_SIZE only on first and last page.
387 * @copied: (temp var) length so far, excluding current p_len.
388 *
389 * A fragment can hold a compound page, in which case per-page
390 * operations, notably kmap_atomic, must be called for each
391 * regular page.
392 */
393 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
394 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
395 p_off = (f_off) & (PAGE_SIZE - 1), \
396 p_len = skb_frag_must_loop(p) ? \
397 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
398 copied = 0; \
399 copied < f_len; \
400 copied += p_len, p++, p_off = 0, \
401 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
402
403 #define HAVE_HW_TIME_STAMP
404
405 /**
406 * struct skb_shared_hwtstamps - hardware time stamps
407 * @hwtstamp: hardware time stamp transformed into duration
408 * since arbitrary point in time
409 *
410 * Software time stamps generated by ktime_get_real() are stored in
411 * skb->tstamp.
412 *
413 * hwtstamps can only be compared against other hwtstamps from
414 * the same device.
415 *
416 * This structure is attached to packets as part of the
417 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
418 */
419 struct skb_shared_hwtstamps {
420 ktime_t hwtstamp;
421 };
422
423 /* Definitions for tx_flags in struct skb_shared_info */
424 enum {
425 /* generate hardware time stamp */
426 SKBTX_HW_TSTAMP = 1 << 0,
427
428 /* generate software time stamp when queueing packet to NIC */
429 SKBTX_SW_TSTAMP = 1 << 1,
430
431 /* device driver is going to provide hardware time stamp */
432 SKBTX_IN_PROGRESS = 1 << 2,
433
434 /* device driver supports TX zero-copy buffers */
435 SKBTX_DEV_ZEROCOPY = 1 << 3,
436
437 /* generate wifi status information (where possible) */
438 SKBTX_WIFI_STATUS = 1 << 4,
439
440 /* This indicates at least one fragment might be overwritten
441 * (as in vmsplice(), sendfile() ...)
442 * If we need to compute a TX checksum, we'll need to copy
443 * all frags to avoid possible bad checksum
444 */
445 SKBTX_SHARED_FRAG = 1 << 5,
446
447 /* generate software time stamp when entering packet scheduling */
448 SKBTX_SCHED_TSTAMP = 1 << 6,
449 };
450
451 #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
452 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
453 SKBTX_SCHED_TSTAMP)
454 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
455
456 /*
457 * The callback notifies userspace to release buffers when skb DMA is done in
458 * lower device, the skb last reference should be 0 when calling this.
459 * The zerocopy_success argument is true if zero copy transmit occurred,
460 * false on data copy or out of memory error caused by data copy attempt.
461 * The ctx field is used to track device context.
462 * The desc field is used to track userspace buffer index.
463 */
464 struct ubuf_info {
465 void (*callback)(struct ubuf_info *, bool zerocopy_success);
466 union {
467 struct {
468 unsigned long desc;
469 void *ctx;
470 };
471 struct {
472 u32 id;
473 u16 len;
474 u16 zerocopy:1;
475 u32 bytelen;
476 };
477 };
478 refcount_t refcnt;
479
480 struct mmpin {
481 struct user_struct *user;
482 unsigned int num_pg;
483 } mmp;
484 };
485
486 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
487
488 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
489 void mm_unaccount_pinned_pages(struct mmpin *mmp);
490
491 struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
492 struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
493 struct ubuf_info *uarg);
494
sock_zerocopy_get(struct ubuf_info * uarg)495 static inline void sock_zerocopy_get(struct ubuf_info *uarg)
496 {
497 refcount_inc(&uarg->refcnt);
498 }
499
500 void sock_zerocopy_put(struct ubuf_info *uarg);
501 void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
502
503 void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
504
505 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
506 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
507 struct msghdr *msg, int len,
508 struct ubuf_info *uarg);
509
510 /* This data is invariant across clones and lives at
511 * the end of the header data, ie. at skb->end.
512 */
513 struct skb_shared_info {
514 __u8 __unused;
515 __u8 meta_len;
516 __u8 nr_frags;
517 __u8 tx_flags;
518 unsigned short gso_size;
519 /* Warning: this field is not always filled in (UFO)! */
520 unsigned short gso_segs;
521 struct sk_buff *frag_list;
522 struct skb_shared_hwtstamps hwtstamps;
523 unsigned int gso_type;
524 u32 tskey;
525
526 /*
527 * Warning : all fields before dataref are cleared in __alloc_skb()
528 */
529 atomic_t dataref;
530
531 /* Intermediate layers must ensure that destructor_arg
532 * remains valid until skb destructor */
533 void * destructor_arg;
534
535 ANDROID_VENDOR_DATA_ARRAY(1, 3);
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 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
620 * @sk: Socket we are owned by
621 * @dev: Device we arrived on/are leaving by
622 * @cb: Control buffer. Free for use by every layer. Put private vars here
623 * @_skb_refdst: destination entry (with norefcount bit)
624 * @sp: the security path, used for xfrm
625 * @len: Length of actual data
626 * @data_len: Data length
627 * @mac_len: Length of link layer header
628 * @hdr_len: writable header length of cloned skb
629 * @csum: Checksum (must include start/offset pair)
630 * @csum_start: Offset from skb->head where checksumming should start
631 * @csum_offset: Offset from csum_start where checksum should be stored
632 * @priority: Packet queueing priority
633 * @ignore_df: allow local fragmentation
634 * @cloned: Head may be cloned (check refcnt to be sure)
635 * @ip_summed: Driver fed us an IP checksum
636 * @nohdr: Payload reference only, must not modify header
637 * @pkt_type: Packet class
638 * @fclone: skbuff clone status
639 * @ipvs_property: skbuff is owned by ipvs
640 * @offload_fwd_mark: Packet was L2-forwarded in hardware
641 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
642 * @tc_skip_classify: do not classify packet. set by IFB device
643 * @tc_at_ingress: used within tc_classify to distinguish in/egress
644 * @redirected: packet was redirected by packet classifier
645 * @from_ingress: packet was redirected from the ingress path
646 * @peeked: this packet has been seen already, so stats have been
647 * done for it, don't do them again
648 * @nf_trace: netfilter packet trace flag
649 * @protocol: Packet protocol from driver
650 * @destructor: Destruct function
651 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
652 * @_nfct: Associated connection, if any (with nfctinfo bits)
653 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
654 * @skb_iif: ifindex of device we arrived on
655 * @tc_index: Traffic control index
656 * @hash: the packet hash
657 * @queue_mapping: Queue mapping for multiqueue devices
658 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
659 * @active_extensions: active extensions (skb_ext_id types)
660 * @ndisc_nodetype: router type (from link layer)
661 * @ooo_okay: allow the mapping of a socket to a queue to be changed
662 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
663 * ports.
664 * @sw_hash: indicates hash was computed in software stack
665 * @wifi_acked_valid: wifi_acked was set
666 * @wifi_acked: whether frame was acked on wifi or not
667 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
668 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
669 * @scm_io_uring: SKB holds io_uring registered files
670 * @dst_pending_confirm: need to confirm neighbour
671 * @decrypted: Decrypted SKB
672 * @napi_id: id of the NAPI struct this skb came from
673 * @secmark: security marking
674 * @mark: Generic packet mark
675 * @vlan_proto: vlan encapsulation protocol
676 * @vlan_tci: vlan tag control information
677 * @inner_protocol: Protocol (encapsulation)
678 * @inner_transport_header: Inner transport layer header (encapsulation)
679 * @inner_network_header: Network layer header (encapsulation)
680 * @inner_mac_header: Link layer header (encapsulation)
681 * @transport_header: Transport layer header
682 * @network_header: Network layer header
683 * @mac_header: Link layer header
684 * @tail: Tail pointer
685 * @end: End pointer
686 * @head: Head of buffer
687 * @data: Data head pointer
688 * @truesize: Buffer size
689 * @users: User count - see {datagram,tcp}.c
690 * @extensions: allocated extensions, valid if active_extensions is nonzero
691 */
692
693 struct sk_buff {
694 union {
695 struct {
696 /* These two members must be first. */
697 struct sk_buff *next;
698 struct sk_buff *prev;
699
700 union {
701 struct net_device *dev;
702 /* Some protocols might use this space to store information,
703 * while device pointer would be NULL.
704 * UDP receive path is one user.
705 */
706 unsigned long dev_scratch;
707 };
708 };
709 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
710 struct list_head list;
711 };
712
713 union {
714 struct sock *sk;
715 int ip_defrag_offset;
716 };
717
718 union {
719 ktime_t tstamp;
720 u64 skb_mstamp_ns; /* earliest departure time */
721 };
722 /*
723 * This is the control buffer. It is free to use for every
724 * layer. Please put your private variables there. If you
725 * want to keep them across layers you have to do a skb_clone()
726 * first. This is owned by whoever has the skb queued ATM.
727 */
728 char cb[48] __aligned(8);
729
730 union {
731 struct {
732 unsigned long _skb_refdst;
733 void (*destructor)(struct sk_buff *skb);
734 };
735 struct list_head tcp_tsorted_anchor;
736 };
737
738 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
739 unsigned long _nfct;
740 #endif
741 unsigned int len,
742 data_len;
743 __u16 mac_len,
744 hdr_len;
745
746 /* Following fields are _not_ copied in __copy_skb_header()
747 * Note that queue_mapping is here mostly to fill a hole.
748 */
749 __u16 queue_mapping;
750
751 /* if you move cloned around you also must adapt those constants */
752 #ifdef __BIG_ENDIAN_BITFIELD
753 #define CLONED_MASK (1 << 7)
754 #else
755 #define CLONED_MASK 1
756 #endif
757 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
758
759 __u8 __cloned_offset[0];
760 __u8 cloned:1,
761 nohdr:1,
762 fclone:2,
763 peeked:1,
764 head_frag:1,
765 pfmemalloc:1;
766 #ifdef CONFIG_SKB_EXTENSIONS
767 __u8 active_extensions;
768 #endif
769 /* fields enclosed in headers_start/headers_end are copied
770 * using a single memcpy() in __copy_skb_header()
771 */
772 /* private: */
773 __u32 headers_start[0];
774 /* public: */
775
776 /* if you move pkt_type around you also must adapt those constants */
777 #ifdef __BIG_ENDIAN_BITFIELD
778 #define PKT_TYPE_MAX (7 << 5)
779 #else
780 #define PKT_TYPE_MAX 7
781 #endif
782 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
783
784 __u8 __pkt_type_offset[0];
785 __u8 pkt_type:3;
786 __u8 ignore_df:1;
787 __u8 nf_trace:1;
788 __u8 ip_summed:2;
789 __u8 ooo_okay:1;
790
791 __u8 l4_hash:1;
792 __u8 sw_hash:1;
793 __u8 wifi_acked_valid:1;
794 __u8 wifi_acked:1;
795 __u8 no_fcs:1;
796 /* Indicates the inner headers are valid in the skbuff. */
797 __u8 encapsulation:1;
798 __u8 encap_hdr_csum:1;
799 __u8 csum_valid:1;
800
801 #ifdef __BIG_ENDIAN_BITFIELD
802 #define PKT_VLAN_PRESENT_BIT 7
803 #else
804 #define PKT_VLAN_PRESENT_BIT 0
805 #endif
806 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
807 __u8 __pkt_vlan_present_offset[0];
808 __u8 vlan_present:1;
809 __u8 csum_complete_sw:1;
810 __u8 csum_level:2;
811 __u8 csum_not_inet:1;
812 __u8 dst_pending_confirm:1;
813 #ifdef CONFIG_IPV6_NDISC_NODETYPE
814 __u8 ndisc_nodetype:2;
815 #endif
816
817 __u8 ipvs_property:1;
818 __u8 inner_protocol_type:1;
819 __u8 remcsum_offload:1;
820 #ifdef CONFIG_NET_SWITCHDEV
821 __u8 offload_fwd_mark:1;
822 __u8 offload_l3_fwd_mark:1;
823 #endif
824 #ifdef CONFIG_NET_CLS_ACT
825 __u8 tc_skip_classify:1;
826 __u8 tc_at_ingress:1;
827 #endif
828 #ifdef CONFIG_NET_REDIRECT
829 __u8 redirected:1;
830 __u8 from_ingress:1;
831 #endif
832 #ifdef CONFIG_TLS_DEVICE
833 __u8 decrypted:1;
834 #endif
835
836 #ifdef CONFIG_NET_SCHED
837 __u16 tc_index; /* traffic control index */
838 #endif
839
840 union {
841 __wsum csum;
842 struct {
843 __u16 csum_start;
844 __u16 csum_offset;
845 };
846 };
847 __u32 priority;
848 int skb_iif;
849 __u32 hash;
850 __be16 vlan_proto;
851 __u16 vlan_tci;
852 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
853 union {
854 unsigned int napi_id;
855 unsigned int sender_cpu;
856 };
857 #endif
858 #ifdef CONFIG_NETWORK_SECMARK
859 __u32 secmark;
860 #endif
861
862 union {
863 __u32 mark;
864 __u32 reserved_tailroom;
865 };
866
867 union {
868 __be16 inner_protocol;
869 __u8 inner_ipproto;
870 };
871
872 __u16 inner_transport_header;
873 __u16 inner_network_header;
874 __u16 inner_mac_header;
875
876 __be16 protocol;
877 __u16 transport_header;
878 __u16 network_header;
879 __u16 mac_header;
880
881 /* private: */
882 __u32 headers_end[0];
883 /* public: */
884
885 /* Android KABI preservation.
886 *
887 * "open coded" version of ANDROID_KABI_USE() to pack more
888 * fields/variables into the space that we have.
889 *
890 * scm_io_uring is from 04df9719df18 ("io_uring/af_unix: defer
891 * registered files gc to io_uring release")
892 */
893 /* NOTE: due to these fields ending up after headers_end, we have to
894 * manually copy them in the __copy_skb_header() call in skbuf.c. Be
895 * very aware of that if you change these fields.
896 */
897 _ANDROID_KABI_REPLACE(_ANDROID_KABI_RESERVE(1),
898 struct {
899 __u8 scm_io_uring:1;
900 __u8 android_kabi_reserved1_padding1;
901 __u16 android_kabi_reserved1_padding2;
902 __u32 android_kabi_reserved1_padding3;
903 });
904 ANDROID_KABI_RESERVE(2);
905
906 /* These elements must be at the end, see alloc_skb() for details. */
907 sk_buff_data_t tail;
908 sk_buff_data_t end;
909 unsigned char *head,
910 *data;
911 unsigned int truesize;
912 refcount_t users;
913
914 #ifdef CONFIG_SKB_EXTENSIONS
915 /* only useable after checking ->active_extensions != 0 */
916 struct skb_ext *extensions;
917 #endif
918 };
919
920 #ifdef __KERNEL__
921 /*
922 * Handling routines are only of interest to the kernel
923 */
924
925 #define SKB_ALLOC_FCLONE 0x01
926 #define SKB_ALLOC_RX 0x02
927 #define SKB_ALLOC_NAPI 0x04
928
929 /**
930 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
931 * @skb: buffer
932 */
skb_pfmemalloc(const struct sk_buff * skb)933 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
934 {
935 return unlikely(skb->pfmemalloc);
936 }
937
938 /*
939 * skb might have a dst pointer attached, refcounted or not.
940 * _skb_refdst low order bit is set if refcount was _not_ taken
941 */
942 #define SKB_DST_NOREF 1UL
943 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
944
945 /**
946 * skb_dst - returns skb dst_entry
947 * @skb: buffer
948 *
949 * Returns skb dst_entry, regardless of reference taken or not.
950 */
skb_dst(const struct sk_buff * skb)951 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
952 {
953 /* If refdst was not refcounted, check we still are in a
954 * rcu_read_lock section
955 */
956 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
957 !rcu_read_lock_held() &&
958 !rcu_read_lock_bh_held());
959 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
960 }
961
962 /**
963 * skb_dst_set - sets skb dst
964 * @skb: buffer
965 * @dst: dst entry
966 *
967 * Sets skb dst, assuming a reference was taken on dst and should
968 * be released by skb_dst_drop()
969 */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)970 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
971 {
972 skb->_skb_refdst = (unsigned long)dst;
973 }
974
975 /**
976 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
977 * @skb: buffer
978 * @dst: dst entry
979 *
980 * Sets skb dst, assuming a reference was not taken on dst.
981 * If dst entry is cached, we do not take reference and dst_release
982 * will be avoided by refdst_drop. If dst entry is not cached, we take
983 * reference, so that last dst_release can destroy the dst immediately.
984 */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)985 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
986 {
987 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
988 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
989 }
990
991 /**
992 * skb_dst_is_noref - Test if skb dst isn't refcounted
993 * @skb: buffer
994 */
skb_dst_is_noref(const struct sk_buff * skb)995 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
996 {
997 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
998 }
999
1000 /**
1001 * skb_rtable - Returns the skb &rtable
1002 * @skb: buffer
1003 */
skb_rtable(const struct sk_buff * skb)1004 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1005 {
1006 return (struct rtable *)skb_dst(skb);
1007 }
1008
1009 /* For mangling skb->pkt_type from user space side from applications
1010 * such as nft, tc, etc, we only allow a conservative subset of
1011 * possible pkt_types to be set.
1012 */
skb_pkt_type_ok(u32 ptype)1013 static inline bool skb_pkt_type_ok(u32 ptype)
1014 {
1015 return ptype <= PACKET_OTHERHOST;
1016 }
1017
1018 /**
1019 * skb_napi_id - Returns the skb's NAPI id
1020 * @skb: buffer
1021 */
skb_napi_id(const struct sk_buff * skb)1022 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1023 {
1024 #ifdef CONFIG_NET_RX_BUSY_POLL
1025 return skb->napi_id;
1026 #else
1027 return 0;
1028 #endif
1029 }
1030
1031 /**
1032 * skb_unref - decrement the skb's reference count
1033 * @skb: buffer
1034 *
1035 * Returns true if we can free the skb.
1036 */
skb_unref(struct sk_buff * skb)1037 static inline bool skb_unref(struct sk_buff *skb)
1038 {
1039 if (unlikely(!skb))
1040 return false;
1041 if (likely(refcount_read(&skb->users) == 1))
1042 smp_rmb();
1043 else if (likely(!refcount_dec_and_test(&skb->users)))
1044 return false;
1045
1046 return true;
1047 }
1048
1049 void skb_release_head_state(struct sk_buff *skb);
1050 void kfree_skb(struct sk_buff *skb);
1051 void kfree_skb_list(struct sk_buff *segs);
1052 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1053 void skb_tx_error(struct sk_buff *skb);
1054 void consume_skb(struct sk_buff *skb);
1055 void __consume_stateless_skb(struct sk_buff *skb);
1056 void __kfree_skb(struct sk_buff *skb);
1057 extern struct kmem_cache *skbuff_head_cache;
1058
1059 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1060 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1061 bool *fragstolen, int *delta_truesize);
1062
1063 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1064 int node);
1065 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1066 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1067 struct sk_buff *build_skb_around(struct sk_buff *skb,
1068 void *data, unsigned int frag_size);
1069
1070 /**
1071 * alloc_skb - allocate a network buffer
1072 * @size: size to allocate
1073 * @priority: allocation mask
1074 *
1075 * This function is a convenient wrapper around __alloc_skb().
1076 */
alloc_skb(unsigned int size,gfp_t priority)1077 static inline struct sk_buff *alloc_skb(unsigned int size,
1078 gfp_t priority)
1079 {
1080 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1081 }
1082
1083 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1084 unsigned long data_len,
1085 int max_page_order,
1086 int *errcode,
1087 gfp_t gfp_mask);
1088 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1089
1090 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1091 struct sk_buff_fclones {
1092 struct sk_buff skb1;
1093
1094 struct sk_buff skb2;
1095
1096 refcount_t fclone_ref;
1097 };
1098
1099 /**
1100 * skb_fclone_busy - check if fclone is busy
1101 * @sk: socket
1102 * @skb: buffer
1103 *
1104 * Returns true if skb is a fast clone, and its clone is not freed.
1105 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1106 * so we also check that this didnt happen.
1107 */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1108 static inline bool skb_fclone_busy(const struct sock *sk,
1109 const struct sk_buff *skb)
1110 {
1111 const struct sk_buff_fclones *fclones;
1112
1113 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1114
1115 return skb->fclone == SKB_FCLONE_ORIG &&
1116 refcount_read(&fclones->fclone_ref) > 1 &&
1117 fclones->skb2.sk == sk;
1118 }
1119
1120 /**
1121 * alloc_skb_fclone - allocate a network buffer from fclone cache
1122 * @size: size to allocate
1123 * @priority: allocation mask
1124 *
1125 * This function is a convenient wrapper around __alloc_skb().
1126 */
alloc_skb_fclone(unsigned int size,gfp_t priority)1127 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1128 gfp_t priority)
1129 {
1130 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1131 }
1132
1133 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1134 void skb_headers_offset_update(struct sk_buff *skb, int off);
1135 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1136 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1137 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1138 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1139 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1140 gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1141 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1142 gfp_t gfp_mask)
1143 {
1144 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1145 }
1146
1147 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1148 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1149 unsigned int headroom);
1150 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1151 int newtailroom, gfp_t priority);
1152 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1153 int offset, int len);
1154 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1155 int offset, int len);
1156 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1157 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1158
1159 /**
1160 * skb_pad - zero pad the tail of an skb
1161 * @skb: buffer to pad
1162 * @pad: space to pad
1163 *
1164 * Ensure that a buffer is followed by a padding area that is zero
1165 * filled. Used by network drivers which may DMA or transfer data
1166 * beyond the buffer end onto the wire.
1167 *
1168 * May return error in out of memory cases. The skb is freed on error.
1169 */
skb_pad(struct sk_buff * skb,int pad)1170 static inline int skb_pad(struct sk_buff *skb, int pad)
1171 {
1172 return __skb_pad(skb, pad, true);
1173 }
1174 #define dev_kfree_skb(a) consume_skb(a)
1175
1176 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1177 int offset, size_t size);
1178
1179 struct skb_seq_state {
1180 __u32 lower_offset;
1181 __u32 upper_offset;
1182 __u32 frag_idx;
1183 __u32 stepped_offset;
1184 struct sk_buff *root_skb;
1185 struct sk_buff *cur_skb;
1186 __u8 *frag_data;
1187 };
1188
1189 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1190 unsigned int to, struct skb_seq_state *st);
1191 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1192 struct skb_seq_state *st);
1193 void skb_abort_seq_read(struct skb_seq_state *st);
1194
1195 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1196 unsigned int to, struct ts_config *config);
1197
1198 /*
1199 * Packet hash types specify the type of hash in skb_set_hash.
1200 *
1201 * Hash types refer to the protocol layer addresses which are used to
1202 * construct a packet's hash. The hashes are used to differentiate or identify
1203 * flows of the protocol layer for the hash type. Hash types are either
1204 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1205 *
1206 * Properties of hashes:
1207 *
1208 * 1) Two packets in different flows have different hash values
1209 * 2) Two packets in the same flow should have the same hash value
1210 *
1211 * A hash at a higher layer is considered to be more specific. A driver should
1212 * set the most specific hash possible.
1213 *
1214 * A driver cannot indicate a more specific hash than the layer at which a hash
1215 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1216 *
1217 * A driver may indicate a hash level which is less specific than the
1218 * actual layer the hash was computed on. For instance, a hash computed
1219 * at L4 may be considered an L3 hash. This should only be done if the
1220 * driver can't unambiguously determine that the HW computed the hash at
1221 * the higher layer. Note that the "should" in the second property above
1222 * permits this.
1223 */
1224 enum pkt_hash_types {
1225 PKT_HASH_TYPE_NONE, /* Undefined type */
1226 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1227 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1228 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1229 };
1230
skb_clear_hash(struct sk_buff * skb)1231 static inline void skb_clear_hash(struct sk_buff *skb)
1232 {
1233 skb->hash = 0;
1234 skb->sw_hash = 0;
1235 skb->l4_hash = 0;
1236 }
1237
skb_clear_hash_if_not_l4(struct sk_buff * skb)1238 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1239 {
1240 if (!skb->l4_hash)
1241 skb_clear_hash(skb);
1242 }
1243
1244 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1245 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1246 {
1247 skb->l4_hash = is_l4;
1248 skb->sw_hash = is_sw;
1249 skb->hash = hash;
1250 }
1251
1252 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1253 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1254 {
1255 /* Used by drivers to set hash from HW */
1256 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1257 }
1258
1259 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1260 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1261 {
1262 __skb_set_hash(skb, hash, true, is_l4);
1263 }
1264
1265 void __skb_get_hash(struct sk_buff *skb);
1266 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1267 u32 skb_get_poff(const struct sk_buff *skb);
1268 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1269 const struct flow_keys_basic *keys, int hlen);
1270 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1271 void *data, int hlen_proto);
1272
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto)1273 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1274 int thoff, u8 ip_proto)
1275 {
1276 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1277 }
1278
1279 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1280 const struct flow_dissector_key *key,
1281 unsigned int key_count);
1282
1283 #ifdef CONFIG_NET
1284 int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1285 union bpf_attr __user *uattr);
1286 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1287 struct bpf_prog *prog);
1288
1289 int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
1290 #else
skb_flow_dissector_prog_query(const union bpf_attr * attr,union bpf_attr __user * uattr)1291 static inline int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1292 union bpf_attr __user *uattr)
1293 {
1294 return -EOPNOTSUPP;
1295 }
1296
skb_flow_dissector_bpf_prog_attach(const union bpf_attr * attr,struct bpf_prog * prog)1297 static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1298 struct bpf_prog *prog)
1299 {
1300 return -EOPNOTSUPP;
1301 }
1302
skb_flow_dissector_bpf_prog_detach(const union bpf_attr * attr)1303 static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
1304 {
1305 return -EOPNOTSUPP;
1306 }
1307 #endif
1308
1309 struct bpf_flow_dissector;
1310 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1311 __be16 proto, int nhoff, int hlen, unsigned int flags);
1312
1313 bool __skb_flow_dissect(const struct net *net,
1314 const struct sk_buff *skb,
1315 struct flow_dissector *flow_dissector,
1316 void *target_container,
1317 void *data, __be16 proto, int nhoff, int hlen,
1318 unsigned int flags);
1319
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1320 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1321 struct flow_dissector *flow_dissector,
1322 void *target_container, unsigned int flags)
1323 {
1324 return __skb_flow_dissect(NULL, skb, flow_dissector,
1325 target_container, NULL, 0, 0, 0, flags);
1326 }
1327
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1328 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1329 struct flow_keys *flow,
1330 unsigned int flags)
1331 {
1332 memset(flow, 0, sizeof(*flow));
1333 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1334 flow, NULL, 0, 0, 0, flags);
1335 }
1336
1337 static inline bool
skb_flow_dissect_flow_keys_basic(const struct net * net,const struct sk_buff * skb,struct flow_keys_basic * flow,void * data,__be16 proto,int nhoff,int hlen,unsigned int flags)1338 skb_flow_dissect_flow_keys_basic(const struct net *net,
1339 const struct sk_buff *skb,
1340 struct flow_keys_basic *flow, void *data,
1341 __be16 proto, int nhoff, int hlen,
1342 unsigned int flags)
1343 {
1344 memset(flow, 0, sizeof(*flow));
1345 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1346 data, proto, nhoff, hlen, flags);
1347 }
1348
1349 void skb_flow_dissect_meta(const struct sk_buff *skb,
1350 struct flow_dissector *flow_dissector,
1351 void *target_container);
1352
1353 /* Gets a skb connection tracking info, ctinfo map should be a
1354 * a map of mapsize to translate enum ip_conntrack_info states
1355 * to user states.
1356 */
1357 void
1358 skb_flow_dissect_ct(const struct sk_buff *skb,
1359 struct flow_dissector *flow_dissector,
1360 void *target_container,
1361 u16 *ctinfo_map,
1362 size_t mapsize);
1363 void
1364 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1365 struct flow_dissector *flow_dissector,
1366 void *target_container);
1367
skb_get_hash(struct sk_buff * skb)1368 static inline __u32 skb_get_hash(struct sk_buff *skb)
1369 {
1370 if (!skb->l4_hash && !skb->sw_hash)
1371 __skb_get_hash(skb);
1372
1373 return skb->hash;
1374 }
1375
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1376 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1377 {
1378 if (!skb->l4_hash && !skb->sw_hash) {
1379 struct flow_keys keys;
1380 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1381
1382 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1383 }
1384
1385 return skb->hash;
1386 }
1387
1388 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1389 const siphash_key_t *perturb);
1390
skb_get_hash_raw(const struct sk_buff * skb)1391 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1392 {
1393 return skb->hash;
1394 }
1395
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1396 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1397 {
1398 to->hash = from->hash;
1399 to->sw_hash = from->sw_hash;
1400 to->l4_hash = from->l4_hash;
1401 };
1402
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1403 static inline void skb_copy_decrypted(struct sk_buff *to,
1404 const struct sk_buff *from)
1405 {
1406 #ifdef CONFIG_TLS_DEVICE
1407 to->decrypted = from->decrypted;
1408 #endif
1409 }
1410
1411 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1412 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1413 {
1414 return skb->head + skb->end;
1415 }
1416
skb_end_offset(const struct sk_buff * skb)1417 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1418 {
1419 return skb->end;
1420 }
1421 #else
skb_end_pointer(const struct sk_buff * skb)1422 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1423 {
1424 return skb->end;
1425 }
1426
skb_end_offset(const struct sk_buff * skb)1427 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1428 {
1429 return skb->end - skb->head;
1430 }
1431 #endif
1432
1433 /* Internal */
1434 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1435
skb_hwtstamps(struct sk_buff * skb)1436 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1437 {
1438 return &skb_shinfo(skb)->hwtstamps;
1439 }
1440
skb_zcopy(struct sk_buff * skb)1441 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1442 {
1443 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1444
1445 return is_zcopy ? skb_uarg(skb) : NULL;
1446 }
1447
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1448 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1449 bool *have_ref)
1450 {
1451 if (skb && uarg && !skb_zcopy(skb)) {
1452 if (unlikely(have_ref && *have_ref))
1453 *have_ref = false;
1454 else
1455 sock_zerocopy_get(uarg);
1456 skb_shinfo(skb)->destructor_arg = uarg;
1457 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1458 }
1459 }
1460
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1461 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1462 {
1463 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1464 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1465 }
1466
skb_zcopy_is_nouarg(struct sk_buff * skb)1467 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1468 {
1469 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1470 }
1471
skb_zcopy_get_nouarg(struct sk_buff * skb)1472 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1473 {
1474 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1475 }
1476
1477 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy)1478 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1479 {
1480 struct ubuf_info *uarg = skb_zcopy(skb);
1481
1482 if (uarg) {
1483 if (skb_zcopy_is_nouarg(skb)) {
1484 /* no notification callback */
1485 } else if (uarg->callback == sock_zerocopy_callback) {
1486 uarg->zerocopy = uarg->zerocopy && zerocopy;
1487 sock_zerocopy_put(uarg);
1488 } else {
1489 uarg->callback(uarg, zerocopy);
1490 }
1491
1492 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1493 }
1494 }
1495
1496 /* Abort a zerocopy operation and revert zckey on error in send syscall */
skb_zcopy_abort(struct sk_buff * skb)1497 static inline void skb_zcopy_abort(struct sk_buff *skb)
1498 {
1499 struct ubuf_info *uarg = skb_zcopy(skb);
1500
1501 if (uarg) {
1502 sock_zerocopy_put_abort(uarg, false);
1503 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1504 }
1505 }
1506
skb_mark_not_on_list(struct sk_buff * skb)1507 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1508 {
1509 skb->next = NULL;
1510 }
1511
1512 /* Iterate through singly-linked GSO fragments of an skb. */
1513 #define skb_list_walk_safe(first, skb, next_skb) \
1514 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1515 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1516
skb_list_del_init(struct sk_buff * skb)1517 static inline void skb_list_del_init(struct sk_buff *skb)
1518 {
1519 __list_del_entry(&skb->list);
1520 skb_mark_not_on_list(skb);
1521 }
1522
1523 /**
1524 * skb_queue_empty - check if a queue is empty
1525 * @list: queue head
1526 *
1527 * Returns true if the queue is empty, false otherwise.
1528 */
skb_queue_empty(const struct sk_buff_head * list)1529 static inline int skb_queue_empty(const struct sk_buff_head *list)
1530 {
1531 return list->next == (const struct sk_buff *) list;
1532 }
1533
1534 /**
1535 * skb_queue_empty_lockless - check if a queue is empty
1536 * @list: queue head
1537 *
1538 * Returns true if the queue is empty, false otherwise.
1539 * This variant can be used in lockless contexts.
1540 */
skb_queue_empty_lockless(const struct sk_buff_head * list)1541 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1542 {
1543 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1544 }
1545
1546
1547 /**
1548 * skb_queue_is_last - check if skb is the last entry in the queue
1549 * @list: queue head
1550 * @skb: buffer
1551 *
1552 * Returns true if @skb is the last buffer on the list.
1553 */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1554 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1555 const struct sk_buff *skb)
1556 {
1557 return skb->next == (const struct sk_buff *) list;
1558 }
1559
1560 /**
1561 * skb_queue_is_first - check if skb is the first entry in the queue
1562 * @list: queue head
1563 * @skb: buffer
1564 *
1565 * Returns true if @skb is the first buffer on the list.
1566 */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1567 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1568 const struct sk_buff *skb)
1569 {
1570 return skb->prev == (const struct sk_buff *) list;
1571 }
1572
1573 /**
1574 * skb_queue_next - return the next packet in the queue
1575 * @list: queue head
1576 * @skb: current buffer
1577 *
1578 * Return the next packet in @list after @skb. It is only valid to
1579 * call this if skb_queue_is_last() evaluates to false.
1580 */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1581 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1582 const struct sk_buff *skb)
1583 {
1584 /* This BUG_ON may seem severe, but if we just return then we
1585 * are going to dereference garbage.
1586 */
1587 BUG_ON(skb_queue_is_last(list, skb));
1588 return skb->next;
1589 }
1590
1591 /**
1592 * skb_queue_prev - return the prev packet in the queue
1593 * @list: queue head
1594 * @skb: current buffer
1595 *
1596 * Return the prev packet in @list before @skb. It is only valid to
1597 * call this if skb_queue_is_first() evaluates to false.
1598 */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1599 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1600 const struct sk_buff *skb)
1601 {
1602 /* This BUG_ON may seem severe, but if we just return then we
1603 * are going to dereference garbage.
1604 */
1605 BUG_ON(skb_queue_is_first(list, skb));
1606 return skb->prev;
1607 }
1608
1609 /**
1610 * skb_get - reference buffer
1611 * @skb: buffer to reference
1612 *
1613 * Makes another reference to a socket buffer and returns a pointer
1614 * to the buffer.
1615 */
skb_get(struct sk_buff * skb)1616 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1617 {
1618 refcount_inc(&skb->users);
1619 return skb;
1620 }
1621
1622 /*
1623 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1624 */
1625
1626 /**
1627 * skb_cloned - is the buffer a clone
1628 * @skb: buffer to check
1629 *
1630 * Returns true if the buffer was generated with skb_clone() and is
1631 * one of multiple shared copies of the buffer. Cloned buffers are
1632 * shared data so must not be written to under normal circumstances.
1633 */
skb_cloned(const struct sk_buff * skb)1634 static inline int skb_cloned(const struct sk_buff *skb)
1635 {
1636 return skb->cloned &&
1637 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1638 }
1639
skb_unclone(struct sk_buff * skb,gfp_t pri)1640 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1641 {
1642 might_sleep_if(gfpflags_allow_blocking(pri));
1643
1644 if (skb_cloned(skb))
1645 return pskb_expand_head(skb, 0, 0, pri);
1646
1647 return 0;
1648 }
1649
1650 /**
1651 * skb_header_cloned - is the header a clone
1652 * @skb: buffer to check
1653 *
1654 * Returns true if modifying the header part of the buffer requires
1655 * the data to be copied.
1656 */
skb_header_cloned(const struct sk_buff * skb)1657 static inline int skb_header_cloned(const struct sk_buff *skb)
1658 {
1659 int dataref;
1660
1661 if (!skb->cloned)
1662 return 0;
1663
1664 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1665 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1666 return dataref != 1;
1667 }
1668
skb_header_unclone(struct sk_buff * skb,gfp_t pri)1669 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1670 {
1671 might_sleep_if(gfpflags_allow_blocking(pri));
1672
1673 if (skb_header_cloned(skb))
1674 return pskb_expand_head(skb, 0, 0, pri);
1675
1676 return 0;
1677 }
1678
1679 /**
1680 * __skb_header_release - release reference to header
1681 * @skb: buffer to operate on
1682 */
__skb_header_release(struct sk_buff * skb)1683 static inline void __skb_header_release(struct sk_buff *skb)
1684 {
1685 skb->nohdr = 1;
1686 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1687 }
1688
1689
1690 /**
1691 * skb_shared - is the buffer shared
1692 * @skb: buffer to check
1693 *
1694 * Returns true if more than one person has a reference to this
1695 * buffer.
1696 */
skb_shared(const struct sk_buff * skb)1697 static inline int skb_shared(const struct sk_buff *skb)
1698 {
1699 return refcount_read(&skb->users) != 1;
1700 }
1701
1702 /**
1703 * skb_share_check - check if buffer is shared and if so clone it
1704 * @skb: buffer to check
1705 * @pri: priority for memory allocation
1706 *
1707 * If the buffer is shared the buffer is cloned and the old copy
1708 * drops a reference. A new clone with a single reference is returned.
1709 * If the buffer is not shared the original buffer is returned. When
1710 * being called from interrupt status or with spinlocks held pri must
1711 * be GFP_ATOMIC.
1712 *
1713 * NULL is returned on a memory allocation failure.
1714 */
skb_share_check(struct sk_buff * skb,gfp_t pri)1715 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1716 {
1717 might_sleep_if(gfpflags_allow_blocking(pri));
1718 if (skb_shared(skb)) {
1719 struct sk_buff *nskb = skb_clone(skb, pri);
1720
1721 if (likely(nskb))
1722 consume_skb(skb);
1723 else
1724 kfree_skb(skb);
1725 skb = nskb;
1726 }
1727 return skb;
1728 }
1729
1730 /*
1731 * Copy shared buffers into a new sk_buff. We effectively do COW on
1732 * packets to handle cases where we have a local reader and forward
1733 * and a couple of other messy ones. The normal one is tcpdumping
1734 * a packet thats being forwarded.
1735 */
1736
1737 /**
1738 * skb_unshare - make a copy of a shared buffer
1739 * @skb: buffer to check
1740 * @pri: priority for memory allocation
1741 *
1742 * If the socket buffer is a clone then this function creates a new
1743 * copy of the data, drops a reference count on the old copy and returns
1744 * the new copy with the reference count at 1. If the buffer is not a clone
1745 * the original buffer is returned. When called with a spinlock held or
1746 * from interrupt state @pri must be %GFP_ATOMIC
1747 *
1748 * %NULL is returned on a memory allocation failure.
1749 */
skb_unshare(struct sk_buff * skb,gfp_t pri)1750 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1751 gfp_t pri)
1752 {
1753 might_sleep_if(gfpflags_allow_blocking(pri));
1754 if (skb_cloned(skb)) {
1755 struct sk_buff *nskb = skb_copy(skb, pri);
1756
1757 /* Free our shared copy */
1758 if (likely(nskb))
1759 consume_skb(skb);
1760 else
1761 kfree_skb(skb);
1762 skb = nskb;
1763 }
1764 return skb;
1765 }
1766
1767 /**
1768 * skb_peek - peek at the head of an &sk_buff_head
1769 * @list_: list to peek at
1770 *
1771 * Peek an &sk_buff. Unlike most other operations you _MUST_
1772 * be careful with this one. A peek leaves the buffer on the
1773 * list and someone else may run off with it. You must hold
1774 * the appropriate locks or have a private queue to do this.
1775 *
1776 * Returns %NULL for an empty list or a pointer to the head element.
1777 * The reference count is not incremented and the reference is therefore
1778 * volatile. Use with caution.
1779 */
skb_peek(const struct sk_buff_head * list_)1780 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1781 {
1782 struct sk_buff *skb = list_->next;
1783
1784 if (skb == (struct sk_buff *)list_)
1785 skb = NULL;
1786 return skb;
1787 }
1788
1789 /**
1790 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1791 * @list_: list to peek at
1792 *
1793 * Like skb_peek(), but the caller knows that the list is not empty.
1794 */
__skb_peek(const struct sk_buff_head * list_)1795 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1796 {
1797 return list_->next;
1798 }
1799
1800 /**
1801 * skb_peek_next - peek skb following the given one from a queue
1802 * @skb: skb to start from
1803 * @list_: list to peek at
1804 *
1805 * Returns %NULL when the end of the list is met or a pointer to the
1806 * next element. The reference count is not incremented and the
1807 * reference is therefore volatile. Use with caution.
1808 */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)1809 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1810 const struct sk_buff_head *list_)
1811 {
1812 struct sk_buff *next = skb->next;
1813
1814 if (next == (struct sk_buff *)list_)
1815 next = NULL;
1816 return next;
1817 }
1818
1819 /**
1820 * skb_peek_tail - peek at the tail of an &sk_buff_head
1821 * @list_: list to peek at
1822 *
1823 * Peek an &sk_buff. Unlike most other operations you _MUST_
1824 * be careful with this one. A peek leaves the buffer on the
1825 * list and someone else may run off with it. You must hold
1826 * the appropriate locks or have a private queue to do this.
1827 *
1828 * Returns %NULL for an empty list or a pointer to the tail element.
1829 * The reference count is not incremented and the reference is therefore
1830 * volatile. Use with caution.
1831 */
skb_peek_tail(const struct sk_buff_head * list_)1832 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1833 {
1834 struct sk_buff *skb = READ_ONCE(list_->prev);
1835
1836 if (skb == (struct sk_buff *)list_)
1837 skb = NULL;
1838 return skb;
1839
1840 }
1841
1842 /**
1843 * skb_queue_len - get queue length
1844 * @list_: list to measure
1845 *
1846 * Return the length of an &sk_buff queue.
1847 */
skb_queue_len(const struct sk_buff_head * list_)1848 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1849 {
1850 return list_->qlen;
1851 }
1852
1853 /**
1854 * skb_queue_len_lockless - get queue length
1855 * @list_: list to measure
1856 *
1857 * Return the length of an &sk_buff queue.
1858 * This variant can be used in lockless contexts.
1859 */
skb_queue_len_lockless(const struct sk_buff_head * list_)1860 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1861 {
1862 return READ_ONCE(list_->qlen);
1863 }
1864
1865 /**
1866 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1867 * @list: queue to initialize
1868 *
1869 * This initializes only the list and queue length aspects of
1870 * an sk_buff_head object. This allows to initialize the list
1871 * aspects of an sk_buff_head without reinitializing things like
1872 * the spinlock. It can also be used for on-stack sk_buff_head
1873 * objects where the spinlock is known to not be used.
1874 */
__skb_queue_head_init(struct sk_buff_head * list)1875 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1876 {
1877 list->prev = list->next = (struct sk_buff *)list;
1878 list->qlen = 0;
1879 }
1880
1881 /*
1882 * This function creates a split out lock class for each invocation;
1883 * this is needed for now since a whole lot of users of the skb-queue
1884 * infrastructure in drivers have different locking usage (in hardirq)
1885 * than the networking core (in softirq only). In the long run either the
1886 * network layer or drivers should need annotation to consolidate the
1887 * main types of usage into 3 classes.
1888 */
skb_queue_head_init(struct sk_buff_head * list)1889 static inline void skb_queue_head_init(struct sk_buff_head *list)
1890 {
1891 spin_lock_init(&list->lock);
1892 __skb_queue_head_init(list);
1893 }
1894
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)1895 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1896 struct lock_class_key *class)
1897 {
1898 skb_queue_head_init(list);
1899 lockdep_set_class(&list->lock, class);
1900 }
1901
1902 /*
1903 * Insert an sk_buff on a list.
1904 *
1905 * The "__skb_xxxx()" functions are the non-atomic ones that
1906 * can only be called with interrupts disabled.
1907 */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)1908 static inline void __skb_insert(struct sk_buff *newsk,
1909 struct sk_buff *prev, struct sk_buff *next,
1910 struct sk_buff_head *list)
1911 {
1912 /* See skb_queue_empty_lockless() and skb_peek_tail()
1913 * for the opposite READ_ONCE()
1914 */
1915 WRITE_ONCE(newsk->next, next);
1916 WRITE_ONCE(newsk->prev, prev);
1917 WRITE_ONCE(next->prev, newsk);
1918 WRITE_ONCE(prev->next, newsk);
1919 WRITE_ONCE(list->qlen, list->qlen + 1);
1920 }
1921
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)1922 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1923 struct sk_buff *prev,
1924 struct sk_buff *next)
1925 {
1926 struct sk_buff *first = list->next;
1927 struct sk_buff *last = list->prev;
1928
1929 WRITE_ONCE(first->prev, prev);
1930 WRITE_ONCE(prev->next, first);
1931
1932 WRITE_ONCE(last->next, next);
1933 WRITE_ONCE(next->prev, last);
1934 }
1935
1936 /**
1937 * skb_queue_splice - join two skb lists, this is designed for stacks
1938 * @list: the new list to add
1939 * @head: the place to add it in the first list
1940 */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)1941 static inline void skb_queue_splice(const struct sk_buff_head *list,
1942 struct sk_buff_head *head)
1943 {
1944 if (!skb_queue_empty(list)) {
1945 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1946 head->qlen += list->qlen;
1947 }
1948 }
1949
1950 /**
1951 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1952 * @list: the new list to add
1953 * @head: the place to add it in the first list
1954 *
1955 * The list at @list is reinitialised
1956 */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)1957 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1958 struct sk_buff_head *head)
1959 {
1960 if (!skb_queue_empty(list)) {
1961 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1962 head->qlen += list->qlen;
1963 __skb_queue_head_init(list);
1964 }
1965 }
1966
1967 /**
1968 * skb_queue_splice_tail - join two skb lists, each list being a queue
1969 * @list: the new list to add
1970 * @head: the place to add it in the first list
1971 */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)1972 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1973 struct sk_buff_head *head)
1974 {
1975 if (!skb_queue_empty(list)) {
1976 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1977 head->qlen += list->qlen;
1978 }
1979 }
1980
1981 /**
1982 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1983 * @list: the new list to add
1984 * @head: the place to add it in the first list
1985 *
1986 * Each of the lists is a queue.
1987 * The list at @list is reinitialised
1988 */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)1989 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1990 struct sk_buff_head *head)
1991 {
1992 if (!skb_queue_empty(list)) {
1993 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1994 head->qlen += list->qlen;
1995 __skb_queue_head_init(list);
1996 }
1997 }
1998
1999 /**
2000 * __skb_queue_after - queue a buffer at the list head
2001 * @list: list to use
2002 * @prev: place after this buffer
2003 * @newsk: buffer to queue
2004 *
2005 * Queue a buffer int the middle of a list. This function takes no locks
2006 * and you must therefore hold required locks before calling it.
2007 *
2008 * A buffer cannot be placed on two lists at the same time.
2009 */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)2010 static inline void __skb_queue_after(struct sk_buff_head *list,
2011 struct sk_buff *prev,
2012 struct sk_buff *newsk)
2013 {
2014 __skb_insert(newsk, prev, prev->next, list);
2015 }
2016
2017 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2018 struct sk_buff_head *list);
2019
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)2020 static inline void __skb_queue_before(struct sk_buff_head *list,
2021 struct sk_buff *next,
2022 struct sk_buff *newsk)
2023 {
2024 __skb_insert(newsk, next->prev, next, list);
2025 }
2026
2027 /**
2028 * __skb_queue_head - queue a buffer at the list head
2029 * @list: list to use
2030 * @newsk: buffer to queue
2031 *
2032 * Queue a buffer at the start of a list. This function takes no locks
2033 * and you must therefore hold required locks before calling it.
2034 *
2035 * A buffer cannot be placed on two lists at the same time.
2036 */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)2037 static inline void __skb_queue_head(struct sk_buff_head *list,
2038 struct sk_buff *newsk)
2039 {
2040 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2041 }
2042 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2043
2044 /**
2045 * __skb_queue_tail - queue a buffer at the list tail
2046 * @list: list to use
2047 * @newsk: buffer to queue
2048 *
2049 * Queue a buffer at the end of a list. This function takes no locks
2050 * and you must therefore hold required locks before calling it.
2051 *
2052 * A buffer cannot be placed on two lists at the same time.
2053 */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2054 static inline void __skb_queue_tail(struct sk_buff_head *list,
2055 struct sk_buff *newsk)
2056 {
2057 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2058 }
2059 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2060
2061 /*
2062 * remove sk_buff from list. _Must_ be called atomically, and with
2063 * the list known..
2064 */
2065 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2066 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2067 {
2068 struct sk_buff *next, *prev;
2069
2070 WRITE_ONCE(list->qlen, list->qlen - 1);
2071 next = skb->next;
2072 prev = skb->prev;
2073 skb->next = skb->prev = NULL;
2074 WRITE_ONCE(next->prev, prev);
2075 WRITE_ONCE(prev->next, next);
2076 }
2077
2078 /**
2079 * __skb_dequeue - remove from the head of the queue
2080 * @list: list to dequeue from
2081 *
2082 * Remove the head of the list. This function does not take any locks
2083 * so must be used with appropriate locks held only. The head item is
2084 * returned or %NULL if the list is empty.
2085 */
__skb_dequeue(struct sk_buff_head * list)2086 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2087 {
2088 struct sk_buff *skb = skb_peek(list);
2089 if (skb)
2090 __skb_unlink(skb, list);
2091 return skb;
2092 }
2093 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2094
2095 /**
2096 * __skb_dequeue_tail - remove from the tail of the queue
2097 * @list: list to dequeue from
2098 *
2099 * Remove the tail of the list. This function does not take any locks
2100 * so must be used with appropriate locks held only. The tail item is
2101 * returned or %NULL if the list is empty.
2102 */
__skb_dequeue_tail(struct sk_buff_head * list)2103 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2104 {
2105 struct sk_buff *skb = skb_peek_tail(list);
2106 if (skb)
2107 __skb_unlink(skb, list);
2108 return skb;
2109 }
2110 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2111
2112
skb_is_nonlinear(const struct sk_buff * skb)2113 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2114 {
2115 return skb->data_len;
2116 }
2117
skb_headlen(const struct sk_buff * skb)2118 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2119 {
2120 return skb->len - skb->data_len;
2121 }
2122
__skb_pagelen(const struct sk_buff * skb)2123 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2124 {
2125 unsigned int i, len = 0;
2126
2127 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2128 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2129 return len;
2130 }
2131
skb_pagelen(const struct sk_buff * skb)2132 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2133 {
2134 return skb_headlen(skb) + __skb_pagelen(skb);
2135 }
2136
2137 /**
2138 * __skb_fill_page_desc - initialise a paged fragment in an skb
2139 * @skb: buffer containing fragment to be initialised
2140 * @i: paged fragment index to initialise
2141 * @page: the page to use for this fragment
2142 * @off: the offset to the data with @page
2143 * @size: the length of the data
2144 *
2145 * Initialises the @i'th fragment of @skb to point to &size bytes at
2146 * offset @off within @page.
2147 *
2148 * Does not take any additional reference on the fragment.
2149 */
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2150 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2151 struct page *page, int off, int size)
2152 {
2153 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2154
2155 /*
2156 * Propagate page pfmemalloc to the skb if we can. The problem is
2157 * that not all callers have unique ownership of the page but rely
2158 * on page_is_pfmemalloc doing the right thing(tm).
2159 */
2160 frag->bv_page = page;
2161 frag->bv_offset = off;
2162 skb_frag_size_set(frag, size);
2163
2164 page = compound_head(page);
2165 if (page_is_pfmemalloc(page))
2166 skb->pfmemalloc = true;
2167 }
2168
2169 /**
2170 * skb_fill_page_desc - initialise a paged fragment in an skb
2171 * @skb: buffer containing fragment to be initialised
2172 * @i: paged fragment index to initialise
2173 * @page: the page to use for this fragment
2174 * @off: the offset to the data with @page
2175 * @size: the length of the data
2176 *
2177 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2178 * @skb to point to @size bytes at offset @off within @page. In
2179 * addition updates @skb such that @i is the last fragment.
2180 *
2181 * Does not take any additional reference on the fragment.
2182 */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2183 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2184 struct page *page, int off, int size)
2185 {
2186 __skb_fill_page_desc(skb, i, page, off, size);
2187 skb_shinfo(skb)->nr_frags = i + 1;
2188 }
2189
2190 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2191 int size, unsigned int truesize);
2192
2193 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2194 unsigned int truesize);
2195
2196 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2197
2198 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2199 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2200 {
2201 return skb->head + skb->tail;
2202 }
2203
skb_reset_tail_pointer(struct sk_buff * skb)2204 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2205 {
2206 skb->tail = skb->data - skb->head;
2207 }
2208
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2209 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2210 {
2211 skb_reset_tail_pointer(skb);
2212 skb->tail += offset;
2213 }
2214
2215 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2216 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2217 {
2218 return skb->tail;
2219 }
2220
skb_reset_tail_pointer(struct sk_buff * skb)2221 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2222 {
2223 skb->tail = skb->data;
2224 }
2225
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2226 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2227 {
2228 skb->tail = skb->data + offset;
2229 }
2230
2231 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2232
skb_assert_len(struct sk_buff * skb)2233 static inline void skb_assert_len(struct sk_buff *skb)
2234 {
2235 #ifdef CONFIG_DEBUG_NET
2236 if (WARN_ONCE(!skb->len, "%s\n", __func__))
2237 DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2238 #endif /* CONFIG_DEBUG_NET */
2239 }
2240
2241 /*
2242 * Add data to an sk_buff
2243 */
2244 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2245 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2246 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2247 {
2248 void *tmp = skb_tail_pointer(skb);
2249 SKB_LINEAR_ASSERT(skb);
2250 skb->tail += len;
2251 skb->len += len;
2252 return tmp;
2253 }
2254
__skb_put_zero(struct sk_buff * skb,unsigned int len)2255 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2256 {
2257 void *tmp = __skb_put(skb, len);
2258
2259 memset(tmp, 0, len);
2260 return tmp;
2261 }
2262
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2263 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2264 unsigned int len)
2265 {
2266 void *tmp = __skb_put(skb, len);
2267
2268 memcpy(tmp, data, len);
2269 return tmp;
2270 }
2271
__skb_put_u8(struct sk_buff * skb,u8 val)2272 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2273 {
2274 *(u8 *)__skb_put(skb, 1) = val;
2275 }
2276
skb_put_zero(struct sk_buff * skb,unsigned int len)2277 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2278 {
2279 void *tmp = skb_put(skb, len);
2280
2281 memset(tmp, 0, len);
2282
2283 return tmp;
2284 }
2285
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2286 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2287 unsigned int len)
2288 {
2289 void *tmp = skb_put(skb, len);
2290
2291 memcpy(tmp, data, len);
2292
2293 return tmp;
2294 }
2295
skb_put_u8(struct sk_buff * skb,u8 val)2296 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2297 {
2298 *(u8 *)skb_put(skb, 1) = val;
2299 }
2300
2301 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2302 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2303 {
2304 skb->data -= len;
2305 skb->len += len;
2306 return skb->data;
2307 }
2308
2309 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2310 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2311 {
2312 skb->len -= len;
2313 BUG_ON(skb->len < skb->data_len);
2314 return skb->data += len;
2315 }
2316
skb_pull_inline(struct sk_buff * skb,unsigned int len)2317 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2318 {
2319 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2320 }
2321
2322 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2323
__pskb_pull(struct sk_buff * skb,unsigned int len)2324 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2325 {
2326 if (len > skb_headlen(skb) &&
2327 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2328 return NULL;
2329 skb->len -= len;
2330 return skb->data += len;
2331 }
2332
pskb_pull(struct sk_buff * skb,unsigned int len)2333 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2334 {
2335 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2336 }
2337
pskb_may_pull(struct sk_buff * skb,unsigned int len)2338 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2339 {
2340 if (likely(len <= skb_headlen(skb)))
2341 return 1;
2342 if (unlikely(len > skb->len))
2343 return 0;
2344 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2345 }
2346
2347 void skb_condense(struct sk_buff *skb);
2348
2349 /**
2350 * skb_headroom - bytes at buffer head
2351 * @skb: buffer to check
2352 *
2353 * Return the number of bytes of free space at the head of an &sk_buff.
2354 */
skb_headroom(const struct sk_buff * skb)2355 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2356 {
2357 return skb->data - skb->head;
2358 }
2359
2360 /**
2361 * skb_tailroom - bytes at buffer end
2362 * @skb: buffer to check
2363 *
2364 * Return the number of bytes of free space at the tail of an sk_buff
2365 */
skb_tailroom(const struct sk_buff * skb)2366 static inline int skb_tailroom(const struct sk_buff *skb)
2367 {
2368 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2369 }
2370
2371 /**
2372 * skb_availroom - 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 * allocated by sk_stream_alloc()
2377 */
skb_availroom(const struct sk_buff * skb)2378 static inline int skb_availroom(const struct sk_buff *skb)
2379 {
2380 if (skb_is_nonlinear(skb))
2381 return 0;
2382
2383 return skb->end - skb->tail - skb->reserved_tailroom;
2384 }
2385
2386 /**
2387 * skb_reserve - adjust headroom
2388 * @skb: buffer to alter
2389 * @len: bytes to move
2390 *
2391 * Increase the headroom of an empty &sk_buff by reducing the tail
2392 * room. This is only allowed for an empty buffer.
2393 */
skb_reserve(struct sk_buff * skb,int len)2394 static inline void skb_reserve(struct sk_buff *skb, int len)
2395 {
2396 skb->data += len;
2397 skb->tail += len;
2398 }
2399
2400 /**
2401 * skb_tailroom_reserve - adjust reserved_tailroom
2402 * @skb: buffer to alter
2403 * @mtu: maximum amount of headlen permitted
2404 * @needed_tailroom: minimum amount of reserved_tailroom
2405 *
2406 * Set reserved_tailroom so that headlen can be as large as possible but
2407 * not larger than mtu and tailroom cannot be smaller than
2408 * needed_tailroom.
2409 * The required headroom should already have been reserved before using
2410 * this function.
2411 */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2412 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2413 unsigned int needed_tailroom)
2414 {
2415 SKB_LINEAR_ASSERT(skb);
2416 if (mtu < skb_tailroom(skb) - needed_tailroom)
2417 /* use at most mtu */
2418 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2419 else
2420 /* use up to all available space */
2421 skb->reserved_tailroom = needed_tailroom;
2422 }
2423
2424 #define ENCAP_TYPE_ETHER 0
2425 #define ENCAP_TYPE_IPPROTO 1
2426
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2427 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2428 __be16 protocol)
2429 {
2430 skb->inner_protocol = protocol;
2431 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2432 }
2433
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2434 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2435 __u8 ipproto)
2436 {
2437 skb->inner_ipproto = ipproto;
2438 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2439 }
2440
skb_reset_inner_headers(struct sk_buff * skb)2441 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2442 {
2443 skb->inner_mac_header = skb->mac_header;
2444 skb->inner_network_header = skb->network_header;
2445 skb->inner_transport_header = skb->transport_header;
2446 }
2447
skb_reset_mac_len(struct sk_buff * skb)2448 static inline void skb_reset_mac_len(struct sk_buff *skb)
2449 {
2450 skb->mac_len = skb->network_header - skb->mac_header;
2451 }
2452
skb_inner_transport_header(const struct sk_buff * skb)2453 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2454 *skb)
2455 {
2456 return skb->head + skb->inner_transport_header;
2457 }
2458
skb_inner_transport_offset(const struct sk_buff * skb)2459 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2460 {
2461 return skb_inner_transport_header(skb) - skb->data;
2462 }
2463
skb_reset_inner_transport_header(struct sk_buff * skb)2464 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2465 {
2466 skb->inner_transport_header = skb->data - skb->head;
2467 }
2468
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2469 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2470 const int offset)
2471 {
2472 skb_reset_inner_transport_header(skb);
2473 skb->inner_transport_header += offset;
2474 }
2475
skb_inner_network_header(const struct sk_buff * skb)2476 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2477 {
2478 return skb->head + skb->inner_network_header;
2479 }
2480
skb_reset_inner_network_header(struct sk_buff * skb)2481 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2482 {
2483 skb->inner_network_header = skb->data - skb->head;
2484 }
2485
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2486 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2487 const int offset)
2488 {
2489 skb_reset_inner_network_header(skb);
2490 skb->inner_network_header += offset;
2491 }
2492
skb_inner_mac_header(const struct sk_buff * skb)2493 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2494 {
2495 return skb->head + skb->inner_mac_header;
2496 }
2497
skb_reset_inner_mac_header(struct sk_buff * skb)2498 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2499 {
2500 skb->inner_mac_header = skb->data - skb->head;
2501 }
2502
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)2503 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2504 const int offset)
2505 {
2506 skb_reset_inner_mac_header(skb);
2507 skb->inner_mac_header += offset;
2508 }
skb_transport_header_was_set(const struct sk_buff * skb)2509 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2510 {
2511 return skb->transport_header != (typeof(skb->transport_header))~0U;
2512 }
2513
skb_transport_header(const struct sk_buff * skb)2514 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2515 {
2516 return skb->head + skb->transport_header;
2517 }
2518
skb_reset_transport_header(struct sk_buff * skb)2519 static inline void skb_reset_transport_header(struct sk_buff *skb)
2520 {
2521 skb->transport_header = skb->data - skb->head;
2522 }
2523
skb_set_transport_header(struct sk_buff * skb,const int offset)2524 static inline void skb_set_transport_header(struct sk_buff *skb,
2525 const int offset)
2526 {
2527 skb_reset_transport_header(skb);
2528 skb->transport_header += offset;
2529 }
2530
skb_network_header(const struct sk_buff * skb)2531 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2532 {
2533 return skb->head + skb->network_header;
2534 }
2535
skb_reset_network_header(struct sk_buff * skb)2536 static inline void skb_reset_network_header(struct sk_buff *skb)
2537 {
2538 skb->network_header = skb->data - skb->head;
2539 }
2540
skb_set_network_header(struct sk_buff * skb,const int offset)2541 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2542 {
2543 skb_reset_network_header(skb);
2544 skb->network_header += offset;
2545 }
2546
skb_mac_header(const struct sk_buff * skb)2547 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2548 {
2549 return skb->head + skb->mac_header;
2550 }
2551
skb_mac_offset(const struct sk_buff * skb)2552 static inline int skb_mac_offset(const struct sk_buff *skb)
2553 {
2554 return skb_mac_header(skb) - skb->data;
2555 }
2556
skb_mac_header_len(const struct sk_buff * skb)2557 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2558 {
2559 return skb->network_header - skb->mac_header;
2560 }
2561
skb_mac_header_was_set(const struct sk_buff * skb)2562 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2563 {
2564 return skb->mac_header != (typeof(skb->mac_header))~0U;
2565 }
2566
skb_reset_mac_header(struct sk_buff * skb)2567 static inline void skb_reset_mac_header(struct sk_buff *skb)
2568 {
2569 skb->mac_header = skb->data - skb->head;
2570 }
2571
skb_set_mac_header(struct sk_buff * skb,const int offset)2572 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2573 {
2574 skb_reset_mac_header(skb);
2575 skb->mac_header += offset;
2576 }
2577
skb_pop_mac_header(struct sk_buff * skb)2578 static inline void skb_pop_mac_header(struct sk_buff *skb)
2579 {
2580 skb->mac_header = skb->network_header;
2581 }
2582
skb_probe_transport_header(struct sk_buff * skb)2583 static inline void skb_probe_transport_header(struct sk_buff *skb)
2584 {
2585 struct flow_keys_basic keys;
2586
2587 if (skb_transport_header_was_set(skb))
2588 return;
2589
2590 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2591 NULL, 0, 0, 0, 0))
2592 skb_set_transport_header(skb, keys.control.thoff);
2593 }
2594
skb_mac_header_rebuild(struct sk_buff * skb)2595 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2596 {
2597 if (skb_mac_header_was_set(skb)) {
2598 const unsigned char *old_mac = skb_mac_header(skb);
2599
2600 skb_set_mac_header(skb, -skb->mac_len);
2601 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2602 }
2603 }
2604
skb_checksum_start_offset(const struct sk_buff * skb)2605 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2606 {
2607 return skb->csum_start - skb_headroom(skb);
2608 }
2609
skb_checksum_start(const struct sk_buff * skb)2610 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2611 {
2612 return skb->head + skb->csum_start;
2613 }
2614
skb_transport_offset(const struct sk_buff * skb)2615 static inline int skb_transport_offset(const struct sk_buff *skb)
2616 {
2617 return skb_transport_header(skb) - skb->data;
2618 }
2619
skb_network_header_len(const struct sk_buff * skb)2620 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2621 {
2622 return skb->transport_header - skb->network_header;
2623 }
2624
skb_inner_network_header_len(const struct sk_buff * skb)2625 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2626 {
2627 return skb->inner_transport_header - skb->inner_network_header;
2628 }
2629
skb_network_offset(const struct sk_buff * skb)2630 static inline int skb_network_offset(const struct sk_buff *skb)
2631 {
2632 return skb_network_header(skb) - skb->data;
2633 }
2634
skb_inner_network_offset(const struct sk_buff * skb)2635 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2636 {
2637 return skb_inner_network_header(skb) - skb->data;
2638 }
2639
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)2640 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2641 {
2642 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2643 }
2644
2645 /*
2646 * CPUs often take a performance hit when accessing unaligned memory
2647 * locations. The actual performance hit varies, it can be small if the
2648 * hardware handles it or large if we have to take an exception and fix it
2649 * in software.
2650 *
2651 * Since an ethernet header is 14 bytes network drivers often end up with
2652 * the IP header at an unaligned offset. The IP header can be aligned by
2653 * shifting the start of the packet by 2 bytes. Drivers should do this
2654 * with:
2655 *
2656 * skb_reserve(skb, NET_IP_ALIGN);
2657 *
2658 * The downside to this alignment of the IP header is that the DMA is now
2659 * unaligned. On some architectures the cost of an unaligned DMA is high
2660 * and this cost outweighs the gains made by aligning the IP header.
2661 *
2662 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2663 * to be overridden.
2664 */
2665 #ifndef NET_IP_ALIGN
2666 #define NET_IP_ALIGN 2
2667 #endif
2668
2669 /*
2670 * The networking layer reserves some headroom in skb data (via
2671 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2672 * the header has to grow. In the default case, if the header has to grow
2673 * 32 bytes or less we avoid the reallocation.
2674 *
2675 * Unfortunately this headroom changes the DMA alignment of the resulting
2676 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2677 * on some architectures. An architecture can override this value,
2678 * perhaps setting it to a cacheline in size (since that will maintain
2679 * cacheline alignment of the DMA). It must be a power of 2.
2680 *
2681 * Various parts of the networking layer expect at least 32 bytes of
2682 * headroom, you should not reduce this.
2683 *
2684 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2685 * to reduce average number of cache lines per packet.
2686 * get_rps_cpus() for example only access one 64 bytes aligned block :
2687 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2688 */
2689 #ifndef NET_SKB_PAD
2690 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2691 #endif
2692
2693 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2694
__skb_set_length(struct sk_buff * skb,unsigned int len)2695 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2696 {
2697 if (WARN_ON(skb_is_nonlinear(skb)))
2698 return;
2699 skb->len = len;
2700 skb_set_tail_pointer(skb, len);
2701 }
2702
__skb_trim(struct sk_buff * skb,unsigned int len)2703 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2704 {
2705 __skb_set_length(skb, len);
2706 }
2707
2708 void skb_trim(struct sk_buff *skb, unsigned int len);
2709
__pskb_trim(struct sk_buff * skb,unsigned int len)2710 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2711 {
2712 if (skb->data_len)
2713 return ___pskb_trim(skb, len);
2714 __skb_trim(skb, len);
2715 return 0;
2716 }
2717
pskb_trim(struct sk_buff * skb,unsigned int len)2718 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2719 {
2720 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2721 }
2722
2723 /**
2724 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2725 * @skb: buffer to alter
2726 * @len: new length
2727 *
2728 * This is identical to pskb_trim except that the caller knows that
2729 * the skb is not cloned so we should never get an error due to out-
2730 * of-memory.
2731 */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)2732 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2733 {
2734 int err = pskb_trim(skb, len);
2735 BUG_ON(err);
2736 }
2737
__skb_grow(struct sk_buff * skb,unsigned int len)2738 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2739 {
2740 unsigned int diff = len - skb->len;
2741
2742 if (skb_tailroom(skb) < diff) {
2743 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2744 GFP_ATOMIC);
2745 if (ret)
2746 return ret;
2747 }
2748 __skb_set_length(skb, len);
2749 return 0;
2750 }
2751
2752 /**
2753 * skb_orphan - orphan a buffer
2754 * @skb: buffer to orphan
2755 *
2756 * If a buffer currently has an owner then we call the owner's
2757 * destructor function and make the @skb unowned. The buffer continues
2758 * to exist but is no longer charged to its former owner.
2759 */
skb_orphan(struct sk_buff * skb)2760 static inline void skb_orphan(struct sk_buff *skb)
2761 {
2762 if (skb->destructor) {
2763 skb->destructor(skb);
2764 skb->destructor = NULL;
2765 skb->sk = NULL;
2766 } else {
2767 BUG_ON(skb->sk);
2768 }
2769 }
2770
2771 /**
2772 * skb_orphan_frags - orphan the frags contained in a buffer
2773 * @skb: buffer to orphan frags from
2774 * @gfp_mask: allocation mask for replacement pages
2775 *
2776 * For each frag in the SKB which needs a destructor (i.e. has an
2777 * owner) create a copy of that frag and release the original
2778 * page by calling the destructor.
2779 */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)2780 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2781 {
2782 if (likely(!skb_zcopy(skb)))
2783 return 0;
2784 if (!skb_zcopy_is_nouarg(skb) &&
2785 skb_uarg(skb)->callback == sock_zerocopy_callback)
2786 return 0;
2787 return skb_copy_ubufs(skb, gfp_mask);
2788 }
2789
2790 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
skb_orphan_frags_rx(struct sk_buff * skb,gfp_t gfp_mask)2791 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2792 {
2793 if (likely(!skb_zcopy(skb)))
2794 return 0;
2795 return skb_copy_ubufs(skb, gfp_mask);
2796 }
2797
2798 /**
2799 * __skb_queue_purge - empty a list
2800 * @list: list to empty
2801 *
2802 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2803 * the list and one reference dropped. This function does not take the
2804 * list lock and the caller must hold the relevant locks to use it.
2805 */
__skb_queue_purge(struct sk_buff_head * list)2806 static inline void __skb_queue_purge(struct sk_buff_head *list)
2807 {
2808 struct sk_buff *skb;
2809 while ((skb = __skb_dequeue(list)) != NULL)
2810 kfree_skb(skb);
2811 }
2812 void skb_queue_purge(struct sk_buff_head *list);
2813
2814 unsigned int skb_rbtree_purge(struct rb_root *root);
2815
2816 void *netdev_alloc_frag(unsigned int fragsz);
2817
2818 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2819 gfp_t gfp_mask);
2820
2821 /**
2822 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2823 * @dev: network device to receive on
2824 * @length: length to allocate
2825 *
2826 * Allocate a new &sk_buff and assign it a usage count of one. The
2827 * buffer has unspecified headroom built in. Users should allocate
2828 * the headroom they think they need without accounting for the
2829 * built in space. The built in space is used for optimisations.
2830 *
2831 * %NULL is returned if there is no free memory. Although this function
2832 * allocates memory it can be called from an interrupt.
2833 */
netdev_alloc_skb(struct net_device * dev,unsigned int length)2834 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2835 unsigned int length)
2836 {
2837 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2838 }
2839
2840 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)2841 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2842 gfp_t gfp_mask)
2843 {
2844 return __netdev_alloc_skb(NULL, length, gfp_mask);
2845 }
2846
2847 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)2848 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2849 {
2850 return netdev_alloc_skb(NULL, length);
2851 }
2852
2853
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)2854 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2855 unsigned int length, gfp_t gfp)
2856 {
2857 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2858
2859 if (NET_IP_ALIGN && skb)
2860 skb_reserve(skb, NET_IP_ALIGN);
2861 return skb;
2862 }
2863
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)2864 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2865 unsigned int length)
2866 {
2867 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2868 }
2869
skb_free_frag(void * addr)2870 static inline void skb_free_frag(void *addr)
2871 {
2872 page_frag_free(addr);
2873 }
2874
2875 void *napi_alloc_frag(unsigned int fragsz);
2876 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2877 unsigned int length, gfp_t gfp_mask);
napi_alloc_skb(struct napi_struct * napi,unsigned int length)2878 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2879 unsigned int length)
2880 {
2881 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2882 }
2883 void napi_consume_skb(struct sk_buff *skb, int budget);
2884
2885 void __kfree_skb_flush(void);
2886 void __kfree_skb_defer(struct sk_buff *skb);
2887
2888 /**
2889 * __dev_alloc_pages - allocate page for network Rx
2890 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2891 * @order: size of the allocation
2892 *
2893 * Allocate a new page.
2894 *
2895 * %NULL is returned if there is no free memory.
2896 */
__dev_alloc_pages(gfp_t gfp_mask,unsigned int order)2897 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2898 unsigned int order)
2899 {
2900 /* This piece of code contains several assumptions.
2901 * 1. This is for device Rx, therefor a cold page is preferred.
2902 * 2. The expectation is the user wants a compound page.
2903 * 3. If requesting a order 0 page it will not be compound
2904 * due to the check to see if order has a value in prep_new_page
2905 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2906 * code in gfp_to_alloc_flags that should be enforcing this.
2907 */
2908 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2909
2910 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2911 }
2912
dev_alloc_pages(unsigned int order)2913 static inline struct page *dev_alloc_pages(unsigned int order)
2914 {
2915 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2916 }
2917
2918 /**
2919 * __dev_alloc_page - allocate a page for network Rx
2920 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2921 *
2922 * Allocate a new page.
2923 *
2924 * %NULL is returned if there is no free memory.
2925 */
__dev_alloc_page(gfp_t gfp_mask)2926 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2927 {
2928 return __dev_alloc_pages(gfp_mask, 0);
2929 }
2930
dev_alloc_page(void)2931 static inline struct page *dev_alloc_page(void)
2932 {
2933 return dev_alloc_pages(0);
2934 }
2935
2936 /**
2937 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2938 * @page: The page that was allocated from skb_alloc_page
2939 * @skb: The skb that may need pfmemalloc set
2940 */
skb_propagate_pfmemalloc(struct page * page,struct sk_buff * skb)2941 static inline void skb_propagate_pfmemalloc(struct page *page,
2942 struct sk_buff *skb)
2943 {
2944 if (page_is_pfmemalloc(page))
2945 skb->pfmemalloc = true;
2946 }
2947
2948 /**
2949 * skb_frag_off() - Returns the offset of a skb fragment
2950 * @frag: the paged fragment
2951 */
skb_frag_off(const skb_frag_t * frag)2952 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2953 {
2954 return frag->bv_offset;
2955 }
2956
2957 /**
2958 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2959 * @frag: skb fragment
2960 * @delta: value to add
2961 */
skb_frag_off_add(skb_frag_t * frag,int delta)2962 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2963 {
2964 frag->bv_offset += delta;
2965 }
2966
2967 /**
2968 * skb_frag_off_set() - Sets the offset of a skb fragment
2969 * @frag: skb fragment
2970 * @offset: offset of fragment
2971 */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)2972 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2973 {
2974 frag->bv_offset = offset;
2975 }
2976
2977 /**
2978 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2979 * @fragto: skb fragment where offset is set
2980 * @fragfrom: skb fragment offset is copied from
2981 */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)2982 static inline void skb_frag_off_copy(skb_frag_t *fragto,
2983 const skb_frag_t *fragfrom)
2984 {
2985 fragto->bv_offset = fragfrom->bv_offset;
2986 }
2987
2988 /**
2989 * skb_frag_page - retrieve the page referred to by a paged fragment
2990 * @frag: the paged fragment
2991 *
2992 * Returns the &struct page associated with @frag.
2993 */
skb_frag_page(const skb_frag_t * frag)2994 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2995 {
2996 return frag->bv_page;
2997 }
2998
2999 /**
3000 * __skb_frag_ref - take an addition reference on a paged fragment.
3001 * @frag: the paged fragment
3002 *
3003 * Takes an additional reference on the paged fragment @frag.
3004 */
__skb_frag_ref(skb_frag_t * frag)3005 static inline void __skb_frag_ref(skb_frag_t *frag)
3006 {
3007 get_page(skb_frag_page(frag));
3008 }
3009
3010 /**
3011 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3012 * @skb: the buffer
3013 * @f: the fragment offset.
3014 *
3015 * Takes an additional reference on the @f'th paged fragment of @skb.
3016 */
skb_frag_ref(struct sk_buff * skb,int f)3017 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3018 {
3019 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3020 }
3021
3022 /**
3023 * __skb_frag_unref - release a reference on a paged fragment.
3024 * @frag: the paged fragment
3025 *
3026 * Releases a reference on the paged fragment @frag.
3027 */
__skb_frag_unref(skb_frag_t * frag)3028 static inline void __skb_frag_unref(skb_frag_t *frag)
3029 {
3030 put_page(skb_frag_page(frag));
3031 }
3032
3033 /**
3034 * skb_frag_unref - release a reference on a paged fragment of an skb.
3035 * @skb: the buffer
3036 * @f: the fragment offset
3037 *
3038 * Releases a reference on the @f'th paged fragment of @skb.
3039 */
skb_frag_unref(struct sk_buff * skb,int f)3040 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3041 {
3042 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3043 }
3044
3045 /**
3046 * skb_frag_address - gets the address of the data contained in a paged fragment
3047 * @frag: the paged fragment buffer
3048 *
3049 * Returns the address of the data within @frag. The page must already
3050 * be mapped.
3051 */
skb_frag_address(const skb_frag_t * frag)3052 static inline void *skb_frag_address(const skb_frag_t *frag)
3053 {
3054 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3055 }
3056
3057 /**
3058 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3059 * @frag: the paged fragment buffer
3060 *
3061 * Returns the address of the data within @frag. Checks that the page
3062 * is mapped and returns %NULL otherwise.
3063 */
skb_frag_address_safe(const skb_frag_t * frag)3064 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3065 {
3066 void *ptr = page_address(skb_frag_page(frag));
3067 if (unlikely(!ptr))
3068 return NULL;
3069
3070 return ptr + skb_frag_off(frag);
3071 }
3072
3073 /**
3074 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3075 * @fragto: skb fragment where page is set
3076 * @fragfrom: skb fragment page is copied from
3077 */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3078 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3079 const skb_frag_t *fragfrom)
3080 {
3081 fragto->bv_page = fragfrom->bv_page;
3082 }
3083
3084 /**
3085 * __skb_frag_set_page - sets the page contained in a paged fragment
3086 * @frag: the paged fragment
3087 * @page: the page to set
3088 *
3089 * Sets the fragment @frag to contain @page.
3090 */
__skb_frag_set_page(skb_frag_t * frag,struct page * page)3091 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3092 {
3093 frag->bv_page = page;
3094 }
3095
3096 /**
3097 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3098 * @skb: the buffer
3099 * @f: the fragment offset
3100 * @page: the page to set
3101 *
3102 * Sets the @f'th fragment of @skb to contain @page.
3103 */
skb_frag_set_page(struct sk_buff * skb,int f,struct page * page)3104 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3105 struct page *page)
3106 {
3107 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3108 }
3109
3110 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3111
3112 /**
3113 * skb_frag_dma_map - maps a paged fragment via the DMA API
3114 * @dev: the device to map the fragment to
3115 * @frag: the paged fragment to map
3116 * @offset: the offset within the fragment (starting at the
3117 * fragment's own offset)
3118 * @size: the number of bytes to map
3119 * @dir: the direction of the mapping (``PCI_DMA_*``)
3120 *
3121 * Maps the page associated with @frag to @device.
3122 */
skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3123 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3124 const skb_frag_t *frag,
3125 size_t offset, size_t size,
3126 enum dma_data_direction dir)
3127 {
3128 return dma_map_page(dev, skb_frag_page(frag),
3129 skb_frag_off(frag) + offset, size, dir);
3130 }
3131
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3132 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3133 gfp_t gfp_mask)
3134 {
3135 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3136 }
3137
3138
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3139 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3140 gfp_t gfp_mask)
3141 {
3142 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3143 }
3144
3145
3146 /**
3147 * skb_clone_writable - is the header of a clone writable
3148 * @skb: buffer to check
3149 * @len: length up to which to write
3150 *
3151 * Returns true if modifying the header part of the cloned buffer
3152 * does not requires the data to be copied.
3153 */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3154 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3155 {
3156 return !skb_header_cloned(skb) &&
3157 skb_headroom(skb) + len <= skb->hdr_len;
3158 }
3159
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3160 static inline int skb_try_make_writable(struct sk_buff *skb,
3161 unsigned int write_len)
3162 {
3163 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3164 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3165 }
3166
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3167 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3168 int cloned)
3169 {
3170 int delta = 0;
3171
3172 if (headroom > skb_headroom(skb))
3173 delta = headroom - skb_headroom(skb);
3174
3175 if (delta || cloned)
3176 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3177 GFP_ATOMIC);
3178 return 0;
3179 }
3180
3181 /**
3182 * skb_cow - copy header of skb when it is required
3183 * @skb: buffer to cow
3184 * @headroom: needed headroom
3185 *
3186 * If the skb passed lacks sufficient headroom or its data part
3187 * is shared, data is reallocated. If reallocation fails, an error
3188 * is returned and original skb is not changed.
3189 *
3190 * The result is skb with writable area skb->head...skb->tail
3191 * and at least @headroom of space at head.
3192 */
skb_cow(struct sk_buff * skb,unsigned int headroom)3193 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3194 {
3195 return __skb_cow(skb, headroom, skb_cloned(skb));
3196 }
3197
3198 /**
3199 * skb_cow_head - skb_cow but only making the head writable
3200 * @skb: buffer to cow
3201 * @headroom: needed headroom
3202 *
3203 * This function is identical to skb_cow except that we replace the
3204 * skb_cloned check by skb_header_cloned. It should be used when
3205 * you only need to push on some header and do not need to modify
3206 * the data.
3207 */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3208 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3209 {
3210 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3211 }
3212
3213 /**
3214 * skb_padto - pad an skbuff up to a minimal size
3215 * @skb: buffer to pad
3216 * @len: minimal length
3217 *
3218 * Pads up a buffer to ensure the trailing bytes exist and are
3219 * blanked. If the buffer already contains sufficient data it
3220 * is untouched. Otherwise it is extended. Returns zero on
3221 * success. The skb is freed on error.
3222 */
skb_padto(struct sk_buff * skb,unsigned int len)3223 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3224 {
3225 unsigned int size = skb->len;
3226 if (likely(size >= len))
3227 return 0;
3228 return skb_pad(skb, len - size);
3229 }
3230
3231 /**
3232 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3233 * @skb: buffer to pad
3234 * @len: minimal length
3235 * @free_on_error: free buffer on error
3236 *
3237 * Pads up a buffer to ensure the trailing bytes exist and are
3238 * blanked. If the buffer already contains sufficient data it
3239 * is untouched. Otherwise it is extended. Returns zero on
3240 * success. The skb is freed on error if @free_on_error is true.
3241 */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3242 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3243 unsigned int len,
3244 bool free_on_error)
3245 {
3246 unsigned int size = skb->len;
3247
3248 if (unlikely(size < len)) {
3249 len -= size;
3250 if (__skb_pad(skb, len, free_on_error))
3251 return -ENOMEM;
3252 __skb_put(skb, len);
3253 }
3254 return 0;
3255 }
3256
3257 /**
3258 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3259 * @skb: buffer to pad
3260 * @len: minimal length
3261 *
3262 * Pads up a buffer to ensure the trailing bytes exist and are
3263 * blanked. If the buffer already contains sufficient data it
3264 * is untouched. Otherwise it is extended. Returns zero on
3265 * success. The skb is freed on error.
3266 */
skb_put_padto(struct sk_buff * skb,unsigned int len)3267 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3268 {
3269 return __skb_put_padto(skb, len, true);
3270 }
3271
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3272 static inline int skb_add_data(struct sk_buff *skb,
3273 struct iov_iter *from, int copy)
3274 {
3275 const int off = skb->len;
3276
3277 if (skb->ip_summed == CHECKSUM_NONE) {
3278 __wsum csum = 0;
3279 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3280 &csum, from)) {
3281 skb->csum = csum_block_add(skb->csum, csum, off);
3282 return 0;
3283 }
3284 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3285 return 0;
3286
3287 __skb_trim(skb, off);
3288 return -EFAULT;
3289 }
3290
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3291 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3292 const struct page *page, int off)
3293 {
3294 if (skb_zcopy(skb))
3295 return false;
3296 if (i) {
3297 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3298
3299 return page == skb_frag_page(frag) &&
3300 off == skb_frag_off(frag) + skb_frag_size(frag);
3301 }
3302 return false;
3303 }
3304
__skb_linearize(struct sk_buff * skb)3305 static inline int __skb_linearize(struct sk_buff *skb)
3306 {
3307 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3308 }
3309
3310 /**
3311 * skb_linearize - convert paged skb to linear one
3312 * @skb: buffer to linarize
3313 *
3314 * If there is no free memory -ENOMEM is returned, otherwise zero
3315 * is returned and the old skb data released.
3316 */
skb_linearize(struct sk_buff * skb)3317 static inline int skb_linearize(struct sk_buff *skb)
3318 {
3319 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3320 }
3321
3322 /**
3323 * skb_has_shared_frag - can any frag be overwritten
3324 * @skb: buffer to test
3325 *
3326 * Return true if the skb has at least one frag that might be modified
3327 * by an external entity (as in vmsplice()/sendfile())
3328 */
skb_has_shared_frag(const struct sk_buff * skb)3329 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3330 {
3331 return skb_is_nonlinear(skb) &&
3332 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3333 }
3334
3335 /**
3336 * skb_linearize_cow - make sure skb is linear and writable
3337 * @skb: buffer to process
3338 *
3339 * If there is no free memory -ENOMEM is returned, otherwise zero
3340 * is returned and the old skb data released.
3341 */
skb_linearize_cow(struct sk_buff * skb)3342 static inline int skb_linearize_cow(struct sk_buff *skb)
3343 {
3344 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3345 __skb_linearize(skb) : 0;
3346 }
3347
3348 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3349 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3350 unsigned int off)
3351 {
3352 if (skb->ip_summed == CHECKSUM_COMPLETE)
3353 skb->csum = csum_block_sub(skb->csum,
3354 csum_partial(start, len, 0), off);
3355 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3356 skb_checksum_start_offset(skb) < 0)
3357 skb->ip_summed = CHECKSUM_NONE;
3358 }
3359
3360 /**
3361 * skb_postpull_rcsum - update checksum for received skb after pull
3362 * @skb: buffer to update
3363 * @start: start of data before pull
3364 * @len: length of data pulled
3365 *
3366 * After doing a pull on a received packet, you need to call this to
3367 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3368 * CHECKSUM_NONE so that it can be recomputed from scratch.
3369 */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3370 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3371 const void *start, unsigned int len)
3372 {
3373 __skb_postpull_rcsum(skb, start, len, 0);
3374 }
3375
3376 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3377 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3378 unsigned int off)
3379 {
3380 if (skb->ip_summed == CHECKSUM_COMPLETE)
3381 skb->csum = csum_block_add(skb->csum,
3382 csum_partial(start, len, 0), off);
3383 }
3384
3385 /**
3386 * skb_postpush_rcsum - update checksum for received skb after push
3387 * @skb: buffer to update
3388 * @start: start of data after push
3389 * @len: length of data pushed
3390 *
3391 * After doing a push on a received packet, you need to call this to
3392 * update the CHECKSUM_COMPLETE checksum.
3393 */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3394 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3395 const void *start, unsigned int len)
3396 {
3397 __skb_postpush_rcsum(skb, start, len, 0);
3398 }
3399
3400 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3401
3402 /**
3403 * skb_push_rcsum - push skb and update receive checksum
3404 * @skb: buffer to update
3405 * @len: length of data pulled
3406 *
3407 * This function performs an skb_push on the packet and updates
3408 * the CHECKSUM_COMPLETE checksum. It should be used on
3409 * receive path processing instead of skb_push unless you know
3410 * that the checksum difference is zero (e.g., a valid IP header)
3411 * or you are setting ip_summed to CHECKSUM_NONE.
3412 */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3413 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3414 {
3415 skb_push(skb, len);
3416 skb_postpush_rcsum(skb, skb->data, len);
3417 return skb->data;
3418 }
3419
3420 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3421 /**
3422 * pskb_trim_rcsum - trim received skb and update checksum
3423 * @skb: buffer to trim
3424 * @len: new length
3425 *
3426 * This is exactly the same as pskb_trim except that it ensures the
3427 * checksum of received packets are still valid after the operation.
3428 * It can change skb pointers.
3429 */
3430
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)3431 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3432 {
3433 if (likely(len >= skb->len))
3434 return 0;
3435 return pskb_trim_rcsum_slow(skb, len);
3436 }
3437
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)3438 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3439 {
3440 if (skb->ip_summed == CHECKSUM_COMPLETE)
3441 skb->ip_summed = CHECKSUM_NONE;
3442 __skb_trim(skb, len);
3443 return 0;
3444 }
3445
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)3446 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3447 {
3448 if (skb->ip_summed == CHECKSUM_COMPLETE)
3449 skb->ip_summed = CHECKSUM_NONE;
3450 return __skb_grow(skb, len);
3451 }
3452
3453 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3454 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3455 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3456 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3457 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3458
3459 #define skb_queue_walk(queue, skb) \
3460 for (skb = (queue)->next; \
3461 skb != (struct sk_buff *)(queue); \
3462 skb = skb->next)
3463
3464 #define skb_queue_walk_safe(queue, skb, tmp) \
3465 for (skb = (queue)->next, tmp = skb->next; \
3466 skb != (struct sk_buff *)(queue); \
3467 skb = tmp, tmp = skb->next)
3468
3469 #define skb_queue_walk_from(queue, skb) \
3470 for (; skb != (struct sk_buff *)(queue); \
3471 skb = skb->next)
3472
3473 #define skb_rbtree_walk(skb, root) \
3474 for (skb = skb_rb_first(root); skb != NULL; \
3475 skb = skb_rb_next(skb))
3476
3477 #define skb_rbtree_walk_from(skb) \
3478 for (; skb != NULL; \
3479 skb = skb_rb_next(skb))
3480
3481 #define skb_rbtree_walk_from_safe(skb, tmp) \
3482 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3483 skb = tmp)
3484
3485 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3486 for (tmp = skb->next; \
3487 skb != (struct sk_buff *)(queue); \
3488 skb = tmp, tmp = skb->next)
3489
3490 #define skb_queue_reverse_walk(queue, skb) \
3491 for (skb = (queue)->prev; \
3492 skb != (struct sk_buff *)(queue); \
3493 skb = skb->prev)
3494
3495 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3496 for (skb = (queue)->prev, tmp = skb->prev; \
3497 skb != (struct sk_buff *)(queue); \
3498 skb = tmp, tmp = skb->prev)
3499
3500 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3501 for (tmp = skb->prev; \
3502 skb != (struct sk_buff *)(queue); \
3503 skb = tmp, tmp = skb->prev)
3504
skb_has_frag_list(const struct sk_buff * skb)3505 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3506 {
3507 return skb_shinfo(skb)->frag_list != NULL;
3508 }
3509
skb_frag_list_init(struct sk_buff * skb)3510 static inline void skb_frag_list_init(struct sk_buff *skb)
3511 {
3512 skb_shinfo(skb)->frag_list = NULL;
3513 }
3514
3515 #define skb_walk_frags(skb, iter) \
3516 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3517
3518
3519 int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3520 const struct sk_buff *skb);
3521 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3522 struct sk_buff_head *queue,
3523 unsigned int flags,
3524 void (*destructor)(struct sock *sk,
3525 struct sk_buff *skb),
3526 int *off, int *err,
3527 struct sk_buff **last);
3528 struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
3529 void (*destructor)(struct sock *sk,
3530 struct sk_buff *skb),
3531 int *off, int *err,
3532 struct sk_buff **last);
3533 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
3534 void (*destructor)(struct sock *sk,
3535 struct sk_buff *skb),
3536 int *off, int *err);
3537 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3538 int *err);
3539 __poll_t datagram_poll(struct file *file, struct socket *sock,
3540 struct poll_table_struct *wait);
3541 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3542 struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)3543 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3544 struct msghdr *msg, int size)
3545 {
3546 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3547 }
3548 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3549 struct msghdr *msg);
3550 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3551 struct iov_iter *to, int len,
3552 struct ahash_request *hash);
3553 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3554 struct iov_iter *from, int len);
3555 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3556 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3557 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
skb_free_datagram_locked(struct sock * sk,struct sk_buff * skb)3558 static inline void skb_free_datagram_locked(struct sock *sk,
3559 struct sk_buff *skb)
3560 {
3561 __skb_free_datagram_locked(sk, skb, 0);
3562 }
3563 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3564 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3565 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3566 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3567 int len, __wsum csum);
3568 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3569 struct pipe_inode_info *pipe, unsigned int len,
3570 unsigned int flags);
3571 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3572 int len);
3573 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3574 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3575 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3576 int len, int hlen);
3577 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3578 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3579 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3580 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3581 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3582 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3583 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3584 unsigned int offset);
3585 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3586 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3587 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3588 int skb_vlan_pop(struct sk_buff *skb);
3589 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3590 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3591 int mac_len, bool ethernet);
3592 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3593 bool ethernet);
3594 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3595 int skb_mpls_dec_ttl(struct sk_buff *skb);
3596 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3597 gfp_t gfp);
3598
memcpy_from_msg(void * data,struct msghdr * msg,int len)3599 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3600 {
3601 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3602 }
3603
memcpy_to_msg(struct msghdr * msg,void * data,int len)3604 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3605 {
3606 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3607 }
3608
3609 struct skb_checksum_ops {
3610 __wsum (*update)(const void *mem, int len, __wsum wsum);
3611 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3612 };
3613
3614 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3615
3616 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3617 __wsum csum, const struct skb_checksum_ops *ops);
3618 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3619 __wsum csum);
3620
3621 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * data,int hlen,void * buffer)3622 __skb_header_pointer(const struct sk_buff *skb, int offset,
3623 int len, void *data, int hlen, void *buffer)
3624 {
3625 if (hlen - offset >= len)
3626 return data + offset;
3627
3628 if (!skb ||
3629 skb_copy_bits(skb, offset, buffer, len) < 0)
3630 return NULL;
3631
3632 return buffer;
3633 }
3634
3635 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)3636 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3637 {
3638 return __skb_header_pointer(skb, offset, len, skb->data,
3639 skb_headlen(skb), buffer);
3640 }
3641
3642 /**
3643 * skb_needs_linearize - check if we need to linearize a given skb
3644 * depending on the given device features.
3645 * @skb: socket buffer to check
3646 * @features: net device features
3647 *
3648 * Returns true if either:
3649 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3650 * 2. skb is fragmented and the device does not support SG.
3651 */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)3652 static inline bool skb_needs_linearize(struct sk_buff *skb,
3653 netdev_features_t features)
3654 {
3655 return skb_is_nonlinear(skb) &&
3656 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3657 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3658 }
3659
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)3660 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3661 void *to,
3662 const unsigned int len)
3663 {
3664 memcpy(to, skb->data, len);
3665 }
3666
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)3667 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3668 const int offset, void *to,
3669 const unsigned int len)
3670 {
3671 memcpy(to, skb->data + offset, len);
3672 }
3673
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)3674 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3675 const void *from,
3676 const unsigned int len)
3677 {
3678 memcpy(skb->data, from, len);
3679 }
3680
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)3681 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3682 const int offset,
3683 const void *from,
3684 const unsigned int len)
3685 {
3686 memcpy(skb->data + offset, from, len);
3687 }
3688
3689 void skb_init(void);
3690
skb_get_ktime(const struct sk_buff * skb)3691 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3692 {
3693 return skb->tstamp;
3694 }
3695
3696 /**
3697 * skb_get_timestamp - get timestamp from a skb
3698 * @skb: skb to get stamp from
3699 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3700 *
3701 * Timestamps are stored in the skb as offsets to a base timestamp.
3702 * This function converts the offset back to a struct timeval and stores
3703 * it in stamp.
3704 */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)3705 static inline void skb_get_timestamp(const struct sk_buff *skb,
3706 struct __kernel_old_timeval *stamp)
3707 {
3708 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3709 }
3710
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)3711 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3712 struct __kernel_sock_timeval *stamp)
3713 {
3714 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3715
3716 stamp->tv_sec = ts.tv_sec;
3717 stamp->tv_usec = ts.tv_nsec / 1000;
3718 }
3719
skb_get_timestampns(const struct sk_buff * skb,struct timespec * stamp)3720 static inline void skb_get_timestampns(const struct sk_buff *skb,
3721 struct timespec *stamp)
3722 {
3723 *stamp = ktime_to_timespec(skb->tstamp);
3724 }
3725
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)3726 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3727 struct __kernel_timespec *stamp)
3728 {
3729 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3730
3731 stamp->tv_sec = ts.tv_sec;
3732 stamp->tv_nsec = ts.tv_nsec;
3733 }
3734
__net_timestamp(struct sk_buff * skb)3735 static inline void __net_timestamp(struct sk_buff *skb)
3736 {
3737 skb->tstamp = ktime_get_real();
3738 }
3739
net_timedelta(ktime_t t)3740 static inline ktime_t net_timedelta(ktime_t t)
3741 {
3742 return ktime_sub(ktime_get_real(), t);
3743 }
3744
net_invalid_timestamp(void)3745 static inline ktime_t net_invalid_timestamp(void)
3746 {
3747 return 0;
3748 }
3749
skb_metadata_len(const struct sk_buff * skb)3750 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3751 {
3752 return skb_shinfo(skb)->meta_len;
3753 }
3754
skb_metadata_end(const struct sk_buff * skb)3755 static inline void *skb_metadata_end(const struct sk_buff *skb)
3756 {
3757 return skb_mac_header(skb);
3758 }
3759
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)3760 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3761 const struct sk_buff *skb_b,
3762 u8 meta_len)
3763 {
3764 const void *a = skb_metadata_end(skb_a);
3765 const void *b = skb_metadata_end(skb_b);
3766 /* Using more efficient varaiant than plain call to memcmp(). */
3767 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3768 u64 diffs = 0;
3769
3770 switch (meta_len) {
3771 #define __it(x, op) (x -= sizeof(u##op))
3772 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3773 case 32: diffs |= __it_diff(a, b, 64);
3774 /* fall through */
3775 case 24: diffs |= __it_diff(a, b, 64);
3776 /* fall through */
3777 case 16: diffs |= __it_diff(a, b, 64);
3778 /* fall through */
3779 case 8: diffs |= __it_diff(a, b, 64);
3780 break;
3781 case 28: diffs |= __it_diff(a, b, 64);
3782 /* fall through */
3783 case 20: diffs |= __it_diff(a, b, 64);
3784 /* fall through */
3785 case 12: diffs |= __it_diff(a, b, 64);
3786 /* fall through */
3787 case 4: diffs |= __it_diff(a, b, 32);
3788 break;
3789 }
3790 return diffs;
3791 #else
3792 return memcmp(a - meta_len, b - meta_len, meta_len);
3793 #endif
3794 }
3795
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)3796 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3797 const struct sk_buff *skb_b)
3798 {
3799 u8 len_a = skb_metadata_len(skb_a);
3800 u8 len_b = skb_metadata_len(skb_b);
3801
3802 if (!(len_a | len_b))
3803 return false;
3804
3805 return len_a != len_b ?
3806 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3807 }
3808
skb_metadata_set(struct sk_buff * skb,u8 meta_len)3809 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3810 {
3811 skb_shinfo(skb)->meta_len = meta_len;
3812 }
3813
skb_metadata_clear(struct sk_buff * skb)3814 static inline void skb_metadata_clear(struct sk_buff *skb)
3815 {
3816 skb_metadata_set(skb, 0);
3817 }
3818
3819 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3820
3821 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3822
3823 void skb_clone_tx_timestamp(struct sk_buff *skb);
3824 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3825
3826 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3827
skb_clone_tx_timestamp(struct sk_buff * skb)3828 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3829 {
3830 }
3831
skb_defer_rx_timestamp(struct sk_buff * skb)3832 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3833 {
3834 return false;
3835 }
3836
3837 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3838
3839 /**
3840 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3841 *
3842 * PHY drivers may accept clones of transmitted packets for
3843 * timestamping via their phy_driver.txtstamp method. These drivers
3844 * must call this function to return the skb back to the stack with a
3845 * timestamp.
3846 *
3847 * @skb: clone of the the original outgoing packet
3848 * @hwtstamps: hardware time stamps
3849 *
3850 */
3851 void skb_complete_tx_timestamp(struct sk_buff *skb,
3852 struct skb_shared_hwtstamps *hwtstamps);
3853
3854 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3855 struct skb_shared_hwtstamps *hwtstamps,
3856 struct sock *sk, int tstype);
3857
3858 /**
3859 * skb_tstamp_tx - queue clone of skb with send time stamps
3860 * @orig_skb: the original outgoing packet
3861 * @hwtstamps: hardware time stamps, may be NULL if not available
3862 *
3863 * If the skb has a socket associated, then this function clones the
3864 * skb (thus sharing the actual data and optional structures), stores
3865 * the optional hardware time stamping information (if non NULL) or
3866 * generates a software time stamp (otherwise), then queues the clone
3867 * to the error queue of the socket. Errors are silently ignored.
3868 */
3869 void skb_tstamp_tx(struct sk_buff *orig_skb,
3870 struct skb_shared_hwtstamps *hwtstamps);
3871
3872 /**
3873 * skb_tx_timestamp() - Driver hook for transmit timestamping
3874 *
3875 * Ethernet MAC Drivers should call this function in their hard_xmit()
3876 * function immediately before giving the sk_buff to the MAC hardware.
3877 *
3878 * Specifically, one should make absolutely sure that this function is
3879 * called before TX completion of this packet can trigger. Otherwise
3880 * the packet could potentially already be freed.
3881 *
3882 * @skb: A socket buffer.
3883 */
skb_tx_timestamp(struct sk_buff * skb)3884 static inline void skb_tx_timestamp(struct sk_buff *skb)
3885 {
3886 skb_clone_tx_timestamp(skb);
3887 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3888 skb_tstamp_tx(skb, NULL);
3889 }
3890
3891 /**
3892 * skb_complete_wifi_ack - deliver skb with wifi status
3893 *
3894 * @skb: the original outgoing packet
3895 * @acked: ack status
3896 *
3897 */
3898 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3899
3900 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3901 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3902
skb_csum_unnecessary(const struct sk_buff * skb)3903 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3904 {
3905 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3906 skb->csum_valid ||
3907 (skb->ip_summed == CHECKSUM_PARTIAL &&
3908 skb_checksum_start_offset(skb) >= 0));
3909 }
3910
3911 /**
3912 * skb_checksum_complete - Calculate checksum of an entire packet
3913 * @skb: packet to process
3914 *
3915 * This function calculates the checksum over the entire packet plus
3916 * the value of skb->csum. The latter can be used to supply the
3917 * checksum of a pseudo header as used by TCP/UDP. It returns the
3918 * checksum.
3919 *
3920 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3921 * this function can be used to verify that checksum on received
3922 * packets. In that case the function should return zero if the
3923 * checksum is correct. In particular, this function will return zero
3924 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3925 * hardware has already verified the correctness of the checksum.
3926 */
skb_checksum_complete(struct sk_buff * skb)3927 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3928 {
3929 return skb_csum_unnecessary(skb) ?
3930 0 : __skb_checksum_complete(skb);
3931 }
3932
__skb_decr_checksum_unnecessary(struct sk_buff * skb)3933 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3934 {
3935 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3936 if (skb->csum_level == 0)
3937 skb->ip_summed = CHECKSUM_NONE;
3938 else
3939 skb->csum_level--;
3940 }
3941 }
3942
__skb_incr_checksum_unnecessary(struct sk_buff * skb)3943 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3944 {
3945 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3946 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3947 skb->csum_level++;
3948 } else if (skb->ip_summed == CHECKSUM_NONE) {
3949 skb->ip_summed = CHECKSUM_UNNECESSARY;
3950 skb->csum_level = 0;
3951 }
3952 }
3953
3954 /* Check if we need to perform checksum complete validation.
3955 *
3956 * Returns true if checksum complete is needed, false otherwise
3957 * (either checksum is unnecessary or zero checksum is allowed).
3958 */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)3959 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3960 bool zero_okay,
3961 __sum16 check)
3962 {
3963 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3964 skb->csum_valid = 1;
3965 __skb_decr_checksum_unnecessary(skb);
3966 return false;
3967 }
3968
3969 return true;
3970 }
3971
3972 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
3973 * in checksum_init.
3974 */
3975 #define CHECKSUM_BREAK 76
3976
3977 /* Unset checksum-complete
3978 *
3979 * Unset checksum complete can be done when packet is being modified
3980 * (uncompressed for instance) and checksum-complete value is
3981 * invalidated.
3982 */
skb_checksum_complete_unset(struct sk_buff * skb)3983 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3984 {
3985 if (skb->ip_summed == CHECKSUM_COMPLETE)
3986 skb->ip_summed = CHECKSUM_NONE;
3987 }
3988
3989 /* Validate (init) checksum based on checksum complete.
3990 *
3991 * Return values:
3992 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3993 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3994 * checksum is stored in skb->csum for use in __skb_checksum_complete
3995 * non-zero: value of invalid checksum
3996 *
3997 */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)3998 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3999 bool complete,
4000 __wsum psum)
4001 {
4002 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4003 if (!csum_fold(csum_add(psum, skb->csum))) {
4004 skb->csum_valid = 1;
4005 return 0;
4006 }
4007 }
4008
4009 skb->csum = psum;
4010
4011 if (complete || skb->len <= CHECKSUM_BREAK) {
4012 __sum16 csum;
4013
4014 csum = __skb_checksum_complete(skb);
4015 skb->csum_valid = !csum;
4016 return csum;
4017 }
4018
4019 return 0;
4020 }
4021
null_compute_pseudo(struct sk_buff * skb,int proto)4022 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4023 {
4024 return 0;
4025 }
4026
4027 /* Perform checksum validate (init). Note that this is a macro since we only
4028 * want to calculate the pseudo header which is an input function if necessary.
4029 * First we try to validate without any computation (checksum unnecessary) and
4030 * then calculate based on checksum complete calling the function to compute
4031 * pseudo header.
4032 *
4033 * Return values:
4034 * 0: checksum is validated or try to in skb_checksum_complete
4035 * non-zero: value of invalid checksum
4036 */
4037 #define __skb_checksum_validate(skb, proto, complete, \
4038 zero_okay, check, compute_pseudo) \
4039 ({ \
4040 __sum16 __ret = 0; \
4041 skb->csum_valid = 0; \
4042 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4043 __ret = __skb_checksum_validate_complete(skb, \
4044 complete, compute_pseudo(skb, proto)); \
4045 __ret; \
4046 })
4047
4048 #define skb_checksum_init(skb, proto, compute_pseudo) \
4049 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4050
4051 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4052 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4053
4054 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4055 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4056
4057 #define skb_checksum_validate_zero_check(skb, proto, check, \
4058 compute_pseudo) \
4059 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4060
4061 #define skb_checksum_simple_validate(skb) \
4062 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4063
__skb_checksum_convert_check(struct sk_buff * skb)4064 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4065 {
4066 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4067 }
4068
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4069 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4070 {
4071 skb->csum = ~pseudo;
4072 skb->ip_summed = CHECKSUM_COMPLETE;
4073 }
4074
4075 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4076 do { \
4077 if (__skb_checksum_convert_check(skb)) \
4078 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4079 } while (0)
4080
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4081 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4082 u16 start, u16 offset)
4083 {
4084 skb->ip_summed = CHECKSUM_PARTIAL;
4085 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4086 skb->csum_offset = offset - start;
4087 }
4088
4089 /* Update skbuf and packet to reflect the remote checksum offload operation.
4090 * When called, ptr indicates the starting point for skb->csum when
4091 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4092 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4093 */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4094 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4095 int start, int offset, bool nopartial)
4096 {
4097 __wsum delta;
4098
4099 if (!nopartial) {
4100 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4101 return;
4102 }
4103
4104 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4105 __skb_checksum_complete(skb);
4106 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4107 }
4108
4109 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4110
4111 /* Adjust skb->csum since we changed the packet */
4112 skb->csum = csum_add(skb->csum, delta);
4113 }
4114
skb_nfct(const struct sk_buff * skb)4115 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4116 {
4117 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4118 return (void *)(skb->_nfct & NFCT_PTRMASK);
4119 #else
4120 return NULL;
4121 #endif
4122 }
4123
skb_get_nfct(const struct sk_buff * skb)4124 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4125 {
4126 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4127 return skb->_nfct;
4128 #else
4129 return 0UL;
4130 #endif
4131 }
4132
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4133 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4134 {
4135 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4136 skb->_nfct = nfct;
4137 #endif
4138 }
4139
4140 #ifdef CONFIG_SKB_EXTENSIONS
4141 enum skb_ext_id {
4142 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4143 SKB_EXT_BRIDGE_NF,
4144 #endif
4145 #ifdef CONFIG_XFRM
4146 SKB_EXT_SEC_PATH,
4147 #endif
4148 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4149 TC_SKB_EXT,
4150 #endif
4151 SKB_EXT_NUM, /* must be last */
4152 };
4153
4154 /**
4155 * struct skb_ext - sk_buff extensions
4156 * @refcnt: 1 on allocation, deallocated on 0
4157 * @offset: offset to add to @data to obtain extension address
4158 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4159 * @data: start of extension data, variable sized
4160 *
4161 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4162 * to use 'u8' types while allowing up to 2kb worth of extension data.
4163 */
4164 struct skb_ext {
4165 refcount_t refcnt;
4166 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4167 u8 chunks; /* same */
4168 char data[0] __aligned(8);
4169 };
4170
4171 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4172 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4173 void __skb_ext_put(struct skb_ext *ext);
4174
skb_ext_put(struct sk_buff * skb)4175 static inline void skb_ext_put(struct sk_buff *skb)
4176 {
4177 if (skb->active_extensions)
4178 __skb_ext_put(skb->extensions);
4179 }
4180
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4181 static inline void __skb_ext_copy(struct sk_buff *dst,
4182 const struct sk_buff *src)
4183 {
4184 dst->active_extensions = src->active_extensions;
4185
4186 if (src->active_extensions) {
4187 struct skb_ext *ext = src->extensions;
4188
4189 refcount_inc(&ext->refcnt);
4190 dst->extensions = ext;
4191 }
4192 }
4193
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4194 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4195 {
4196 skb_ext_put(dst);
4197 __skb_ext_copy(dst, src);
4198 }
4199
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4200 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4201 {
4202 return !!ext->offset[i];
4203 }
4204
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4205 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4206 {
4207 return skb->active_extensions & (1 << id);
4208 }
4209
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4210 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4211 {
4212 if (skb_ext_exist(skb, id))
4213 __skb_ext_del(skb, id);
4214 }
4215
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4216 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4217 {
4218 if (skb_ext_exist(skb, id)) {
4219 struct skb_ext *ext = skb->extensions;
4220
4221 return (void *)ext + (ext->offset[id] << 3);
4222 }
4223
4224 return NULL;
4225 }
4226
skb_ext_reset(struct sk_buff * skb)4227 static inline void skb_ext_reset(struct sk_buff *skb)
4228 {
4229 if (unlikely(skb->active_extensions)) {
4230 __skb_ext_put(skb->extensions);
4231 skb->active_extensions = 0;
4232 }
4233 }
4234
skb_has_extensions(struct sk_buff * skb)4235 static inline bool skb_has_extensions(struct sk_buff *skb)
4236 {
4237 return unlikely(skb->active_extensions);
4238 }
4239 #else
skb_ext_put(struct sk_buff * skb)4240 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4241 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4242 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4243 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * s)4244 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4245 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4246 #endif /* CONFIG_SKB_EXTENSIONS */
4247
nf_reset_ct(struct sk_buff * skb)4248 static inline void nf_reset_ct(struct sk_buff *skb)
4249 {
4250 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4251 nf_conntrack_put(skb_nfct(skb));
4252 skb->_nfct = 0;
4253 #endif
4254 }
4255
nf_reset_trace(struct sk_buff * skb)4256 static inline void nf_reset_trace(struct sk_buff *skb)
4257 {
4258 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4259 skb->nf_trace = 0;
4260 #endif
4261 }
4262
ipvs_reset(struct sk_buff * skb)4263 static inline void ipvs_reset(struct sk_buff *skb)
4264 {
4265 #if IS_ENABLED(CONFIG_IP_VS)
4266 skb->ipvs_property = 0;
4267 #endif
4268 }
4269
4270 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4271 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4272 bool copy)
4273 {
4274 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4275 dst->_nfct = src->_nfct;
4276 nf_conntrack_get(skb_nfct(src));
4277 #endif
4278 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4279 if (copy)
4280 dst->nf_trace = src->nf_trace;
4281 #endif
4282 }
4283
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4284 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4285 {
4286 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4287 nf_conntrack_put(skb_nfct(dst));
4288 #endif
4289 __nf_copy(dst, src, true);
4290 }
4291
4292 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4293 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4294 {
4295 to->secmark = from->secmark;
4296 }
4297
skb_init_secmark(struct sk_buff * skb)4298 static inline void skb_init_secmark(struct sk_buff *skb)
4299 {
4300 skb->secmark = 0;
4301 }
4302 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4303 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4304 { }
4305
skb_init_secmark(struct sk_buff * skb)4306 static inline void skb_init_secmark(struct sk_buff *skb)
4307 { }
4308 #endif
4309
secpath_exists(const struct sk_buff * skb)4310 static inline int secpath_exists(const struct sk_buff *skb)
4311 {
4312 #ifdef CONFIG_XFRM
4313 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4314 #else
4315 return 0;
4316 #endif
4317 }
4318
skb_irq_freeable(const struct sk_buff * skb)4319 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4320 {
4321 return !skb->destructor &&
4322 !secpath_exists(skb) &&
4323 !skb_nfct(skb) &&
4324 !skb->_skb_refdst &&
4325 !skb_has_frag_list(skb);
4326 }
4327
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)4328 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4329 {
4330 skb->queue_mapping = queue_mapping;
4331 }
4332
skb_get_queue_mapping(const struct sk_buff * skb)4333 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4334 {
4335 return skb->queue_mapping;
4336 }
4337
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)4338 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4339 {
4340 to->queue_mapping = from->queue_mapping;
4341 }
4342
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)4343 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4344 {
4345 skb->queue_mapping = rx_queue + 1;
4346 }
4347
skb_get_rx_queue(const struct sk_buff * skb)4348 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4349 {
4350 return skb->queue_mapping - 1;
4351 }
4352
skb_rx_queue_recorded(const struct sk_buff * skb)4353 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4354 {
4355 return skb->queue_mapping != 0;
4356 }
4357
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)4358 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4359 {
4360 skb->dst_pending_confirm = val;
4361 }
4362
skb_get_dst_pending_confirm(const struct sk_buff * skb)4363 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4364 {
4365 return skb->dst_pending_confirm != 0;
4366 }
4367
skb_sec_path(const struct sk_buff * skb)4368 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4369 {
4370 #ifdef CONFIG_XFRM
4371 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4372 #else
4373 return NULL;
4374 #endif
4375 }
4376
4377 /* Keeps track of mac header offset relative to skb->head.
4378 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4379 * For non-tunnel skb it points to skb_mac_header() and for
4380 * tunnel skb it points to outer mac header.
4381 * Keeps track of level of encapsulation of network headers.
4382 */
4383 struct skb_gso_cb {
4384 union {
4385 int mac_offset;
4386 int data_offset;
4387 };
4388 int encap_level;
4389 __wsum csum;
4390 __u16 csum_start;
4391 };
4392 #define SKB_SGO_CB_OFFSET 32
4393 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
4394
skb_tnl_header_len(const struct sk_buff * inner_skb)4395 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4396 {
4397 return (skb_mac_header(inner_skb) - inner_skb->head) -
4398 SKB_GSO_CB(inner_skb)->mac_offset;
4399 }
4400
gso_pskb_expand_head(struct sk_buff * skb,int extra)4401 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4402 {
4403 int new_headroom, headroom;
4404 int ret;
4405
4406 headroom = skb_headroom(skb);
4407 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4408 if (ret)
4409 return ret;
4410
4411 new_headroom = skb_headroom(skb);
4412 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4413 return 0;
4414 }
4415
gso_reset_checksum(struct sk_buff * skb,__wsum res)4416 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4417 {
4418 /* Do not update partial checksums if remote checksum is enabled. */
4419 if (skb->remcsum_offload)
4420 return;
4421
4422 SKB_GSO_CB(skb)->csum = res;
4423 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4424 }
4425
4426 /* Compute the checksum for a gso segment. First compute the checksum value
4427 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4428 * then add in skb->csum (checksum from csum_start to end of packet).
4429 * skb->csum and csum_start are then updated to reflect the checksum of the
4430 * resultant packet starting from the transport header-- the resultant checksum
4431 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4432 * header.
4433 */
gso_make_checksum(struct sk_buff * skb,__wsum res)4434 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4435 {
4436 unsigned char *csum_start = skb_transport_header(skb);
4437 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4438 __wsum partial = SKB_GSO_CB(skb)->csum;
4439
4440 SKB_GSO_CB(skb)->csum = res;
4441 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4442
4443 return csum_fold(csum_partial(csum_start, plen, partial));
4444 }
4445
skb_is_gso(const struct sk_buff * skb)4446 static inline bool skb_is_gso(const struct sk_buff *skb)
4447 {
4448 return skb_shinfo(skb)->gso_size;
4449 }
4450
4451 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)4452 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4453 {
4454 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4455 }
4456
4457 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)4458 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4459 {
4460 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4461 }
4462
4463 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)4464 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4465 {
4466 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4467 }
4468
skb_gso_reset(struct sk_buff * skb)4469 static inline void skb_gso_reset(struct sk_buff *skb)
4470 {
4471 skb_shinfo(skb)->gso_size = 0;
4472 skb_shinfo(skb)->gso_segs = 0;
4473 skb_shinfo(skb)->gso_type = 0;
4474 }
4475
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)4476 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4477 u16 increment)
4478 {
4479 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4480 return;
4481 shinfo->gso_size += increment;
4482 }
4483
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)4484 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4485 u16 decrement)
4486 {
4487 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4488 return;
4489 shinfo->gso_size -= decrement;
4490 }
4491
4492 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4493
skb_warn_if_lro(const struct sk_buff * skb)4494 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4495 {
4496 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4497 * wanted then gso_type will be set. */
4498 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4499
4500 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4501 unlikely(shinfo->gso_type == 0)) {
4502 __skb_warn_lro_forwarding(skb);
4503 return true;
4504 }
4505 return false;
4506 }
4507
skb_forward_csum(struct sk_buff * skb)4508 static inline void skb_forward_csum(struct sk_buff *skb)
4509 {
4510 /* Unfortunately we don't support this one. Any brave souls? */
4511 if (skb->ip_summed == CHECKSUM_COMPLETE)
4512 skb->ip_summed = CHECKSUM_NONE;
4513 }
4514
4515 /**
4516 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4517 * @skb: skb to check
4518 *
4519 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4520 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4521 * use this helper, to document places where we make this assertion.
4522 */
skb_checksum_none_assert(const struct sk_buff * skb)4523 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4524 {
4525 #ifdef DEBUG
4526 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4527 #endif
4528 }
4529
4530 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4531
4532 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4533 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4534 unsigned int transport_len,
4535 __sum16(*skb_chkf)(struct sk_buff *skb));
4536
4537 /**
4538 * skb_head_is_locked - Determine if the skb->head is locked down
4539 * @skb: skb to check
4540 *
4541 * The head on skbs build around a head frag can be removed if they are
4542 * not cloned. This function returns true if the skb head is locked down
4543 * due to either being allocated via kmalloc, or by being a clone with
4544 * multiple references to the head.
4545 */
skb_head_is_locked(const struct sk_buff * skb)4546 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4547 {
4548 return !skb->head_frag || skb_cloned(skb);
4549 }
4550
4551 /* Local Checksum Offload.
4552 * Compute outer checksum based on the assumption that the
4553 * inner checksum will be offloaded later.
4554 * See Documentation/networking/checksum-offloads.rst for
4555 * explanation of how this works.
4556 * Fill in outer checksum adjustment (e.g. with sum of outer
4557 * pseudo-header) before calling.
4558 * Also ensure that inner checksum is in linear data area.
4559 */
lco_csum(struct sk_buff * skb)4560 static inline __wsum lco_csum(struct sk_buff *skb)
4561 {
4562 unsigned char *csum_start = skb_checksum_start(skb);
4563 unsigned char *l4_hdr = skb_transport_header(skb);
4564 __wsum partial;
4565
4566 /* Start with complement of inner checksum adjustment */
4567 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4568 skb->csum_offset));
4569
4570 /* Add in checksum of our headers (incl. outer checksum
4571 * adjustment filled in by caller) and return result.
4572 */
4573 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4574 }
4575
skb_is_redirected(const struct sk_buff * skb)4576 static inline bool skb_is_redirected(const struct sk_buff *skb)
4577 {
4578 #ifdef CONFIG_NET_REDIRECT
4579 return skb->redirected;
4580 #else
4581 return false;
4582 #endif
4583 }
4584
skb_set_redirected(struct sk_buff * skb,bool from_ingress)4585 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4586 {
4587 #ifdef CONFIG_NET_REDIRECT
4588 skb->redirected = 1;
4589 skb->from_ingress = from_ingress;
4590 if (skb->from_ingress)
4591 skb->tstamp = 0;
4592 #endif
4593 }
4594
skb_reset_redirect(struct sk_buff * skb)4595 static inline void skb_reset_redirect(struct sk_buff *skb)
4596 {
4597 #ifdef CONFIG_NET_REDIRECT
4598 skb->redirected = 0;
4599 #endif
4600 }
4601
4602 #endif /* __KERNEL__ */
4603 #endif /* _LINUX_SKBUFF_H */
4604