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 * IPv6|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 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 set in 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 or 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 in skb->csum. This means 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, but it 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 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, it should treat the packet as if CHECKSUM_NONE were 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 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 that there is no indication in the skbuff that the
206 * CHECKSUM_PARTIAL refers to an FCOE checksum, so 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, and both csum_start and
217 * csum_offset are set to refer to the outermost checksum being offloaded
218 * (two offloaded 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 ahash_request;
244 struct net_device;
245 struct scatterlist;
246 struct pipe_inode_info;
247 struct iov_iter;
248 struct napi_struct;
249 struct bpf_prog;
250 union bpf_attr;
251 struct skb_ext;
252
253 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
254 struct nf_bridge_info {
255 enum {
256 BRNF_PROTO_UNCHANGED,
257 BRNF_PROTO_8021Q,
258 BRNF_PROTO_PPPOE
259 } orig_proto:8;
260 u8 pkt_otherhost:1;
261 u8 in_prerouting:1;
262 u8 bridged_dnat: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 __u16 mru;
290 };
291 #endif
292
293 struct sk_buff_head {
294 /* These two members must be first. */
295 struct sk_buff *next;
296 struct sk_buff *prev;
297
298 __u32 qlen;
299 spinlock_t lock;
300 };
301
302 struct sk_buff;
303
304 /* To allow 64K frame to be packed as single skb without frag_list we
305 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
306 * buffers which do not start on a page boundary.
307 *
308 * Since GRO uses frags we allocate at least 16 regardless of page
309 * size.
310 */
311 #if (65536/PAGE_SIZE + 1) < 16
312 #define MAX_SKB_FRAGS 16UL
313 #else
314 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
315 #endif
316 extern int sysctl_max_skb_frags;
317
318 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
319 * segment using its current segmentation instead.
320 */
321 #define GSO_BY_FRAGS 0xFFFF
322
323 typedef struct bio_vec skb_frag_t;
324
325 /**
326 * skb_frag_size() - Returns the size of a skb fragment
327 * @frag: skb fragment
328 */
skb_frag_size(const skb_frag_t * frag)329 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
330 {
331 return frag->bv_len;
332 }
333
334 /**
335 * skb_frag_size_set() - Sets the size of a skb fragment
336 * @frag: skb fragment
337 * @size: size of fragment
338 */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)339 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
340 {
341 frag->bv_len = size;
342 }
343
344 /**
345 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
346 * @frag: skb fragment
347 * @delta: value to add
348 */
skb_frag_size_add(skb_frag_t * frag,int delta)349 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
350 {
351 frag->bv_len += delta;
352 }
353
354 /**
355 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
356 * @frag: skb fragment
357 * @delta: value to subtract
358 */
skb_frag_size_sub(skb_frag_t * frag,int delta)359 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
360 {
361 frag->bv_len -= delta;
362 }
363
364 /**
365 * skb_frag_must_loop - Test if %p is a high memory page
366 * @p: fragment's page
367 */
skb_frag_must_loop(struct page * p)368 static inline bool skb_frag_must_loop(struct page *p)
369 {
370 #if defined(CONFIG_HIGHMEM)
371 if (PageHighMem(p))
372 return true;
373 #endif
374 return false;
375 }
376
377 /**
378 * skb_frag_foreach_page - loop over pages in a fragment
379 *
380 * @f: skb frag to operate on
381 * @f_off: offset from start of f->bv_page
382 * @f_len: length from f_off to loop over
383 * @p: (temp var) current page
384 * @p_off: (temp var) offset from start of current page,
385 * non-zero only on first page.
386 * @p_len: (temp var) length in current page,
387 * < PAGE_SIZE only on first and last page.
388 * @copied: (temp var) length so far, excluding current p_len.
389 *
390 * A fragment can hold a compound page, in which case per-page
391 * operations, notably kmap_atomic, must be called for each
392 * regular page.
393 */
394 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
395 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
396 p_off = (f_off) & (PAGE_SIZE - 1), \
397 p_len = skb_frag_must_loop(p) ? \
398 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
399 copied = 0; \
400 copied < f_len; \
401 copied += p_len, p++, p_off = 0, \
402 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
403
404 #define HAVE_HW_TIME_STAMP
405
406 /**
407 * struct skb_shared_hwtstamps - hardware time stamps
408 * @hwtstamp: hardware time stamp transformed into duration
409 * since arbitrary point in time
410 *
411 * Software time stamps generated by ktime_get_real() are stored in
412 * skb->tstamp.
413 *
414 * hwtstamps can only be compared against other hwtstamps from
415 * the same device.
416 *
417 * This structure is attached to packets as part of the
418 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
419 */
420 struct skb_shared_hwtstamps {
421 ktime_t hwtstamp;
422 };
423
424 /* Definitions for tx_flags in struct skb_shared_info */
425 enum {
426 /* generate hardware time stamp */
427 SKBTX_HW_TSTAMP = 1 << 0,
428
429 /* generate software time stamp when queueing packet to NIC */
430 SKBTX_SW_TSTAMP = 1 << 1,
431
432 /* device driver is going to provide hardware time stamp */
433 SKBTX_IN_PROGRESS = 1 << 2,
434
435 /* device driver supports TX zero-copy buffers */
436 SKBTX_DEV_ZEROCOPY = 1 << 3,
437
438 /* generate wifi status information (where possible) */
439 SKBTX_WIFI_STATUS = 1 << 4,
440
441 /* This indicates at least one fragment might be overwritten
442 * (as in vmsplice(), sendfile() ...)
443 * If we need to compute a TX checksum, we'll need to copy
444 * all frags to avoid possible bad checksum
445 */
446 SKBTX_SHARED_FRAG = 1 << 5,
447
448 /* generate software time stamp when entering packet scheduling */
449 SKBTX_SCHED_TSTAMP = 1 << 6,
450 };
451
452 #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
453 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
454 SKBTX_SCHED_TSTAMP)
455 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
456
457 /*
458 * The callback notifies userspace to release buffers when skb DMA is done in
459 * lower device, the skb last reference should be 0 when calling this.
460 * The zerocopy_success argument is true if zero copy transmit occurred,
461 * false on data copy or out of memory error caused by data copy attempt.
462 * The ctx field is used to track device context.
463 * The desc field is used to track userspace buffer index.
464 */
465 struct ubuf_info {
466 void (*callback)(struct ubuf_info *, bool zerocopy_success);
467 union {
468 struct {
469 unsigned long desc;
470 void *ctx;
471 };
472 struct {
473 u32 id;
474 u16 len;
475 u16 zerocopy:1;
476 u32 bytelen;
477 };
478 };
479 refcount_t refcnt;
480
481 struct mmpin {
482 struct user_struct *user;
483 unsigned int num_pg;
484 } mmp;
485 };
486
487 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
488
489 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
490 void mm_unaccount_pinned_pages(struct mmpin *mmp);
491
492 struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
493 struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
494 struct ubuf_info *uarg);
495
sock_zerocopy_get(struct ubuf_info * uarg)496 static inline void sock_zerocopy_get(struct ubuf_info *uarg)
497 {
498 refcount_inc(&uarg->refcnt);
499 }
500
501 void sock_zerocopy_put(struct ubuf_info *uarg);
502 void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
503
504 void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
505
506 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
507 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
508 struct msghdr *msg, int len,
509 struct ubuf_info *uarg);
510
511 /* This data is invariant across clones and lives at
512 * the end of the header data, ie. at skb->end.
513 */
514 struct skb_shared_info {
515 __u8 __unused;
516 __u8 meta_len;
517 __u8 nr_frags;
518 __u8 tx_flags;
519 unsigned short gso_size;
520 /* Warning: this field is not always filled in (UFO)! */
521 unsigned short gso_segs;
522 struct sk_buff *frag_list;
523 struct skb_shared_hwtstamps hwtstamps;
524 unsigned int gso_type;
525 u32 tskey;
526
527 /*
528 * Warning : all fields before dataref are cleared in __alloc_skb()
529 */
530 atomic_t dataref;
531
532 /* Intermediate layers must ensure that destructor_arg
533 * remains valid until skb destructor */
534 void * destructor_arg;
535
536 ANDROID_OEM_DATA_ARRAY(1, 3);
537
538 /* must be last field, see pskb_expand_head() */
539 skb_frag_t frags[MAX_SKB_FRAGS];
540 };
541
542 /* We divide dataref into two halves. The higher 16 bits hold references
543 * to the payload part of skb->data. The lower 16 bits hold references to
544 * the entire skb->data. A clone of a headerless skb holds the length of
545 * the header in skb->hdr_len.
546 *
547 * All users must obey the rule that the skb->data reference count must be
548 * greater than or equal to the payload reference count.
549 *
550 * Holding a reference to the payload part means that the user does not
551 * care about modifications to the header part of skb->data.
552 */
553 #define SKB_DATAREF_SHIFT 16
554 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
555
556
557 enum {
558 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
559 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
560 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
561 };
562
563 enum {
564 SKB_GSO_TCPV4 = 1 << 0,
565
566 /* This indicates the skb is from an untrusted source. */
567 SKB_GSO_DODGY = 1 << 1,
568
569 /* This indicates the tcp segment has CWR set. */
570 SKB_GSO_TCP_ECN = 1 << 2,
571
572 SKB_GSO_TCP_FIXEDID = 1 << 3,
573
574 SKB_GSO_TCPV6 = 1 << 4,
575
576 SKB_GSO_FCOE = 1 << 5,
577
578 SKB_GSO_GRE = 1 << 6,
579
580 SKB_GSO_GRE_CSUM = 1 << 7,
581
582 SKB_GSO_IPXIP4 = 1 << 8,
583
584 SKB_GSO_IPXIP6 = 1 << 9,
585
586 SKB_GSO_UDP_TUNNEL = 1 << 10,
587
588 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
589
590 SKB_GSO_PARTIAL = 1 << 12,
591
592 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
593
594 SKB_GSO_SCTP = 1 << 14,
595
596 SKB_GSO_ESP = 1 << 15,
597
598 SKB_GSO_UDP = 1 << 16,
599
600 SKB_GSO_UDP_L4 = 1 << 17,
601
602 SKB_GSO_FRAGLIST = 1 << 18,
603 };
604
605 #if BITS_PER_LONG > 32
606 #define NET_SKBUFF_DATA_USES_OFFSET 1
607 #endif
608
609 #ifdef NET_SKBUFF_DATA_USES_OFFSET
610 typedef unsigned int sk_buff_data_t;
611 #else
612 typedef unsigned char *sk_buff_data_t;
613 #endif
614
615 /**
616 * struct sk_buff - socket buffer
617 * @next: Next buffer in list
618 * @prev: Previous buffer in list
619 * @tstamp: Time we arrived/left
620 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
621 * for retransmit timer
622 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
623 * @list: queue head
624 * @sk: Socket we are owned by
625 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
626 * fragmentation management
627 * @dev: Device we arrived on/are leaving by
628 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
629 * @cb: Control buffer. Free for use by every layer. Put private vars here
630 * @_skb_refdst: destination entry (with norefcount bit)
631 * @sp: the security path, used for xfrm
632 * @len: Length of actual data
633 * @data_len: Data length
634 * @mac_len: Length of link layer header
635 * @hdr_len: writable header length of cloned skb
636 * @csum: Checksum (must include start/offset pair)
637 * @csum_start: Offset from skb->head where checksumming should start
638 * @csum_offset: Offset from csum_start where checksum should be stored
639 * @priority: Packet queueing priority
640 * @ignore_df: allow local fragmentation
641 * @cloned: Head may be cloned (check refcnt to be sure)
642 * @ip_summed: Driver fed us an IP checksum
643 * @nohdr: Payload reference only, must not modify header
644 * @pkt_type: Packet class
645 * @fclone: skbuff clone status
646 * @ipvs_property: skbuff is owned by ipvs
647 * @inner_protocol_type: whether the inner protocol is
648 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
649 * @remcsum_offload: remote checksum offload is enabled
650 * @offload_fwd_mark: Packet was L2-forwarded in hardware
651 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
652 * @tc_skip_classify: do not classify packet. set by IFB device
653 * @tc_at_ingress: used within tc_classify to distinguish in/egress
654 * @redirected: packet was redirected by packet classifier
655 * @from_ingress: packet was redirected from the ingress path
656 * @peeked: this packet has been seen already, so stats have been
657 * done for it, don't do them again
658 * @nf_trace: netfilter packet trace flag
659 * @protocol: Packet protocol from driver
660 * @destructor: Destruct function
661 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
662 * @_nfct: Associated connection, if any (with nfctinfo bits)
663 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
664 * @skb_iif: ifindex of device we arrived on
665 * @tc_index: Traffic control index
666 * @hash: the packet hash
667 * @queue_mapping: Queue mapping for multiqueue devices
668 * @head_frag: skb was allocated from page fragments,
669 * not allocated by kmalloc() or vmalloc().
670 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
671 * @active_extensions: active extensions (skb_ext_id types)
672 * @ndisc_nodetype: router type (from link layer)
673 * @ooo_okay: allow the mapping of a socket to a queue to be changed
674 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
675 * ports.
676 * @sw_hash: indicates hash was computed in software stack
677 * @wifi_acked_valid: wifi_acked was set
678 * @wifi_acked: whether frame was acked on wifi or not
679 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
680 * @encapsulation: indicates the inner headers in the skbuff are valid
681 * @encap_hdr_csum: software checksum is needed
682 * @csum_valid: checksum is already valid
683 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
684 * @csum_complete_sw: checksum was completed by software
685 * @csum_level: indicates the number of consecutive checksums found in
686 * the packet minus one that have been verified as
687 * CHECKSUM_UNNECESSARY (max 3)
688 * @scm_io_uring: SKB holds io_uring registered files
689 * @dst_pending_confirm: need to confirm neighbour
690 * @decrypted: Decrypted SKB
691 * @napi_id: id of the NAPI struct this skb came from
692 * @sender_cpu: (aka @napi_id) source CPU in XPS
693 * @secmark: security marking
694 * @mark: Generic packet mark
695 * @reserved_tailroom: (aka @mark) number of bytes of free space available
696 * at the tail of an sk_buff
697 * @vlan_present: VLAN tag is present
698 * @vlan_proto: vlan encapsulation protocol
699 * @vlan_tci: vlan tag control information
700 * @inner_protocol: Protocol (encapsulation)
701 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
702 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
703 * @inner_transport_header: Inner transport layer header (encapsulation)
704 * @inner_network_header: Network layer header (encapsulation)
705 * @inner_mac_header: Link layer header (encapsulation)
706 * @transport_header: Transport layer header
707 * @network_header: Network layer header
708 * @mac_header: Link layer header
709 * @tail: Tail pointer
710 * @end: End pointer
711 * @head: Head of buffer
712 * @data: Data head pointer
713 * @truesize: Buffer size
714 * @users: User count - see {datagram,tcp}.c
715 * @extensions: allocated extensions, valid if active_extensions is nonzero
716 */
717
718 struct sk_buff {
719 union {
720 struct {
721 /* These two members must be first. */
722 struct sk_buff *next;
723 struct sk_buff *prev;
724
725 union {
726 struct net_device *dev;
727 /* Some protocols might use this space to store information,
728 * while device pointer would be NULL.
729 * UDP receive path is one user.
730 */
731 unsigned long dev_scratch;
732 };
733 };
734 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
735 struct list_head list;
736 };
737
738 union {
739 struct sock *sk;
740 int ip_defrag_offset;
741 };
742
743 union {
744 ktime_t tstamp;
745 u64 skb_mstamp_ns; /* earliest departure time */
746 };
747 /*
748 * This is the control buffer. It is free to use for every
749 * layer. Please put your private variables there. If you
750 * want to keep them across layers you have to do a skb_clone()
751 * first. This is owned by whoever has the skb queued ATM.
752 */
753 char cb[48] __aligned(8);
754
755 union {
756 struct {
757 unsigned long _skb_refdst;
758 void (*destructor)(struct sk_buff *skb);
759 };
760 struct list_head tcp_tsorted_anchor;
761 };
762
763 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
764 unsigned long _nfct;
765 #endif
766 unsigned int len,
767 data_len;
768 __u16 mac_len,
769 hdr_len;
770
771 /* Following fields are _not_ copied in __copy_skb_header()
772 * Note that queue_mapping is here mostly to fill a hole.
773 */
774 __u16 queue_mapping;
775
776 /* if you move cloned around you also must adapt those constants */
777 #ifdef __BIG_ENDIAN_BITFIELD
778 #define CLONED_MASK (1 << 7)
779 #else
780 #define CLONED_MASK 1
781 #endif
782 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
783
784 /* private: */
785 __u8 __cloned_offset[0];
786 /* public: */
787 __u8 cloned:1,
788 nohdr:1,
789 fclone:2,
790 peeked:1,
791 head_frag:1,
792 pfmemalloc:1;
793 #ifdef CONFIG_SKB_EXTENSIONS
794 __u8 active_extensions;
795 #endif
796 /* fields enclosed in headers_start/headers_end are copied
797 * using a single memcpy() in __copy_skb_header()
798 */
799 /* private: */
800 __u32 headers_start[0];
801 /* public: */
802
803 /* if you move pkt_type around you also must adapt those constants */
804 #ifdef __BIG_ENDIAN_BITFIELD
805 #define PKT_TYPE_MAX (7 << 5)
806 #else
807 #define PKT_TYPE_MAX 7
808 #endif
809 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
810
811 /* private: */
812 __u8 __pkt_type_offset[0];
813 /* public: */
814 __u8 pkt_type:3;
815 __u8 ignore_df:1;
816 __u8 nf_trace:1;
817 __u8 ip_summed:2;
818 __u8 ooo_okay:1;
819
820 __u8 l4_hash:1;
821 __u8 sw_hash:1;
822 __u8 wifi_acked_valid:1;
823 __u8 wifi_acked:1;
824 __u8 no_fcs:1;
825 /* Indicates the inner headers are valid in the skbuff. */
826 __u8 encapsulation:1;
827 __u8 encap_hdr_csum:1;
828 __u8 csum_valid:1;
829
830 #ifdef __BIG_ENDIAN_BITFIELD
831 #define PKT_VLAN_PRESENT_BIT 7
832 #else
833 #define PKT_VLAN_PRESENT_BIT 0
834 #endif
835 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
836 /* private: */
837 __u8 __pkt_vlan_present_offset[0];
838 /* public: */
839 __u8 vlan_present:1;
840 __u8 csum_complete_sw:1;
841 __u8 csum_level:2;
842 __u8 csum_not_inet:1;
843 __u8 dst_pending_confirm:1;
844 #ifdef CONFIG_IPV6_NDISC_NODETYPE
845 __u8 ndisc_nodetype:2;
846 #endif
847
848 __u8 ipvs_property:1;
849 __u8 inner_protocol_type:1;
850 __u8 remcsum_offload:1;
851 #ifdef CONFIG_NET_SWITCHDEV
852 __u8 offload_fwd_mark:1;
853 __u8 offload_l3_fwd_mark:1;
854 #endif
855 #ifdef CONFIG_NET_CLS_ACT
856 __u8 tc_skip_classify:1;
857 __u8 tc_at_ingress:1;
858 #endif
859 #ifdef CONFIG_NET_REDIRECT
860 __u8 redirected:1;
861 __u8 from_ingress:1;
862 #endif
863 #ifdef CONFIG_TLS_DEVICE
864 __u8 decrypted:1;
865 #endif
866
867 #ifdef CONFIG_NET_SCHED
868 __u16 tc_index; /* traffic control index */
869 #endif
870
871 union {
872 __wsum csum;
873 struct {
874 __u16 csum_start;
875 __u16 csum_offset;
876 };
877 };
878 __u32 priority;
879 int skb_iif;
880 __u32 hash;
881 __be16 vlan_proto;
882 __u16 vlan_tci;
883 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
884 union {
885 unsigned int napi_id;
886 unsigned int sender_cpu;
887 };
888 #endif
889 #ifdef CONFIG_NETWORK_SECMARK
890 __u32 secmark;
891 #endif
892
893 union {
894 __u32 mark;
895 __u32 reserved_tailroom;
896 };
897
898 union {
899 __be16 inner_protocol;
900 __u8 inner_ipproto;
901 };
902
903 __u16 inner_transport_header;
904 __u16 inner_network_header;
905 __u16 inner_mac_header;
906
907 __be16 protocol;
908 __u16 transport_header;
909 __u16 network_header;
910 __u16 mac_header;
911
912 /* private: */
913 __u32 headers_end[0];
914 /* public: */
915
916 /* Android KABI preservation.
917 *
918 * "open coded" version of ANDROID_KABI_USE() to pack more
919 * fields/variables into the space that we have.
920 *
921 * scm_io_uring is from 04df9719df18 ("io_uring/af_unix: defer
922 * registered files gc to io_uring release")
923 */
924 _ANDROID_KABI_REPLACE(_ANDROID_KABI_RESERVE(1),
925 struct {
926 __u8 scm_io_uring:1;
927 __u8 android_kabi_reserved1_padding1;
928 __u16 android_kabi_reserved1_padding2;
929 __u32 android_kabi_reserved1_padding3;
930 });
931 ANDROID_KABI_RESERVE(2);
932
933 /* These elements must be at the end, see alloc_skb() for details. */
934 sk_buff_data_t tail;
935 sk_buff_data_t end;
936 unsigned char *head,
937 *data;
938 unsigned int truesize;
939 refcount_t users;
940
941 #ifdef CONFIG_SKB_EXTENSIONS
942 /* only useable after checking ->active_extensions != 0 */
943 struct skb_ext *extensions;
944 #endif
945 };
946
947 #ifdef __KERNEL__
948 /*
949 * Handling routines are only of interest to the kernel
950 */
951
952 #define SKB_ALLOC_FCLONE 0x01
953 #define SKB_ALLOC_RX 0x02
954 #define SKB_ALLOC_NAPI 0x04
955
956 /**
957 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
958 * @skb: buffer
959 */
skb_pfmemalloc(const struct sk_buff * skb)960 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
961 {
962 return unlikely(skb->pfmemalloc);
963 }
964
965 /*
966 * skb might have a dst pointer attached, refcounted or not.
967 * _skb_refdst low order bit is set if refcount was _not_ taken
968 */
969 #define SKB_DST_NOREF 1UL
970 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
971
972 /**
973 * skb_dst - returns skb dst_entry
974 * @skb: buffer
975 *
976 * Returns skb dst_entry, regardless of reference taken or not.
977 */
skb_dst(const struct sk_buff * skb)978 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
979 {
980 /* If refdst was not refcounted, check we still are in a
981 * rcu_read_lock section
982 */
983 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
984 !rcu_read_lock_held() &&
985 !rcu_read_lock_bh_held());
986 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
987 }
988
989 /**
990 * skb_dst_set - sets skb dst
991 * @skb: buffer
992 * @dst: dst entry
993 *
994 * Sets skb dst, assuming a reference was taken on dst and should
995 * be released by skb_dst_drop()
996 */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)997 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
998 {
999 skb->_skb_refdst = (unsigned long)dst;
1000 }
1001
1002 /**
1003 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1004 * @skb: buffer
1005 * @dst: dst entry
1006 *
1007 * Sets skb dst, assuming a reference was not taken on dst.
1008 * If dst entry is cached, we do not take reference and dst_release
1009 * will be avoided by refdst_drop. If dst entry is not cached, we take
1010 * reference, so that last dst_release can destroy the dst immediately.
1011 */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)1012 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1013 {
1014 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1015 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1016 }
1017
1018 /**
1019 * skb_dst_is_noref - Test if skb dst isn't refcounted
1020 * @skb: buffer
1021 */
skb_dst_is_noref(const struct sk_buff * skb)1022 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1023 {
1024 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1025 }
1026
1027 /**
1028 * skb_rtable - Returns the skb &rtable
1029 * @skb: buffer
1030 */
skb_rtable(const struct sk_buff * skb)1031 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1032 {
1033 return (struct rtable *)skb_dst(skb);
1034 }
1035
1036 /* For mangling skb->pkt_type from user space side from applications
1037 * such as nft, tc, etc, we only allow a conservative subset of
1038 * possible pkt_types to be set.
1039 */
skb_pkt_type_ok(u32 ptype)1040 static inline bool skb_pkt_type_ok(u32 ptype)
1041 {
1042 return ptype <= PACKET_OTHERHOST;
1043 }
1044
1045 /**
1046 * skb_napi_id - Returns the skb's NAPI id
1047 * @skb: buffer
1048 */
skb_napi_id(const struct sk_buff * skb)1049 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1050 {
1051 #ifdef CONFIG_NET_RX_BUSY_POLL
1052 return skb->napi_id;
1053 #else
1054 return 0;
1055 #endif
1056 }
1057
1058 /**
1059 * skb_unref - decrement the skb's reference count
1060 * @skb: buffer
1061 *
1062 * Returns true if we can free the skb.
1063 */
skb_unref(struct sk_buff * skb)1064 static inline bool skb_unref(struct sk_buff *skb)
1065 {
1066 if (unlikely(!skb))
1067 return false;
1068 if (likely(refcount_read(&skb->users) == 1))
1069 smp_rmb();
1070 else if (likely(!refcount_dec_and_test(&skb->users)))
1071 return false;
1072
1073 return true;
1074 }
1075
1076 void skb_release_head_state(struct sk_buff *skb);
1077 void kfree_skb(struct sk_buff *skb);
1078 void kfree_skb_list(struct sk_buff *segs);
1079 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1080 void skb_tx_error(struct sk_buff *skb);
1081
1082 #ifdef CONFIG_TRACEPOINTS
1083 void consume_skb(struct sk_buff *skb);
1084 #else
consume_skb(struct sk_buff * skb)1085 static inline void consume_skb(struct sk_buff *skb)
1086 {
1087 return kfree_skb(skb);
1088 }
1089 #endif
1090
1091 void __consume_stateless_skb(struct sk_buff *skb);
1092 void __kfree_skb(struct sk_buff *skb);
1093 extern struct kmem_cache *skbuff_head_cache;
1094
1095 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1096 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1097 bool *fragstolen, int *delta_truesize);
1098
1099 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1100 int node);
1101 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1102 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1103 struct sk_buff *build_skb_around(struct sk_buff *skb,
1104 void *data, unsigned int frag_size);
1105
1106 /**
1107 * alloc_skb - allocate a network buffer
1108 * @size: size to allocate
1109 * @priority: allocation mask
1110 *
1111 * This function is a convenient wrapper around __alloc_skb().
1112 */
alloc_skb(unsigned int size,gfp_t priority)1113 static inline struct sk_buff *alloc_skb(unsigned int size,
1114 gfp_t priority)
1115 {
1116 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1117 }
1118
1119 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1120 unsigned long data_len,
1121 int max_page_order,
1122 int *errcode,
1123 gfp_t gfp_mask);
1124 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1125
1126 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1127 struct sk_buff_fclones {
1128 struct sk_buff skb1;
1129
1130 struct sk_buff skb2;
1131
1132 refcount_t fclone_ref;
1133 };
1134
1135 /**
1136 * skb_fclone_busy - check if fclone is busy
1137 * @sk: socket
1138 * @skb: buffer
1139 *
1140 * Returns true if skb is a fast clone, and its clone is not freed.
1141 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1142 * so we also check that this didnt happen.
1143 */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1144 static inline bool skb_fclone_busy(const struct sock *sk,
1145 const struct sk_buff *skb)
1146 {
1147 const struct sk_buff_fclones *fclones;
1148
1149 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1150
1151 return skb->fclone == SKB_FCLONE_ORIG &&
1152 refcount_read(&fclones->fclone_ref) > 1 &&
1153 fclones->skb2.sk == sk;
1154 }
1155
1156 /**
1157 * alloc_skb_fclone - allocate a network buffer from fclone cache
1158 * @size: size to allocate
1159 * @priority: allocation mask
1160 *
1161 * This function is a convenient wrapper around __alloc_skb().
1162 */
alloc_skb_fclone(unsigned int size,gfp_t priority)1163 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1164 gfp_t priority)
1165 {
1166 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1167 }
1168
1169 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1170 void skb_headers_offset_update(struct sk_buff *skb, int off);
1171 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1172 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1173 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1174 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1175 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1176 gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1177 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1178 gfp_t gfp_mask)
1179 {
1180 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1181 }
1182
1183 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1184 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1185 unsigned int headroom);
1186 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1187 int newtailroom, gfp_t priority);
1188 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1189 int offset, int len);
1190 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1191 int offset, int len);
1192 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1193 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1194
1195 /**
1196 * skb_pad - zero pad the tail of an skb
1197 * @skb: buffer to pad
1198 * @pad: space to pad
1199 *
1200 * Ensure that a buffer is followed by a padding area that is zero
1201 * filled. Used by network drivers which may DMA or transfer data
1202 * beyond the buffer end onto the wire.
1203 *
1204 * May return error in out of memory cases. The skb is freed on error.
1205 */
skb_pad(struct sk_buff * skb,int pad)1206 static inline int skb_pad(struct sk_buff *skb, int pad)
1207 {
1208 return __skb_pad(skb, pad, true);
1209 }
1210 #define dev_kfree_skb(a) consume_skb(a)
1211
1212 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1213 int offset, size_t size);
1214
1215 struct skb_seq_state {
1216 __u32 lower_offset;
1217 __u32 upper_offset;
1218 __u32 frag_idx;
1219 __u32 stepped_offset;
1220 struct sk_buff *root_skb;
1221 struct sk_buff *cur_skb;
1222 __u8 *frag_data;
1223 };
1224
1225 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1226 unsigned int to, struct skb_seq_state *st);
1227 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1228 struct skb_seq_state *st);
1229 void skb_abort_seq_read(struct skb_seq_state *st);
1230
1231 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1232 unsigned int to, struct ts_config *config);
1233
1234 /*
1235 * Packet hash types specify the type of hash in skb_set_hash.
1236 *
1237 * Hash types refer to the protocol layer addresses which are used to
1238 * construct a packet's hash. The hashes are used to differentiate or identify
1239 * flows of the protocol layer for the hash type. Hash types are either
1240 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1241 *
1242 * Properties of hashes:
1243 *
1244 * 1) Two packets in different flows have different hash values
1245 * 2) Two packets in the same flow should have the same hash value
1246 *
1247 * A hash at a higher layer is considered to be more specific. A driver should
1248 * set the most specific hash possible.
1249 *
1250 * A driver cannot indicate a more specific hash than the layer at which a hash
1251 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1252 *
1253 * A driver may indicate a hash level which is less specific than the
1254 * actual layer the hash was computed on. For instance, a hash computed
1255 * at L4 may be considered an L3 hash. This should only be done if the
1256 * driver can't unambiguously determine that the HW computed the hash at
1257 * the higher layer. Note that the "should" in the second property above
1258 * permits this.
1259 */
1260 enum pkt_hash_types {
1261 PKT_HASH_TYPE_NONE, /* Undefined type */
1262 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1263 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1264 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1265 };
1266
skb_clear_hash(struct sk_buff * skb)1267 static inline void skb_clear_hash(struct sk_buff *skb)
1268 {
1269 skb->hash = 0;
1270 skb->sw_hash = 0;
1271 skb->l4_hash = 0;
1272 }
1273
skb_clear_hash_if_not_l4(struct sk_buff * skb)1274 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1275 {
1276 if (!skb->l4_hash)
1277 skb_clear_hash(skb);
1278 }
1279
1280 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1281 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1282 {
1283 skb->l4_hash = is_l4;
1284 skb->sw_hash = is_sw;
1285 skb->hash = hash;
1286 }
1287
1288 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1289 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1290 {
1291 /* Used by drivers to set hash from HW */
1292 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1293 }
1294
1295 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1296 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1297 {
1298 __skb_set_hash(skb, hash, true, is_l4);
1299 }
1300
1301 void __skb_get_hash(struct sk_buff *skb);
1302 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1303 u32 skb_get_poff(const struct sk_buff *skb);
1304 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1305 const struct flow_keys_basic *keys, int hlen);
1306 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1307 void *data, int hlen_proto);
1308
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto)1309 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1310 int thoff, u8 ip_proto)
1311 {
1312 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1313 }
1314
1315 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1316 const struct flow_dissector_key *key,
1317 unsigned int key_count);
1318
1319 struct bpf_flow_dissector;
1320 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1321 __be16 proto, int nhoff, int hlen, unsigned int flags);
1322
1323 bool __skb_flow_dissect(const struct net *net,
1324 const struct sk_buff *skb,
1325 struct flow_dissector *flow_dissector,
1326 void *target_container,
1327 void *data, __be16 proto, int nhoff, int hlen,
1328 unsigned int flags);
1329
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1330 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1331 struct flow_dissector *flow_dissector,
1332 void *target_container, unsigned int flags)
1333 {
1334 return __skb_flow_dissect(NULL, skb, flow_dissector,
1335 target_container, NULL, 0, 0, 0, flags);
1336 }
1337
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1338 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1339 struct flow_keys *flow,
1340 unsigned int flags)
1341 {
1342 memset(flow, 0, sizeof(*flow));
1343 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1344 flow, NULL, 0, 0, 0, flags);
1345 }
1346
1347 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)1348 skb_flow_dissect_flow_keys_basic(const struct net *net,
1349 const struct sk_buff *skb,
1350 struct flow_keys_basic *flow, void *data,
1351 __be16 proto, int nhoff, int hlen,
1352 unsigned int flags)
1353 {
1354 memset(flow, 0, sizeof(*flow));
1355 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1356 data, proto, nhoff, hlen, flags);
1357 }
1358
1359 void skb_flow_dissect_meta(const struct sk_buff *skb,
1360 struct flow_dissector *flow_dissector,
1361 void *target_container);
1362
1363 /* Gets a skb connection tracking info, ctinfo map should be a
1364 * map of mapsize to translate enum ip_conntrack_info states
1365 * to user states.
1366 */
1367 void
1368 skb_flow_dissect_ct(const struct sk_buff *skb,
1369 struct flow_dissector *flow_dissector,
1370 void *target_container,
1371 u16 *ctinfo_map,
1372 size_t mapsize);
1373 void
1374 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1375 struct flow_dissector *flow_dissector,
1376 void *target_container);
1377
1378 void skb_flow_dissect_hash(const struct sk_buff *skb,
1379 struct flow_dissector *flow_dissector,
1380 void *target_container);
1381
skb_get_hash(struct sk_buff * skb)1382 static inline __u32 skb_get_hash(struct sk_buff *skb)
1383 {
1384 if (!skb->l4_hash && !skb->sw_hash)
1385 __skb_get_hash(skb);
1386
1387 return skb->hash;
1388 }
1389
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1390 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1391 {
1392 if (!skb->l4_hash && !skb->sw_hash) {
1393 struct flow_keys keys;
1394 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1395
1396 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1397 }
1398
1399 return skb->hash;
1400 }
1401
1402 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1403 const siphash_key_t *perturb);
1404
skb_get_hash_raw(const struct sk_buff * skb)1405 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1406 {
1407 return skb->hash;
1408 }
1409
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1410 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1411 {
1412 to->hash = from->hash;
1413 to->sw_hash = from->sw_hash;
1414 to->l4_hash = from->l4_hash;
1415 };
1416
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1417 static inline void skb_copy_decrypted(struct sk_buff *to,
1418 const struct sk_buff *from)
1419 {
1420 #ifdef CONFIG_TLS_DEVICE
1421 to->decrypted = from->decrypted;
1422 #endif
1423 }
1424
1425 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1426 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1427 {
1428 return skb->head + skb->end;
1429 }
1430
skb_end_offset(const struct sk_buff * skb)1431 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1432 {
1433 return skb->end;
1434 }
1435
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1436 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1437 {
1438 skb->end = offset;
1439 }
1440 #else
skb_end_pointer(const struct sk_buff * skb)1441 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1442 {
1443 return skb->end;
1444 }
1445
skb_end_offset(const struct sk_buff * skb)1446 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1447 {
1448 return skb->end - skb->head;
1449 }
1450
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1451 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1452 {
1453 skb->end = skb->head + offset;
1454 }
1455 #endif
1456
1457 /* Internal */
1458 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1459
skb_hwtstamps(struct sk_buff * skb)1460 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1461 {
1462 return &skb_shinfo(skb)->hwtstamps;
1463 }
1464
skb_zcopy(struct sk_buff * skb)1465 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1466 {
1467 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1468
1469 return is_zcopy ? skb_uarg(skb) : NULL;
1470 }
1471
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1472 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1473 bool *have_ref)
1474 {
1475 if (skb && uarg && !skb_zcopy(skb)) {
1476 if (unlikely(have_ref && *have_ref))
1477 *have_ref = false;
1478 else
1479 sock_zerocopy_get(uarg);
1480 skb_shinfo(skb)->destructor_arg = uarg;
1481 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1482 }
1483 }
1484
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1485 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1486 {
1487 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1488 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1489 }
1490
skb_zcopy_is_nouarg(struct sk_buff * skb)1491 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1492 {
1493 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1494 }
1495
skb_zcopy_get_nouarg(struct sk_buff * skb)1496 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1497 {
1498 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1499 }
1500
1501 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy)1502 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1503 {
1504 struct ubuf_info *uarg = skb_zcopy(skb);
1505
1506 if (uarg) {
1507 if (skb_zcopy_is_nouarg(skb)) {
1508 /* no notification callback */
1509 } else if (uarg->callback == sock_zerocopy_callback) {
1510 uarg->zerocopy = uarg->zerocopy && zerocopy;
1511 sock_zerocopy_put(uarg);
1512 } else {
1513 uarg->callback(uarg, zerocopy);
1514 }
1515
1516 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1517 }
1518 }
1519
1520 /* Abort a zerocopy operation and revert zckey on error in send syscall */
skb_zcopy_abort(struct sk_buff * skb)1521 static inline void skb_zcopy_abort(struct sk_buff *skb)
1522 {
1523 struct ubuf_info *uarg = skb_zcopy(skb);
1524
1525 if (uarg) {
1526 sock_zerocopy_put_abort(uarg, false);
1527 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1528 }
1529 }
1530
skb_mark_not_on_list(struct sk_buff * skb)1531 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1532 {
1533 skb->next = NULL;
1534 }
1535
1536 /* Iterate through singly-linked GSO fragments of an skb. */
1537 #define skb_list_walk_safe(first, skb, next_skb) \
1538 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1539 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1540
skb_list_del_init(struct sk_buff * skb)1541 static inline void skb_list_del_init(struct sk_buff *skb)
1542 {
1543 __list_del_entry(&skb->list);
1544 skb_mark_not_on_list(skb);
1545 }
1546
1547 /**
1548 * skb_queue_empty - check if a queue is empty
1549 * @list: queue head
1550 *
1551 * Returns true if the queue is empty, false otherwise.
1552 */
skb_queue_empty(const struct sk_buff_head * list)1553 static inline int skb_queue_empty(const struct sk_buff_head *list)
1554 {
1555 return list->next == (const struct sk_buff *) list;
1556 }
1557
1558 /**
1559 * skb_queue_empty_lockless - check if a queue is empty
1560 * @list: queue head
1561 *
1562 * Returns true if the queue is empty, false otherwise.
1563 * This variant can be used in lockless contexts.
1564 */
skb_queue_empty_lockless(const struct sk_buff_head * list)1565 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1566 {
1567 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1568 }
1569
1570
1571 /**
1572 * skb_queue_is_last - check if skb is the last entry in the queue
1573 * @list: queue head
1574 * @skb: buffer
1575 *
1576 * Returns true if @skb is the last buffer on the list.
1577 */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1578 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1579 const struct sk_buff *skb)
1580 {
1581 return skb->next == (const struct sk_buff *) list;
1582 }
1583
1584 /**
1585 * skb_queue_is_first - check if skb is the first entry in the queue
1586 * @list: queue head
1587 * @skb: buffer
1588 *
1589 * Returns true if @skb is the first buffer on the list.
1590 */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1591 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1592 const struct sk_buff *skb)
1593 {
1594 return skb->prev == (const struct sk_buff *) list;
1595 }
1596
1597 /**
1598 * skb_queue_next - return the next packet in the queue
1599 * @list: queue head
1600 * @skb: current buffer
1601 *
1602 * Return the next packet in @list after @skb. It is only valid to
1603 * call this if skb_queue_is_last() evaluates to false.
1604 */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1605 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1606 const struct sk_buff *skb)
1607 {
1608 /* This BUG_ON may seem severe, but if we just return then we
1609 * are going to dereference garbage.
1610 */
1611 BUG_ON(skb_queue_is_last(list, skb));
1612 return skb->next;
1613 }
1614
1615 /**
1616 * skb_queue_prev - return the prev packet in the queue
1617 * @list: queue head
1618 * @skb: current buffer
1619 *
1620 * Return the prev packet in @list before @skb. It is only valid to
1621 * call this if skb_queue_is_first() evaluates to false.
1622 */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1623 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1624 const struct sk_buff *skb)
1625 {
1626 /* This BUG_ON may seem severe, but if we just return then we
1627 * are going to dereference garbage.
1628 */
1629 BUG_ON(skb_queue_is_first(list, skb));
1630 return skb->prev;
1631 }
1632
1633 /**
1634 * skb_get - reference buffer
1635 * @skb: buffer to reference
1636 *
1637 * Makes another reference to a socket buffer and returns a pointer
1638 * to the buffer.
1639 */
skb_get(struct sk_buff * skb)1640 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1641 {
1642 refcount_inc(&skb->users);
1643 return skb;
1644 }
1645
1646 /*
1647 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1648 */
1649
1650 /**
1651 * skb_cloned - is the buffer a clone
1652 * @skb: buffer to check
1653 *
1654 * Returns true if the buffer was generated with skb_clone() and is
1655 * one of multiple shared copies of the buffer. Cloned buffers are
1656 * shared data so must not be written to under normal circumstances.
1657 */
skb_cloned(const struct sk_buff * skb)1658 static inline int skb_cloned(const struct sk_buff *skb)
1659 {
1660 return skb->cloned &&
1661 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1662 }
1663
skb_unclone(struct sk_buff * skb,gfp_t pri)1664 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1665 {
1666 might_sleep_if(gfpflags_allow_blocking(pri));
1667
1668 if (skb_cloned(skb))
1669 return pskb_expand_head(skb, 0, 0, pri);
1670
1671 return 0;
1672 }
1673
1674 /* This variant of skb_unclone() makes sure skb->truesize
1675 * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1676 *
1677 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
1678 * when various debugging features are in place.
1679 */
1680 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
skb_unclone_keeptruesize(struct sk_buff * skb,gfp_t pri)1681 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1682 {
1683 might_sleep_if(gfpflags_allow_blocking(pri));
1684
1685 if (skb_cloned(skb))
1686 return __skb_unclone_keeptruesize(skb, pri);
1687 return 0;
1688 }
1689
1690 /**
1691 * skb_header_cloned - is the header a clone
1692 * @skb: buffer to check
1693 *
1694 * Returns true if modifying the header part of the buffer requires
1695 * the data to be copied.
1696 */
skb_header_cloned(const struct sk_buff * skb)1697 static inline int skb_header_cloned(const struct sk_buff *skb)
1698 {
1699 int dataref;
1700
1701 if (!skb->cloned)
1702 return 0;
1703
1704 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1705 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1706 return dataref != 1;
1707 }
1708
skb_header_unclone(struct sk_buff * skb,gfp_t pri)1709 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1710 {
1711 might_sleep_if(gfpflags_allow_blocking(pri));
1712
1713 if (skb_header_cloned(skb))
1714 return pskb_expand_head(skb, 0, 0, pri);
1715
1716 return 0;
1717 }
1718
1719 /**
1720 * __skb_header_release - release reference to header
1721 * @skb: buffer to operate on
1722 */
__skb_header_release(struct sk_buff * skb)1723 static inline void __skb_header_release(struct sk_buff *skb)
1724 {
1725 skb->nohdr = 1;
1726 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1727 }
1728
1729
1730 /**
1731 * skb_shared - is the buffer shared
1732 * @skb: buffer to check
1733 *
1734 * Returns true if more than one person has a reference to this
1735 * buffer.
1736 */
skb_shared(const struct sk_buff * skb)1737 static inline int skb_shared(const struct sk_buff *skb)
1738 {
1739 return refcount_read(&skb->users) != 1;
1740 }
1741
1742 /**
1743 * skb_share_check - check if buffer is shared and if so clone it
1744 * @skb: buffer to check
1745 * @pri: priority for memory allocation
1746 *
1747 * If the buffer is shared the buffer is cloned and the old copy
1748 * drops a reference. A new clone with a single reference is returned.
1749 * If the buffer is not shared the original buffer is returned. When
1750 * being called from interrupt status or with spinlocks held pri must
1751 * be GFP_ATOMIC.
1752 *
1753 * NULL is returned on a memory allocation failure.
1754 */
skb_share_check(struct sk_buff * skb,gfp_t pri)1755 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1756 {
1757 might_sleep_if(gfpflags_allow_blocking(pri));
1758 if (skb_shared(skb)) {
1759 struct sk_buff *nskb = skb_clone(skb, pri);
1760
1761 if (likely(nskb))
1762 consume_skb(skb);
1763 else
1764 kfree_skb(skb);
1765 skb = nskb;
1766 }
1767 return skb;
1768 }
1769
1770 /*
1771 * Copy shared buffers into a new sk_buff. We effectively do COW on
1772 * packets to handle cases where we have a local reader and forward
1773 * and a couple of other messy ones. The normal one is tcpdumping
1774 * a packet thats being forwarded.
1775 */
1776
1777 /**
1778 * skb_unshare - make a copy of a shared buffer
1779 * @skb: buffer to check
1780 * @pri: priority for memory allocation
1781 *
1782 * If the socket buffer is a clone then this function creates a new
1783 * copy of the data, drops a reference count on the old copy and returns
1784 * the new copy with the reference count at 1. If the buffer is not a clone
1785 * the original buffer is returned. When called with a spinlock held or
1786 * from interrupt state @pri must be %GFP_ATOMIC
1787 *
1788 * %NULL is returned on a memory allocation failure.
1789 */
skb_unshare(struct sk_buff * skb,gfp_t pri)1790 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1791 gfp_t pri)
1792 {
1793 might_sleep_if(gfpflags_allow_blocking(pri));
1794 if (skb_cloned(skb)) {
1795 struct sk_buff *nskb = skb_copy(skb, pri);
1796
1797 /* Free our shared copy */
1798 if (likely(nskb))
1799 consume_skb(skb);
1800 else
1801 kfree_skb(skb);
1802 skb = nskb;
1803 }
1804 return skb;
1805 }
1806
1807 /**
1808 * skb_peek - peek at the head of an &sk_buff_head
1809 * @list_: list to peek at
1810 *
1811 * Peek an &sk_buff. Unlike most other operations you _MUST_
1812 * be careful with this one. A peek leaves the buffer on the
1813 * list and someone else may run off with it. You must hold
1814 * the appropriate locks or have a private queue to do this.
1815 *
1816 * Returns %NULL for an empty list or a pointer to the head element.
1817 * The reference count is not incremented and the reference is therefore
1818 * volatile. Use with caution.
1819 */
skb_peek(const struct sk_buff_head * list_)1820 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1821 {
1822 struct sk_buff *skb = list_->next;
1823
1824 if (skb == (struct sk_buff *)list_)
1825 skb = NULL;
1826 return skb;
1827 }
1828
1829 /**
1830 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1831 * @list_: list to peek at
1832 *
1833 * Like skb_peek(), but the caller knows that the list is not empty.
1834 */
__skb_peek(const struct sk_buff_head * list_)1835 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1836 {
1837 return list_->next;
1838 }
1839
1840 /**
1841 * skb_peek_next - peek skb following the given one from a queue
1842 * @skb: skb to start from
1843 * @list_: list to peek at
1844 *
1845 * Returns %NULL when the end of the list is met or a pointer to the
1846 * next element. The reference count is not incremented and the
1847 * reference is therefore volatile. Use with caution.
1848 */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)1849 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1850 const struct sk_buff_head *list_)
1851 {
1852 struct sk_buff *next = skb->next;
1853
1854 if (next == (struct sk_buff *)list_)
1855 next = NULL;
1856 return next;
1857 }
1858
1859 /**
1860 * skb_peek_tail - peek at the tail of an &sk_buff_head
1861 * @list_: list to peek at
1862 *
1863 * Peek an &sk_buff. Unlike most other operations you _MUST_
1864 * be careful with this one. A peek leaves the buffer on the
1865 * list and someone else may run off with it. You must hold
1866 * the appropriate locks or have a private queue to do this.
1867 *
1868 * Returns %NULL for an empty list or a pointer to the tail element.
1869 * The reference count is not incremented and the reference is therefore
1870 * volatile. Use with caution.
1871 */
skb_peek_tail(const struct sk_buff_head * list_)1872 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1873 {
1874 struct sk_buff *skb = READ_ONCE(list_->prev);
1875
1876 if (skb == (struct sk_buff *)list_)
1877 skb = NULL;
1878 return skb;
1879
1880 }
1881
1882 /**
1883 * skb_queue_len - get queue length
1884 * @list_: list to measure
1885 *
1886 * Return the length of an &sk_buff queue.
1887 */
skb_queue_len(const struct sk_buff_head * list_)1888 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1889 {
1890 return list_->qlen;
1891 }
1892
1893 /**
1894 * skb_queue_len_lockless - get queue length
1895 * @list_: list to measure
1896 *
1897 * Return the length of an &sk_buff queue.
1898 * This variant can be used in lockless contexts.
1899 */
skb_queue_len_lockless(const struct sk_buff_head * list_)1900 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1901 {
1902 return READ_ONCE(list_->qlen);
1903 }
1904
1905 /**
1906 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1907 * @list: queue to initialize
1908 *
1909 * This initializes only the list and queue length aspects of
1910 * an sk_buff_head object. This allows to initialize the list
1911 * aspects of an sk_buff_head without reinitializing things like
1912 * the spinlock. It can also be used for on-stack sk_buff_head
1913 * objects where the spinlock is known to not be used.
1914 */
__skb_queue_head_init(struct sk_buff_head * list)1915 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1916 {
1917 list->prev = list->next = (struct sk_buff *)list;
1918 list->qlen = 0;
1919 }
1920
1921 /*
1922 * This function creates a split out lock class for each invocation;
1923 * this is needed for now since a whole lot of users of the skb-queue
1924 * infrastructure in drivers have different locking usage (in hardirq)
1925 * than the networking core (in softirq only). In the long run either the
1926 * network layer or drivers should need annotation to consolidate the
1927 * main types of usage into 3 classes.
1928 */
skb_queue_head_init(struct sk_buff_head * list)1929 static inline void skb_queue_head_init(struct sk_buff_head *list)
1930 {
1931 spin_lock_init(&list->lock);
1932 __skb_queue_head_init(list);
1933 }
1934
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)1935 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1936 struct lock_class_key *class)
1937 {
1938 skb_queue_head_init(list);
1939 lockdep_set_class(&list->lock, class);
1940 }
1941
1942 /*
1943 * Insert an sk_buff on a list.
1944 *
1945 * The "__skb_xxxx()" functions are the non-atomic ones that
1946 * can only be called with interrupts disabled.
1947 */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)1948 static inline void __skb_insert(struct sk_buff *newsk,
1949 struct sk_buff *prev, struct sk_buff *next,
1950 struct sk_buff_head *list)
1951 {
1952 /* See skb_queue_empty_lockless() and skb_peek_tail()
1953 * for the opposite READ_ONCE()
1954 */
1955 WRITE_ONCE(newsk->next, next);
1956 WRITE_ONCE(newsk->prev, prev);
1957 WRITE_ONCE(next->prev, newsk);
1958 WRITE_ONCE(prev->next, newsk);
1959 WRITE_ONCE(list->qlen, list->qlen + 1);
1960 }
1961
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)1962 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1963 struct sk_buff *prev,
1964 struct sk_buff *next)
1965 {
1966 struct sk_buff *first = list->next;
1967 struct sk_buff *last = list->prev;
1968
1969 WRITE_ONCE(first->prev, prev);
1970 WRITE_ONCE(prev->next, first);
1971
1972 WRITE_ONCE(last->next, next);
1973 WRITE_ONCE(next->prev, last);
1974 }
1975
1976 /**
1977 * skb_queue_splice - join two skb lists, this is designed for stacks
1978 * @list: the new list to add
1979 * @head: the place to add it in the first list
1980 */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)1981 static inline void skb_queue_splice(const struct sk_buff_head *list,
1982 struct sk_buff_head *head)
1983 {
1984 if (!skb_queue_empty(list)) {
1985 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1986 head->qlen += list->qlen;
1987 }
1988 }
1989
1990 /**
1991 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1992 * @list: the new list to add
1993 * @head: the place to add it in the first list
1994 *
1995 * The list at @list is reinitialised
1996 */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)1997 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1998 struct sk_buff_head *head)
1999 {
2000 if (!skb_queue_empty(list)) {
2001 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2002 head->qlen += list->qlen;
2003 __skb_queue_head_init(list);
2004 }
2005 }
2006
2007 /**
2008 * skb_queue_splice_tail - join two skb lists, each list being a queue
2009 * @list: the new list to add
2010 * @head: the place to add it in the first list
2011 */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)2012 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2013 struct sk_buff_head *head)
2014 {
2015 if (!skb_queue_empty(list)) {
2016 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2017 head->qlen += list->qlen;
2018 }
2019 }
2020
2021 /**
2022 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2023 * @list: the new list to add
2024 * @head: the place to add it in the first list
2025 *
2026 * Each of the lists is a queue.
2027 * The list at @list is reinitialised
2028 */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)2029 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2030 struct sk_buff_head *head)
2031 {
2032 if (!skb_queue_empty(list)) {
2033 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2034 head->qlen += list->qlen;
2035 __skb_queue_head_init(list);
2036 }
2037 }
2038
2039 /**
2040 * __skb_queue_after - queue a buffer at the list head
2041 * @list: list to use
2042 * @prev: place after this buffer
2043 * @newsk: buffer to queue
2044 *
2045 * Queue a buffer int the middle of a list. This function takes no locks
2046 * and you must therefore hold required locks before calling it.
2047 *
2048 * A buffer cannot be placed on two lists at the same time.
2049 */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)2050 static inline void __skb_queue_after(struct sk_buff_head *list,
2051 struct sk_buff *prev,
2052 struct sk_buff *newsk)
2053 {
2054 __skb_insert(newsk, prev, prev->next, list);
2055 }
2056
2057 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2058 struct sk_buff_head *list);
2059
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)2060 static inline void __skb_queue_before(struct sk_buff_head *list,
2061 struct sk_buff *next,
2062 struct sk_buff *newsk)
2063 {
2064 __skb_insert(newsk, next->prev, next, list);
2065 }
2066
2067 /**
2068 * __skb_queue_head - queue a buffer at the list head
2069 * @list: list to use
2070 * @newsk: buffer to queue
2071 *
2072 * Queue a buffer at the start of a list. This function takes no locks
2073 * and you must therefore hold required locks before calling it.
2074 *
2075 * A buffer cannot be placed on two lists at the same time.
2076 */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)2077 static inline void __skb_queue_head(struct sk_buff_head *list,
2078 struct sk_buff *newsk)
2079 {
2080 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2081 }
2082 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2083
2084 /**
2085 * __skb_queue_tail - queue a buffer at the list tail
2086 * @list: list to use
2087 * @newsk: buffer to queue
2088 *
2089 * Queue a buffer at the end of a list. This function takes no locks
2090 * and you must therefore hold required locks before calling it.
2091 *
2092 * A buffer cannot be placed on two lists at the same time.
2093 */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2094 static inline void __skb_queue_tail(struct sk_buff_head *list,
2095 struct sk_buff *newsk)
2096 {
2097 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2098 }
2099 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2100
2101 /*
2102 * remove sk_buff from list. _Must_ be called atomically, and with
2103 * the list known..
2104 */
2105 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2106 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2107 {
2108 struct sk_buff *next, *prev;
2109
2110 WRITE_ONCE(list->qlen, list->qlen - 1);
2111 next = skb->next;
2112 prev = skb->prev;
2113 skb->next = skb->prev = NULL;
2114 WRITE_ONCE(next->prev, prev);
2115 WRITE_ONCE(prev->next, next);
2116 }
2117
2118 /**
2119 * __skb_dequeue - remove from the head of the queue
2120 * @list: list to dequeue from
2121 *
2122 * Remove the head of the list. This function does not take any locks
2123 * so must be used with appropriate locks held only. The head item is
2124 * returned or %NULL if the list is empty.
2125 */
__skb_dequeue(struct sk_buff_head * list)2126 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2127 {
2128 struct sk_buff *skb = skb_peek(list);
2129 if (skb)
2130 __skb_unlink(skb, list);
2131 return skb;
2132 }
2133 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2134
2135 /**
2136 * __skb_dequeue_tail - remove from the tail of the queue
2137 * @list: list to dequeue from
2138 *
2139 * Remove the tail of the list. This function does not take any locks
2140 * so must be used with appropriate locks held only. The tail item is
2141 * returned or %NULL if the list is empty.
2142 */
__skb_dequeue_tail(struct sk_buff_head * list)2143 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2144 {
2145 struct sk_buff *skb = skb_peek_tail(list);
2146 if (skb)
2147 __skb_unlink(skb, list);
2148 return skb;
2149 }
2150 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2151
2152
skb_is_nonlinear(const struct sk_buff * skb)2153 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2154 {
2155 return skb->data_len;
2156 }
2157
skb_headlen(const struct sk_buff * skb)2158 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2159 {
2160 return skb->len - skb->data_len;
2161 }
2162
__skb_pagelen(const struct sk_buff * skb)2163 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2164 {
2165 unsigned int i, len = 0;
2166
2167 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2168 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2169 return len;
2170 }
2171
skb_pagelen(const struct sk_buff * skb)2172 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2173 {
2174 return skb_headlen(skb) + __skb_pagelen(skb);
2175 }
2176
2177 /**
2178 * __skb_fill_page_desc - initialise a paged fragment in an skb
2179 * @skb: buffer containing fragment to be initialised
2180 * @i: paged fragment index to initialise
2181 * @page: the page to use for this fragment
2182 * @off: the offset to the data with @page
2183 * @size: the length of the data
2184 *
2185 * Initialises the @i'th fragment of @skb to point to &size bytes at
2186 * offset @off within @page.
2187 *
2188 * Does not take any additional reference on the fragment.
2189 */
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2190 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2191 struct page *page, int off, int size)
2192 {
2193 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2194
2195 /*
2196 * Propagate page pfmemalloc to the skb if we can. The problem is
2197 * that not all callers have unique ownership of the page but rely
2198 * on page_is_pfmemalloc doing the right thing(tm).
2199 */
2200 frag->bv_page = page;
2201 frag->bv_offset = off;
2202 skb_frag_size_set(frag, size);
2203
2204 page = compound_head(page);
2205 if (page_is_pfmemalloc(page))
2206 skb->pfmemalloc = true;
2207 }
2208
2209 /**
2210 * skb_fill_page_desc - initialise a paged fragment in an skb
2211 * @skb: buffer containing fragment to be initialised
2212 * @i: paged fragment index to initialise
2213 * @page: the page to use for this fragment
2214 * @off: the offset to the data with @page
2215 * @size: the length of the data
2216 *
2217 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2218 * @skb to point to @size bytes at offset @off within @page. In
2219 * addition updates @skb such that @i is the last fragment.
2220 *
2221 * Does not take any additional reference on the fragment.
2222 */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2223 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2224 struct page *page, int off, int size)
2225 {
2226 __skb_fill_page_desc(skb, i, page, off, size);
2227 skb_shinfo(skb)->nr_frags = i + 1;
2228 }
2229
2230 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2231 int size, unsigned int truesize);
2232
2233 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2234 unsigned int truesize);
2235
2236 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2237
2238 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2239 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2240 {
2241 return skb->head + skb->tail;
2242 }
2243
skb_reset_tail_pointer(struct sk_buff * skb)2244 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2245 {
2246 skb->tail = skb->data - skb->head;
2247 }
2248
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2249 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2250 {
2251 skb_reset_tail_pointer(skb);
2252 skb->tail += offset;
2253 }
2254
2255 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2256 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2257 {
2258 return skb->tail;
2259 }
2260
skb_reset_tail_pointer(struct sk_buff * skb)2261 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2262 {
2263 skb->tail = skb->data;
2264 }
2265
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2266 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2267 {
2268 skb->tail = skb->data + offset;
2269 }
2270
2271 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2272
skb_assert_len(struct sk_buff * skb)2273 static inline void skb_assert_len(struct sk_buff *skb)
2274 {
2275 #ifdef CONFIG_DEBUG_NET
2276 if (WARN_ONCE(!skb->len, "%s\n", __func__))
2277 DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2278 #endif /* CONFIG_DEBUG_NET */
2279 }
2280
2281 /*
2282 * Add data to an sk_buff
2283 */
2284 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2285 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2286 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2287 {
2288 void *tmp = skb_tail_pointer(skb);
2289 SKB_LINEAR_ASSERT(skb);
2290 skb->tail += len;
2291 skb->len += len;
2292 return tmp;
2293 }
2294
__skb_put_zero(struct sk_buff * skb,unsigned int len)2295 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2296 {
2297 void *tmp = __skb_put(skb, len);
2298
2299 memset(tmp, 0, len);
2300 return tmp;
2301 }
2302
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2303 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2304 unsigned int len)
2305 {
2306 void *tmp = __skb_put(skb, len);
2307
2308 memcpy(tmp, data, len);
2309 return tmp;
2310 }
2311
__skb_put_u8(struct sk_buff * skb,u8 val)2312 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2313 {
2314 *(u8 *)__skb_put(skb, 1) = val;
2315 }
2316
skb_put_zero(struct sk_buff * skb,unsigned int len)2317 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2318 {
2319 void *tmp = skb_put(skb, len);
2320
2321 memset(tmp, 0, len);
2322
2323 return tmp;
2324 }
2325
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2326 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2327 unsigned int len)
2328 {
2329 void *tmp = skb_put(skb, len);
2330
2331 memcpy(tmp, data, len);
2332
2333 return tmp;
2334 }
2335
skb_put_u8(struct sk_buff * skb,u8 val)2336 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2337 {
2338 *(u8 *)skb_put(skb, 1) = val;
2339 }
2340
2341 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2342 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2343 {
2344 skb->data -= len;
2345 skb->len += len;
2346 return skb->data;
2347 }
2348
2349 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2350 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2351 {
2352 skb->len -= len;
2353 BUG_ON(skb->len < skb->data_len);
2354 return skb->data += len;
2355 }
2356
skb_pull_inline(struct sk_buff * skb,unsigned int len)2357 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2358 {
2359 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2360 }
2361
2362 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2363
__pskb_pull(struct sk_buff * skb,unsigned int len)2364 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2365 {
2366 if (len > skb_headlen(skb) &&
2367 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2368 return NULL;
2369 skb->len -= len;
2370 return skb->data += len;
2371 }
2372
pskb_pull(struct sk_buff * skb,unsigned int len)2373 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2374 {
2375 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2376 }
2377
pskb_may_pull(struct sk_buff * skb,unsigned int len)2378 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2379 {
2380 if (likely(len <= skb_headlen(skb)))
2381 return true;
2382 if (unlikely(len > skb->len))
2383 return false;
2384 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2385 }
2386
2387 void skb_condense(struct sk_buff *skb);
2388
2389 /**
2390 * skb_headroom - bytes at buffer head
2391 * @skb: buffer to check
2392 *
2393 * Return the number of bytes of free space at the head of an &sk_buff.
2394 */
skb_headroom(const struct sk_buff * skb)2395 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2396 {
2397 return skb->data - skb->head;
2398 }
2399
2400 /**
2401 * skb_tailroom - bytes at buffer end
2402 * @skb: buffer to check
2403 *
2404 * Return the number of bytes of free space at the tail of an sk_buff
2405 */
skb_tailroom(const struct sk_buff * skb)2406 static inline int skb_tailroom(const struct sk_buff *skb)
2407 {
2408 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2409 }
2410
2411 /**
2412 * skb_availroom - bytes at buffer end
2413 * @skb: buffer to check
2414 *
2415 * Return the number of bytes of free space at the tail of an sk_buff
2416 * allocated by sk_stream_alloc()
2417 */
skb_availroom(const struct sk_buff * skb)2418 static inline int skb_availroom(const struct sk_buff *skb)
2419 {
2420 if (skb_is_nonlinear(skb))
2421 return 0;
2422
2423 return skb->end - skb->tail - skb->reserved_tailroom;
2424 }
2425
2426 /**
2427 * skb_reserve - adjust headroom
2428 * @skb: buffer to alter
2429 * @len: bytes to move
2430 *
2431 * Increase the headroom of an empty &sk_buff by reducing the tail
2432 * room. This is only allowed for an empty buffer.
2433 */
skb_reserve(struct sk_buff * skb,int len)2434 static inline void skb_reserve(struct sk_buff *skb, int len)
2435 {
2436 skb->data += len;
2437 skb->tail += len;
2438 }
2439
2440 /**
2441 * skb_tailroom_reserve - adjust reserved_tailroom
2442 * @skb: buffer to alter
2443 * @mtu: maximum amount of headlen permitted
2444 * @needed_tailroom: minimum amount of reserved_tailroom
2445 *
2446 * Set reserved_tailroom so that headlen can be as large as possible but
2447 * not larger than mtu and tailroom cannot be smaller than
2448 * needed_tailroom.
2449 * The required headroom should already have been reserved before using
2450 * this function.
2451 */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2452 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2453 unsigned int needed_tailroom)
2454 {
2455 SKB_LINEAR_ASSERT(skb);
2456 if (mtu < skb_tailroom(skb) - needed_tailroom)
2457 /* use at most mtu */
2458 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2459 else
2460 /* use up to all available space */
2461 skb->reserved_tailroom = needed_tailroom;
2462 }
2463
2464 #define ENCAP_TYPE_ETHER 0
2465 #define ENCAP_TYPE_IPPROTO 1
2466
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2467 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2468 __be16 protocol)
2469 {
2470 skb->inner_protocol = protocol;
2471 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2472 }
2473
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2474 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2475 __u8 ipproto)
2476 {
2477 skb->inner_ipproto = ipproto;
2478 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2479 }
2480
skb_reset_inner_headers(struct sk_buff * skb)2481 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2482 {
2483 skb->inner_mac_header = skb->mac_header;
2484 skb->inner_network_header = skb->network_header;
2485 skb->inner_transport_header = skb->transport_header;
2486 }
2487
skb_reset_mac_len(struct sk_buff * skb)2488 static inline void skb_reset_mac_len(struct sk_buff *skb)
2489 {
2490 skb->mac_len = skb->network_header - skb->mac_header;
2491 }
2492
skb_inner_transport_header(const struct sk_buff * skb)2493 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2494 *skb)
2495 {
2496 return skb->head + skb->inner_transport_header;
2497 }
2498
skb_inner_transport_offset(const struct sk_buff * skb)2499 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2500 {
2501 return skb_inner_transport_header(skb) - skb->data;
2502 }
2503
skb_reset_inner_transport_header(struct sk_buff * skb)2504 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2505 {
2506 skb->inner_transport_header = skb->data - skb->head;
2507 }
2508
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2509 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2510 const int offset)
2511 {
2512 skb_reset_inner_transport_header(skb);
2513 skb->inner_transport_header += offset;
2514 }
2515
skb_inner_network_header(const struct sk_buff * skb)2516 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2517 {
2518 return skb->head + skb->inner_network_header;
2519 }
2520
skb_reset_inner_network_header(struct sk_buff * skb)2521 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2522 {
2523 skb->inner_network_header = skb->data - skb->head;
2524 }
2525
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2526 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2527 const int offset)
2528 {
2529 skb_reset_inner_network_header(skb);
2530 skb->inner_network_header += offset;
2531 }
2532
skb_inner_mac_header(const struct sk_buff * skb)2533 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2534 {
2535 return skb->head + skb->inner_mac_header;
2536 }
2537
skb_reset_inner_mac_header(struct sk_buff * skb)2538 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2539 {
2540 skb->inner_mac_header = skb->data - skb->head;
2541 }
2542
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)2543 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2544 const int offset)
2545 {
2546 skb_reset_inner_mac_header(skb);
2547 skb->inner_mac_header += offset;
2548 }
skb_transport_header_was_set(const struct sk_buff * skb)2549 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2550 {
2551 return skb->transport_header != (typeof(skb->transport_header))~0U;
2552 }
2553
skb_transport_header(const struct sk_buff * skb)2554 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2555 {
2556 return skb->head + skb->transport_header;
2557 }
2558
skb_reset_transport_header(struct sk_buff * skb)2559 static inline void skb_reset_transport_header(struct sk_buff *skb)
2560 {
2561 skb->transport_header = skb->data - skb->head;
2562 }
2563
skb_set_transport_header(struct sk_buff * skb,const int offset)2564 static inline void skb_set_transport_header(struct sk_buff *skb,
2565 const int offset)
2566 {
2567 skb_reset_transport_header(skb);
2568 skb->transport_header += offset;
2569 }
2570
skb_network_header(const struct sk_buff * skb)2571 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2572 {
2573 return skb->head + skb->network_header;
2574 }
2575
skb_reset_network_header(struct sk_buff * skb)2576 static inline void skb_reset_network_header(struct sk_buff *skb)
2577 {
2578 skb->network_header = skb->data - skb->head;
2579 }
2580
skb_set_network_header(struct sk_buff * skb,const int offset)2581 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2582 {
2583 skb_reset_network_header(skb);
2584 skb->network_header += offset;
2585 }
2586
skb_mac_header(const struct sk_buff * skb)2587 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2588 {
2589 return skb->head + skb->mac_header;
2590 }
2591
skb_mac_offset(const struct sk_buff * skb)2592 static inline int skb_mac_offset(const struct sk_buff *skb)
2593 {
2594 return skb_mac_header(skb) - skb->data;
2595 }
2596
skb_mac_header_len(const struct sk_buff * skb)2597 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2598 {
2599 return skb->network_header - skb->mac_header;
2600 }
2601
skb_mac_header_was_set(const struct sk_buff * skb)2602 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2603 {
2604 return skb->mac_header != (typeof(skb->mac_header))~0U;
2605 }
2606
skb_unset_mac_header(struct sk_buff * skb)2607 static inline void skb_unset_mac_header(struct sk_buff *skb)
2608 {
2609 skb->mac_header = (typeof(skb->mac_header))~0U;
2610 }
2611
skb_reset_mac_header(struct sk_buff * skb)2612 static inline void skb_reset_mac_header(struct sk_buff *skb)
2613 {
2614 skb->mac_header = skb->data - skb->head;
2615 }
2616
skb_set_mac_header(struct sk_buff * skb,const int offset)2617 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2618 {
2619 skb_reset_mac_header(skb);
2620 skb->mac_header += offset;
2621 }
2622
skb_pop_mac_header(struct sk_buff * skb)2623 static inline void skb_pop_mac_header(struct sk_buff *skb)
2624 {
2625 skb->mac_header = skb->network_header;
2626 }
2627
skb_probe_transport_header(struct sk_buff * skb)2628 static inline void skb_probe_transport_header(struct sk_buff *skb)
2629 {
2630 struct flow_keys_basic keys;
2631
2632 if (skb_transport_header_was_set(skb))
2633 return;
2634
2635 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2636 NULL, 0, 0, 0, 0))
2637 skb_set_transport_header(skb, keys.control.thoff);
2638 }
2639
skb_mac_header_rebuild(struct sk_buff * skb)2640 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2641 {
2642 if (skb_mac_header_was_set(skb)) {
2643 const unsigned char *old_mac = skb_mac_header(skb);
2644
2645 skb_set_mac_header(skb, -skb->mac_len);
2646 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2647 }
2648 }
2649
skb_checksum_start_offset(const struct sk_buff * skb)2650 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2651 {
2652 return skb->csum_start - skb_headroom(skb);
2653 }
2654
skb_checksum_start(const struct sk_buff * skb)2655 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2656 {
2657 return skb->head + skb->csum_start;
2658 }
2659
skb_transport_offset(const struct sk_buff * skb)2660 static inline int skb_transport_offset(const struct sk_buff *skb)
2661 {
2662 return skb_transport_header(skb) - skb->data;
2663 }
2664
skb_network_header_len(const struct sk_buff * skb)2665 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2666 {
2667 return skb->transport_header - skb->network_header;
2668 }
2669
skb_inner_network_header_len(const struct sk_buff * skb)2670 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2671 {
2672 return skb->inner_transport_header - skb->inner_network_header;
2673 }
2674
skb_network_offset(const struct sk_buff * skb)2675 static inline int skb_network_offset(const struct sk_buff *skb)
2676 {
2677 return skb_network_header(skb) - skb->data;
2678 }
2679
skb_inner_network_offset(const struct sk_buff * skb)2680 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2681 {
2682 return skb_inner_network_header(skb) - skb->data;
2683 }
2684
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)2685 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2686 {
2687 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2688 }
2689
2690 /*
2691 * CPUs often take a performance hit when accessing unaligned memory
2692 * locations. The actual performance hit varies, it can be small if the
2693 * hardware handles it or large if we have to take an exception and fix it
2694 * in software.
2695 *
2696 * Since an ethernet header is 14 bytes network drivers often end up with
2697 * the IP header at an unaligned offset. The IP header can be aligned by
2698 * shifting the start of the packet by 2 bytes. Drivers should do this
2699 * with:
2700 *
2701 * skb_reserve(skb, NET_IP_ALIGN);
2702 *
2703 * The downside to this alignment of the IP header is that the DMA is now
2704 * unaligned. On some architectures the cost of an unaligned DMA is high
2705 * and this cost outweighs the gains made by aligning the IP header.
2706 *
2707 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2708 * to be overridden.
2709 */
2710 #ifndef NET_IP_ALIGN
2711 #define NET_IP_ALIGN 2
2712 #endif
2713
2714 /*
2715 * The networking layer reserves some headroom in skb data (via
2716 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2717 * the header has to grow. In the default case, if the header has to grow
2718 * 32 bytes or less we avoid the reallocation.
2719 *
2720 * Unfortunately this headroom changes the DMA alignment of the resulting
2721 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2722 * on some architectures. An architecture can override this value,
2723 * perhaps setting it to a cacheline in size (since that will maintain
2724 * cacheline alignment of the DMA). It must be a power of 2.
2725 *
2726 * Various parts of the networking layer expect at least 32 bytes of
2727 * headroom, you should not reduce this.
2728 *
2729 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2730 * to reduce average number of cache lines per packet.
2731 * get_rps_cpu() for example only access one 64 bytes aligned block :
2732 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2733 */
2734 #ifndef NET_SKB_PAD
2735 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2736 #endif
2737
2738 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2739
__skb_set_length(struct sk_buff * skb,unsigned int len)2740 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2741 {
2742 if (WARN_ON(skb_is_nonlinear(skb)))
2743 return;
2744 skb->len = len;
2745 skb_set_tail_pointer(skb, len);
2746 }
2747
__skb_trim(struct sk_buff * skb,unsigned int len)2748 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2749 {
2750 __skb_set_length(skb, len);
2751 }
2752
2753 void skb_trim(struct sk_buff *skb, unsigned int len);
2754
__pskb_trim(struct sk_buff * skb,unsigned int len)2755 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2756 {
2757 if (skb->data_len)
2758 return ___pskb_trim(skb, len);
2759 __skb_trim(skb, len);
2760 return 0;
2761 }
2762
pskb_trim(struct sk_buff * skb,unsigned int len)2763 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2764 {
2765 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2766 }
2767
2768 /**
2769 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2770 * @skb: buffer to alter
2771 * @len: new length
2772 *
2773 * This is identical to pskb_trim except that the caller knows that
2774 * the skb is not cloned so we should never get an error due to out-
2775 * of-memory.
2776 */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)2777 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2778 {
2779 int err = pskb_trim(skb, len);
2780 BUG_ON(err);
2781 }
2782
__skb_grow(struct sk_buff * skb,unsigned int len)2783 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2784 {
2785 unsigned int diff = len - skb->len;
2786
2787 if (skb_tailroom(skb) < diff) {
2788 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2789 GFP_ATOMIC);
2790 if (ret)
2791 return ret;
2792 }
2793 __skb_set_length(skb, len);
2794 return 0;
2795 }
2796
2797 /**
2798 * skb_orphan - orphan a buffer
2799 * @skb: buffer to orphan
2800 *
2801 * If a buffer currently has an owner then we call the owner's
2802 * destructor function and make the @skb unowned. The buffer continues
2803 * to exist but is no longer charged to its former owner.
2804 */
skb_orphan(struct sk_buff * skb)2805 static inline void skb_orphan(struct sk_buff *skb)
2806 {
2807 if (skb->destructor) {
2808 skb->destructor(skb);
2809 skb->destructor = NULL;
2810 skb->sk = NULL;
2811 } else {
2812 BUG_ON(skb->sk);
2813 }
2814 }
2815
2816 /**
2817 * skb_orphan_frags - orphan the frags contained in a buffer
2818 * @skb: buffer to orphan frags from
2819 * @gfp_mask: allocation mask for replacement pages
2820 *
2821 * For each frag in the SKB which needs a destructor (i.e. has an
2822 * owner) create a copy of that frag and release the original
2823 * page by calling the destructor.
2824 */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)2825 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2826 {
2827 if (likely(!skb_zcopy(skb)))
2828 return 0;
2829 if (!skb_zcopy_is_nouarg(skb) &&
2830 skb_uarg(skb)->callback == sock_zerocopy_callback)
2831 return 0;
2832 return skb_copy_ubufs(skb, gfp_mask);
2833 }
2834
2835 /* 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)2836 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2837 {
2838 if (likely(!skb_zcopy(skb)))
2839 return 0;
2840 return skb_copy_ubufs(skb, gfp_mask);
2841 }
2842
2843 /**
2844 * __skb_queue_purge - empty a list
2845 * @list: list to empty
2846 *
2847 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2848 * the list and one reference dropped. This function does not take the
2849 * list lock and the caller must hold the relevant locks to use it.
2850 */
__skb_queue_purge(struct sk_buff_head * list)2851 static inline void __skb_queue_purge(struct sk_buff_head *list)
2852 {
2853 struct sk_buff *skb;
2854 while ((skb = __skb_dequeue(list)) != NULL)
2855 kfree_skb(skb);
2856 }
2857 void skb_queue_purge(struct sk_buff_head *list);
2858
2859 unsigned int skb_rbtree_purge(struct rb_root *root);
2860
2861 void *netdev_alloc_frag(unsigned int fragsz);
2862
2863 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2864 gfp_t gfp_mask);
2865
2866 /**
2867 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2868 * @dev: network device to receive on
2869 * @length: length to allocate
2870 *
2871 * Allocate a new &sk_buff and assign it a usage count of one. The
2872 * buffer has unspecified headroom built in. Users should allocate
2873 * the headroom they think they need without accounting for the
2874 * built in space. The built in space is used for optimisations.
2875 *
2876 * %NULL is returned if there is no free memory. Although this function
2877 * allocates memory it can be called from an interrupt.
2878 */
netdev_alloc_skb(struct net_device * dev,unsigned int length)2879 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2880 unsigned int length)
2881 {
2882 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2883 }
2884
2885 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)2886 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2887 gfp_t gfp_mask)
2888 {
2889 return __netdev_alloc_skb(NULL, length, gfp_mask);
2890 }
2891
2892 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)2893 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2894 {
2895 return netdev_alloc_skb(NULL, length);
2896 }
2897
2898
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)2899 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2900 unsigned int length, gfp_t gfp)
2901 {
2902 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2903
2904 if (NET_IP_ALIGN && skb)
2905 skb_reserve(skb, NET_IP_ALIGN);
2906 return skb;
2907 }
2908
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)2909 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2910 unsigned int length)
2911 {
2912 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2913 }
2914
skb_free_frag(void * addr)2915 static inline void skb_free_frag(void *addr)
2916 {
2917 page_frag_free(addr);
2918 }
2919
2920 void *napi_alloc_frag(unsigned int fragsz);
2921 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2922 unsigned int length, gfp_t gfp_mask);
napi_alloc_skb(struct napi_struct * napi,unsigned int length)2923 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2924 unsigned int length)
2925 {
2926 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2927 }
2928 void napi_consume_skb(struct sk_buff *skb, int budget);
2929
2930 void __kfree_skb_flush(void);
2931 void __kfree_skb_defer(struct sk_buff *skb);
2932
2933 /**
2934 * __dev_alloc_pages - allocate page for network Rx
2935 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2936 * @order: size of the allocation
2937 *
2938 * Allocate a new page.
2939 *
2940 * %NULL is returned if there is no free memory.
2941 */
__dev_alloc_pages(gfp_t gfp_mask,unsigned int order)2942 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2943 unsigned int order)
2944 {
2945 /* This piece of code contains several assumptions.
2946 * 1. This is for device Rx, therefor a cold page is preferred.
2947 * 2. The expectation is the user wants a compound page.
2948 * 3. If requesting a order 0 page it will not be compound
2949 * due to the check to see if order has a value in prep_new_page
2950 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2951 * code in gfp_to_alloc_flags that should be enforcing this.
2952 */
2953 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2954
2955 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2956 }
2957
dev_alloc_pages(unsigned int order)2958 static inline struct page *dev_alloc_pages(unsigned int order)
2959 {
2960 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2961 }
2962
2963 /**
2964 * __dev_alloc_page - allocate a page for network Rx
2965 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2966 *
2967 * Allocate a new page.
2968 *
2969 * %NULL is returned if there is no free memory.
2970 */
__dev_alloc_page(gfp_t gfp_mask)2971 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2972 {
2973 return __dev_alloc_pages(gfp_mask, 0);
2974 }
2975
dev_alloc_page(void)2976 static inline struct page *dev_alloc_page(void)
2977 {
2978 return dev_alloc_pages(0);
2979 }
2980
2981 /**
2982 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2983 * @page: The page that was allocated from skb_alloc_page
2984 * @skb: The skb that may need pfmemalloc set
2985 */
skb_propagate_pfmemalloc(struct page * page,struct sk_buff * skb)2986 static inline void skb_propagate_pfmemalloc(struct page *page,
2987 struct sk_buff *skb)
2988 {
2989 if (page_is_pfmemalloc(page))
2990 skb->pfmemalloc = true;
2991 }
2992
2993 /**
2994 * skb_frag_off() - Returns the offset of a skb fragment
2995 * @frag: the paged fragment
2996 */
skb_frag_off(const skb_frag_t * frag)2997 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2998 {
2999 return frag->bv_offset;
3000 }
3001
3002 /**
3003 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3004 * @frag: skb fragment
3005 * @delta: value to add
3006 */
skb_frag_off_add(skb_frag_t * frag,int delta)3007 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3008 {
3009 frag->bv_offset += delta;
3010 }
3011
3012 /**
3013 * skb_frag_off_set() - Sets the offset of a skb fragment
3014 * @frag: skb fragment
3015 * @offset: offset of fragment
3016 */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)3017 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3018 {
3019 frag->bv_offset = offset;
3020 }
3021
3022 /**
3023 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3024 * @fragto: skb fragment where offset is set
3025 * @fragfrom: skb fragment offset is copied from
3026 */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3027 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3028 const skb_frag_t *fragfrom)
3029 {
3030 fragto->bv_offset = fragfrom->bv_offset;
3031 }
3032
3033 /**
3034 * skb_frag_page - retrieve the page referred to by a paged fragment
3035 * @frag: the paged fragment
3036 *
3037 * Returns the &struct page associated with @frag.
3038 */
skb_frag_page(const skb_frag_t * frag)3039 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3040 {
3041 return frag->bv_page;
3042 }
3043
3044 /**
3045 * __skb_frag_ref - take an addition reference on a paged fragment.
3046 * @frag: the paged fragment
3047 *
3048 * Takes an additional reference on the paged fragment @frag.
3049 */
__skb_frag_ref(skb_frag_t * frag)3050 static inline void __skb_frag_ref(skb_frag_t *frag)
3051 {
3052 get_page(skb_frag_page(frag));
3053 }
3054
3055 /**
3056 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3057 * @skb: the buffer
3058 * @f: the fragment offset.
3059 *
3060 * Takes an additional reference on the @f'th paged fragment of @skb.
3061 */
skb_frag_ref(struct sk_buff * skb,int f)3062 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3063 {
3064 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3065 }
3066
3067 /**
3068 * __skb_frag_unref - release a reference on a paged fragment.
3069 * @frag: the paged fragment
3070 *
3071 * Releases a reference on the paged fragment @frag.
3072 */
__skb_frag_unref(skb_frag_t * frag)3073 static inline void __skb_frag_unref(skb_frag_t *frag)
3074 {
3075 put_page(skb_frag_page(frag));
3076 }
3077
3078 /**
3079 * skb_frag_unref - release a reference on a paged fragment of an skb.
3080 * @skb: the buffer
3081 * @f: the fragment offset
3082 *
3083 * Releases a reference on the @f'th paged fragment of @skb.
3084 */
skb_frag_unref(struct sk_buff * skb,int f)3085 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3086 {
3087 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3088 }
3089
3090 /**
3091 * skb_frag_address - gets the address of the data contained in a paged fragment
3092 * @frag: the paged fragment buffer
3093 *
3094 * Returns the address of the data within @frag. The page must already
3095 * be mapped.
3096 */
skb_frag_address(const skb_frag_t * frag)3097 static inline void *skb_frag_address(const skb_frag_t *frag)
3098 {
3099 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3100 }
3101
3102 /**
3103 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3104 * @frag: the paged fragment buffer
3105 *
3106 * Returns the address of the data within @frag. Checks that the page
3107 * is mapped and returns %NULL otherwise.
3108 */
skb_frag_address_safe(const skb_frag_t * frag)3109 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3110 {
3111 void *ptr = page_address(skb_frag_page(frag));
3112 if (unlikely(!ptr))
3113 return NULL;
3114
3115 return ptr + skb_frag_off(frag);
3116 }
3117
3118 /**
3119 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3120 * @fragto: skb fragment where page is set
3121 * @fragfrom: skb fragment page is copied from
3122 */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3123 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3124 const skb_frag_t *fragfrom)
3125 {
3126 fragto->bv_page = fragfrom->bv_page;
3127 }
3128
3129 /**
3130 * __skb_frag_set_page - sets the page contained in a paged fragment
3131 * @frag: the paged fragment
3132 * @page: the page to set
3133 *
3134 * Sets the fragment @frag to contain @page.
3135 */
__skb_frag_set_page(skb_frag_t * frag,struct page * page)3136 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3137 {
3138 frag->bv_page = page;
3139 }
3140
3141 /**
3142 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3143 * @skb: the buffer
3144 * @f: the fragment offset
3145 * @page: the page to set
3146 *
3147 * Sets the @f'th fragment of @skb to contain @page.
3148 */
skb_frag_set_page(struct sk_buff * skb,int f,struct page * page)3149 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3150 struct page *page)
3151 {
3152 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3153 }
3154
3155 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3156
3157 /**
3158 * skb_frag_dma_map - maps a paged fragment via the DMA API
3159 * @dev: the device to map the fragment to
3160 * @frag: the paged fragment to map
3161 * @offset: the offset within the fragment (starting at the
3162 * fragment's own offset)
3163 * @size: the number of bytes to map
3164 * @dir: the direction of the mapping (``PCI_DMA_*``)
3165 *
3166 * Maps the page associated with @frag to @device.
3167 */
skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3168 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3169 const skb_frag_t *frag,
3170 size_t offset, size_t size,
3171 enum dma_data_direction dir)
3172 {
3173 return dma_map_page(dev, skb_frag_page(frag),
3174 skb_frag_off(frag) + offset, size, dir);
3175 }
3176
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3177 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3178 gfp_t gfp_mask)
3179 {
3180 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3181 }
3182
3183
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3184 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3185 gfp_t gfp_mask)
3186 {
3187 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3188 }
3189
3190
3191 /**
3192 * skb_clone_writable - is the header of a clone writable
3193 * @skb: buffer to check
3194 * @len: length up to which to write
3195 *
3196 * Returns true if modifying the header part of the cloned buffer
3197 * does not requires the data to be copied.
3198 */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3199 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3200 {
3201 return !skb_header_cloned(skb) &&
3202 skb_headroom(skb) + len <= skb->hdr_len;
3203 }
3204
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3205 static inline int skb_try_make_writable(struct sk_buff *skb,
3206 unsigned int write_len)
3207 {
3208 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3209 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3210 }
3211
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3212 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3213 int cloned)
3214 {
3215 int delta = 0;
3216
3217 if (headroom > skb_headroom(skb))
3218 delta = headroom - skb_headroom(skb);
3219
3220 if (delta || cloned)
3221 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3222 GFP_ATOMIC);
3223 return 0;
3224 }
3225
3226 /**
3227 * skb_cow - copy header of skb when it is required
3228 * @skb: buffer to cow
3229 * @headroom: needed headroom
3230 *
3231 * If the skb passed lacks sufficient headroom or its data part
3232 * is shared, data is reallocated. If reallocation fails, an error
3233 * is returned and original skb is not changed.
3234 *
3235 * The result is skb with writable area skb->head...skb->tail
3236 * and at least @headroom of space at head.
3237 */
skb_cow(struct sk_buff * skb,unsigned int headroom)3238 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3239 {
3240 return __skb_cow(skb, headroom, skb_cloned(skb));
3241 }
3242
3243 /**
3244 * skb_cow_head - skb_cow but only making the head writable
3245 * @skb: buffer to cow
3246 * @headroom: needed headroom
3247 *
3248 * This function is identical to skb_cow except that we replace the
3249 * skb_cloned check by skb_header_cloned. It should be used when
3250 * you only need to push on some header and do not need to modify
3251 * the data.
3252 */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3253 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3254 {
3255 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3256 }
3257
3258 /**
3259 * skb_padto - pad an skbuff up to a minimal size
3260 * @skb: buffer to pad
3261 * @len: minimal length
3262 *
3263 * Pads up a buffer to ensure the trailing bytes exist and are
3264 * blanked. If the buffer already contains sufficient data it
3265 * is untouched. Otherwise it is extended. Returns zero on
3266 * success. The skb is freed on error.
3267 */
skb_padto(struct sk_buff * skb,unsigned int len)3268 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3269 {
3270 unsigned int size = skb->len;
3271 if (likely(size >= len))
3272 return 0;
3273 return skb_pad(skb, len - size);
3274 }
3275
3276 /**
3277 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3278 * @skb: buffer to pad
3279 * @len: minimal length
3280 * @free_on_error: free buffer on error
3281 *
3282 * Pads up a buffer to ensure the trailing bytes exist and are
3283 * blanked. If the buffer already contains sufficient data it
3284 * is untouched. Otherwise it is extended. Returns zero on
3285 * success. The skb is freed on error if @free_on_error is true.
3286 */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3287 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3288 unsigned int len,
3289 bool free_on_error)
3290 {
3291 unsigned int size = skb->len;
3292
3293 if (unlikely(size < len)) {
3294 len -= size;
3295 if (__skb_pad(skb, len, free_on_error))
3296 return -ENOMEM;
3297 __skb_put(skb, len);
3298 }
3299 return 0;
3300 }
3301
3302 /**
3303 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3304 * @skb: buffer to pad
3305 * @len: minimal length
3306 *
3307 * Pads up a buffer to ensure the trailing bytes exist and are
3308 * blanked. If the buffer already contains sufficient data it
3309 * is untouched. Otherwise it is extended. Returns zero on
3310 * success. The skb is freed on error.
3311 */
skb_put_padto(struct sk_buff * skb,unsigned int len)3312 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3313 {
3314 return __skb_put_padto(skb, len, true);
3315 }
3316
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3317 static inline int skb_add_data(struct sk_buff *skb,
3318 struct iov_iter *from, int copy)
3319 {
3320 const int off = skb->len;
3321
3322 if (skb->ip_summed == CHECKSUM_NONE) {
3323 __wsum csum = 0;
3324 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3325 &csum, from)) {
3326 skb->csum = csum_block_add(skb->csum, csum, off);
3327 return 0;
3328 }
3329 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3330 return 0;
3331
3332 __skb_trim(skb, off);
3333 return -EFAULT;
3334 }
3335
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3336 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3337 const struct page *page, int off)
3338 {
3339 if (skb_zcopy(skb))
3340 return false;
3341 if (i) {
3342 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3343
3344 return page == skb_frag_page(frag) &&
3345 off == skb_frag_off(frag) + skb_frag_size(frag);
3346 }
3347 return false;
3348 }
3349
__skb_linearize(struct sk_buff * skb)3350 static inline int __skb_linearize(struct sk_buff *skb)
3351 {
3352 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3353 }
3354
3355 /**
3356 * skb_linearize - convert paged skb to linear one
3357 * @skb: buffer to linarize
3358 *
3359 * If there is no free memory -ENOMEM is returned, otherwise zero
3360 * is returned and the old skb data released.
3361 */
skb_linearize(struct sk_buff * skb)3362 static inline int skb_linearize(struct sk_buff *skb)
3363 {
3364 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3365 }
3366
3367 /**
3368 * skb_has_shared_frag - can any frag be overwritten
3369 * @skb: buffer to test
3370 *
3371 * Return true if the skb has at least one frag that might be modified
3372 * by an external entity (as in vmsplice()/sendfile())
3373 */
skb_has_shared_frag(const struct sk_buff * skb)3374 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3375 {
3376 return skb_is_nonlinear(skb) &&
3377 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3378 }
3379
3380 /**
3381 * skb_linearize_cow - make sure skb is linear and writable
3382 * @skb: buffer to process
3383 *
3384 * If there is no free memory -ENOMEM is returned, otherwise zero
3385 * is returned and the old skb data released.
3386 */
skb_linearize_cow(struct sk_buff * skb)3387 static inline int skb_linearize_cow(struct sk_buff *skb)
3388 {
3389 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3390 __skb_linearize(skb) : 0;
3391 }
3392
3393 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3394 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3395 unsigned int off)
3396 {
3397 if (skb->ip_summed == CHECKSUM_COMPLETE)
3398 skb->csum = csum_block_sub(skb->csum,
3399 csum_partial(start, len, 0), off);
3400 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3401 skb_checksum_start_offset(skb) < 0)
3402 skb->ip_summed = CHECKSUM_NONE;
3403 }
3404
3405 /**
3406 * skb_postpull_rcsum - update checksum for received skb after pull
3407 * @skb: buffer to update
3408 * @start: start of data before pull
3409 * @len: length of data pulled
3410 *
3411 * After doing a pull on a received packet, you need to call this to
3412 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3413 * CHECKSUM_NONE so that it can be recomputed from scratch.
3414 */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3415 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3416 const void *start, unsigned int len)
3417 {
3418 __skb_postpull_rcsum(skb, start, len, 0);
3419 }
3420
3421 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3422 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3423 unsigned int off)
3424 {
3425 if (skb->ip_summed == CHECKSUM_COMPLETE)
3426 skb->csum = csum_block_add(skb->csum,
3427 csum_partial(start, len, 0), off);
3428 }
3429
3430 /**
3431 * skb_postpush_rcsum - update checksum for received skb after push
3432 * @skb: buffer to update
3433 * @start: start of data after push
3434 * @len: length of data pushed
3435 *
3436 * After doing a push on a received packet, you need to call this to
3437 * update the CHECKSUM_COMPLETE checksum.
3438 */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3439 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3440 const void *start, unsigned int len)
3441 {
3442 __skb_postpush_rcsum(skb, start, len, 0);
3443 }
3444
3445 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3446
3447 /**
3448 * skb_push_rcsum - push skb and update receive checksum
3449 * @skb: buffer to update
3450 * @len: length of data pulled
3451 *
3452 * This function performs an skb_push on the packet and updates
3453 * the CHECKSUM_COMPLETE checksum. It should be used on
3454 * receive path processing instead of skb_push unless you know
3455 * that the checksum difference is zero (e.g., a valid IP header)
3456 * or you are setting ip_summed to CHECKSUM_NONE.
3457 */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3458 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3459 {
3460 skb_push(skb, len);
3461 skb_postpush_rcsum(skb, skb->data, len);
3462 return skb->data;
3463 }
3464
3465 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3466 /**
3467 * pskb_trim_rcsum - trim received skb and update checksum
3468 * @skb: buffer to trim
3469 * @len: new length
3470 *
3471 * This is exactly the same as pskb_trim except that it ensures the
3472 * checksum of received packets are still valid after the operation.
3473 * It can change skb pointers.
3474 */
3475
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)3476 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3477 {
3478 if (likely(len >= skb->len))
3479 return 0;
3480 return pskb_trim_rcsum_slow(skb, len);
3481 }
3482
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)3483 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3484 {
3485 if (skb->ip_summed == CHECKSUM_COMPLETE)
3486 skb->ip_summed = CHECKSUM_NONE;
3487 __skb_trim(skb, len);
3488 return 0;
3489 }
3490
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)3491 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3492 {
3493 if (skb->ip_summed == CHECKSUM_COMPLETE)
3494 skb->ip_summed = CHECKSUM_NONE;
3495 return __skb_grow(skb, len);
3496 }
3497
3498 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3499 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3500 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3501 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3502 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3503
3504 #define skb_queue_walk(queue, skb) \
3505 for (skb = (queue)->next; \
3506 skb != (struct sk_buff *)(queue); \
3507 skb = skb->next)
3508
3509 #define skb_queue_walk_safe(queue, skb, tmp) \
3510 for (skb = (queue)->next, tmp = skb->next; \
3511 skb != (struct sk_buff *)(queue); \
3512 skb = tmp, tmp = skb->next)
3513
3514 #define skb_queue_walk_from(queue, skb) \
3515 for (; skb != (struct sk_buff *)(queue); \
3516 skb = skb->next)
3517
3518 #define skb_rbtree_walk(skb, root) \
3519 for (skb = skb_rb_first(root); skb != NULL; \
3520 skb = skb_rb_next(skb))
3521
3522 #define skb_rbtree_walk_from(skb) \
3523 for (; skb != NULL; \
3524 skb = skb_rb_next(skb))
3525
3526 #define skb_rbtree_walk_from_safe(skb, tmp) \
3527 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3528 skb = tmp)
3529
3530 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3531 for (tmp = skb->next; \
3532 skb != (struct sk_buff *)(queue); \
3533 skb = tmp, tmp = skb->next)
3534
3535 #define skb_queue_reverse_walk(queue, skb) \
3536 for (skb = (queue)->prev; \
3537 skb != (struct sk_buff *)(queue); \
3538 skb = skb->prev)
3539
3540 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3541 for (skb = (queue)->prev, tmp = skb->prev; \
3542 skb != (struct sk_buff *)(queue); \
3543 skb = tmp, tmp = skb->prev)
3544
3545 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3546 for (tmp = skb->prev; \
3547 skb != (struct sk_buff *)(queue); \
3548 skb = tmp, tmp = skb->prev)
3549
skb_has_frag_list(const struct sk_buff * skb)3550 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3551 {
3552 return skb_shinfo(skb)->frag_list != NULL;
3553 }
3554
skb_frag_list_init(struct sk_buff * skb)3555 static inline void skb_frag_list_init(struct sk_buff *skb)
3556 {
3557 skb_shinfo(skb)->frag_list = NULL;
3558 }
3559
3560 #define skb_walk_frags(skb, iter) \
3561 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3562
3563
3564 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3565 int *err, long *timeo_p,
3566 const struct sk_buff *skb);
3567 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3568 struct sk_buff_head *queue,
3569 unsigned int flags,
3570 int *off, int *err,
3571 struct sk_buff **last);
3572 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3573 struct sk_buff_head *queue,
3574 unsigned int flags, int *off, int *err,
3575 struct sk_buff **last);
3576 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3577 struct sk_buff_head *sk_queue,
3578 unsigned int flags, int *off, int *err);
3579 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3580 int *err);
3581 __poll_t datagram_poll(struct file *file, struct socket *sock,
3582 struct poll_table_struct *wait);
3583 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3584 struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)3585 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3586 struct msghdr *msg, int size)
3587 {
3588 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3589 }
3590 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3591 struct msghdr *msg);
3592 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3593 struct iov_iter *to, int len,
3594 struct ahash_request *hash);
3595 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3596 struct iov_iter *from, int len);
3597 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3598 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3599 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)3600 static inline void skb_free_datagram_locked(struct sock *sk,
3601 struct sk_buff *skb)
3602 {
3603 __skb_free_datagram_locked(sk, skb, 0);
3604 }
3605 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3606 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3607 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3608 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3609 int len);
3610 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3611 struct pipe_inode_info *pipe, unsigned int len,
3612 unsigned int flags);
3613 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3614 int len);
3615 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3616 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3617 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3618 int len, int hlen);
3619 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3620 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3621 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3622 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3623 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3624 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3625 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3626 unsigned int offset);
3627 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3628 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3629 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3630 int skb_vlan_pop(struct sk_buff *skb);
3631 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3632 int skb_eth_pop(struct sk_buff *skb);
3633 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3634 const unsigned char *src);
3635 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3636 int mac_len, bool ethernet);
3637 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3638 bool ethernet);
3639 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3640 int skb_mpls_dec_ttl(struct sk_buff *skb);
3641 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3642 gfp_t gfp);
3643
memcpy_from_msg(void * data,struct msghdr * msg,int len)3644 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3645 {
3646 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3647 }
3648
memcpy_to_msg(struct msghdr * msg,void * data,int len)3649 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3650 {
3651 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3652 }
3653
3654 struct skb_checksum_ops {
3655 __wsum (*update)(const void *mem, int len, __wsum wsum);
3656 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3657 };
3658
3659 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3660
3661 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3662 __wsum csum, const struct skb_checksum_ops *ops);
3663 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3664 __wsum csum);
3665
3666 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * data,int hlen,void * buffer)3667 __skb_header_pointer(const struct sk_buff *skb, int offset,
3668 int len, void *data, int hlen, void *buffer)
3669 {
3670 if (hlen - offset >= len)
3671 return data + offset;
3672
3673 if (!skb ||
3674 skb_copy_bits(skb, offset, buffer, len) < 0)
3675 return NULL;
3676
3677 return buffer;
3678 }
3679
3680 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)3681 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3682 {
3683 return __skb_header_pointer(skb, offset, len, skb->data,
3684 skb_headlen(skb), buffer);
3685 }
3686
3687 /**
3688 * skb_needs_linearize - check if we need to linearize a given skb
3689 * depending on the given device features.
3690 * @skb: socket buffer to check
3691 * @features: net device features
3692 *
3693 * Returns true if either:
3694 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3695 * 2. skb is fragmented and the device does not support SG.
3696 */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)3697 static inline bool skb_needs_linearize(struct sk_buff *skb,
3698 netdev_features_t features)
3699 {
3700 return skb_is_nonlinear(skb) &&
3701 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3702 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3703 }
3704
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)3705 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3706 void *to,
3707 const unsigned int len)
3708 {
3709 memcpy(to, skb->data, len);
3710 }
3711
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)3712 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3713 const int offset, void *to,
3714 const unsigned int len)
3715 {
3716 memcpy(to, skb->data + offset, len);
3717 }
3718
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)3719 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3720 const void *from,
3721 const unsigned int len)
3722 {
3723 memcpy(skb->data, from, len);
3724 }
3725
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)3726 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3727 const int offset,
3728 const void *from,
3729 const unsigned int len)
3730 {
3731 memcpy(skb->data + offset, from, len);
3732 }
3733
3734 void skb_init(void);
3735
skb_get_ktime(const struct sk_buff * skb)3736 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3737 {
3738 return skb->tstamp;
3739 }
3740
3741 /**
3742 * skb_get_timestamp - get timestamp from a skb
3743 * @skb: skb to get stamp from
3744 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3745 *
3746 * Timestamps are stored in the skb as offsets to a base timestamp.
3747 * This function converts the offset back to a struct timeval and stores
3748 * it in stamp.
3749 */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)3750 static inline void skb_get_timestamp(const struct sk_buff *skb,
3751 struct __kernel_old_timeval *stamp)
3752 {
3753 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3754 }
3755
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)3756 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3757 struct __kernel_sock_timeval *stamp)
3758 {
3759 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3760
3761 stamp->tv_sec = ts.tv_sec;
3762 stamp->tv_usec = ts.tv_nsec / 1000;
3763 }
3764
skb_get_timestampns(const struct sk_buff * skb,struct __kernel_old_timespec * stamp)3765 static inline void skb_get_timestampns(const struct sk_buff *skb,
3766 struct __kernel_old_timespec *stamp)
3767 {
3768 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3769
3770 stamp->tv_sec = ts.tv_sec;
3771 stamp->tv_nsec = ts.tv_nsec;
3772 }
3773
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)3774 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3775 struct __kernel_timespec *stamp)
3776 {
3777 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3778
3779 stamp->tv_sec = ts.tv_sec;
3780 stamp->tv_nsec = ts.tv_nsec;
3781 }
3782
__net_timestamp(struct sk_buff * skb)3783 static inline void __net_timestamp(struct sk_buff *skb)
3784 {
3785 skb->tstamp = ktime_get_real();
3786 }
3787
net_timedelta(ktime_t t)3788 static inline ktime_t net_timedelta(ktime_t t)
3789 {
3790 return ktime_sub(ktime_get_real(), t);
3791 }
3792
net_invalid_timestamp(void)3793 static inline ktime_t net_invalid_timestamp(void)
3794 {
3795 return 0;
3796 }
3797
skb_metadata_len(const struct sk_buff * skb)3798 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3799 {
3800 return skb_shinfo(skb)->meta_len;
3801 }
3802
skb_metadata_end(const struct sk_buff * skb)3803 static inline void *skb_metadata_end(const struct sk_buff *skb)
3804 {
3805 return skb_mac_header(skb);
3806 }
3807
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)3808 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3809 const struct sk_buff *skb_b,
3810 u8 meta_len)
3811 {
3812 const void *a = skb_metadata_end(skb_a);
3813 const void *b = skb_metadata_end(skb_b);
3814 /* Using more efficient varaiant than plain call to memcmp(). */
3815 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3816 u64 diffs = 0;
3817
3818 switch (meta_len) {
3819 #define __it(x, op) (x -= sizeof(u##op))
3820 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3821 case 32: diffs |= __it_diff(a, b, 64);
3822 fallthrough;
3823 case 24: diffs |= __it_diff(a, b, 64);
3824 fallthrough;
3825 case 16: diffs |= __it_diff(a, b, 64);
3826 fallthrough;
3827 case 8: diffs |= __it_diff(a, b, 64);
3828 break;
3829 case 28: diffs |= __it_diff(a, b, 64);
3830 fallthrough;
3831 case 20: diffs |= __it_diff(a, b, 64);
3832 fallthrough;
3833 case 12: diffs |= __it_diff(a, b, 64);
3834 fallthrough;
3835 case 4: diffs |= __it_diff(a, b, 32);
3836 break;
3837 }
3838 return diffs;
3839 #else
3840 return memcmp(a - meta_len, b - meta_len, meta_len);
3841 #endif
3842 }
3843
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)3844 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3845 const struct sk_buff *skb_b)
3846 {
3847 u8 len_a = skb_metadata_len(skb_a);
3848 u8 len_b = skb_metadata_len(skb_b);
3849
3850 if (!(len_a | len_b))
3851 return false;
3852
3853 return len_a != len_b ?
3854 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3855 }
3856
skb_metadata_set(struct sk_buff * skb,u8 meta_len)3857 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3858 {
3859 skb_shinfo(skb)->meta_len = meta_len;
3860 }
3861
skb_metadata_clear(struct sk_buff * skb)3862 static inline void skb_metadata_clear(struct sk_buff *skb)
3863 {
3864 skb_metadata_set(skb, 0);
3865 }
3866
3867 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3868
3869 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3870
3871 void skb_clone_tx_timestamp(struct sk_buff *skb);
3872 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3873
3874 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3875
skb_clone_tx_timestamp(struct sk_buff * skb)3876 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3877 {
3878 }
3879
skb_defer_rx_timestamp(struct sk_buff * skb)3880 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3881 {
3882 return false;
3883 }
3884
3885 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3886
3887 /**
3888 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3889 *
3890 * PHY drivers may accept clones of transmitted packets for
3891 * timestamping via their phy_driver.txtstamp method. These drivers
3892 * must call this function to return the skb back to the stack with a
3893 * timestamp.
3894 *
3895 * @skb: clone of the original outgoing packet
3896 * @hwtstamps: hardware time stamps
3897 *
3898 */
3899 void skb_complete_tx_timestamp(struct sk_buff *skb,
3900 struct skb_shared_hwtstamps *hwtstamps);
3901
3902 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3903 struct skb_shared_hwtstamps *hwtstamps,
3904 struct sock *sk, int tstype);
3905
3906 /**
3907 * skb_tstamp_tx - queue clone of skb with send time stamps
3908 * @orig_skb: the original outgoing packet
3909 * @hwtstamps: hardware time stamps, may be NULL if not available
3910 *
3911 * If the skb has a socket associated, then this function clones the
3912 * skb (thus sharing the actual data and optional structures), stores
3913 * the optional hardware time stamping information (if non NULL) or
3914 * generates a software time stamp (otherwise), then queues the clone
3915 * to the error queue of the socket. Errors are silently ignored.
3916 */
3917 void skb_tstamp_tx(struct sk_buff *orig_skb,
3918 struct skb_shared_hwtstamps *hwtstamps);
3919
3920 /**
3921 * skb_tx_timestamp() - Driver hook for transmit timestamping
3922 *
3923 * Ethernet MAC Drivers should call this function in their hard_xmit()
3924 * function immediately before giving the sk_buff to the MAC hardware.
3925 *
3926 * Specifically, one should make absolutely sure that this function is
3927 * called before TX completion of this packet can trigger. Otherwise
3928 * the packet could potentially already be freed.
3929 *
3930 * @skb: A socket buffer.
3931 */
skb_tx_timestamp(struct sk_buff * skb)3932 static inline void skb_tx_timestamp(struct sk_buff *skb)
3933 {
3934 skb_clone_tx_timestamp(skb);
3935 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3936 skb_tstamp_tx(skb, NULL);
3937 }
3938
3939 /**
3940 * skb_complete_wifi_ack - deliver skb with wifi status
3941 *
3942 * @skb: the original outgoing packet
3943 * @acked: ack status
3944 *
3945 */
3946 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3947
3948 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3949 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3950
skb_csum_unnecessary(const struct sk_buff * skb)3951 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3952 {
3953 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3954 skb->csum_valid ||
3955 (skb->ip_summed == CHECKSUM_PARTIAL &&
3956 skb_checksum_start_offset(skb) >= 0));
3957 }
3958
3959 /**
3960 * skb_checksum_complete - Calculate checksum of an entire packet
3961 * @skb: packet to process
3962 *
3963 * This function calculates the checksum over the entire packet plus
3964 * the value of skb->csum. The latter can be used to supply the
3965 * checksum of a pseudo header as used by TCP/UDP. It returns the
3966 * checksum.
3967 *
3968 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3969 * this function can be used to verify that checksum on received
3970 * packets. In that case the function should return zero if the
3971 * checksum is correct. In particular, this function will return zero
3972 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3973 * hardware has already verified the correctness of the checksum.
3974 */
skb_checksum_complete(struct sk_buff * skb)3975 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3976 {
3977 return skb_csum_unnecessary(skb) ?
3978 0 : __skb_checksum_complete(skb);
3979 }
3980
__skb_decr_checksum_unnecessary(struct sk_buff * skb)3981 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3982 {
3983 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3984 if (skb->csum_level == 0)
3985 skb->ip_summed = CHECKSUM_NONE;
3986 else
3987 skb->csum_level--;
3988 }
3989 }
3990
__skb_incr_checksum_unnecessary(struct sk_buff * skb)3991 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3992 {
3993 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3994 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3995 skb->csum_level++;
3996 } else if (skb->ip_summed == CHECKSUM_NONE) {
3997 skb->ip_summed = CHECKSUM_UNNECESSARY;
3998 skb->csum_level = 0;
3999 }
4000 }
4001
__skb_reset_checksum_unnecessary(struct sk_buff * skb)4002 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4003 {
4004 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4005 skb->ip_summed = CHECKSUM_NONE;
4006 skb->csum_level = 0;
4007 }
4008 }
4009
4010 /* Check if we need to perform checksum complete validation.
4011 *
4012 * Returns true if checksum complete is needed, false otherwise
4013 * (either checksum is unnecessary or zero checksum is allowed).
4014 */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)4015 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4016 bool zero_okay,
4017 __sum16 check)
4018 {
4019 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4020 skb->csum_valid = 1;
4021 __skb_decr_checksum_unnecessary(skb);
4022 return false;
4023 }
4024
4025 return true;
4026 }
4027
4028 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4029 * in checksum_init.
4030 */
4031 #define CHECKSUM_BREAK 76
4032
4033 /* Unset checksum-complete
4034 *
4035 * Unset checksum complete can be done when packet is being modified
4036 * (uncompressed for instance) and checksum-complete value is
4037 * invalidated.
4038 */
skb_checksum_complete_unset(struct sk_buff * skb)4039 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4040 {
4041 if (skb->ip_summed == CHECKSUM_COMPLETE)
4042 skb->ip_summed = CHECKSUM_NONE;
4043 }
4044
4045 /* Validate (init) checksum based on checksum complete.
4046 *
4047 * Return values:
4048 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4049 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4050 * checksum is stored in skb->csum for use in __skb_checksum_complete
4051 * non-zero: value of invalid checksum
4052 *
4053 */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)4054 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4055 bool complete,
4056 __wsum psum)
4057 {
4058 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4059 if (!csum_fold(csum_add(psum, skb->csum))) {
4060 skb->csum_valid = 1;
4061 return 0;
4062 }
4063 }
4064
4065 skb->csum = psum;
4066
4067 if (complete || skb->len <= CHECKSUM_BREAK) {
4068 __sum16 csum;
4069
4070 csum = __skb_checksum_complete(skb);
4071 skb->csum_valid = !csum;
4072 return csum;
4073 }
4074
4075 return 0;
4076 }
4077
null_compute_pseudo(struct sk_buff * skb,int proto)4078 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4079 {
4080 return 0;
4081 }
4082
4083 /* Perform checksum validate (init). Note that this is a macro since we only
4084 * want to calculate the pseudo header which is an input function if necessary.
4085 * First we try to validate without any computation (checksum unnecessary) and
4086 * then calculate based on checksum complete calling the function to compute
4087 * pseudo header.
4088 *
4089 * Return values:
4090 * 0: checksum is validated or try to in skb_checksum_complete
4091 * non-zero: value of invalid checksum
4092 */
4093 #define __skb_checksum_validate(skb, proto, complete, \
4094 zero_okay, check, compute_pseudo) \
4095 ({ \
4096 __sum16 __ret = 0; \
4097 skb->csum_valid = 0; \
4098 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4099 __ret = __skb_checksum_validate_complete(skb, \
4100 complete, compute_pseudo(skb, proto)); \
4101 __ret; \
4102 })
4103
4104 #define skb_checksum_init(skb, proto, compute_pseudo) \
4105 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4106
4107 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4108 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4109
4110 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4111 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4112
4113 #define skb_checksum_validate_zero_check(skb, proto, check, \
4114 compute_pseudo) \
4115 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4116
4117 #define skb_checksum_simple_validate(skb) \
4118 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4119
__skb_checksum_convert_check(struct sk_buff * skb)4120 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4121 {
4122 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4123 }
4124
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4125 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4126 {
4127 skb->csum = ~pseudo;
4128 skb->ip_summed = CHECKSUM_COMPLETE;
4129 }
4130
4131 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4132 do { \
4133 if (__skb_checksum_convert_check(skb)) \
4134 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4135 } while (0)
4136
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4137 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4138 u16 start, u16 offset)
4139 {
4140 skb->ip_summed = CHECKSUM_PARTIAL;
4141 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4142 skb->csum_offset = offset - start;
4143 }
4144
4145 /* Update skbuf and packet to reflect the remote checksum offload operation.
4146 * When called, ptr indicates the starting point for skb->csum when
4147 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4148 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4149 */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4150 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4151 int start, int offset, bool nopartial)
4152 {
4153 __wsum delta;
4154
4155 if (!nopartial) {
4156 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4157 return;
4158 }
4159
4160 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4161 __skb_checksum_complete(skb);
4162 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4163 }
4164
4165 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4166
4167 /* Adjust skb->csum since we changed the packet */
4168 skb->csum = csum_add(skb->csum, delta);
4169 }
4170
skb_nfct(const struct sk_buff * skb)4171 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4172 {
4173 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4174 return (void *)(skb->_nfct & NFCT_PTRMASK);
4175 #else
4176 return NULL;
4177 #endif
4178 }
4179
skb_get_nfct(const struct sk_buff * skb)4180 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4181 {
4182 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4183 return skb->_nfct;
4184 #else
4185 return 0UL;
4186 #endif
4187 }
4188
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4189 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4190 {
4191 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4192 skb->_nfct = nfct;
4193 #endif
4194 }
4195
4196 #ifdef CONFIG_SKB_EXTENSIONS
4197 enum skb_ext_id {
4198 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4199 SKB_EXT_BRIDGE_NF,
4200 #endif
4201 #ifdef CONFIG_XFRM
4202 SKB_EXT_SEC_PATH,
4203 #endif
4204 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4205 TC_SKB_EXT,
4206 #endif
4207 #if IS_ENABLED(CONFIG_MPTCP)
4208 SKB_EXT_MPTCP,
4209 #endif
4210 #if IS_ENABLED(CONFIG_KCOV)
4211 SKB_EXT_KCOV_HANDLE,
4212 #endif
4213 SKB_EXT_NUM, /* must be last */
4214 };
4215
4216 /**
4217 * struct skb_ext - sk_buff extensions
4218 * @refcnt: 1 on allocation, deallocated on 0
4219 * @offset: offset to add to @data to obtain extension address
4220 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4221 * @data: start of extension data, variable sized
4222 *
4223 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4224 * to use 'u8' types while allowing up to 2kb worth of extension data.
4225 */
4226 struct skb_ext {
4227 refcount_t refcnt;
4228 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4229 u8 chunks; /* same */
4230 char data[] __aligned(8);
4231 };
4232
4233 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4234 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4235 struct skb_ext *ext);
4236 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4237 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4238 void __skb_ext_put(struct skb_ext *ext);
4239
skb_ext_put(struct sk_buff * skb)4240 static inline void skb_ext_put(struct sk_buff *skb)
4241 {
4242 if (skb->active_extensions)
4243 __skb_ext_put(skb->extensions);
4244 }
4245
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4246 static inline void __skb_ext_copy(struct sk_buff *dst,
4247 const struct sk_buff *src)
4248 {
4249 dst->active_extensions = src->active_extensions;
4250
4251 if (src->active_extensions) {
4252 struct skb_ext *ext = src->extensions;
4253
4254 refcount_inc(&ext->refcnt);
4255 dst->extensions = ext;
4256 }
4257 }
4258
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4259 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4260 {
4261 skb_ext_put(dst);
4262 __skb_ext_copy(dst, src);
4263 }
4264
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4265 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4266 {
4267 return !!ext->offset[i];
4268 }
4269
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4270 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4271 {
4272 return skb->active_extensions & (1 << id);
4273 }
4274
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4275 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4276 {
4277 if (skb_ext_exist(skb, id))
4278 __skb_ext_del(skb, id);
4279 }
4280
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4281 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4282 {
4283 if (skb_ext_exist(skb, id)) {
4284 struct skb_ext *ext = skb->extensions;
4285
4286 return (void *)ext + (ext->offset[id] << 3);
4287 }
4288
4289 return NULL;
4290 }
4291
skb_ext_reset(struct sk_buff * skb)4292 static inline void skb_ext_reset(struct sk_buff *skb)
4293 {
4294 if (unlikely(skb->active_extensions)) {
4295 __skb_ext_put(skb->extensions);
4296 skb->active_extensions = 0;
4297 }
4298 }
4299
skb_has_extensions(struct sk_buff * skb)4300 static inline bool skb_has_extensions(struct sk_buff *skb)
4301 {
4302 return unlikely(skb->active_extensions);
4303 }
4304 #else
skb_ext_put(struct sk_buff * skb)4305 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4306 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4307 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4308 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)4309 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4310 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4311 #endif /* CONFIG_SKB_EXTENSIONS */
4312
nf_reset_ct(struct sk_buff * skb)4313 static inline void nf_reset_ct(struct sk_buff *skb)
4314 {
4315 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4316 nf_conntrack_put(skb_nfct(skb));
4317 skb->_nfct = 0;
4318 #endif
4319 }
4320
nf_reset_trace(struct sk_buff * skb)4321 static inline void nf_reset_trace(struct sk_buff *skb)
4322 {
4323 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4324 skb->nf_trace = 0;
4325 #endif
4326 }
4327
ipvs_reset(struct sk_buff * skb)4328 static inline void ipvs_reset(struct sk_buff *skb)
4329 {
4330 #if IS_ENABLED(CONFIG_IP_VS)
4331 skb->ipvs_property = 0;
4332 #endif
4333 }
4334
4335 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4336 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4337 bool copy)
4338 {
4339 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4340 dst->_nfct = src->_nfct;
4341 nf_conntrack_get(skb_nfct(src));
4342 #endif
4343 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4344 if (copy)
4345 dst->nf_trace = src->nf_trace;
4346 #endif
4347 }
4348
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4349 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4350 {
4351 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4352 nf_conntrack_put(skb_nfct(dst));
4353 #endif
4354 __nf_copy(dst, src, true);
4355 }
4356
4357 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4358 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4359 {
4360 to->secmark = from->secmark;
4361 }
4362
skb_init_secmark(struct sk_buff * skb)4363 static inline void skb_init_secmark(struct sk_buff *skb)
4364 {
4365 skb->secmark = 0;
4366 }
4367 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4368 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4369 { }
4370
skb_init_secmark(struct sk_buff * skb)4371 static inline void skb_init_secmark(struct sk_buff *skb)
4372 { }
4373 #endif
4374
secpath_exists(const struct sk_buff * skb)4375 static inline int secpath_exists(const struct sk_buff *skb)
4376 {
4377 #ifdef CONFIG_XFRM
4378 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4379 #else
4380 return 0;
4381 #endif
4382 }
4383
skb_irq_freeable(const struct sk_buff * skb)4384 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4385 {
4386 return !skb->destructor &&
4387 !secpath_exists(skb) &&
4388 !skb_nfct(skb) &&
4389 !skb->_skb_refdst &&
4390 !skb_has_frag_list(skb);
4391 }
4392
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)4393 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4394 {
4395 skb->queue_mapping = queue_mapping;
4396 }
4397
skb_get_queue_mapping(const struct sk_buff * skb)4398 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4399 {
4400 return skb->queue_mapping;
4401 }
4402
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)4403 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4404 {
4405 to->queue_mapping = from->queue_mapping;
4406 }
4407
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)4408 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4409 {
4410 skb->queue_mapping = rx_queue + 1;
4411 }
4412
skb_get_rx_queue(const struct sk_buff * skb)4413 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4414 {
4415 return skb->queue_mapping - 1;
4416 }
4417
skb_rx_queue_recorded(const struct sk_buff * skb)4418 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4419 {
4420 return skb->queue_mapping != 0;
4421 }
4422
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)4423 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4424 {
4425 skb->dst_pending_confirm = val;
4426 }
4427
skb_get_dst_pending_confirm(const struct sk_buff * skb)4428 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4429 {
4430 return skb->dst_pending_confirm != 0;
4431 }
4432
skb_sec_path(const struct sk_buff * skb)4433 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4434 {
4435 #ifdef CONFIG_XFRM
4436 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4437 #else
4438 return NULL;
4439 #endif
4440 }
4441
4442 /* Keeps track of mac header offset relative to skb->head.
4443 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4444 * For non-tunnel skb it points to skb_mac_header() and for
4445 * tunnel skb it points to outer mac header.
4446 * Keeps track of level of encapsulation of network headers.
4447 */
4448 struct skb_gso_cb {
4449 union {
4450 int mac_offset;
4451 int data_offset;
4452 };
4453 int encap_level;
4454 __wsum csum;
4455 __u16 csum_start;
4456 };
4457 #define SKB_GSO_CB_OFFSET 32
4458 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4459
skb_tnl_header_len(const struct sk_buff * inner_skb)4460 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4461 {
4462 return (skb_mac_header(inner_skb) - inner_skb->head) -
4463 SKB_GSO_CB(inner_skb)->mac_offset;
4464 }
4465
gso_pskb_expand_head(struct sk_buff * skb,int extra)4466 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4467 {
4468 int new_headroom, headroom;
4469 int ret;
4470
4471 headroom = skb_headroom(skb);
4472 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4473 if (ret)
4474 return ret;
4475
4476 new_headroom = skb_headroom(skb);
4477 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4478 return 0;
4479 }
4480
gso_reset_checksum(struct sk_buff * skb,__wsum res)4481 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4482 {
4483 /* Do not update partial checksums if remote checksum is enabled. */
4484 if (skb->remcsum_offload)
4485 return;
4486
4487 SKB_GSO_CB(skb)->csum = res;
4488 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4489 }
4490
4491 /* Compute the checksum for a gso segment. First compute the checksum value
4492 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4493 * then add in skb->csum (checksum from csum_start to end of packet).
4494 * skb->csum and csum_start are then updated to reflect the checksum of the
4495 * resultant packet starting from the transport header-- the resultant checksum
4496 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4497 * header.
4498 */
gso_make_checksum(struct sk_buff * skb,__wsum res)4499 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4500 {
4501 unsigned char *csum_start = skb_transport_header(skb);
4502 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4503 __wsum partial = SKB_GSO_CB(skb)->csum;
4504
4505 SKB_GSO_CB(skb)->csum = res;
4506 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4507
4508 return csum_fold(csum_partial(csum_start, plen, partial));
4509 }
4510
skb_is_gso(const struct sk_buff * skb)4511 static inline bool skb_is_gso(const struct sk_buff *skb)
4512 {
4513 return skb_shinfo(skb)->gso_size;
4514 }
4515
4516 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)4517 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4518 {
4519 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4520 }
4521
4522 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)4523 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4524 {
4525 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4526 }
4527
4528 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)4529 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4530 {
4531 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4532 }
4533
skb_gso_reset(struct sk_buff * skb)4534 static inline void skb_gso_reset(struct sk_buff *skb)
4535 {
4536 skb_shinfo(skb)->gso_size = 0;
4537 skb_shinfo(skb)->gso_segs = 0;
4538 skb_shinfo(skb)->gso_type = 0;
4539 }
4540
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)4541 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4542 u16 increment)
4543 {
4544 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4545 return;
4546 shinfo->gso_size += increment;
4547 }
4548
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)4549 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4550 u16 decrement)
4551 {
4552 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4553 return;
4554 shinfo->gso_size -= decrement;
4555 }
4556
4557 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4558
skb_warn_if_lro(const struct sk_buff * skb)4559 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4560 {
4561 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4562 * wanted then gso_type will be set. */
4563 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4564
4565 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4566 unlikely(shinfo->gso_type == 0)) {
4567 __skb_warn_lro_forwarding(skb);
4568 return true;
4569 }
4570 return false;
4571 }
4572
skb_forward_csum(struct sk_buff * skb)4573 static inline void skb_forward_csum(struct sk_buff *skb)
4574 {
4575 /* Unfortunately we don't support this one. Any brave souls? */
4576 if (skb->ip_summed == CHECKSUM_COMPLETE)
4577 skb->ip_summed = CHECKSUM_NONE;
4578 }
4579
4580 /**
4581 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4582 * @skb: skb to check
4583 *
4584 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4585 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4586 * use this helper, to document places where we make this assertion.
4587 */
skb_checksum_none_assert(const struct sk_buff * skb)4588 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4589 {
4590 #ifdef DEBUG
4591 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4592 #endif
4593 }
4594
4595 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4596
4597 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4598 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4599 unsigned int transport_len,
4600 __sum16(*skb_chkf)(struct sk_buff *skb));
4601
4602 /**
4603 * skb_head_is_locked - Determine if the skb->head is locked down
4604 * @skb: skb to check
4605 *
4606 * The head on skbs build around a head frag can be removed if they are
4607 * not cloned. This function returns true if the skb head is locked down
4608 * due to either being allocated via kmalloc, or by being a clone with
4609 * multiple references to the head.
4610 */
skb_head_is_locked(const struct sk_buff * skb)4611 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4612 {
4613 return !skb->head_frag || skb_cloned(skb);
4614 }
4615
4616 /* Local Checksum Offload.
4617 * Compute outer checksum based on the assumption that the
4618 * inner checksum will be offloaded later.
4619 * See Documentation/networking/checksum-offloads.rst for
4620 * explanation of how this works.
4621 * Fill in outer checksum adjustment (e.g. with sum of outer
4622 * pseudo-header) before calling.
4623 * Also ensure that inner checksum is in linear data area.
4624 */
lco_csum(struct sk_buff * skb)4625 static inline __wsum lco_csum(struct sk_buff *skb)
4626 {
4627 unsigned char *csum_start = skb_checksum_start(skb);
4628 unsigned char *l4_hdr = skb_transport_header(skb);
4629 __wsum partial;
4630
4631 /* Start with complement of inner checksum adjustment */
4632 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4633 skb->csum_offset));
4634
4635 /* Add in checksum of our headers (incl. outer checksum
4636 * adjustment filled in by caller) and return result.
4637 */
4638 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4639 }
4640
skb_is_redirected(const struct sk_buff * skb)4641 static inline bool skb_is_redirected(const struct sk_buff *skb)
4642 {
4643 #ifdef CONFIG_NET_REDIRECT
4644 return skb->redirected;
4645 #else
4646 return false;
4647 #endif
4648 }
4649
skb_set_redirected(struct sk_buff * skb,bool from_ingress)4650 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4651 {
4652 #ifdef CONFIG_NET_REDIRECT
4653 skb->redirected = 1;
4654 skb->from_ingress = from_ingress;
4655 if (skb->from_ingress)
4656 skb->tstamp = 0;
4657 #endif
4658 }
4659
skb_reset_redirect(struct sk_buff * skb)4660 static inline void skb_reset_redirect(struct sk_buff *skb)
4661 {
4662 #ifdef CONFIG_NET_REDIRECT
4663 skb->redirected = 0;
4664 #endif
4665 }
4666
4667 #if IS_ENABLED(CONFIG_KCOV) && IS_ENABLED(CONFIG_SKB_EXTENSIONS)
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)4668 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4669 const u64 kcov_handle)
4670 {
4671 /* Do not allocate skb extensions only to set kcov_handle to zero
4672 * (as it is zero by default). However, if the extensions are
4673 * already allocated, update kcov_handle anyway since
4674 * skb_set_kcov_handle can be called to zero a previously set
4675 * value.
4676 */
4677 if (skb_has_extensions(skb) || kcov_handle) {
4678 u64 *kcov_handle_ptr = skb_ext_add(skb, SKB_EXT_KCOV_HANDLE);
4679
4680 if (kcov_handle_ptr)
4681 *kcov_handle_ptr = kcov_handle;
4682 }
4683 }
4684
skb_get_kcov_handle(struct sk_buff * skb)4685 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4686 {
4687 u64 *kcov_handle = skb_ext_find(skb, SKB_EXT_KCOV_HANDLE);
4688
4689 return kcov_handle ? *kcov_handle : 0;
4690 }
4691 #else
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)4692 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4693 const u64 kcov_handle) { }
skb_get_kcov_handle(struct sk_buff * skb)4694 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) { return 0; }
4695 #endif /* CONFIG_KCOV && CONFIG_SKB_EXTENSIONS */
4696
4697 #endif /* __KERNEL__ */
4698 #endif /* _LINUX_SKBUFF_H */
4699