1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
8 */
9
10 #ifndef _LINUX_SKBUFF_H
11 #define _LINUX_SKBUFF_H
12
13 #include <linux/kernel.h>
14 #include <linux/compiler.h>
15 #include <linux/time.h>
16 #include <linux/bug.h>
17 #include <linux/bvec.h>
18 #include <linux/cache.h>
19 #include <linux/rbtree.h>
20 #include <linux/socket.h>
21 #include <linux/refcount.h>
22
23 #include <linux/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/hrtimer.h>
31 #include <linux/dma-mapping.h>
32 #include <linux/netdev_features.h>
33 #include <linux/sched.h>
34 #include <linux/sched/clock.h>
35 #include <net/flow_dissector.h>
36 #include <linux/splice.h>
37 #include <linux/in6.h>
38 #include <linux/if_packet.h>
39 #include <net/flow.h>
40 #include <net/page_pool.h>
41 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
42 #include <linux/netfilter/nf_conntrack_common.h>
43 #endif
44 #include <linux/android_kabi.h>
45 #include <linux/android_vendor.h>
46
47 /* The interface for checksum offload between the stack and networking drivers
48 * is as follows...
49 *
50 * A. IP checksum related features
51 *
52 * Drivers advertise checksum offload capabilities in the features of a device.
53 * From the stack's point of view these are capabilities offered by the driver.
54 * A driver typically only advertises features that it is capable of offloading
55 * to its device.
56 *
57 * The checksum related features are:
58 *
59 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
60 * IP (one's complement) checksum for any combination
61 * of protocols or protocol layering. The checksum is
62 * computed and set in a packet per the CHECKSUM_PARTIAL
63 * interface (see below).
64 *
65 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
66 * TCP or UDP packets over IPv4. These are specifically
67 * unencapsulated packets of the form IPv4|TCP or
68 * IPv4|UDP where the Protocol field in the IPv4 header
69 * is TCP or UDP. The IPv4 header may contain IP options.
70 * This feature cannot be set in features for a device
71 * with NETIF_F_HW_CSUM also set. This feature is being
72 * DEPRECATED (see below).
73 *
74 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
75 * TCP or UDP packets over IPv6. These are specifically
76 * unencapsulated packets of the form IPv6|TCP or
77 * IPv6|UDP where the Next Header field in the IPv6
78 * header is either TCP or UDP. IPv6 extension headers
79 * are not supported with this feature. This feature
80 * cannot be set in features for a device with
81 * NETIF_F_HW_CSUM also set. This feature is being
82 * DEPRECATED (see below).
83 *
84 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
85 * This flag is only used to disable the RX checksum
86 * feature for a device. The stack will accept receive
87 * checksum indication in packets received on a device
88 * regardless of whether NETIF_F_RXCSUM is set.
89 *
90 * B. Checksumming of received packets by device. Indication of checksum
91 * verification is set in skb->ip_summed. Possible values are:
92 *
93 * CHECKSUM_NONE:
94 *
95 * Device did not checksum this packet e.g. due to lack of capabilities.
96 * The packet contains full (though not verified) checksum in packet but
97 * not in skb->csum. Thus, skb->csum is undefined in this case.
98 *
99 * CHECKSUM_UNNECESSARY:
100 *
101 * The hardware you're dealing with doesn't calculate the full checksum
102 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
103 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
104 * if their checksums are okay. skb->csum is still undefined in this case
105 * though. A driver or device must never modify the checksum field in the
106 * packet even if checksum is verified.
107 *
108 * CHECKSUM_UNNECESSARY is applicable to following protocols:
109 * TCP: IPv6 and IPv4.
110 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
111 * zero UDP checksum for either IPv4 or IPv6, the networking stack
112 * may perform further validation in this case.
113 * GRE: only if the checksum is present in the header.
114 * SCTP: indicates the CRC in SCTP header has been validated.
115 * FCOE: indicates the CRC in FC frame has been validated.
116 *
117 * skb->csum_level indicates the number of consecutive checksums found in
118 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
119 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
120 * and a device is able to verify the checksums for UDP (possibly zero),
121 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
122 * two. If the device were only able to verify the UDP checksum and not
123 * GRE, either because it doesn't support GRE checksum or because GRE
124 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
125 * not considered in this case).
126 *
127 * CHECKSUM_COMPLETE:
128 *
129 * This is the most generic way. The device supplied checksum of the _whole_
130 * packet as seen by netif_rx() and fills in skb->csum. This means the
131 * hardware doesn't need to parse L3/L4 headers to implement this.
132 *
133 * Notes:
134 * - Even if device supports only some protocols, but is able to produce
135 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
136 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
137 *
138 * CHECKSUM_PARTIAL:
139 *
140 * A checksum is set up to be offloaded to a device as described in the
141 * output description for CHECKSUM_PARTIAL. This may occur on a packet
142 * received directly from another Linux OS, e.g., a virtualized Linux kernel
143 * on the same host, or it may be set in the input path in GRO or remote
144 * checksum offload. For the purposes of checksum verification, the checksum
145 * referred to by skb->csum_start + skb->csum_offset and any preceding
146 * checksums in the packet are considered verified. Any checksums in the
147 * packet that are after the checksum being offloaded are not considered to
148 * be verified.
149 *
150 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
151 * in the skb->ip_summed for a packet. Values are:
152 *
153 * CHECKSUM_PARTIAL:
154 *
155 * The driver is required to checksum the packet as seen by hard_start_xmit()
156 * from skb->csum_start up to the end, and to record/write the checksum at
157 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
158 * csum_start and csum_offset values are valid values given the length and
159 * offset of the packet, but it should not attempt to validate that the
160 * checksum refers to a legitimate transport layer checksum -- it is the
161 * purview of the stack to validate that csum_start and csum_offset are set
162 * correctly.
163 *
164 * When the stack requests checksum offload for a packet, the driver MUST
165 * ensure that the checksum is set correctly. A driver can either offload the
166 * checksum calculation to the device, or call skb_checksum_help (in the case
167 * that the device does not support offload for a particular checksum).
168 *
169 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
170 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
171 * checksum offload capability.
172 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
173 * on network device checksumming capabilities: if a packet does not match
174 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
175 * csum_not_inet, see item D.) is called to resolve the checksum.
176 *
177 * CHECKSUM_NONE:
178 *
179 * The skb was already checksummed by the protocol, or a checksum is not
180 * required.
181 *
182 * CHECKSUM_UNNECESSARY:
183 *
184 * This has the same meaning as CHECKSUM_NONE for checksum offload on
185 * output.
186 *
187 * CHECKSUM_COMPLETE:
188 * Not used in checksum output. If a driver observes a packet with this value
189 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
190 *
191 * D. Non-IP checksum (CRC) offloads
192 *
193 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
194 * offloading the SCTP CRC in a packet. To perform this offload the stack
195 * will set csum_start and csum_offset accordingly, set ip_summed to
196 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
197 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
198 * A driver that supports both IP checksum offload and SCTP CRC32c offload
199 * must verify which offload is configured for a packet by testing the
200 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
201 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
202 *
203 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
204 * offloading the FCOE CRC in a packet. To perform this offload the stack
205 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
206 * accordingly. Note that there is no indication in the skbuff that the
207 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
208 * both IP checksum offload and FCOE CRC offload must verify which offload
209 * is configured for a packet, presumably by inspecting packet headers.
210 *
211 * E. Checksumming on output with GSO.
212 *
213 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
214 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
215 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
216 * part of the GSO operation is implied. If a checksum is being offloaded
217 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
218 * csum_offset are set to refer to the outermost checksum being offloaded
219 * (two offloaded checksums are possible with UDP encapsulation).
220 */
221
222 /* Don't change this without changing skb_csum_unnecessary! */
223 #define CHECKSUM_NONE 0
224 #define CHECKSUM_UNNECESSARY 1
225 #define CHECKSUM_COMPLETE 2
226 #define CHECKSUM_PARTIAL 3
227
228 /* Maximum value in skb->csum_level */
229 #define SKB_MAX_CSUM_LEVEL 3
230
231 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
232 #define SKB_WITH_OVERHEAD(X) \
233 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
234 #define SKB_MAX_ORDER(X, ORDER) \
235 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
236 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
237 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
238
239 /* return minimum truesize of one skb containing X bytes of data */
240 #define SKB_TRUESIZE(X) ((X) + \
241 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
242 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
243
244 struct ahash_request;
245 struct net_device;
246 struct scatterlist;
247 struct pipe_inode_info;
248 struct iov_iter;
249 struct napi_struct;
250 struct bpf_prog;
251 union bpf_attr;
252 struct skb_ext;
253
254 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
255 struct nf_bridge_info {
256 enum {
257 BRNF_PROTO_UNCHANGED,
258 BRNF_PROTO_8021Q,
259 BRNF_PROTO_PPPOE
260 } orig_proto:8;
261 u8 pkt_otherhost:1;
262 u8 in_prerouting:1;
263 u8 bridged_dnat:1;
264 u8 sabotage_in_done:1;
265 __u16 frag_max_size;
266 struct net_device *physindev;
267
268 /* always valid & non-NULL from FORWARD on, for physdev match */
269 struct net_device *physoutdev;
270 union {
271 /* prerouting: detect dnat in orig/reply direction */
272 __be32 ipv4_daddr;
273 struct in6_addr ipv6_daddr;
274
275 /* after prerouting + nat detected: store original source
276 * mac since neigh resolution overwrites it, only used while
277 * skb is out in neigh layer.
278 */
279 char neigh_header[8];
280 };
281 };
282 #endif
283
284 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
285 /* Chain in tc_skb_ext will be used to share the tc chain with
286 * ovs recirc_id. It will be set to the current chain by tc
287 * and read by ovs to recirc_id.
288 */
289 struct tc_skb_ext {
290 __u32 chain;
291 __u16 mru;
292 __u16 zone;
293 u8 post_ct:1;
294 u8 post_ct_snat:1;
295 u8 post_ct_dnat:1;
296 };
297 #endif
298
299 struct sk_buff_head {
300 /* These two members must be first. */
301 struct sk_buff *next;
302 struct sk_buff *prev;
303
304 __u32 qlen;
305 spinlock_t lock;
306 };
307
308 struct sk_buff;
309
310 /* The reason of skb drop, which is used in kfree_skb_reason().
311 * en...maybe they should be splited by group?
312 *
313 * Each item here should also be in 'TRACE_SKB_DROP_REASON', which is
314 * used to translate the reason to string.
315 */
316 enum skb_drop_reason {
317 SKB_DROP_REASON_NOT_SPECIFIED, /* drop reason is not specified */
318 SKB_DROP_REASON_NO_SOCKET, /* socket not found */
319 SKB_DROP_REASON_PKT_TOO_SMALL, /* packet size is too small */
320 SKB_DROP_REASON_TCP_CSUM, /* TCP checksum error */
321 SKB_DROP_REASON_SOCKET_FILTER, /* dropped by socket filter */
322 SKB_DROP_REASON_UDP_CSUM, /* UDP checksum error */
323 SKB_DROP_REASON_NETFILTER_DROP, /* dropped by netfilter */
324 SKB_DROP_REASON_OTHERHOST, /* packet don't belong to current
325 * host (interface is in promisc
326 * mode)
327 */
328 SKB_DROP_REASON_IP_CSUM, /* IP checksum error */
329 SKB_DROP_REASON_IP_INHDR, /* there is something wrong with
330 * IP header (see
331 * IPSTATS_MIB_INHDRERRORS)
332 */
333 SKB_DROP_REASON_IP_RPFILTER, /* IP rpfilter validate failed.
334 * see the document for rp_filter
335 * in ip-sysctl.rst for more
336 * information
337 */
338 SKB_DROP_REASON_UNICAST_IN_L2_MULTICAST, /* destination address of L2
339 * is multicast, but L3 is
340 * unicast.
341 */
342 SKB_DROP_REASON_MAX,
343 };
344
345 /* To allow 64K frame to be packed as single skb without frag_list we
346 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
347 * buffers which do not start on a page boundary.
348 *
349 * Since GRO uses frags we allocate at least 16 regardless of page
350 * size.
351 */
352 #if (65536/PAGE_SIZE + 1) < 16
353 #define MAX_SKB_FRAGS 16UL
354 #else
355 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
356 #endif
357 extern int sysctl_max_skb_frags;
358
359 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
360 * segment using its current segmentation instead.
361 */
362 #define GSO_BY_FRAGS 0xFFFF
363
364 typedef struct bio_vec skb_frag_t;
365
366 /**
367 * skb_frag_size() - Returns the size of a skb fragment
368 * @frag: skb fragment
369 */
skb_frag_size(const skb_frag_t * frag)370 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
371 {
372 return frag->bv_len;
373 }
374
375 /**
376 * skb_frag_size_set() - Sets the size of a skb fragment
377 * @frag: skb fragment
378 * @size: size of fragment
379 */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)380 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
381 {
382 frag->bv_len = size;
383 }
384
385 /**
386 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
387 * @frag: skb fragment
388 * @delta: value to add
389 */
skb_frag_size_add(skb_frag_t * frag,int delta)390 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
391 {
392 frag->bv_len += delta;
393 }
394
395 /**
396 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
397 * @frag: skb fragment
398 * @delta: value to subtract
399 */
skb_frag_size_sub(skb_frag_t * frag,int delta)400 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
401 {
402 frag->bv_len -= delta;
403 }
404
405 /**
406 * skb_frag_must_loop - Test if %p is a high memory page
407 * @p: fragment's page
408 */
skb_frag_must_loop(struct page * p)409 static inline bool skb_frag_must_loop(struct page *p)
410 {
411 #if defined(CONFIG_HIGHMEM)
412 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
413 return true;
414 #endif
415 return false;
416 }
417
418 /**
419 * skb_frag_foreach_page - loop over pages in a fragment
420 *
421 * @f: skb frag to operate on
422 * @f_off: offset from start of f->bv_page
423 * @f_len: length from f_off to loop over
424 * @p: (temp var) current page
425 * @p_off: (temp var) offset from start of current page,
426 * non-zero only on first page.
427 * @p_len: (temp var) length in current page,
428 * < PAGE_SIZE only on first and last page.
429 * @copied: (temp var) length so far, excluding current p_len.
430 *
431 * A fragment can hold a compound page, in which case per-page
432 * operations, notably kmap_atomic, must be called for each
433 * regular page.
434 */
435 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
436 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
437 p_off = (f_off) & (PAGE_SIZE - 1), \
438 p_len = skb_frag_must_loop(p) ? \
439 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
440 copied = 0; \
441 copied < f_len; \
442 copied += p_len, p++, p_off = 0, \
443 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
444
445 #define HAVE_HW_TIME_STAMP
446
447 /**
448 * struct skb_shared_hwtstamps - hardware time stamps
449 * @hwtstamp: hardware time stamp transformed into duration
450 * since arbitrary point in time
451 *
452 * Software time stamps generated by ktime_get_real() are stored in
453 * skb->tstamp.
454 *
455 * hwtstamps can only be compared against other hwtstamps from
456 * the same device.
457 *
458 * This structure is attached to packets as part of the
459 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
460 */
461 struct skb_shared_hwtstamps {
462 ktime_t hwtstamp;
463 };
464
465 /* Definitions for tx_flags in struct skb_shared_info */
466 enum {
467 /* generate hardware time stamp */
468 SKBTX_HW_TSTAMP = 1 << 0,
469
470 /* generate software time stamp when queueing packet to NIC */
471 SKBTX_SW_TSTAMP = 1 << 1,
472
473 /* device driver is going to provide hardware time stamp */
474 SKBTX_IN_PROGRESS = 1 << 2,
475
476 /* generate wifi status information (where possible) */
477 SKBTX_WIFI_STATUS = 1 << 4,
478
479 /* generate software time stamp when entering packet scheduling */
480 SKBTX_SCHED_TSTAMP = 1 << 6,
481 };
482
483 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
484 SKBTX_SCHED_TSTAMP)
485 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
486
487 /* Definitions for flags in struct skb_shared_info */
488 enum {
489 /* use zcopy routines */
490 SKBFL_ZEROCOPY_ENABLE = BIT(0),
491
492 /* This indicates at least one fragment might be overwritten
493 * (as in vmsplice(), sendfile() ...)
494 * If we need to compute a TX checksum, we'll need to copy
495 * all frags to avoid possible bad checksum
496 */
497 SKBFL_SHARED_FRAG = BIT(1),
498 };
499
500 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
501
502 /*
503 * The callback notifies userspace to release buffers when skb DMA is done in
504 * lower device, the skb last reference should be 0 when calling this.
505 * The zerocopy_success argument is true if zero copy transmit occurred,
506 * false on data copy or out of memory error caused by data copy attempt.
507 * The ctx field is used to track device context.
508 * The desc field is used to track userspace buffer index.
509 */
510 struct ubuf_info {
511 void (*callback)(struct sk_buff *, struct ubuf_info *,
512 bool zerocopy_success);
513 union {
514 struct {
515 unsigned long desc;
516 void *ctx;
517 };
518 struct {
519 u32 id;
520 u16 len;
521 u16 zerocopy:1;
522 u32 bytelen;
523 };
524 };
525 refcount_t refcnt;
526 u8 flags;
527
528 struct mmpin {
529 struct user_struct *user;
530 unsigned int num_pg;
531 } mmp;
532 };
533
534 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
535
536 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
537 void mm_unaccount_pinned_pages(struct mmpin *mmp);
538
539 struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size);
540 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
541 struct ubuf_info *uarg);
542
543 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
544
545 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
546 bool success);
547
548 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
549 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
550 struct msghdr *msg, int len,
551 struct ubuf_info *uarg);
552
553 /* This data is invariant across clones and lives at
554 * the end of the header data, ie. at skb->end.
555 */
556 struct skb_shared_info {
557 __u8 flags;
558 __u8 meta_len;
559 __u8 nr_frags;
560 __u8 tx_flags;
561 unsigned short gso_size;
562 /* Warning: this field is not always filled in (UFO)! */
563 unsigned short gso_segs;
564 struct sk_buff *frag_list;
565 struct skb_shared_hwtstamps hwtstamps;
566 unsigned int gso_type;
567 u32 tskey;
568
569 /*
570 * Warning : all fields before dataref are cleared in __alloc_skb()
571 */
572 atomic_t dataref;
573
574 /* Intermediate layers must ensure that destructor_arg
575 * remains valid until skb destructor */
576 void * destructor_arg;
577
578 ANDROID_OEM_DATA_ARRAY(1, 3);
579
580 /* must be last field, see pskb_expand_head() */
581 skb_frag_t frags[MAX_SKB_FRAGS];
582 };
583
584 /* We divide dataref into two halves. The higher 16 bits hold references
585 * to the payload part of skb->data. The lower 16 bits hold references to
586 * the entire skb->data. A clone of a headerless skb holds the length of
587 * the header in skb->hdr_len.
588 *
589 * All users must obey the rule that the skb->data reference count must be
590 * greater than or equal to the payload reference count.
591 *
592 * Holding a reference to the payload part means that the user does not
593 * care about modifications to the header part of skb->data.
594 */
595 #define SKB_DATAREF_SHIFT 16
596 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
597
598
599 enum {
600 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
601 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
602 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
603 };
604
605 enum {
606 SKB_GSO_TCPV4 = 1 << 0,
607
608 /* This indicates the skb is from an untrusted source. */
609 SKB_GSO_DODGY = 1 << 1,
610
611 /* This indicates the tcp segment has CWR set. */
612 SKB_GSO_TCP_ECN = 1 << 2,
613
614 SKB_GSO_TCP_FIXEDID = 1 << 3,
615
616 SKB_GSO_TCPV6 = 1 << 4,
617
618 SKB_GSO_FCOE = 1 << 5,
619
620 SKB_GSO_GRE = 1 << 6,
621
622 SKB_GSO_GRE_CSUM = 1 << 7,
623
624 SKB_GSO_IPXIP4 = 1 << 8,
625
626 SKB_GSO_IPXIP6 = 1 << 9,
627
628 SKB_GSO_UDP_TUNNEL = 1 << 10,
629
630 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
631
632 SKB_GSO_PARTIAL = 1 << 12,
633
634 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
635
636 SKB_GSO_SCTP = 1 << 14,
637
638 SKB_GSO_ESP = 1 << 15,
639
640 SKB_GSO_UDP = 1 << 16,
641
642 SKB_GSO_UDP_L4 = 1 << 17,
643
644 SKB_GSO_FRAGLIST = 1 << 18,
645 };
646
647 #if BITS_PER_LONG > 32
648 #define NET_SKBUFF_DATA_USES_OFFSET 1
649 #endif
650
651 #ifdef NET_SKBUFF_DATA_USES_OFFSET
652 typedef unsigned int sk_buff_data_t;
653 #else
654 typedef unsigned char *sk_buff_data_t;
655 #endif
656
657 /**
658 * struct sk_buff - socket buffer
659 * @next: Next buffer in list
660 * @prev: Previous buffer in list
661 * @tstamp: Time we arrived/left
662 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
663 * for retransmit timer
664 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
665 * @list: queue head
666 * @sk: Socket we are owned by
667 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
668 * fragmentation management
669 * @dev: Device we arrived on/are leaving by
670 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
671 * @cb: Control buffer. Free for use by every layer. Put private vars here
672 * @_skb_refdst: destination entry (with norefcount bit)
673 * @sp: the security path, used for xfrm
674 * @len: Length of actual data
675 * @data_len: Data length
676 * @mac_len: Length of link layer header
677 * @hdr_len: writable header length of cloned skb
678 * @csum: Checksum (must include start/offset pair)
679 * @csum_start: Offset from skb->head where checksumming should start
680 * @csum_offset: Offset from csum_start where checksum should be stored
681 * @priority: Packet queueing priority
682 * @ignore_df: allow local fragmentation
683 * @cloned: Head may be cloned (check refcnt to be sure)
684 * @ip_summed: Driver fed us an IP checksum
685 * @nohdr: Payload reference only, must not modify header
686 * @pkt_type: Packet class
687 * @fclone: skbuff clone status
688 * @ipvs_property: skbuff is owned by ipvs
689 * @inner_protocol_type: whether the inner protocol is
690 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
691 * @remcsum_offload: remote checksum offload is enabled
692 * @offload_fwd_mark: Packet was L2-forwarded in hardware
693 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
694 * @tc_skip_classify: do not classify packet. set by IFB device
695 * @tc_at_ingress: used within tc_classify to distinguish in/egress
696 * @redirected: packet was redirected by packet classifier
697 * @from_ingress: packet was redirected from the ingress path
698 * @peeked: this packet has been seen already, so stats have been
699 * done for it, don't do them again
700 * @nf_trace: netfilter packet trace flag
701 * @protocol: Packet protocol from driver
702 * @destructor: Destruct function
703 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
704 * @_sk_redir: socket redirection information for skmsg
705 * @_nfct: Associated connection, if any (with nfctinfo bits)
706 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
707 * @skb_iif: ifindex of device we arrived on
708 * @tc_index: Traffic control index
709 * @hash: the packet hash
710 * @queue_mapping: Queue mapping for multiqueue devices
711 * @head_frag: skb was allocated from page fragments,
712 * not allocated by kmalloc() or vmalloc().
713 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
714 * @pp_recycle: mark the packet for recycling instead of freeing (implies
715 * page_pool support on driver)
716 * @active_extensions: active extensions (skb_ext_id types)
717 * @ndisc_nodetype: router type (from link layer)
718 * @ooo_okay: allow the mapping of a socket to a queue to be changed
719 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
720 * ports.
721 * @sw_hash: indicates hash was computed in software stack
722 * @wifi_acked_valid: wifi_acked was set
723 * @wifi_acked: whether frame was acked on wifi or not
724 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
725 * @encapsulation: indicates the inner headers in the skbuff are valid
726 * @encap_hdr_csum: software checksum is needed
727 * @csum_valid: checksum is already valid
728 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
729 * @csum_complete_sw: checksum was completed by software
730 * @csum_level: indicates the number of consecutive checksums found in
731 * the packet minus one that have been verified as
732 * CHECKSUM_UNNECESSARY (max 3)
733 * @scm_io_uring: SKB holds io_uring registered files
734 * @dst_pending_confirm: need to confirm neighbour
735 * @decrypted: Decrypted SKB
736 * @slow_gro: state present at GRO time, slower prepare step required
737 * @napi_id: id of the NAPI struct this skb came from
738 * @sender_cpu: (aka @napi_id) source CPU in XPS
739 * @secmark: security marking
740 * @mark: Generic packet mark
741 * @reserved_tailroom: (aka @mark) number of bytes of free space available
742 * at the tail of an sk_buff
743 * @vlan_present: VLAN tag is present
744 * @vlan_proto: vlan encapsulation protocol
745 * @vlan_tci: vlan tag control information
746 * @inner_protocol: Protocol (encapsulation)
747 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
748 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
749 * @inner_transport_header: Inner transport layer header (encapsulation)
750 * @inner_network_header: Network layer header (encapsulation)
751 * @inner_mac_header: Link layer header (encapsulation)
752 * @transport_header: Transport layer header
753 * @network_header: Network layer header
754 * @mac_header: Link layer header
755 * @kcov_handle: KCOV remote handle for remote coverage collection
756 * @tail: Tail pointer
757 * @end: End pointer
758 * @head: Head of buffer
759 * @data: Data head pointer
760 * @truesize: Buffer size
761 * @users: User count - see {datagram,tcp}.c
762 * @extensions: allocated extensions, valid if active_extensions is nonzero
763 */
764
765 struct sk_buff {
766 union {
767 struct {
768 /* These two members must be first. */
769 struct sk_buff *next;
770 struct sk_buff *prev;
771
772 union {
773 struct net_device *dev;
774 /* Some protocols might use this space to store information,
775 * while device pointer would be NULL.
776 * UDP receive path is one user.
777 */
778 unsigned long dev_scratch;
779 };
780 };
781 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
782 struct list_head list;
783 };
784
785 union {
786 struct sock *sk;
787 int ip_defrag_offset;
788 };
789
790 union {
791 ktime_t tstamp;
792 u64 skb_mstamp_ns; /* earliest departure time */
793 };
794 /*
795 * This is the control buffer. It is free to use for every
796 * layer. Please put your private variables there. If you
797 * want to keep them across layers you have to do a skb_clone()
798 * first. This is owned by whoever has the skb queued ATM.
799 */
800 char cb[48] __aligned(8);
801
802 union {
803 struct {
804 unsigned long _skb_refdst;
805 void (*destructor)(struct sk_buff *skb);
806 };
807 struct list_head tcp_tsorted_anchor;
808 #ifdef CONFIG_NET_SOCK_MSG
809 unsigned long _sk_redir;
810 #endif
811 };
812
813 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
814 unsigned long _nfct;
815 #endif
816 unsigned int len,
817 data_len;
818 __u16 mac_len,
819 hdr_len;
820
821 /* Following fields are _not_ copied in __copy_skb_header()
822 * Note that queue_mapping is here mostly to fill a hole.
823 */
824 __u16 queue_mapping;
825
826 /* if you move cloned around you also must adapt those constants */
827 #ifdef __BIG_ENDIAN_BITFIELD
828 #define CLONED_MASK (1 << 7)
829 #else
830 #define CLONED_MASK 1
831 #endif
832 #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
833
834 /* private: */
835 __u8 __cloned_offset[0];
836 /* public: */
837 __u8 cloned:1,
838 nohdr:1,
839 fclone:2,
840 peeked:1,
841 head_frag:1,
842 pfmemalloc:1,
843 pp_recycle:1; /* page_pool recycle indicator */
844 #ifdef CONFIG_SKB_EXTENSIONS
845 __u8 active_extensions;
846 #endif
847
848 /* fields enclosed in headers_start/headers_end are copied
849 * using a single memcpy() in __copy_skb_header()
850 */
851 /* private: */
852 __u32 headers_start[0];
853 /* public: */
854
855 /* if you move pkt_type around you also must adapt those constants */
856 #ifdef __BIG_ENDIAN_BITFIELD
857 #define PKT_TYPE_MAX (7 << 5)
858 #else
859 #define PKT_TYPE_MAX 7
860 #endif
861 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
862
863 /* private: */
864 __u8 __pkt_type_offset[0];
865 /* public: */
866 __u8 pkt_type:3;
867 __u8 ignore_df:1;
868 __u8 nf_trace:1;
869 __u8 ip_summed:2;
870 __u8 ooo_okay:1;
871
872 __u8 l4_hash:1;
873 __u8 sw_hash:1;
874 __u8 wifi_acked_valid:1;
875 __u8 wifi_acked:1;
876 __u8 no_fcs:1;
877 /* Indicates the inner headers are valid in the skbuff. */
878 __u8 encapsulation:1;
879 __u8 encap_hdr_csum:1;
880 __u8 csum_valid:1;
881
882 #ifdef __BIG_ENDIAN_BITFIELD
883 #define PKT_VLAN_PRESENT_BIT 7
884 #else
885 #define PKT_VLAN_PRESENT_BIT 0
886 #endif
887 #define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
888 /* private: */
889 __u8 __pkt_vlan_present_offset[0];
890 /* public: */
891 __u8 vlan_present:1;
892 __u8 csum_complete_sw:1;
893 __u8 csum_level:2;
894 __u8 csum_not_inet:1;
895 __u8 dst_pending_confirm:1;
896 #ifdef CONFIG_IPV6_NDISC_NODETYPE
897 __u8 ndisc_nodetype:2;
898 #endif
899
900 __u8 ipvs_property:1;
901 __u8 inner_protocol_type:1;
902 __u8 remcsum_offload:1;
903 #ifdef CONFIG_NET_SWITCHDEV
904 __u8 offload_fwd_mark:1;
905 __u8 offload_l3_fwd_mark:1;
906 #endif
907 #ifdef CONFIG_NET_CLS_ACT
908 __u8 tc_skip_classify:1;
909 __u8 tc_at_ingress:1;
910 #endif
911 __u8 redirected:1;
912 #ifdef CONFIG_NET_REDIRECT
913 __u8 from_ingress:1;
914 #endif
915 #ifdef CONFIG_TLS_DEVICE
916 __u8 decrypted:1;
917 #endif
918 __u8 slow_gro:1;
919
920 #ifdef CONFIG_NET_SCHED
921 __u16 tc_index; /* traffic control index */
922 #endif
923
924 union {
925 __wsum csum;
926 struct {
927 __u16 csum_start;
928 __u16 csum_offset;
929 };
930 };
931 __u32 priority;
932 int skb_iif;
933 __u32 hash;
934 __be16 vlan_proto;
935 __u16 vlan_tci;
936 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
937 union {
938 unsigned int napi_id;
939 unsigned int sender_cpu;
940 };
941 #endif
942 #ifdef CONFIG_NETWORK_SECMARK
943 __u32 secmark;
944 #endif
945
946 union {
947 __u32 mark;
948 __u32 reserved_tailroom;
949 };
950
951 union {
952 __be16 inner_protocol;
953 __u8 inner_ipproto;
954 };
955
956 __u16 inner_transport_header;
957 __u16 inner_network_header;
958 __u16 inner_mac_header;
959
960 __be16 protocol;
961 __u16 transport_header;
962 __u16 network_header;
963 __u16 mac_header;
964
965 #ifdef CONFIG_KCOV
966 u64 kcov_handle;
967 #endif
968
969 /* private: */
970 __u32 headers_end[0];
971 /* public: */
972
973 /* Android KABI preservation.
974 *
975 * "open coded" version of ANDROID_KABI_USE() to pack more
976 * fields/variables into the space that we have.
977 *
978 * scm_io_uring is from 04df9719df18 ("io_uring/af_unix: defer
979 * registered files gc to io_uring release")
980 */
981 /* NOTE: due to these fields ending up after headers_end, we have to
982 * manually copy them in the __copy_skb_header() call in skbuf.c. Be
983 * very aware of that if you change these fields.
984 */
985 _ANDROID_KABI_REPLACE(_ANDROID_KABI_RESERVE(1),
986 struct {
987 __u8 scm_io_uring:1;
988 __u8 android_kabi_reserved1_padding1;
989 __u16 android_kabi_reserved1_padding2;
990 __u32 android_kabi_reserved1_padding3;
991 });
992 ANDROID_KABI_RESERVE(2);
993
994 /* These elements must be at the end, see alloc_skb() for details. */
995 sk_buff_data_t tail;
996 sk_buff_data_t end;
997 unsigned char *head,
998 *data;
999 unsigned int truesize;
1000 refcount_t users;
1001
1002 #ifdef CONFIG_SKB_EXTENSIONS
1003 /* only useable after checking ->active_extensions != 0 */
1004 struct skb_ext *extensions;
1005 #endif
1006 };
1007
1008 #ifdef __KERNEL__
1009 /*
1010 * Handling routines are only of interest to the kernel
1011 */
1012
1013 #define SKB_ALLOC_FCLONE 0x01
1014 #define SKB_ALLOC_RX 0x02
1015 #define SKB_ALLOC_NAPI 0x04
1016
1017 /**
1018 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
1019 * @skb: buffer
1020 */
skb_pfmemalloc(const struct sk_buff * skb)1021 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
1022 {
1023 return unlikely(skb->pfmemalloc);
1024 }
1025
1026 /*
1027 * skb might have a dst pointer attached, refcounted or not.
1028 * _skb_refdst low order bit is set if refcount was _not_ taken
1029 */
1030 #define SKB_DST_NOREF 1UL
1031 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
1032
1033 /**
1034 * skb_dst - returns skb dst_entry
1035 * @skb: buffer
1036 *
1037 * Returns skb dst_entry, regardless of reference taken or not.
1038 */
skb_dst(const struct sk_buff * skb)1039 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1040 {
1041 /* If refdst was not refcounted, check we still are in a
1042 * rcu_read_lock section
1043 */
1044 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1045 !rcu_read_lock_held() &&
1046 !rcu_read_lock_bh_held());
1047 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
1048 }
1049
1050 /**
1051 * skb_dst_set - sets skb dst
1052 * @skb: buffer
1053 * @dst: dst entry
1054 *
1055 * Sets skb dst, assuming a reference was taken on dst and should
1056 * be released by skb_dst_drop()
1057 */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)1058 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1059 {
1060 skb->slow_gro |= !!dst;
1061 skb->_skb_refdst = (unsigned long)dst;
1062 }
1063
1064 /**
1065 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1066 * @skb: buffer
1067 * @dst: dst entry
1068 *
1069 * Sets skb dst, assuming a reference was not taken on dst.
1070 * If dst entry is cached, we do not take reference and dst_release
1071 * will be avoided by refdst_drop. If dst entry is not cached, we take
1072 * reference, so that last dst_release can destroy the dst immediately.
1073 */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)1074 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1075 {
1076 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1077 skb->slow_gro |= !!dst;
1078 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1079 }
1080
1081 /**
1082 * skb_dst_is_noref - Test if skb dst isn't refcounted
1083 * @skb: buffer
1084 */
skb_dst_is_noref(const struct sk_buff * skb)1085 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1086 {
1087 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1088 }
1089
1090 /**
1091 * skb_rtable - Returns the skb &rtable
1092 * @skb: buffer
1093 */
skb_rtable(const struct sk_buff * skb)1094 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1095 {
1096 return (struct rtable *)skb_dst(skb);
1097 }
1098
1099 /* For mangling skb->pkt_type from user space side from applications
1100 * such as nft, tc, etc, we only allow a conservative subset of
1101 * possible pkt_types to be set.
1102 */
skb_pkt_type_ok(u32 ptype)1103 static inline bool skb_pkt_type_ok(u32 ptype)
1104 {
1105 return ptype <= PACKET_OTHERHOST;
1106 }
1107
1108 /**
1109 * skb_napi_id - Returns the skb's NAPI id
1110 * @skb: buffer
1111 */
skb_napi_id(const struct sk_buff * skb)1112 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1113 {
1114 #ifdef CONFIG_NET_RX_BUSY_POLL
1115 return skb->napi_id;
1116 #else
1117 return 0;
1118 #endif
1119 }
1120
1121 /**
1122 * skb_unref - decrement the skb's reference count
1123 * @skb: buffer
1124 *
1125 * Returns true if we can free the skb.
1126 */
skb_unref(struct sk_buff * skb)1127 static inline bool skb_unref(struct sk_buff *skb)
1128 {
1129 if (unlikely(!skb))
1130 return false;
1131 if (likely(refcount_read(&skb->users) == 1))
1132 smp_rmb();
1133 else if (likely(!refcount_dec_and_test(&skb->users)))
1134 return false;
1135
1136 return true;
1137 }
1138
1139 void kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason);
1140 void kfree_skb(struct sk_buff *skb);
1141
1142 void skb_release_head_state(struct sk_buff *skb);
1143 void kfree_skb_list(struct sk_buff *segs);
1144 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1145 void skb_tx_error(struct sk_buff *skb);
1146
1147 #ifdef CONFIG_TRACEPOINTS
1148 void consume_skb(struct sk_buff *skb);
1149 #else
consume_skb(struct sk_buff * skb)1150 static inline void consume_skb(struct sk_buff *skb)
1151 {
1152 return kfree_skb(skb);
1153 }
1154 #endif
1155
1156 void __consume_stateless_skb(struct sk_buff *skb);
1157 void __kfree_skb(struct sk_buff *skb);
1158 extern struct kmem_cache *skbuff_head_cache;
1159
1160 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1161 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1162 bool *fragstolen, int *delta_truesize);
1163
1164 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1165 int node);
1166 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1167 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1168 struct sk_buff *build_skb_around(struct sk_buff *skb,
1169 void *data, unsigned int frag_size);
1170
1171 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1172
1173 /**
1174 * alloc_skb - allocate a network buffer
1175 * @size: size to allocate
1176 * @priority: allocation mask
1177 *
1178 * This function is a convenient wrapper around __alloc_skb().
1179 */
alloc_skb(unsigned int size,gfp_t priority)1180 static inline struct sk_buff *alloc_skb(unsigned int size,
1181 gfp_t priority)
1182 {
1183 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1184 }
1185
1186 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1187 unsigned long data_len,
1188 int max_page_order,
1189 int *errcode,
1190 gfp_t gfp_mask);
1191 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1192
1193 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1194 struct sk_buff_fclones {
1195 struct sk_buff skb1;
1196
1197 struct sk_buff skb2;
1198
1199 refcount_t fclone_ref;
1200 };
1201
1202 /**
1203 * skb_fclone_busy - check if fclone is busy
1204 * @sk: socket
1205 * @skb: buffer
1206 *
1207 * Returns true if skb is a fast clone, and its clone is not freed.
1208 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1209 * so we also check that this didnt happen.
1210 */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1211 static inline bool skb_fclone_busy(const struct sock *sk,
1212 const struct sk_buff *skb)
1213 {
1214 const struct sk_buff_fclones *fclones;
1215
1216 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1217
1218 return skb->fclone == SKB_FCLONE_ORIG &&
1219 refcount_read(&fclones->fclone_ref) > 1 &&
1220 READ_ONCE(fclones->skb2.sk) == sk;
1221 }
1222
1223 /**
1224 * alloc_skb_fclone - allocate a network buffer from fclone cache
1225 * @size: size to allocate
1226 * @priority: allocation mask
1227 *
1228 * This function is a convenient wrapper around __alloc_skb().
1229 */
alloc_skb_fclone(unsigned int size,gfp_t priority)1230 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1231 gfp_t priority)
1232 {
1233 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1234 }
1235
1236 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1237 void skb_headers_offset_update(struct sk_buff *skb, int off);
1238 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1239 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1240 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1241 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1242 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1243 gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1244 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1245 gfp_t gfp_mask)
1246 {
1247 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1248 }
1249
1250 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1251 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1252 unsigned int headroom);
1253 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1254 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1255 int newtailroom, gfp_t priority);
1256 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1257 int offset, int len);
1258 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1259 int offset, int len);
1260 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1261 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1262
1263 /**
1264 * skb_pad - zero pad the tail of an skb
1265 * @skb: buffer to pad
1266 * @pad: space to pad
1267 *
1268 * Ensure that a buffer is followed by a padding area that is zero
1269 * filled. Used by network drivers which may DMA or transfer data
1270 * beyond the buffer end onto the wire.
1271 *
1272 * May return error in out of memory cases. The skb is freed on error.
1273 */
skb_pad(struct sk_buff * skb,int pad)1274 static inline int skb_pad(struct sk_buff *skb, int pad)
1275 {
1276 return __skb_pad(skb, pad, true);
1277 }
1278 #define dev_kfree_skb(a) consume_skb(a)
1279
1280 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1281 int offset, size_t size);
1282
1283 struct skb_seq_state {
1284 __u32 lower_offset;
1285 __u32 upper_offset;
1286 __u32 frag_idx;
1287 __u32 stepped_offset;
1288 struct sk_buff *root_skb;
1289 struct sk_buff *cur_skb;
1290 __u8 *frag_data;
1291 __u32 frag_off;
1292 };
1293
1294 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1295 unsigned int to, struct skb_seq_state *st);
1296 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1297 struct skb_seq_state *st);
1298 void skb_abort_seq_read(struct skb_seq_state *st);
1299
1300 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1301 unsigned int to, struct ts_config *config);
1302
1303 /*
1304 * Packet hash types specify the type of hash in skb_set_hash.
1305 *
1306 * Hash types refer to the protocol layer addresses which are used to
1307 * construct a packet's hash. The hashes are used to differentiate or identify
1308 * flows of the protocol layer for the hash type. Hash types are either
1309 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1310 *
1311 * Properties of hashes:
1312 *
1313 * 1) Two packets in different flows have different hash values
1314 * 2) Two packets in the same flow should have the same hash value
1315 *
1316 * A hash at a higher layer is considered to be more specific. A driver should
1317 * set the most specific hash possible.
1318 *
1319 * A driver cannot indicate a more specific hash than the layer at which a hash
1320 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1321 *
1322 * A driver may indicate a hash level which is less specific than the
1323 * actual layer the hash was computed on. For instance, a hash computed
1324 * at L4 may be considered an L3 hash. This should only be done if the
1325 * driver can't unambiguously determine that the HW computed the hash at
1326 * the higher layer. Note that the "should" in the second property above
1327 * permits this.
1328 */
1329 enum pkt_hash_types {
1330 PKT_HASH_TYPE_NONE, /* Undefined type */
1331 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1332 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1333 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1334 };
1335
skb_clear_hash(struct sk_buff * skb)1336 static inline void skb_clear_hash(struct sk_buff *skb)
1337 {
1338 skb->hash = 0;
1339 skb->sw_hash = 0;
1340 skb->l4_hash = 0;
1341 }
1342
skb_clear_hash_if_not_l4(struct sk_buff * skb)1343 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1344 {
1345 if (!skb->l4_hash)
1346 skb_clear_hash(skb);
1347 }
1348
1349 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1350 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1351 {
1352 skb->l4_hash = is_l4;
1353 skb->sw_hash = is_sw;
1354 skb->hash = hash;
1355 }
1356
1357 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1358 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1359 {
1360 /* Used by drivers to set hash from HW */
1361 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1362 }
1363
1364 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1365 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1366 {
1367 __skb_set_hash(skb, hash, true, is_l4);
1368 }
1369
1370 void __skb_get_hash(struct sk_buff *skb);
1371 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1372 u32 skb_get_poff(const struct sk_buff *skb);
1373 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1374 const struct flow_keys_basic *keys, int hlen);
1375 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1376 const void *data, int hlen_proto);
1377
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto)1378 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1379 int thoff, u8 ip_proto)
1380 {
1381 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1382 }
1383
1384 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1385 const struct flow_dissector_key *key,
1386 unsigned int key_count);
1387
1388 struct bpf_flow_dissector;
1389 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1390 __be16 proto, int nhoff, int hlen, unsigned int flags);
1391
1392 bool __skb_flow_dissect(const struct net *net,
1393 const struct sk_buff *skb,
1394 struct flow_dissector *flow_dissector,
1395 void *target_container, const void *data,
1396 __be16 proto, int nhoff, int hlen, unsigned int flags);
1397
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1398 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1399 struct flow_dissector *flow_dissector,
1400 void *target_container, unsigned int flags)
1401 {
1402 return __skb_flow_dissect(NULL, skb, flow_dissector,
1403 target_container, NULL, 0, 0, 0, flags);
1404 }
1405
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1406 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1407 struct flow_keys *flow,
1408 unsigned int flags)
1409 {
1410 memset(flow, 0, sizeof(*flow));
1411 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1412 flow, NULL, 0, 0, 0, flags);
1413 }
1414
1415 static inline bool
skb_flow_dissect_flow_keys_basic(const struct net * net,const struct sk_buff * skb,struct flow_keys_basic * flow,const void * data,__be16 proto,int nhoff,int hlen,unsigned int flags)1416 skb_flow_dissect_flow_keys_basic(const struct net *net,
1417 const struct sk_buff *skb,
1418 struct flow_keys_basic *flow,
1419 const void *data, __be16 proto,
1420 int nhoff, int hlen, unsigned int flags)
1421 {
1422 memset(flow, 0, sizeof(*flow));
1423 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1424 data, proto, nhoff, hlen, flags);
1425 }
1426
1427 void skb_flow_dissect_meta(const struct sk_buff *skb,
1428 struct flow_dissector *flow_dissector,
1429 void *target_container);
1430
1431 /* Gets a skb connection tracking info, ctinfo map should be a
1432 * map of mapsize to translate enum ip_conntrack_info states
1433 * to user states.
1434 */
1435 void
1436 skb_flow_dissect_ct(const struct sk_buff *skb,
1437 struct flow_dissector *flow_dissector,
1438 void *target_container,
1439 u16 *ctinfo_map, size_t mapsize,
1440 bool post_ct, u16 zone);
1441 void
1442 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1443 struct flow_dissector *flow_dissector,
1444 void *target_container);
1445
1446 void skb_flow_dissect_hash(const struct sk_buff *skb,
1447 struct flow_dissector *flow_dissector,
1448 void *target_container);
1449
skb_get_hash(struct sk_buff * skb)1450 static inline __u32 skb_get_hash(struct sk_buff *skb)
1451 {
1452 if (!skb->l4_hash && !skb->sw_hash)
1453 __skb_get_hash(skb);
1454
1455 return skb->hash;
1456 }
1457
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1458 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1459 {
1460 if (!skb->l4_hash && !skb->sw_hash) {
1461 struct flow_keys keys;
1462 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1463
1464 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1465 }
1466
1467 return skb->hash;
1468 }
1469
1470 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1471 const siphash_key_t *perturb);
1472
skb_get_hash_raw(const struct sk_buff * skb)1473 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1474 {
1475 return skb->hash;
1476 }
1477
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1478 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1479 {
1480 to->hash = from->hash;
1481 to->sw_hash = from->sw_hash;
1482 to->l4_hash = from->l4_hash;
1483 };
1484
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1485 static inline void skb_copy_decrypted(struct sk_buff *to,
1486 const struct sk_buff *from)
1487 {
1488 #ifdef CONFIG_TLS_DEVICE
1489 to->decrypted = from->decrypted;
1490 #endif
1491 }
1492
1493 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1494 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1495 {
1496 return skb->head + skb->end;
1497 }
1498
skb_end_offset(const struct sk_buff * skb)1499 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1500 {
1501 return skb->end;
1502 }
1503
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1504 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1505 {
1506 skb->end = offset;
1507 }
1508 #else
skb_end_pointer(const struct sk_buff * skb)1509 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1510 {
1511 return skb->end;
1512 }
1513
skb_end_offset(const struct sk_buff * skb)1514 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1515 {
1516 return skb->end - skb->head;
1517 }
1518
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1519 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1520 {
1521 skb->end = skb->head + offset;
1522 }
1523 #endif
1524
1525 /* Internal */
1526 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1527
skb_hwtstamps(struct sk_buff * skb)1528 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1529 {
1530 return &skb_shinfo(skb)->hwtstamps;
1531 }
1532
skb_zcopy(struct sk_buff * skb)1533 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1534 {
1535 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1536
1537 return is_zcopy ? skb_uarg(skb) : NULL;
1538 }
1539
net_zcopy_get(struct ubuf_info * uarg)1540 static inline void net_zcopy_get(struct ubuf_info *uarg)
1541 {
1542 refcount_inc(&uarg->refcnt);
1543 }
1544
skb_zcopy_init(struct sk_buff * skb,struct ubuf_info * uarg)1545 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1546 {
1547 skb_shinfo(skb)->destructor_arg = uarg;
1548 skb_shinfo(skb)->flags |= uarg->flags;
1549 }
1550
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1551 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1552 bool *have_ref)
1553 {
1554 if (skb && uarg && !skb_zcopy(skb)) {
1555 if (unlikely(have_ref && *have_ref))
1556 *have_ref = false;
1557 else
1558 net_zcopy_get(uarg);
1559 skb_zcopy_init(skb, uarg);
1560 }
1561 }
1562
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1563 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1564 {
1565 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1566 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1567 }
1568
skb_zcopy_is_nouarg(struct sk_buff * skb)1569 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1570 {
1571 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1572 }
1573
skb_zcopy_get_nouarg(struct sk_buff * skb)1574 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1575 {
1576 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1577 }
1578
net_zcopy_put(struct ubuf_info * uarg)1579 static inline void net_zcopy_put(struct ubuf_info *uarg)
1580 {
1581 if (uarg)
1582 uarg->callback(NULL, uarg, true);
1583 }
1584
net_zcopy_put_abort(struct ubuf_info * uarg,bool have_uref)1585 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1586 {
1587 if (uarg) {
1588 if (uarg->callback == msg_zerocopy_callback)
1589 msg_zerocopy_put_abort(uarg, have_uref);
1590 else if (have_uref)
1591 net_zcopy_put(uarg);
1592 }
1593 }
1594
1595 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy_success)1596 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1597 {
1598 struct ubuf_info *uarg = skb_zcopy(skb);
1599
1600 if (uarg) {
1601 if (!skb_zcopy_is_nouarg(skb))
1602 uarg->callback(skb, uarg, zerocopy_success);
1603
1604 skb_shinfo(skb)->flags &= ~SKBFL_ZEROCOPY_FRAG;
1605 }
1606 }
1607
skb_mark_not_on_list(struct sk_buff * skb)1608 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1609 {
1610 skb->next = NULL;
1611 }
1612
1613 /* Iterate through singly-linked GSO fragments of an skb. */
1614 #define skb_list_walk_safe(first, skb, next_skb) \
1615 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1616 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1617
skb_list_del_init(struct sk_buff * skb)1618 static inline void skb_list_del_init(struct sk_buff *skb)
1619 {
1620 __list_del_entry(&skb->list);
1621 skb_mark_not_on_list(skb);
1622 }
1623
1624 /**
1625 * skb_queue_empty - check if a queue is empty
1626 * @list: queue head
1627 *
1628 * Returns true if the queue is empty, false otherwise.
1629 */
skb_queue_empty(const struct sk_buff_head * list)1630 static inline int skb_queue_empty(const struct sk_buff_head *list)
1631 {
1632 return list->next == (const struct sk_buff *) list;
1633 }
1634
1635 /**
1636 * skb_queue_empty_lockless - check if a queue is empty
1637 * @list: queue head
1638 *
1639 * Returns true if the queue is empty, false otherwise.
1640 * This variant can be used in lockless contexts.
1641 */
skb_queue_empty_lockless(const struct sk_buff_head * list)1642 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1643 {
1644 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1645 }
1646
1647
1648 /**
1649 * skb_queue_is_last - check if skb is the last entry in the queue
1650 * @list: queue head
1651 * @skb: buffer
1652 *
1653 * Returns true if @skb is the last buffer on the list.
1654 */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1655 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1656 const struct sk_buff *skb)
1657 {
1658 return skb->next == (const struct sk_buff *) list;
1659 }
1660
1661 /**
1662 * skb_queue_is_first - check if skb is the first entry in the queue
1663 * @list: queue head
1664 * @skb: buffer
1665 *
1666 * Returns true if @skb is the first buffer on the list.
1667 */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1668 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1669 const struct sk_buff *skb)
1670 {
1671 return skb->prev == (const struct sk_buff *) list;
1672 }
1673
1674 /**
1675 * skb_queue_next - return the next packet in the queue
1676 * @list: queue head
1677 * @skb: current buffer
1678 *
1679 * Return the next packet in @list after @skb. It is only valid to
1680 * call this if skb_queue_is_last() evaluates to false.
1681 */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1682 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1683 const struct sk_buff *skb)
1684 {
1685 /* This BUG_ON may seem severe, but if we just return then we
1686 * are going to dereference garbage.
1687 */
1688 BUG_ON(skb_queue_is_last(list, skb));
1689 return skb->next;
1690 }
1691
1692 /**
1693 * skb_queue_prev - return the prev packet in the queue
1694 * @list: queue head
1695 * @skb: current buffer
1696 *
1697 * Return the prev packet in @list before @skb. It is only valid to
1698 * call this if skb_queue_is_first() evaluates to false.
1699 */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1700 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1701 const struct sk_buff *skb)
1702 {
1703 /* This BUG_ON may seem severe, but if we just return then we
1704 * are going to dereference garbage.
1705 */
1706 BUG_ON(skb_queue_is_first(list, skb));
1707 return skb->prev;
1708 }
1709
1710 /**
1711 * skb_get - reference buffer
1712 * @skb: buffer to reference
1713 *
1714 * Makes another reference to a socket buffer and returns a pointer
1715 * to the buffer.
1716 */
skb_get(struct sk_buff * skb)1717 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1718 {
1719 refcount_inc(&skb->users);
1720 return skb;
1721 }
1722
1723 /*
1724 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1725 */
1726
1727 /**
1728 * skb_cloned - is the buffer a clone
1729 * @skb: buffer to check
1730 *
1731 * Returns true if the buffer was generated with skb_clone() and is
1732 * one of multiple shared copies of the buffer. Cloned buffers are
1733 * shared data so must not be written to under normal circumstances.
1734 */
skb_cloned(const struct sk_buff * skb)1735 static inline int skb_cloned(const struct sk_buff *skb)
1736 {
1737 return skb->cloned &&
1738 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1739 }
1740
skb_unclone(struct sk_buff * skb,gfp_t pri)1741 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1742 {
1743 might_sleep_if(gfpflags_allow_blocking(pri));
1744
1745 if (skb_cloned(skb))
1746 return pskb_expand_head(skb, 0, 0, pri);
1747
1748 return 0;
1749 }
1750
1751 /* This variant of skb_unclone() makes sure skb->truesize
1752 * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1753 *
1754 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
1755 * when various debugging features are in place.
1756 */
1757 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
skb_unclone_keeptruesize(struct sk_buff * skb,gfp_t pri)1758 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1759 {
1760 might_sleep_if(gfpflags_allow_blocking(pri));
1761
1762 if (skb_cloned(skb))
1763 return __skb_unclone_keeptruesize(skb, pri);
1764 return 0;
1765 }
1766
1767 /**
1768 * skb_header_cloned - is the header a clone
1769 * @skb: buffer to check
1770 *
1771 * Returns true if modifying the header part of the buffer requires
1772 * the data to be copied.
1773 */
skb_header_cloned(const struct sk_buff * skb)1774 static inline int skb_header_cloned(const struct sk_buff *skb)
1775 {
1776 int dataref;
1777
1778 if (!skb->cloned)
1779 return 0;
1780
1781 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1782 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1783 return dataref != 1;
1784 }
1785
skb_header_unclone(struct sk_buff * skb,gfp_t pri)1786 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1787 {
1788 might_sleep_if(gfpflags_allow_blocking(pri));
1789
1790 if (skb_header_cloned(skb))
1791 return pskb_expand_head(skb, 0, 0, pri);
1792
1793 return 0;
1794 }
1795
1796 /**
1797 * __skb_header_release - release reference to header
1798 * @skb: buffer to operate on
1799 */
__skb_header_release(struct sk_buff * skb)1800 static inline void __skb_header_release(struct sk_buff *skb)
1801 {
1802 skb->nohdr = 1;
1803 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1804 }
1805
1806
1807 /**
1808 * skb_shared - is the buffer shared
1809 * @skb: buffer to check
1810 *
1811 * Returns true if more than one person has a reference to this
1812 * buffer.
1813 */
skb_shared(const struct sk_buff * skb)1814 static inline int skb_shared(const struct sk_buff *skb)
1815 {
1816 return refcount_read(&skb->users) != 1;
1817 }
1818
1819 /**
1820 * skb_share_check - check if buffer is shared and if so clone it
1821 * @skb: buffer to check
1822 * @pri: priority for memory allocation
1823 *
1824 * If the buffer is shared the buffer is cloned and the old copy
1825 * drops a reference. A new clone with a single reference is returned.
1826 * If the buffer is not shared the original buffer is returned. When
1827 * being called from interrupt status or with spinlocks held pri must
1828 * be GFP_ATOMIC.
1829 *
1830 * NULL is returned on a memory allocation failure.
1831 */
skb_share_check(struct sk_buff * skb,gfp_t pri)1832 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1833 {
1834 might_sleep_if(gfpflags_allow_blocking(pri));
1835 if (skb_shared(skb)) {
1836 struct sk_buff *nskb = skb_clone(skb, pri);
1837
1838 if (likely(nskb))
1839 consume_skb(skb);
1840 else
1841 kfree_skb(skb);
1842 skb = nskb;
1843 }
1844 return skb;
1845 }
1846
1847 /*
1848 * Copy shared buffers into a new sk_buff. We effectively do COW on
1849 * packets to handle cases where we have a local reader and forward
1850 * and a couple of other messy ones. The normal one is tcpdumping
1851 * a packet thats being forwarded.
1852 */
1853
1854 /**
1855 * skb_unshare - make a copy of a shared buffer
1856 * @skb: buffer to check
1857 * @pri: priority for memory allocation
1858 *
1859 * If the socket buffer is a clone then this function creates a new
1860 * copy of the data, drops a reference count on the old copy and returns
1861 * the new copy with the reference count at 1. If the buffer is not a clone
1862 * the original buffer is returned. When called with a spinlock held or
1863 * from interrupt state @pri must be %GFP_ATOMIC
1864 *
1865 * %NULL is returned on a memory allocation failure.
1866 */
skb_unshare(struct sk_buff * skb,gfp_t pri)1867 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1868 gfp_t pri)
1869 {
1870 might_sleep_if(gfpflags_allow_blocking(pri));
1871 if (skb_cloned(skb)) {
1872 struct sk_buff *nskb = skb_copy(skb, pri);
1873
1874 /* Free our shared copy */
1875 if (likely(nskb))
1876 consume_skb(skb);
1877 else
1878 kfree_skb(skb);
1879 skb = nskb;
1880 }
1881 return skb;
1882 }
1883
1884 /**
1885 * skb_peek - peek at the head of an &sk_buff_head
1886 * @list_: list to peek at
1887 *
1888 * Peek an &sk_buff. Unlike most other operations you _MUST_
1889 * be careful with this one. A peek leaves the buffer on the
1890 * list and someone else may run off with it. You must hold
1891 * the appropriate locks or have a private queue to do this.
1892 *
1893 * Returns %NULL for an empty list or a pointer to the head element.
1894 * The reference count is not incremented and the reference is therefore
1895 * volatile. Use with caution.
1896 */
skb_peek(const struct sk_buff_head * list_)1897 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1898 {
1899 struct sk_buff *skb = list_->next;
1900
1901 if (skb == (struct sk_buff *)list_)
1902 skb = NULL;
1903 return skb;
1904 }
1905
1906 /**
1907 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1908 * @list_: list to peek at
1909 *
1910 * Like skb_peek(), but the caller knows that the list is not empty.
1911 */
__skb_peek(const struct sk_buff_head * list_)1912 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1913 {
1914 return list_->next;
1915 }
1916
1917 /**
1918 * skb_peek_next - peek skb following the given one from a queue
1919 * @skb: skb to start from
1920 * @list_: list to peek at
1921 *
1922 * Returns %NULL when the end of the list is met or a pointer to the
1923 * next element. The reference count is not incremented and the
1924 * reference is therefore volatile. Use with caution.
1925 */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)1926 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1927 const struct sk_buff_head *list_)
1928 {
1929 struct sk_buff *next = skb->next;
1930
1931 if (next == (struct sk_buff *)list_)
1932 next = NULL;
1933 return next;
1934 }
1935
1936 /**
1937 * skb_peek_tail - peek at the tail of an &sk_buff_head
1938 * @list_: list to peek at
1939 *
1940 * Peek an &sk_buff. Unlike most other operations you _MUST_
1941 * be careful with this one. A peek leaves the buffer on the
1942 * list and someone else may run off with it. You must hold
1943 * the appropriate locks or have a private queue to do this.
1944 *
1945 * Returns %NULL for an empty list or a pointer to the tail element.
1946 * The reference count is not incremented and the reference is therefore
1947 * volatile. Use with caution.
1948 */
skb_peek_tail(const struct sk_buff_head * list_)1949 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1950 {
1951 struct sk_buff *skb = READ_ONCE(list_->prev);
1952
1953 if (skb == (struct sk_buff *)list_)
1954 skb = NULL;
1955 return skb;
1956
1957 }
1958
1959 /**
1960 * skb_queue_len - get queue length
1961 * @list_: list to measure
1962 *
1963 * Return the length of an &sk_buff queue.
1964 */
skb_queue_len(const struct sk_buff_head * list_)1965 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1966 {
1967 return list_->qlen;
1968 }
1969
1970 /**
1971 * skb_queue_len_lockless - get queue length
1972 * @list_: list to measure
1973 *
1974 * Return the length of an &sk_buff queue.
1975 * This variant can be used in lockless contexts.
1976 */
skb_queue_len_lockless(const struct sk_buff_head * list_)1977 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1978 {
1979 return READ_ONCE(list_->qlen);
1980 }
1981
1982 /**
1983 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1984 * @list: queue to initialize
1985 *
1986 * This initializes only the list and queue length aspects of
1987 * an sk_buff_head object. This allows to initialize the list
1988 * aspects of an sk_buff_head without reinitializing things like
1989 * the spinlock. It can also be used for on-stack sk_buff_head
1990 * objects where the spinlock is known to not be used.
1991 */
__skb_queue_head_init(struct sk_buff_head * list)1992 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1993 {
1994 list->prev = list->next = (struct sk_buff *)list;
1995 list->qlen = 0;
1996 }
1997
1998 /*
1999 * This function creates a split out lock class for each invocation;
2000 * this is needed for now since a whole lot of users of the skb-queue
2001 * infrastructure in drivers have different locking usage (in hardirq)
2002 * than the networking core (in softirq only). In the long run either the
2003 * network layer or drivers should need annotation to consolidate the
2004 * main types of usage into 3 classes.
2005 */
skb_queue_head_init(struct sk_buff_head * list)2006 static inline void skb_queue_head_init(struct sk_buff_head *list)
2007 {
2008 spin_lock_init(&list->lock);
2009 __skb_queue_head_init(list);
2010 }
2011
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)2012 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
2013 struct lock_class_key *class)
2014 {
2015 skb_queue_head_init(list);
2016 lockdep_set_class(&list->lock, class);
2017 }
2018
2019 /*
2020 * Insert an sk_buff on a list.
2021 *
2022 * The "__skb_xxxx()" functions are the non-atomic ones that
2023 * can only be called with interrupts disabled.
2024 */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)2025 static inline void __skb_insert(struct sk_buff *newsk,
2026 struct sk_buff *prev, struct sk_buff *next,
2027 struct sk_buff_head *list)
2028 {
2029 /* See skb_queue_empty_lockless() and skb_peek_tail()
2030 * for the opposite READ_ONCE()
2031 */
2032 WRITE_ONCE(newsk->next, next);
2033 WRITE_ONCE(newsk->prev, prev);
2034 WRITE_ONCE(next->prev, newsk);
2035 WRITE_ONCE(prev->next, newsk);
2036 WRITE_ONCE(list->qlen, list->qlen + 1);
2037 }
2038
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)2039 static inline void __skb_queue_splice(const struct sk_buff_head *list,
2040 struct sk_buff *prev,
2041 struct sk_buff *next)
2042 {
2043 struct sk_buff *first = list->next;
2044 struct sk_buff *last = list->prev;
2045
2046 WRITE_ONCE(first->prev, prev);
2047 WRITE_ONCE(prev->next, first);
2048
2049 WRITE_ONCE(last->next, next);
2050 WRITE_ONCE(next->prev, last);
2051 }
2052
2053 /**
2054 * skb_queue_splice - join two skb lists, this is designed for stacks
2055 * @list: the new list to add
2056 * @head: the place to add it in the first list
2057 */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)2058 static inline void skb_queue_splice(const struct sk_buff_head *list,
2059 struct sk_buff_head *head)
2060 {
2061 if (!skb_queue_empty(list)) {
2062 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2063 head->qlen += list->qlen;
2064 }
2065 }
2066
2067 /**
2068 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2069 * @list: the new list to add
2070 * @head: the place to add it in the first list
2071 *
2072 * The list at @list is reinitialised
2073 */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)2074 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2075 struct sk_buff_head *head)
2076 {
2077 if (!skb_queue_empty(list)) {
2078 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2079 head->qlen += list->qlen;
2080 __skb_queue_head_init(list);
2081 }
2082 }
2083
2084 /**
2085 * skb_queue_splice_tail - join two skb lists, each list being a queue
2086 * @list: the new list to add
2087 * @head: the place to add it in the first list
2088 */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)2089 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2090 struct sk_buff_head *head)
2091 {
2092 if (!skb_queue_empty(list)) {
2093 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2094 head->qlen += list->qlen;
2095 }
2096 }
2097
2098 /**
2099 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2100 * @list: the new list to add
2101 * @head: the place to add it in the first list
2102 *
2103 * Each of the lists is a queue.
2104 * The list at @list is reinitialised
2105 */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)2106 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2107 struct sk_buff_head *head)
2108 {
2109 if (!skb_queue_empty(list)) {
2110 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2111 head->qlen += list->qlen;
2112 __skb_queue_head_init(list);
2113 }
2114 }
2115
2116 /**
2117 * __skb_queue_after - queue a buffer at the list head
2118 * @list: list to use
2119 * @prev: place after this buffer
2120 * @newsk: buffer to queue
2121 *
2122 * Queue a buffer int the middle of a list. This function takes no locks
2123 * and you must therefore hold required locks before calling it.
2124 *
2125 * A buffer cannot be placed on two lists at the same time.
2126 */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)2127 static inline void __skb_queue_after(struct sk_buff_head *list,
2128 struct sk_buff *prev,
2129 struct sk_buff *newsk)
2130 {
2131 __skb_insert(newsk, prev, prev->next, list);
2132 }
2133
2134 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2135 struct sk_buff_head *list);
2136
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)2137 static inline void __skb_queue_before(struct sk_buff_head *list,
2138 struct sk_buff *next,
2139 struct sk_buff *newsk)
2140 {
2141 __skb_insert(newsk, next->prev, next, list);
2142 }
2143
2144 /**
2145 * __skb_queue_head - queue a buffer at the list head
2146 * @list: list to use
2147 * @newsk: buffer to queue
2148 *
2149 * Queue a buffer at the start of a list. This function takes no locks
2150 * and you must therefore hold required locks before calling it.
2151 *
2152 * A buffer cannot be placed on two lists at the same time.
2153 */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)2154 static inline void __skb_queue_head(struct sk_buff_head *list,
2155 struct sk_buff *newsk)
2156 {
2157 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2158 }
2159 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2160
2161 /**
2162 * __skb_queue_tail - queue a buffer at the list tail
2163 * @list: list to use
2164 * @newsk: buffer to queue
2165 *
2166 * Queue a buffer at the end of a list. This function takes no locks
2167 * and you must therefore hold required locks before calling it.
2168 *
2169 * A buffer cannot be placed on two lists at the same time.
2170 */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2171 static inline void __skb_queue_tail(struct sk_buff_head *list,
2172 struct sk_buff *newsk)
2173 {
2174 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2175 }
2176 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2177
2178 /*
2179 * remove sk_buff from list. _Must_ be called atomically, and with
2180 * the list known..
2181 */
2182 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2183 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2184 {
2185 struct sk_buff *next, *prev;
2186
2187 WRITE_ONCE(list->qlen, list->qlen - 1);
2188 next = skb->next;
2189 prev = skb->prev;
2190 skb->next = skb->prev = NULL;
2191 WRITE_ONCE(next->prev, prev);
2192 WRITE_ONCE(prev->next, next);
2193 }
2194
2195 /**
2196 * __skb_dequeue - remove from the head of the queue
2197 * @list: list to dequeue from
2198 *
2199 * Remove the head of the list. This function does not take any locks
2200 * so must be used with appropriate locks held only. The head item is
2201 * returned or %NULL if the list is empty.
2202 */
__skb_dequeue(struct sk_buff_head * list)2203 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2204 {
2205 struct sk_buff *skb = skb_peek(list);
2206 if (skb)
2207 __skb_unlink(skb, list);
2208 return skb;
2209 }
2210 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2211
2212 /**
2213 * __skb_dequeue_tail - remove from the tail of the queue
2214 * @list: list to dequeue from
2215 *
2216 * Remove the tail of the list. This function does not take any locks
2217 * so must be used with appropriate locks held only. The tail item is
2218 * returned or %NULL if the list is empty.
2219 */
__skb_dequeue_tail(struct sk_buff_head * list)2220 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2221 {
2222 struct sk_buff *skb = skb_peek_tail(list);
2223 if (skb)
2224 __skb_unlink(skb, list);
2225 return skb;
2226 }
2227 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2228
2229
skb_is_nonlinear(const struct sk_buff * skb)2230 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2231 {
2232 return skb->data_len;
2233 }
2234
skb_headlen(const struct sk_buff * skb)2235 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2236 {
2237 return skb->len - skb->data_len;
2238 }
2239
__skb_pagelen(const struct sk_buff * skb)2240 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2241 {
2242 unsigned int i, len = 0;
2243
2244 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2245 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2246 return len;
2247 }
2248
skb_pagelen(const struct sk_buff * skb)2249 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2250 {
2251 return skb_headlen(skb) + __skb_pagelen(skb);
2252 }
2253
__skb_fill_page_desc_noacc(struct skb_shared_info * shinfo,int i,struct page * page,int off,int size)2254 static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo,
2255 int i, struct page *page,
2256 int off, int size)
2257 {
2258 skb_frag_t *frag = &shinfo->frags[i];
2259
2260 /*
2261 * Propagate page pfmemalloc to the skb if we can. The problem is
2262 * that not all callers have unique ownership of the page but rely
2263 * on page_is_pfmemalloc doing the right thing(tm).
2264 */
2265 frag->bv_page = page;
2266 frag->bv_offset = off;
2267 skb_frag_size_set(frag, size);
2268 }
2269
2270 /**
2271 * __skb_fill_page_desc - initialise a paged fragment in an skb
2272 * @skb: buffer containing fragment to be initialised
2273 * @i: paged fragment index to initialise
2274 * @page: the page to use for this fragment
2275 * @off: the offset to the data with @page
2276 * @size: the length of the data
2277 *
2278 * Initialises the @i'th fragment of @skb to point to &size bytes at
2279 * offset @off within @page.
2280 *
2281 * Does not take any additional reference on the fragment.
2282 */
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2283 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2284 struct page *page, int off, int size)
2285 {
2286 __skb_fill_page_desc_noacc(skb_shinfo(skb), i, page, off, size);
2287 page = compound_head(page);
2288 if (page_is_pfmemalloc(page))
2289 skb->pfmemalloc = true;
2290 }
2291
2292 /**
2293 * skb_fill_page_desc - initialise a paged fragment in an skb
2294 * @skb: buffer containing fragment to be initialised
2295 * @i: paged fragment index to initialise
2296 * @page: the page to use for this fragment
2297 * @off: the offset to the data with @page
2298 * @size: the length of the data
2299 *
2300 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2301 * @skb to point to @size bytes at offset @off within @page. In
2302 * addition updates @skb such that @i is the last fragment.
2303 *
2304 * Does not take any additional reference on the fragment.
2305 */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2306 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2307 struct page *page, int off, int size)
2308 {
2309 __skb_fill_page_desc(skb, i, page, off, size);
2310 skb_shinfo(skb)->nr_frags = i + 1;
2311 }
2312
2313 /**
2314 * skb_fill_page_desc_noacc - initialise a paged fragment in an skb
2315 * @skb: buffer containing fragment to be initialised
2316 * @i: paged fragment index to initialise
2317 * @page: the page to use for this fragment
2318 * @off: the offset to the data with @page
2319 * @size: the length of the data
2320 *
2321 * Variant of skb_fill_page_desc() which does not deal with
2322 * pfmemalloc, if page is not owned by us.
2323 */
skb_fill_page_desc_noacc(struct sk_buff * skb,int i,struct page * page,int off,int size)2324 static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i,
2325 struct page *page, int off,
2326 int size)
2327 {
2328 struct skb_shared_info *shinfo = skb_shinfo(skb);
2329
2330 __skb_fill_page_desc_noacc(shinfo, i, page, off, size);
2331 shinfo->nr_frags = i + 1;
2332 }
2333
2334 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2335 int size, unsigned int truesize);
2336
2337 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2338 unsigned int truesize);
2339
2340 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2341
2342 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2343 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2344 {
2345 return skb->head + skb->tail;
2346 }
2347
skb_reset_tail_pointer(struct sk_buff * skb)2348 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2349 {
2350 skb->tail = skb->data - skb->head;
2351 }
2352
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2353 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2354 {
2355 skb_reset_tail_pointer(skb);
2356 skb->tail += offset;
2357 }
2358
2359 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2360 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2361 {
2362 return skb->tail;
2363 }
2364
skb_reset_tail_pointer(struct sk_buff * skb)2365 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2366 {
2367 skb->tail = skb->data;
2368 }
2369
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2370 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2371 {
2372 skb->tail = skb->data + offset;
2373 }
2374
2375 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2376
skb_assert_len(struct sk_buff * skb)2377 static inline void skb_assert_len(struct sk_buff *skb)
2378 {
2379 #ifdef CONFIG_DEBUG_NET
2380 if (WARN_ONCE(!skb->len, "%s\n", __func__))
2381 DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2382 #endif /* CONFIG_DEBUG_NET */
2383 }
2384
2385 /*
2386 * Add data to an sk_buff
2387 */
2388 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2389 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2390 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2391 {
2392 void *tmp = skb_tail_pointer(skb);
2393 SKB_LINEAR_ASSERT(skb);
2394 skb->tail += len;
2395 skb->len += len;
2396 return tmp;
2397 }
2398
__skb_put_zero(struct sk_buff * skb,unsigned int len)2399 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2400 {
2401 void *tmp = __skb_put(skb, len);
2402
2403 memset(tmp, 0, len);
2404 return tmp;
2405 }
2406
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2407 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2408 unsigned int len)
2409 {
2410 void *tmp = __skb_put(skb, len);
2411
2412 memcpy(tmp, data, len);
2413 return tmp;
2414 }
2415
__skb_put_u8(struct sk_buff * skb,u8 val)2416 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2417 {
2418 *(u8 *)__skb_put(skb, 1) = val;
2419 }
2420
skb_put_zero(struct sk_buff * skb,unsigned int len)2421 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2422 {
2423 void *tmp = skb_put(skb, len);
2424
2425 memset(tmp, 0, len);
2426
2427 return tmp;
2428 }
2429
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2430 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2431 unsigned int len)
2432 {
2433 void *tmp = skb_put(skb, len);
2434
2435 memcpy(tmp, data, len);
2436
2437 return tmp;
2438 }
2439
skb_put_u8(struct sk_buff * skb,u8 val)2440 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2441 {
2442 *(u8 *)skb_put(skb, 1) = val;
2443 }
2444
2445 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2446 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2447 {
2448 skb->data -= len;
2449 skb->len += len;
2450 return skb->data;
2451 }
2452
2453 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2454 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2455 {
2456 skb->len -= len;
2457 BUG_ON(skb->len < skb->data_len);
2458 return skb->data += len;
2459 }
2460
skb_pull_inline(struct sk_buff * skb,unsigned int len)2461 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2462 {
2463 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2464 }
2465
2466 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2467
__pskb_pull(struct sk_buff * skb,unsigned int len)2468 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2469 {
2470 if (len > skb_headlen(skb) &&
2471 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2472 return NULL;
2473 skb->len -= len;
2474 return skb->data += len;
2475 }
2476
pskb_pull(struct sk_buff * skb,unsigned int len)2477 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2478 {
2479 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2480 }
2481
pskb_may_pull(struct sk_buff * skb,unsigned int len)2482 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2483 {
2484 if (likely(len <= skb_headlen(skb)))
2485 return true;
2486 if (unlikely(len > skb->len))
2487 return false;
2488 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2489 }
2490
2491 void skb_condense(struct sk_buff *skb);
2492
2493 /**
2494 * skb_headroom - bytes at buffer head
2495 * @skb: buffer to check
2496 *
2497 * Return the number of bytes of free space at the head of an &sk_buff.
2498 */
skb_headroom(const struct sk_buff * skb)2499 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2500 {
2501 return skb->data - skb->head;
2502 }
2503
2504 /**
2505 * skb_tailroom - bytes at buffer end
2506 * @skb: buffer to check
2507 *
2508 * Return the number of bytes of free space at the tail of an sk_buff
2509 */
skb_tailroom(const struct sk_buff * skb)2510 static inline int skb_tailroom(const struct sk_buff *skb)
2511 {
2512 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2513 }
2514
2515 /**
2516 * skb_availroom - bytes at buffer end
2517 * @skb: buffer to check
2518 *
2519 * Return the number of bytes of free space at the tail of an sk_buff
2520 * allocated by sk_stream_alloc()
2521 */
skb_availroom(const struct sk_buff * skb)2522 static inline int skb_availroom(const struct sk_buff *skb)
2523 {
2524 if (skb_is_nonlinear(skb))
2525 return 0;
2526
2527 return skb->end - skb->tail - skb->reserved_tailroom;
2528 }
2529
2530 /**
2531 * skb_reserve - adjust headroom
2532 * @skb: buffer to alter
2533 * @len: bytes to move
2534 *
2535 * Increase the headroom of an empty &sk_buff by reducing the tail
2536 * room. This is only allowed for an empty buffer.
2537 */
skb_reserve(struct sk_buff * skb,int len)2538 static inline void skb_reserve(struct sk_buff *skb, int len)
2539 {
2540 skb->data += len;
2541 skb->tail += len;
2542 }
2543
2544 /**
2545 * skb_tailroom_reserve - adjust reserved_tailroom
2546 * @skb: buffer to alter
2547 * @mtu: maximum amount of headlen permitted
2548 * @needed_tailroom: minimum amount of reserved_tailroom
2549 *
2550 * Set reserved_tailroom so that headlen can be as large as possible but
2551 * not larger than mtu and tailroom cannot be smaller than
2552 * needed_tailroom.
2553 * The required headroom should already have been reserved before using
2554 * this function.
2555 */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2556 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2557 unsigned int needed_tailroom)
2558 {
2559 SKB_LINEAR_ASSERT(skb);
2560 if (mtu < skb_tailroom(skb) - needed_tailroom)
2561 /* use at most mtu */
2562 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2563 else
2564 /* use up to all available space */
2565 skb->reserved_tailroom = needed_tailroom;
2566 }
2567
2568 #define ENCAP_TYPE_ETHER 0
2569 #define ENCAP_TYPE_IPPROTO 1
2570
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2571 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2572 __be16 protocol)
2573 {
2574 skb->inner_protocol = protocol;
2575 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2576 }
2577
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2578 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2579 __u8 ipproto)
2580 {
2581 skb->inner_ipproto = ipproto;
2582 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2583 }
2584
skb_reset_inner_headers(struct sk_buff * skb)2585 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2586 {
2587 skb->inner_mac_header = skb->mac_header;
2588 skb->inner_network_header = skb->network_header;
2589 skb->inner_transport_header = skb->transport_header;
2590 }
2591
skb_reset_mac_len(struct sk_buff * skb)2592 static inline void skb_reset_mac_len(struct sk_buff *skb)
2593 {
2594 skb->mac_len = skb->network_header - skb->mac_header;
2595 }
2596
skb_inner_transport_header(const struct sk_buff * skb)2597 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2598 *skb)
2599 {
2600 return skb->head + skb->inner_transport_header;
2601 }
2602
skb_inner_transport_offset(const struct sk_buff * skb)2603 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2604 {
2605 return skb_inner_transport_header(skb) - skb->data;
2606 }
2607
skb_reset_inner_transport_header(struct sk_buff * skb)2608 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2609 {
2610 skb->inner_transport_header = skb->data - skb->head;
2611 }
2612
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2613 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2614 const int offset)
2615 {
2616 skb_reset_inner_transport_header(skb);
2617 skb->inner_transport_header += offset;
2618 }
2619
skb_inner_network_header(const struct sk_buff * skb)2620 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2621 {
2622 return skb->head + skb->inner_network_header;
2623 }
2624
skb_reset_inner_network_header(struct sk_buff * skb)2625 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2626 {
2627 skb->inner_network_header = skb->data - skb->head;
2628 }
2629
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2630 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2631 const int offset)
2632 {
2633 skb_reset_inner_network_header(skb);
2634 skb->inner_network_header += offset;
2635 }
2636
skb_inner_mac_header(const struct sk_buff * skb)2637 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2638 {
2639 return skb->head + skb->inner_mac_header;
2640 }
2641
skb_reset_inner_mac_header(struct sk_buff * skb)2642 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2643 {
2644 skb->inner_mac_header = skb->data - skb->head;
2645 }
2646
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)2647 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2648 const int offset)
2649 {
2650 skb_reset_inner_mac_header(skb);
2651 skb->inner_mac_header += offset;
2652 }
skb_transport_header_was_set(const struct sk_buff * skb)2653 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2654 {
2655 return skb->transport_header != (typeof(skb->transport_header))~0U;
2656 }
2657
skb_transport_header(const struct sk_buff * skb)2658 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2659 {
2660 return skb->head + skb->transport_header;
2661 }
2662
skb_reset_transport_header(struct sk_buff * skb)2663 static inline void skb_reset_transport_header(struct sk_buff *skb)
2664 {
2665 skb->transport_header = skb->data - skb->head;
2666 }
2667
skb_set_transport_header(struct sk_buff * skb,const int offset)2668 static inline void skb_set_transport_header(struct sk_buff *skb,
2669 const int offset)
2670 {
2671 skb_reset_transport_header(skb);
2672 skb->transport_header += offset;
2673 }
2674
skb_network_header(const struct sk_buff * skb)2675 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2676 {
2677 return skb->head + skb->network_header;
2678 }
2679
skb_reset_network_header(struct sk_buff * skb)2680 static inline void skb_reset_network_header(struct sk_buff *skb)
2681 {
2682 skb->network_header = skb->data - skb->head;
2683 }
2684
skb_set_network_header(struct sk_buff * skb,const int offset)2685 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2686 {
2687 skb_reset_network_header(skb);
2688 skb->network_header += offset;
2689 }
2690
skb_mac_header(const struct sk_buff * skb)2691 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2692 {
2693 return skb->head + skb->mac_header;
2694 }
2695
skb_mac_offset(const struct sk_buff * skb)2696 static inline int skb_mac_offset(const struct sk_buff *skb)
2697 {
2698 return skb_mac_header(skb) - skb->data;
2699 }
2700
skb_mac_header_len(const struct sk_buff * skb)2701 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2702 {
2703 return skb->network_header - skb->mac_header;
2704 }
2705
skb_mac_header_was_set(const struct sk_buff * skb)2706 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2707 {
2708 return skb->mac_header != (typeof(skb->mac_header))~0U;
2709 }
2710
skb_unset_mac_header(struct sk_buff * skb)2711 static inline void skb_unset_mac_header(struct sk_buff *skb)
2712 {
2713 skb->mac_header = (typeof(skb->mac_header))~0U;
2714 }
2715
skb_reset_mac_header(struct sk_buff * skb)2716 static inline void skb_reset_mac_header(struct sk_buff *skb)
2717 {
2718 skb->mac_header = skb->data - skb->head;
2719 }
2720
skb_set_mac_header(struct sk_buff * skb,const int offset)2721 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2722 {
2723 skb_reset_mac_header(skb);
2724 skb->mac_header += offset;
2725 }
2726
skb_pop_mac_header(struct sk_buff * skb)2727 static inline void skb_pop_mac_header(struct sk_buff *skb)
2728 {
2729 skb->mac_header = skb->network_header;
2730 }
2731
skb_probe_transport_header(struct sk_buff * skb)2732 static inline void skb_probe_transport_header(struct sk_buff *skb)
2733 {
2734 struct flow_keys_basic keys;
2735
2736 if (skb_transport_header_was_set(skb))
2737 return;
2738
2739 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2740 NULL, 0, 0, 0, 0))
2741 skb_set_transport_header(skb, keys.control.thoff);
2742 }
2743
skb_mac_header_rebuild(struct sk_buff * skb)2744 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2745 {
2746 if (skb_mac_header_was_set(skb)) {
2747 const unsigned char *old_mac = skb_mac_header(skb);
2748
2749 skb_set_mac_header(skb, -skb->mac_len);
2750 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2751 }
2752 }
2753
skb_checksum_start_offset(const struct sk_buff * skb)2754 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2755 {
2756 return skb->csum_start - skb_headroom(skb);
2757 }
2758
skb_checksum_start(const struct sk_buff * skb)2759 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2760 {
2761 return skb->head + skb->csum_start;
2762 }
2763
skb_transport_offset(const struct sk_buff * skb)2764 static inline int skb_transport_offset(const struct sk_buff *skb)
2765 {
2766 return skb_transport_header(skb) - skb->data;
2767 }
2768
skb_network_header_len(const struct sk_buff * skb)2769 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2770 {
2771 return skb->transport_header - skb->network_header;
2772 }
2773
skb_inner_network_header_len(const struct sk_buff * skb)2774 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2775 {
2776 return skb->inner_transport_header - skb->inner_network_header;
2777 }
2778
skb_network_offset(const struct sk_buff * skb)2779 static inline int skb_network_offset(const struct sk_buff *skb)
2780 {
2781 return skb_network_header(skb) - skb->data;
2782 }
2783
skb_inner_network_offset(const struct sk_buff * skb)2784 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2785 {
2786 return skb_inner_network_header(skb) - skb->data;
2787 }
2788
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)2789 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2790 {
2791 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2792 }
2793
2794 /*
2795 * CPUs often take a performance hit when accessing unaligned memory
2796 * locations. The actual performance hit varies, it can be small if the
2797 * hardware handles it or large if we have to take an exception and fix it
2798 * in software.
2799 *
2800 * Since an ethernet header is 14 bytes network drivers often end up with
2801 * the IP header at an unaligned offset. The IP header can be aligned by
2802 * shifting the start of the packet by 2 bytes. Drivers should do this
2803 * with:
2804 *
2805 * skb_reserve(skb, NET_IP_ALIGN);
2806 *
2807 * The downside to this alignment of the IP header is that the DMA is now
2808 * unaligned. On some architectures the cost of an unaligned DMA is high
2809 * and this cost outweighs the gains made by aligning the IP header.
2810 *
2811 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2812 * to be overridden.
2813 */
2814 #ifndef NET_IP_ALIGN
2815 #define NET_IP_ALIGN 2
2816 #endif
2817
2818 /*
2819 * The networking layer reserves some headroom in skb data (via
2820 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2821 * the header has to grow. In the default case, if the header has to grow
2822 * 32 bytes or less we avoid the reallocation.
2823 *
2824 * Unfortunately this headroom changes the DMA alignment of the resulting
2825 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2826 * on some architectures. An architecture can override this value,
2827 * perhaps setting it to a cacheline in size (since that will maintain
2828 * cacheline alignment of the DMA). It must be a power of 2.
2829 *
2830 * Various parts of the networking layer expect at least 32 bytes of
2831 * headroom, you should not reduce this.
2832 *
2833 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2834 * to reduce average number of cache lines per packet.
2835 * get_rps_cpu() for example only access one 64 bytes aligned block :
2836 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2837 */
2838 #ifndef NET_SKB_PAD
2839 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2840 #endif
2841
2842 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2843
__skb_set_length(struct sk_buff * skb,unsigned int len)2844 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2845 {
2846 if (WARN_ON(skb_is_nonlinear(skb)))
2847 return;
2848 skb->len = len;
2849 skb_set_tail_pointer(skb, len);
2850 }
2851
__skb_trim(struct sk_buff * skb,unsigned int len)2852 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2853 {
2854 __skb_set_length(skb, len);
2855 }
2856
2857 void skb_trim(struct sk_buff *skb, unsigned int len);
2858
__pskb_trim(struct sk_buff * skb,unsigned int len)2859 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2860 {
2861 if (skb->data_len)
2862 return ___pskb_trim(skb, len);
2863 __skb_trim(skb, len);
2864 return 0;
2865 }
2866
pskb_trim(struct sk_buff * skb,unsigned int len)2867 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2868 {
2869 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2870 }
2871
2872 /**
2873 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2874 * @skb: buffer to alter
2875 * @len: new length
2876 *
2877 * This is identical to pskb_trim except that the caller knows that
2878 * the skb is not cloned so we should never get an error due to out-
2879 * of-memory.
2880 */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)2881 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2882 {
2883 int err = pskb_trim(skb, len);
2884 BUG_ON(err);
2885 }
2886
__skb_grow(struct sk_buff * skb,unsigned int len)2887 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2888 {
2889 unsigned int diff = len - skb->len;
2890
2891 if (skb_tailroom(skb) < diff) {
2892 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2893 GFP_ATOMIC);
2894 if (ret)
2895 return ret;
2896 }
2897 __skb_set_length(skb, len);
2898 return 0;
2899 }
2900
2901 /**
2902 * skb_orphan - orphan a buffer
2903 * @skb: buffer to orphan
2904 *
2905 * If a buffer currently has an owner then we call the owner's
2906 * destructor function and make the @skb unowned. The buffer continues
2907 * to exist but is no longer charged to its former owner.
2908 */
skb_orphan(struct sk_buff * skb)2909 static inline void skb_orphan(struct sk_buff *skb)
2910 {
2911 if (skb->destructor) {
2912 skb->destructor(skb);
2913 skb->destructor = NULL;
2914 skb->sk = NULL;
2915 } else {
2916 BUG_ON(skb->sk);
2917 }
2918 }
2919
2920 /**
2921 * skb_orphan_frags - orphan the frags contained in a buffer
2922 * @skb: buffer to orphan frags from
2923 * @gfp_mask: allocation mask for replacement pages
2924 *
2925 * For each frag in the SKB which needs a destructor (i.e. has an
2926 * owner) create a copy of that frag and release the original
2927 * page by calling the destructor.
2928 */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)2929 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2930 {
2931 if (likely(!skb_zcopy(skb)))
2932 return 0;
2933 if (!skb_zcopy_is_nouarg(skb) &&
2934 skb_uarg(skb)->callback == msg_zerocopy_callback)
2935 return 0;
2936 return skb_copy_ubufs(skb, gfp_mask);
2937 }
2938
2939 /* 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)2940 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2941 {
2942 if (likely(!skb_zcopy(skb)))
2943 return 0;
2944 return skb_copy_ubufs(skb, gfp_mask);
2945 }
2946
2947 /**
2948 * __skb_queue_purge - empty a list
2949 * @list: list to empty
2950 *
2951 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2952 * the list and one reference dropped. This function does not take the
2953 * list lock and the caller must hold the relevant locks to use it.
2954 */
__skb_queue_purge(struct sk_buff_head * list)2955 static inline void __skb_queue_purge(struct sk_buff_head *list)
2956 {
2957 struct sk_buff *skb;
2958 while ((skb = __skb_dequeue(list)) != NULL)
2959 kfree_skb(skb);
2960 }
2961 void skb_queue_purge(struct sk_buff_head *list);
2962
2963 unsigned int skb_rbtree_purge(struct rb_root *root);
2964
2965 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
2966
2967 /**
2968 * netdev_alloc_frag - allocate a page fragment
2969 * @fragsz: fragment size
2970 *
2971 * Allocates a frag from a page for receive buffer.
2972 * Uses GFP_ATOMIC allocations.
2973 */
netdev_alloc_frag(unsigned int fragsz)2974 static inline void *netdev_alloc_frag(unsigned int fragsz)
2975 {
2976 return __netdev_alloc_frag_align(fragsz, ~0u);
2977 }
2978
netdev_alloc_frag_align(unsigned int fragsz,unsigned int align)2979 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
2980 unsigned int align)
2981 {
2982 WARN_ON_ONCE(!is_power_of_2(align));
2983 return __netdev_alloc_frag_align(fragsz, -align);
2984 }
2985
2986 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2987 gfp_t gfp_mask);
2988
2989 /**
2990 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2991 * @dev: network device to receive on
2992 * @length: length to allocate
2993 *
2994 * Allocate a new &sk_buff and assign it a usage count of one. The
2995 * buffer has unspecified headroom built in. Users should allocate
2996 * the headroom they think they need without accounting for the
2997 * built in space. The built in space is used for optimisations.
2998 *
2999 * %NULL is returned if there is no free memory. Although this function
3000 * allocates memory it can be called from an interrupt.
3001 */
netdev_alloc_skb(struct net_device * dev,unsigned int length)3002 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
3003 unsigned int length)
3004 {
3005 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
3006 }
3007
3008 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)3009 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
3010 gfp_t gfp_mask)
3011 {
3012 return __netdev_alloc_skb(NULL, length, gfp_mask);
3013 }
3014
3015 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)3016 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
3017 {
3018 return netdev_alloc_skb(NULL, length);
3019 }
3020
3021
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)3022 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
3023 unsigned int length, gfp_t gfp)
3024 {
3025 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
3026
3027 if (NET_IP_ALIGN && skb)
3028 skb_reserve(skb, NET_IP_ALIGN);
3029 return skb;
3030 }
3031
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)3032 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
3033 unsigned int length)
3034 {
3035 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
3036 }
3037
skb_free_frag(void * addr)3038 static inline void skb_free_frag(void *addr)
3039 {
3040 page_frag_free(addr);
3041 }
3042
3043 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3044
napi_alloc_frag(unsigned int fragsz)3045 static inline void *napi_alloc_frag(unsigned int fragsz)
3046 {
3047 return __napi_alloc_frag_align(fragsz, ~0u);
3048 }
3049
napi_alloc_frag_align(unsigned int fragsz,unsigned int align)3050 static inline void *napi_alloc_frag_align(unsigned int fragsz,
3051 unsigned int align)
3052 {
3053 WARN_ON_ONCE(!is_power_of_2(align));
3054 return __napi_alloc_frag_align(fragsz, -align);
3055 }
3056
3057 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
3058 unsigned int length, gfp_t gfp_mask);
napi_alloc_skb(struct napi_struct * napi,unsigned int length)3059 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
3060 unsigned int length)
3061 {
3062 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
3063 }
3064 void napi_consume_skb(struct sk_buff *skb, int budget);
3065
3066 void napi_skb_free_stolen_head(struct sk_buff *skb);
3067 void __kfree_skb_defer(struct sk_buff *skb);
3068
3069 /**
3070 * __dev_alloc_pages - allocate page for network Rx
3071 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3072 * @order: size of the allocation
3073 *
3074 * Allocate a new page.
3075 *
3076 * %NULL is returned if there is no free memory.
3077 */
__dev_alloc_pages(gfp_t gfp_mask,unsigned int order)3078 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
3079 unsigned int order)
3080 {
3081 /* This piece of code contains several assumptions.
3082 * 1. This is for device Rx, therefor a cold page is preferred.
3083 * 2. The expectation is the user wants a compound page.
3084 * 3. If requesting a order 0 page it will not be compound
3085 * due to the check to see if order has a value in prep_new_page
3086 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3087 * code in gfp_to_alloc_flags that should be enforcing this.
3088 */
3089 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3090
3091 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3092 }
3093
dev_alloc_pages(unsigned int order)3094 static inline struct page *dev_alloc_pages(unsigned int order)
3095 {
3096 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
3097 }
3098
3099 /**
3100 * __dev_alloc_page - allocate a page for network Rx
3101 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3102 *
3103 * Allocate a new page.
3104 *
3105 * %NULL is returned if there is no free memory.
3106 */
__dev_alloc_page(gfp_t gfp_mask)3107 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3108 {
3109 return __dev_alloc_pages(gfp_mask, 0);
3110 }
3111
dev_alloc_page(void)3112 static inline struct page *dev_alloc_page(void)
3113 {
3114 return dev_alloc_pages(0);
3115 }
3116
3117 /**
3118 * dev_page_is_reusable - check whether a page can be reused for network Rx
3119 * @page: the page to test
3120 *
3121 * A page shouldn't be considered for reusing/recycling if it was allocated
3122 * under memory pressure or at a distant memory node.
3123 *
3124 * Returns false if this page should be returned to page allocator, true
3125 * otherwise.
3126 */
dev_page_is_reusable(const struct page * page)3127 static inline bool dev_page_is_reusable(const struct page *page)
3128 {
3129 return likely(page_to_nid(page) == numa_mem_id() &&
3130 !page_is_pfmemalloc(page));
3131 }
3132
3133 /**
3134 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3135 * @page: The page that was allocated from skb_alloc_page
3136 * @skb: The skb that may need pfmemalloc set
3137 */
skb_propagate_pfmemalloc(const struct page * page,struct sk_buff * skb)3138 static inline void skb_propagate_pfmemalloc(const struct page *page,
3139 struct sk_buff *skb)
3140 {
3141 if (page_is_pfmemalloc(page))
3142 skb->pfmemalloc = true;
3143 }
3144
3145 /**
3146 * skb_frag_off() - Returns the offset of a skb fragment
3147 * @frag: the paged fragment
3148 */
skb_frag_off(const skb_frag_t * frag)3149 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3150 {
3151 return frag->bv_offset;
3152 }
3153
3154 /**
3155 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3156 * @frag: skb fragment
3157 * @delta: value to add
3158 */
skb_frag_off_add(skb_frag_t * frag,int delta)3159 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3160 {
3161 frag->bv_offset += delta;
3162 }
3163
3164 /**
3165 * skb_frag_off_set() - Sets the offset of a skb fragment
3166 * @frag: skb fragment
3167 * @offset: offset of fragment
3168 */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)3169 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3170 {
3171 frag->bv_offset = offset;
3172 }
3173
3174 /**
3175 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3176 * @fragto: skb fragment where offset is set
3177 * @fragfrom: skb fragment offset is copied from
3178 */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3179 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3180 const skb_frag_t *fragfrom)
3181 {
3182 fragto->bv_offset = fragfrom->bv_offset;
3183 }
3184
3185 /**
3186 * skb_frag_page - retrieve the page referred to by a paged fragment
3187 * @frag: the paged fragment
3188 *
3189 * Returns the &struct page associated with @frag.
3190 */
skb_frag_page(const skb_frag_t * frag)3191 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3192 {
3193 return frag->bv_page;
3194 }
3195
3196 /**
3197 * __skb_frag_ref - take an addition reference on a paged fragment.
3198 * @frag: the paged fragment
3199 *
3200 * Takes an additional reference on the paged fragment @frag.
3201 */
__skb_frag_ref(skb_frag_t * frag)3202 static inline void __skb_frag_ref(skb_frag_t *frag)
3203 {
3204 get_page(skb_frag_page(frag));
3205 }
3206
3207 /**
3208 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3209 * @skb: the buffer
3210 * @f: the fragment offset.
3211 *
3212 * Takes an additional reference on the @f'th paged fragment of @skb.
3213 */
skb_frag_ref(struct sk_buff * skb,int f)3214 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3215 {
3216 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3217 }
3218
3219 /**
3220 * __skb_frag_unref - release a reference on a paged fragment.
3221 * @frag: the paged fragment
3222 * @recycle: recycle the page if allocated via page_pool
3223 *
3224 * Releases a reference on the paged fragment @frag
3225 * or recycles the page via the page_pool API.
3226 */
__skb_frag_unref(skb_frag_t * frag,bool recycle)3227 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3228 {
3229 struct page *page = skb_frag_page(frag);
3230
3231 #ifdef CONFIG_PAGE_POOL
3232 if (recycle && page_pool_return_skb_page(page))
3233 return;
3234 #endif
3235 put_page(page);
3236 }
3237
3238 /**
3239 * skb_frag_unref - release a reference on a paged fragment of an skb.
3240 * @skb: the buffer
3241 * @f: the fragment offset
3242 *
3243 * Releases a reference on the @f'th paged fragment of @skb.
3244 */
skb_frag_unref(struct sk_buff * skb,int f)3245 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3246 {
3247 __skb_frag_unref(&skb_shinfo(skb)->frags[f], skb->pp_recycle);
3248 }
3249
3250 /**
3251 * skb_frag_address - gets the address of the data contained in a paged fragment
3252 * @frag: the paged fragment buffer
3253 *
3254 * Returns the address of the data within @frag. The page must already
3255 * be mapped.
3256 */
skb_frag_address(const skb_frag_t * frag)3257 static inline void *skb_frag_address(const skb_frag_t *frag)
3258 {
3259 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3260 }
3261
3262 /**
3263 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3264 * @frag: the paged fragment buffer
3265 *
3266 * Returns the address of the data within @frag. Checks that the page
3267 * is mapped and returns %NULL otherwise.
3268 */
skb_frag_address_safe(const skb_frag_t * frag)3269 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3270 {
3271 void *ptr = page_address(skb_frag_page(frag));
3272 if (unlikely(!ptr))
3273 return NULL;
3274
3275 return ptr + skb_frag_off(frag);
3276 }
3277
3278 /**
3279 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3280 * @fragto: skb fragment where page is set
3281 * @fragfrom: skb fragment page is copied from
3282 */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3283 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3284 const skb_frag_t *fragfrom)
3285 {
3286 fragto->bv_page = fragfrom->bv_page;
3287 }
3288
3289 /**
3290 * __skb_frag_set_page - sets the page contained in a paged fragment
3291 * @frag: the paged fragment
3292 * @page: the page to set
3293 *
3294 * Sets the fragment @frag to contain @page.
3295 */
__skb_frag_set_page(skb_frag_t * frag,struct page * page)3296 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3297 {
3298 frag->bv_page = page;
3299 }
3300
3301 /**
3302 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3303 * @skb: the buffer
3304 * @f: the fragment offset
3305 * @page: the page to set
3306 *
3307 * Sets the @f'th fragment of @skb to contain @page.
3308 */
skb_frag_set_page(struct sk_buff * skb,int f,struct page * page)3309 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3310 struct page *page)
3311 {
3312 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3313 }
3314
3315 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3316
3317 /**
3318 * skb_frag_dma_map - maps a paged fragment via the DMA API
3319 * @dev: the device to map the fragment to
3320 * @frag: the paged fragment to map
3321 * @offset: the offset within the fragment (starting at the
3322 * fragment's own offset)
3323 * @size: the number of bytes to map
3324 * @dir: the direction of the mapping (``PCI_DMA_*``)
3325 *
3326 * Maps the page associated with @frag to @device.
3327 */
skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3328 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3329 const skb_frag_t *frag,
3330 size_t offset, size_t size,
3331 enum dma_data_direction dir)
3332 {
3333 return dma_map_page(dev, skb_frag_page(frag),
3334 skb_frag_off(frag) + offset, size, dir);
3335 }
3336
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3337 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3338 gfp_t gfp_mask)
3339 {
3340 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3341 }
3342
3343
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3344 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3345 gfp_t gfp_mask)
3346 {
3347 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3348 }
3349
3350
3351 /**
3352 * skb_clone_writable - is the header of a clone writable
3353 * @skb: buffer to check
3354 * @len: length up to which to write
3355 *
3356 * Returns true if modifying the header part of the cloned buffer
3357 * does not requires the data to be copied.
3358 */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3359 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3360 {
3361 return !skb_header_cloned(skb) &&
3362 skb_headroom(skb) + len <= skb->hdr_len;
3363 }
3364
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3365 static inline int skb_try_make_writable(struct sk_buff *skb,
3366 unsigned int write_len)
3367 {
3368 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3369 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3370 }
3371
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3372 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3373 int cloned)
3374 {
3375 int delta = 0;
3376
3377 if (headroom > skb_headroom(skb))
3378 delta = headroom - skb_headroom(skb);
3379
3380 if (delta || cloned)
3381 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3382 GFP_ATOMIC);
3383 return 0;
3384 }
3385
3386 /**
3387 * skb_cow - copy header of skb when it is required
3388 * @skb: buffer to cow
3389 * @headroom: needed headroom
3390 *
3391 * If the skb passed lacks sufficient headroom or its data part
3392 * is shared, data is reallocated. If reallocation fails, an error
3393 * is returned and original skb is not changed.
3394 *
3395 * The result is skb with writable area skb->head...skb->tail
3396 * and at least @headroom of space at head.
3397 */
skb_cow(struct sk_buff * skb,unsigned int headroom)3398 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3399 {
3400 return __skb_cow(skb, headroom, skb_cloned(skb));
3401 }
3402
3403 /**
3404 * skb_cow_head - skb_cow but only making the head writable
3405 * @skb: buffer to cow
3406 * @headroom: needed headroom
3407 *
3408 * This function is identical to skb_cow except that we replace the
3409 * skb_cloned check by skb_header_cloned. It should be used when
3410 * you only need to push on some header and do not need to modify
3411 * the data.
3412 */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3413 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3414 {
3415 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3416 }
3417
3418 /**
3419 * skb_padto - pad an skbuff up to a minimal size
3420 * @skb: buffer to pad
3421 * @len: minimal length
3422 *
3423 * Pads up a buffer to ensure the trailing bytes exist and are
3424 * blanked. If the buffer already contains sufficient data it
3425 * is untouched. Otherwise it is extended. Returns zero on
3426 * success. The skb is freed on error.
3427 */
skb_padto(struct sk_buff * skb,unsigned int len)3428 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3429 {
3430 unsigned int size = skb->len;
3431 if (likely(size >= len))
3432 return 0;
3433 return skb_pad(skb, len - size);
3434 }
3435
3436 /**
3437 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3438 * @skb: buffer to pad
3439 * @len: minimal length
3440 * @free_on_error: free buffer on error
3441 *
3442 * Pads up a buffer to ensure the trailing bytes exist and are
3443 * blanked. If the buffer already contains sufficient data it
3444 * is untouched. Otherwise it is extended. Returns zero on
3445 * success. The skb is freed on error if @free_on_error is true.
3446 */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3447 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3448 unsigned int len,
3449 bool free_on_error)
3450 {
3451 unsigned int size = skb->len;
3452
3453 if (unlikely(size < len)) {
3454 len -= size;
3455 if (__skb_pad(skb, len, free_on_error))
3456 return -ENOMEM;
3457 __skb_put(skb, len);
3458 }
3459 return 0;
3460 }
3461
3462 /**
3463 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3464 * @skb: buffer to pad
3465 * @len: minimal length
3466 *
3467 * Pads up a buffer to ensure the trailing bytes exist and are
3468 * blanked. If the buffer already contains sufficient data it
3469 * is untouched. Otherwise it is extended. Returns zero on
3470 * success. The skb is freed on error.
3471 */
skb_put_padto(struct sk_buff * skb,unsigned int len)3472 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3473 {
3474 return __skb_put_padto(skb, len, true);
3475 }
3476
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3477 static inline int skb_add_data(struct sk_buff *skb,
3478 struct iov_iter *from, int copy)
3479 {
3480 const int off = skb->len;
3481
3482 if (skb->ip_summed == CHECKSUM_NONE) {
3483 __wsum csum = 0;
3484 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3485 &csum, from)) {
3486 skb->csum = csum_block_add(skb->csum, csum, off);
3487 return 0;
3488 }
3489 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3490 return 0;
3491
3492 __skb_trim(skb, off);
3493 return -EFAULT;
3494 }
3495
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3496 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3497 const struct page *page, int off)
3498 {
3499 if (skb_zcopy(skb))
3500 return false;
3501 if (i) {
3502 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3503
3504 return page == skb_frag_page(frag) &&
3505 off == skb_frag_off(frag) + skb_frag_size(frag);
3506 }
3507 return false;
3508 }
3509
__skb_linearize(struct sk_buff * skb)3510 static inline int __skb_linearize(struct sk_buff *skb)
3511 {
3512 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3513 }
3514
3515 /**
3516 * skb_linearize - convert paged skb to linear one
3517 * @skb: buffer to linarize
3518 *
3519 * If there is no free memory -ENOMEM is returned, otherwise zero
3520 * is returned and the old skb data released.
3521 */
skb_linearize(struct sk_buff * skb)3522 static inline int skb_linearize(struct sk_buff *skb)
3523 {
3524 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3525 }
3526
3527 /**
3528 * skb_has_shared_frag - can any frag be overwritten
3529 * @skb: buffer to test
3530 *
3531 * Return true if the skb has at least one frag that might be modified
3532 * by an external entity (as in vmsplice()/sendfile())
3533 */
skb_has_shared_frag(const struct sk_buff * skb)3534 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3535 {
3536 return skb_is_nonlinear(skb) &&
3537 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3538 }
3539
3540 /**
3541 * skb_linearize_cow - make sure skb is linear and writable
3542 * @skb: buffer to process
3543 *
3544 * If there is no free memory -ENOMEM is returned, otherwise zero
3545 * is returned and the old skb data released.
3546 */
skb_linearize_cow(struct sk_buff * skb)3547 static inline int skb_linearize_cow(struct sk_buff *skb)
3548 {
3549 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3550 __skb_linearize(skb) : 0;
3551 }
3552
3553 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3554 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3555 unsigned int off)
3556 {
3557 if (skb->ip_summed == CHECKSUM_COMPLETE)
3558 skb->csum = csum_block_sub(skb->csum,
3559 csum_partial(start, len, 0), off);
3560 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3561 skb_checksum_start_offset(skb) < 0)
3562 skb->ip_summed = CHECKSUM_NONE;
3563 }
3564
3565 /**
3566 * skb_postpull_rcsum - update checksum for received skb after pull
3567 * @skb: buffer to update
3568 * @start: start of data before pull
3569 * @len: length of data pulled
3570 *
3571 * After doing a pull on a received packet, you need to call this to
3572 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3573 * CHECKSUM_NONE so that it can be recomputed from scratch.
3574 */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3575 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3576 const void *start, unsigned int len)
3577 {
3578 __skb_postpull_rcsum(skb, start, len, 0);
3579 }
3580
3581 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3582 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3583 unsigned int off)
3584 {
3585 if (skb->ip_summed == CHECKSUM_COMPLETE)
3586 skb->csum = csum_block_add(skb->csum,
3587 csum_partial(start, len, 0), off);
3588 }
3589
3590 /**
3591 * skb_postpush_rcsum - update checksum for received skb after push
3592 * @skb: buffer to update
3593 * @start: start of data after push
3594 * @len: length of data pushed
3595 *
3596 * After doing a push on a received packet, you need to call this to
3597 * update the CHECKSUM_COMPLETE checksum.
3598 */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3599 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3600 const void *start, unsigned int len)
3601 {
3602 __skb_postpush_rcsum(skb, start, len, 0);
3603 }
3604
3605 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3606
3607 /**
3608 * skb_push_rcsum - push skb and update receive checksum
3609 * @skb: buffer to update
3610 * @len: length of data pulled
3611 *
3612 * This function performs an skb_push on the packet and updates
3613 * the CHECKSUM_COMPLETE checksum. It should be used on
3614 * receive path processing instead of skb_push unless you know
3615 * that the checksum difference is zero (e.g., a valid IP header)
3616 * or you are setting ip_summed to CHECKSUM_NONE.
3617 */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3618 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3619 {
3620 skb_push(skb, len);
3621 skb_postpush_rcsum(skb, skb->data, len);
3622 return skb->data;
3623 }
3624
3625 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3626 /**
3627 * pskb_trim_rcsum - trim received skb and update checksum
3628 * @skb: buffer to trim
3629 * @len: new length
3630 *
3631 * This is exactly the same as pskb_trim except that it ensures the
3632 * checksum of received packets are still valid after the operation.
3633 * It can change skb pointers.
3634 */
3635
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)3636 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3637 {
3638 if (likely(len >= skb->len))
3639 return 0;
3640 return pskb_trim_rcsum_slow(skb, len);
3641 }
3642
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)3643 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3644 {
3645 if (skb->ip_summed == CHECKSUM_COMPLETE)
3646 skb->ip_summed = CHECKSUM_NONE;
3647 __skb_trim(skb, len);
3648 return 0;
3649 }
3650
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)3651 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3652 {
3653 if (skb->ip_summed == CHECKSUM_COMPLETE)
3654 skb->ip_summed = CHECKSUM_NONE;
3655 return __skb_grow(skb, len);
3656 }
3657
3658 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3659 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3660 #define skb_rb_last(root) rb_to_skb(rb_last(root))
3661 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3662 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3663
3664 #define skb_queue_walk(queue, skb) \
3665 for (skb = (queue)->next; \
3666 skb != (struct sk_buff *)(queue); \
3667 skb = skb->next)
3668
3669 #define skb_queue_walk_safe(queue, skb, tmp) \
3670 for (skb = (queue)->next, tmp = skb->next; \
3671 skb != (struct sk_buff *)(queue); \
3672 skb = tmp, tmp = skb->next)
3673
3674 #define skb_queue_walk_from(queue, skb) \
3675 for (; skb != (struct sk_buff *)(queue); \
3676 skb = skb->next)
3677
3678 #define skb_rbtree_walk(skb, root) \
3679 for (skb = skb_rb_first(root); skb != NULL; \
3680 skb = skb_rb_next(skb))
3681
3682 #define skb_rbtree_walk_from(skb) \
3683 for (; skb != NULL; \
3684 skb = skb_rb_next(skb))
3685
3686 #define skb_rbtree_walk_from_safe(skb, tmp) \
3687 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3688 skb = tmp)
3689
3690 #define skb_queue_walk_from_safe(queue, skb, tmp) \
3691 for (tmp = skb->next; \
3692 skb != (struct sk_buff *)(queue); \
3693 skb = tmp, tmp = skb->next)
3694
3695 #define skb_queue_reverse_walk(queue, skb) \
3696 for (skb = (queue)->prev; \
3697 skb != (struct sk_buff *)(queue); \
3698 skb = skb->prev)
3699
3700 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3701 for (skb = (queue)->prev, tmp = skb->prev; \
3702 skb != (struct sk_buff *)(queue); \
3703 skb = tmp, tmp = skb->prev)
3704
3705 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3706 for (tmp = skb->prev; \
3707 skb != (struct sk_buff *)(queue); \
3708 skb = tmp, tmp = skb->prev)
3709
skb_has_frag_list(const struct sk_buff * skb)3710 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3711 {
3712 return skb_shinfo(skb)->frag_list != NULL;
3713 }
3714
skb_frag_list_init(struct sk_buff * skb)3715 static inline void skb_frag_list_init(struct sk_buff *skb)
3716 {
3717 skb_shinfo(skb)->frag_list = NULL;
3718 }
3719
3720 #define skb_walk_frags(skb, iter) \
3721 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3722
3723
3724 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3725 int *err, long *timeo_p,
3726 const struct sk_buff *skb);
3727 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3728 struct sk_buff_head *queue,
3729 unsigned int flags,
3730 int *off, int *err,
3731 struct sk_buff **last);
3732 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3733 struct sk_buff_head *queue,
3734 unsigned int flags, int *off, int *err,
3735 struct sk_buff **last);
3736 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3737 struct sk_buff_head *sk_queue,
3738 unsigned int flags, int *off, int *err);
3739 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3740 int *err);
3741 __poll_t datagram_poll(struct file *file, struct socket *sock,
3742 struct poll_table_struct *wait);
3743 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3744 struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)3745 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3746 struct msghdr *msg, int size)
3747 {
3748 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3749 }
3750 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3751 struct msghdr *msg);
3752 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3753 struct iov_iter *to, int len,
3754 struct ahash_request *hash);
3755 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3756 struct iov_iter *from, int len);
3757 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3758 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3759 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)3760 static inline void skb_free_datagram_locked(struct sock *sk,
3761 struct sk_buff *skb)
3762 {
3763 __skb_free_datagram_locked(sk, skb, 0);
3764 }
3765 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3766 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3767 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3768 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3769 int len);
3770 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3771 struct pipe_inode_info *pipe, unsigned int len,
3772 unsigned int flags);
3773 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3774 int len);
3775 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3776 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3777 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3778 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3779 int len, int hlen);
3780 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3781 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3782 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3783 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3784 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3785 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3786 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3787 unsigned int offset);
3788 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3789 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3790 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3791 int skb_vlan_pop(struct sk_buff *skb);
3792 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3793 int skb_eth_pop(struct sk_buff *skb);
3794 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3795 const unsigned char *src);
3796 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3797 int mac_len, bool ethernet);
3798 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3799 bool ethernet);
3800 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3801 int skb_mpls_dec_ttl(struct sk_buff *skb);
3802 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3803 gfp_t gfp);
3804
memcpy_from_msg(void * data,struct msghdr * msg,int len)3805 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3806 {
3807 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3808 }
3809
memcpy_to_msg(struct msghdr * msg,void * data,int len)3810 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3811 {
3812 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3813 }
3814
3815 struct skb_checksum_ops {
3816 __wsum (*update)(const void *mem, int len, __wsum wsum);
3817 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3818 };
3819
3820 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3821
3822 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3823 __wsum csum, const struct skb_checksum_ops *ops);
3824 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3825 __wsum csum);
3826
3827 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,const void * data,int hlen,void * buffer)3828 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3829 const void *data, int hlen, void *buffer)
3830 {
3831 if (likely(hlen - offset >= len))
3832 return (void *)data + offset;
3833
3834 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3835 return NULL;
3836
3837 return buffer;
3838 }
3839
3840 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)3841 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3842 {
3843 return __skb_header_pointer(skb, offset, len, skb->data,
3844 skb_headlen(skb), buffer);
3845 }
3846
3847 /**
3848 * skb_needs_linearize - check if we need to linearize a given skb
3849 * depending on the given device features.
3850 * @skb: socket buffer to check
3851 * @features: net device features
3852 *
3853 * Returns true if either:
3854 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3855 * 2. skb is fragmented and the device does not support SG.
3856 */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)3857 static inline bool skb_needs_linearize(struct sk_buff *skb,
3858 netdev_features_t features)
3859 {
3860 return skb_is_nonlinear(skb) &&
3861 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3862 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3863 }
3864
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)3865 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3866 void *to,
3867 const unsigned int len)
3868 {
3869 memcpy(to, skb->data, len);
3870 }
3871
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)3872 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3873 const int offset, void *to,
3874 const unsigned int len)
3875 {
3876 memcpy(to, skb->data + offset, len);
3877 }
3878
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)3879 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3880 const void *from,
3881 const unsigned int len)
3882 {
3883 memcpy(skb->data, from, len);
3884 }
3885
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)3886 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3887 const int offset,
3888 const void *from,
3889 const unsigned int len)
3890 {
3891 memcpy(skb->data + offset, from, len);
3892 }
3893
3894 void skb_init(void);
3895
skb_get_ktime(const struct sk_buff * skb)3896 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3897 {
3898 return skb->tstamp;
3899 }
3900
3901 /**
3902 * skb_get_timestamp - get timestamp from a skb
3903 * @skb: skb to get stamp from
3904 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
3905 *
3906 * Timestamps are stored in the skb as offsets to a base timestamp.
3907 * This function converts the offset back to a struct timeval and stores
3908 * it in stamp.
3909 */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)3910 static inline void skb_get_timestamp(const struct sk_buff *skb,
3911 struct __kernel_old_timeval *stamp)
3912 {
3913 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3914 }
3915
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)3916 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3917 struct __kernel_sock_timeval *stamp)
3918 {
3919 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3920
3921 stamp->tv_sec = ts.tv_sec;
3922 stamp->tv_usec = ts.tv_nsec / 1000;
3923 }
3924
skb_get_timestampns(const struct sk_buff * skb,struct __kernel_old_timespec * stamp)3925 static inline void skb_get_timestampns(const struct sk_buff *skb,
3926 struct __kernel_old_timespec *stamp)
3927 {
3928 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3929
3930 stamp->tv_sec = ts.tv_sec;
3931 stamp->tv_nsec = ts.tv_nsec;
3932 }
3933
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)3934 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3935 struct __kernel_timespec *stamp)
3936 {
3937 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3938
3939 stamp->tv_sec = ts.tv_sec;
3940 stamp->tv_nsec = ts.tv_nsec;
3941 }
3942
__net_timestamp(struct sk_buff * skb)3943 static inline void __net_timestamp(struct sk_buff *skb)
3944 {
3945 skb->tstamp = ktime_get_real();
3946 }
3947
net_timedelta(ktime_t t)3948 static inline ktime_t net_timedelta(ktime_t t)
3949 {
3950 return ktime_sub(ktime_get_real(), t);
3951 }
3952
net_invalid_timestamp(void)3953 static inline ktime_t net_invalid_timestamp(void)
3954 {
3955 return 0;
3956 }
3957
skb_metadata_len(const struct sk_buff * skb)3958 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3959 {
3960 return skb_shinfo(skb)->meta_len;
3961 }
3962
skb_metadata_end(const struct sk_buff * skb)3963 static inline void *skb_metadata_end(const struct sk_buff *skb)
3964 {
3965 return skb_mac_header(skb);
3966 }
3967
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)3968 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3969 const struct sk_buff *skb_b,
3970 u8 meta_len)
3971 {
3972 const void *a = skb_metadata_end(skb_a);
3973 const void *b = skb_metadata_end(skb_b);
3974 /* Using more efficient varaiant than plain call to memcmp(). */
3975 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3976 u64 diffs = 0;
3977
3978 switch (meta_len) {
3979 #define __it(x, op) (x -= sizeof(u##op))
3980 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3981 case 32: diffs |= __it_diff(a, b, 64);
3982 fallthrough;
3983 case 24: diffs |= __it_diff(a, b, 64);
3984 fallthrough;
3985 case 16: diffs |= __it_diff(a, b, 64);
3986 fallthrough;
3987 case 8: diffs |= __it_diff(a, b, 64);
3988 break;
3989 case 28: diffs |= __it_diff(a, b, 64);
3990 fallthrough;
3991 case 20: diffs |= __it_diff(a, b, 64);
3992 fallthrough;
3993 case 12: diffs |= __it_diff(a, b, 64);
3994 fallthrough;
3995 case 4: diffs |= __it_diff(a, b, 32);
3996 break;
3997 }
3998 return diffs;
3999 #else
4000 return memcmp(a - meta_len, b - meta_len, meta_len);
4001 #endif
4002 }
4003
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)4004 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
4005 const struct sk_buff *skb_b)
4006 {
4007 u8 len_a = skb_metadata_len(skb_a);
4008 u8 len_b = skb_metadata_len(skb_b);
4009
4010 if (!(len_a | len_b))
4011 return false;
4012
4013 return len_a != len_b ?
4014 true : __skb_metadata_differs(skb_a, skb_b, len_a);
4015 }
4016
skb_metadata_set(struct sk_buff * skb,u8 meta_len)4017 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
4018 {
4019 skb_shinfo(skb)->meta_len = meta_len;
4020 }
4021
skb_metadata_clear(struct sk_buff * skb)4022 static inline void skb_metadata_clear(struct sk_buff *skb)
4023 {
4024 skb_metadata_set(skb, 0);
4025 }
4026
4027 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
4028
4029 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
4030
4031 void skb_clone_tx_timestamp(struct sk_buff *skb);
4032 bool skb_defer_rx_timestamp(struct sk_buff *skb);
4033
4034 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
4035
skb_clone_tx_timestamp(struct sk_buff * skb)4036 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
4037 {
4038 }
4039
skb_defer_rx_timestamp(struct sk_buff * skb)4040 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
4041 {
4042 return false;
4043 }
4044
4045 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
4046
4047 /**
4048 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
4049 *
4050 * PHY drivers may accept clones of transmitted packets for
4051 * timestamping via their phy_driver.txtstamp method. These drivers
4052 * must call this function to return the skb back to the stack with a
4053 * timestamp.
4054 *
4055 * @skb: clone of the original outgoing packet
4056 * @hwtstamps: hardware time stamps
4057 *
4058 */
4059 void skb_complete_tx_timestamp(struct sk_buff *skb,
4060 struct skb_shared_hwtstamps *hwtstamps);
4061
4062 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4063 struct skb_shared_hwtstamps *hwtstamps,
4064 struct sock *sk, int tstype);
4065
4066 /**
4067 * skb_tstamp_tx - queue clone of skb with send time stamps
4068 * @orig_skb: the original outgoing packet
4069 * @hwtstamps: hardware time stamps, may be NULL if not available
4070 *
4071 * If the skb has a socket associated, then this function clones the
4072 * skb (thus sharing the actual data and optional structures), stores
4073 * the optional hardware time stamping information (if non NULL) or
4074 * generates a software time stamp (otherwise), then queues the clone
4075 * to the error queue of the socket. Errors are silently ignored.
4076 */
4077 void skb_tstamp_tx(struct sk_buff *orig_skb,
4078 struct skb_shared_hwtstamps *hwtstamps);
4079
4080 /**
4081 * skb_tx_timestamp() - Driver hook for transmit timestamping
4082 *
4083 * Ethernet MAC Drivers should call this function in their hard_xmit()
4084 * function immediately before giving the sk_buff to the MAC hardware.
4085 *
4086 * Specifically, one should make absolutely sure that this function is
4087 * called before TX completion of this packet can trigger. Otherwise
4088 * the packet could potentially already be freed.
4089 *
4090 * @skb: A socket buffer.
4091 */
skb_tx_timestamp(struct sk_buff * skb)4092 static inline void skb_tx_timestamp(struct sk_buff *skb)
4093 {
4094 skb_clone_tx_timestamp(skb);
4095 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4096 skb_tstamp_tx(skb, NULL);
4097 }
4098
4099 /**
4100 * skb_complete_wifi_ack - deliver skb with wifi status
4101 *
4102 * @skb: the original outgoing packet
4103 * @acked: ack status
4104 *
4105 */
4106 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4107
4108 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4109 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4110
skb_csum_unnecessary(const struct sk_buff * skb)4111 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4112 {
4113 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4114 skb->csum_valid ||
4115 (skb->ip_summed == CHECKSUM_PARTIAL &&
4116 skb_checksum_start_offset(skb) >= 0));
4117 }
4118
4119 /**
4120 * skb_checksum_complete - Calculate checksum of an entire packet
4121 * @skb: packet to process
4122 *
4123 * This function calculates the checksum over the entire packet plus
4124 * the value of skb->csum. The latter can be used to supply the
4125 * checksum of a pseudo header as used by TCP/UDP. It returns the
4126 * checksum.
4127 *
4128 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4129 * this function can be used to verify that checksum on received
4130 * packets. In that case the function should return zero if the
4131 * checksum is correct. In particular, this function will return zero
4132 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4133 * hardware has already verified the correctness of the checksum.
4134 */
skb_checksum_complete(struct sk_buff * skb)4135 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4136 {
4137 return skb_csum_unnecessary(skb) ?
4138 0 : __skb_checksum_complete(skb);
4139 }
4140
__skb_decr_checksum_unnecessary(struct sk_buff * skb)4141 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4142 {
4143 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4144 if (skb->csum_level == 0)
4145 skb->ip_summed = CHECKSUM_NONE;
4146 else
4147 skb->csum_level--;
4148 }
4149 }
4150
__skb_incr_checksum_unnecessary(struct sk_buff * skb)4151 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4152 {
4153 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4154 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4155 skb->csum_level++;
4156 } else if (skb->ip_summed == CHECKSUM_NONE) {
4157 skb->ip_summed = CHECKSUM_UNNECESSARY;
4158 skb->csum_level = 0;
4159 }
4160 }
4161
__skb_reset_checksum_unnecessary(struct sk_buff * skb)4162 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4163 {
4164 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4165 skb->ip_summed = CHECKSUM_NONE;
4166 skb->csum_level = 0;
4167 }
4168 }
4169
4170 /* Check if we need to perform checksum complete validation.
4171 *
4172 * Returns true if checksum complete is needed, false otherwise
4173 * (either checksum is unnecessary or zero checksum is allowed).
4174 */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)4175 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4176 bool zero_okay,
4177 __sum16 check)
4178 {
4179 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4180 skb->csum_valid = 1;
4181 __skb_decr_checksum_unnecessary(skb);
4182 return false;
4183 }
4184
4185 return true;
4186 }
4187
4188 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4189 * in checksum_init.
4190 */
4191 #define CHECKSUM_BREAK 76
4192
4193 /* Unset checksum-complete
4194 *
4195 * Unset checksum complete can be done when packet is being modified
4196 * (uncompressed for instance) and checksum-complete value is
4197 * invalidated.
4198 */
skb_checksum_complete_unset(struct sk_buff * skb)4199 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4200 {
4201 if (skb->ip_summed == CHECKSUM_COMPLETE)
4202 skb->ip_summed = CHECKSUM_NONE;
4203 }
4204
4205 /* Validate (init) checksum based on checksum complete.
4206 *
4207 * Return values:
4208 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4209 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4210 * checksum is stored in skb->csum for use in __skb_checksum_complete
4211 * non-zero: value of invalid checksum
4212 *
4213 */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)4214 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4215 bool complete,
4216 __wsum psum)
4217 {
4218 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4219 if (!csum_fold(csum_add(psum, skb->csum))) {
4220 skb->csum_valid = 1;
4221 return 0;
4222 }
4223 }
4224
4225 skb->csum = psum;
4226
4227 if (complete || skb->len <= CHECKSUM_BREAK) {
4228 __sum16 csum;
4229
4230 csum = __skb_checksum_complete(skb);
4231 skb->csum_valid = !csum;
4232 return csum;
4233 }
4234
4235 return 0;
4236 }
4237
null_compute_pseudo(struct sk_buff * skb,int proto)4238 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4239 {
4240 return 0;
4241 }
4242
4243 /* Perform checksum validate (init). Note that this is a macro since we only
4244 * want to calculate the pseudo header which is an input function if necessary.
4245 * First we try to validate without any computation (checksum unnecessary) and
4246 * then calculate based on checksum complete calling the function to compute
4247 * pseudo header.
4248 *
4249 * Return values:
4250 * 0: checksum is validated or try to in skb_checksum_complete
4251 * non-zero: value of invalid checksum
4252 */
4253 #define __skb_checksum_validate(skb, proto, complete, \
4254 zero_okay, check, compute_pseudo) \
4255 ({ \
4256 __sum16 __ret = 0; \
4257 skb->csum_valid = 0; \
4258 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4259 __ret = __skb_checksum_validate_complete(skb, \
4260 complete, compute_pseudo(skb, proto)); \
4261 __ret; \
4262 })
4263
4264 #define skb_checksum_init(skb, proto, compute_pseudo) \
4265 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4266
4267 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4268 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4269
4270 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4271 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4272
4273 #define skb_checksum_validate_zero_check(skb, proto, check, \
4274 compute_pseudo) \
4275 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4276
4277 #define skb_checksum_simple_validate(skb) \
4278 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4279
__skb_checksum_convert_check(struct sk_buff * skb)4280 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4281 {
4282 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4283 }
4284
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4285 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4286 {
4287 skb->csum = ~pseudo;
4288 skb->ip_summed = CHECKSUM_COMPLETE;
4289 }
4290
4291 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4292 do { \
4293 if (__skb_checksum_convert_check(skb)) \
4294 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4295 } while (0)
4296
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4297 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4298 u16 start, u16 offset)
4299 {
4300 skb->ip_summed = CHECKSUM_PARTIAL;
4301 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4302 skb->csum_offset = offset - start;
4303 }
4304
4305 /* Update skbuf and packet to reflect the remote checksum offload operation.
4306 * When called, ptr indicates the starting point for skb->csum when
4307 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4308 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4309 */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4310 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4311 int start, int offset, bool nopartial)
4312 {
4313 __wsum delta;
4314
4315 if (!nopartial) {
4316 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4317 return;
4318 }
4319
4320 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4321 __skb_checksum_complete(skb);
4322 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4323 }
4324
4325 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4326
4327 /* Adjust skb->csum since we changed the packet */
4328 skb->csum = csum_add(skb->csum, delta);
4329 }
4330
skb_nfct(const struct sk_buff * skb)4331 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4332 {
4333 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4334 return (void *)(skb->_nfct & NFCT_PTRMASK);
4335 #else
4336 return NULL;
4337 #endif
4338 }
4339
skb_get_nfct(const struct sk_buff * skb)4340 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4341 {
4342 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4343 return skb->_nfct;
4344 #else
4345 return 0UL;
4346 #endif
4347 }
4348
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4349 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4350 {
4351 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4352 skb->slow_gro |= !!nfct;
4353 skb->_nfct = nfct;
4354 #endif
4355 }
4356
4357 #ifdef CONFIG_SKB_EXTENSIONS
4358 enum skb_ext_id {
4359 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4360 SKB_EXT_BRIDGE_NF,
4361 #endif
4362 #ifdef CONFIG_XFRM
4363 SKB_EXT_SEC_PATH,
4364 #endif
4365 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4366 TC_SKB_EXT,
4367 #endif
4368 #if IS_ENABLED(CONFIG_MPTCP)
4369 SKB_EXT_MPTCP,
4370 #endif
4371 SKB_EXT_NUM, /* must be last */
4372 };
4373
4374 /**
4375 * struct skb_ext - sk_buff extensions
4376 * @refcnt: 1 on allocation, deallocated on 0
4377 * @offset: offset to add to @data to obtain extension address
4378 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4379 * @data: start of extension data, variable sized
4380 *
4381 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4382 * to use 'u8' types while allowing up to 2kb worth of extension data.
4383 */
4384 struct skb_ext {
4385 refcount_t refcnt;
4386 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4387 u8 chunks; /* same */
4388 char data[] __aligned(8);
4389 };
4390
4391 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4392 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4393 struct skb_ext *ext);
4394 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4395 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4396 void __skb_ext_put(struct skb_ext *ext);
4397
skb_ext_put(struct sk_buff * skb)4398 static inline void skb_ext_put(struct sk_buff *skb)
4399 {
4400 if (skb->active_extensions)
4401 __skb_ext_put(skb->extensions);
4402 }
4403
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4404 static inline void __skb_ext_copy(struct sk_buff *dst,
4405 const struct sk_buff *src)
4406 {
4407 dst->active_extensions = src->active_extensions;
4408
4409 if (src->active_extensions) {
4410 struct skb_ext *ext = src->extensions;
4411
4412 refcount_inc(&ext->refcnt);
4413 dst->extensions = ext;
4414 }
4415 }
4416
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4417 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4418 {
4419 skb_ext_put(dst);
4420 __skb_ext_copy(dst, src);
4421 }
4422
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4423 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4424 {
4425 return !!ext->offset[i];
4426 }
4427
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4428 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4429 {
4430 return skb->active_extensions & (1 << id);
4431 }
4432
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4433 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4434 {
4435 if (skb_ext_exist(skb, id))
4436 __skb_ext_del(skb, id);
4437 }
4438
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4439 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4440 {
4441 if (skb_ext_exist(skb, id)) {
4442 struct skb_ext *ext = skb->extensions;
4443
4444 return (void *)ext + (ext->offset[id] << 3);
4445 }
4446
4447 return NULL;
4448 }
4449
skb_ext_reset(struct sk_buff * skb)4450 static inline void skb_ext_reset(struct sk_buff *skb)
4451 {
4452 if (unlikely(skb->active_extensions)) {
4453 __skb_ext_put(skb->extensions);
4454 skb->active_extensions = 0;
4455 }
4456 }
4457
skb_has_extensions(struct sk_buff * skb)4458 static inline bool skb_has_extensions(struct sk_buff *skb)
4459 {
4460 return unlikely(skb->active_extensions);
4461 }
4462 #else
skb_ext_put(struct sk_buff * skb)4463 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4464 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4465 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4466 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)4467 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4468 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4469 #endif /* CONFIG_SKB_EXTENSIONS */
4470
nf_reset_ct(struct sk_buff * skb)4471 static inline void nf_reset_ct(struct sk_buff *skb)
4472 {
4473 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4474 nf_conntrack_put(skb_nfct(skb));
4475 skb->_nfct = 0;
4476 #endif
4477 }
4478
nf_reset_trace(struct sk_buff * skb)4479 static inline void nf_reset_trace(struct sk_buff *skb)
4480 {
4481 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4482 skb->nf_trace = 0;
4483 #endif
4484 }
4485
ipvs_reset(struct sk_buff * skb)4486 static inline void ipvs_reset(struct sk_buff *skb)
4487 {
4488 #if IS_ENABLED(CONFIG_IP_VS)
4489 skb->ipvs_property = 0;
4490 #endif
4491 }
4492
4493 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4494 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4495 bool copy)
4496 {
4497 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4498 dst->_nfct = src->_nfct;
4499 nf_conntrack_get(skb_nfct(src));
4500 #endif
4501 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4502 if (copy)
4503 dst->nf_trace = src->nf_trace;
4504 #endif
4505 }
4506
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4507 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4508 {
4509 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4510 nf_conntrack_put(skb_nfct(dst));
4511 #endif
4512 dst->slow_gro = src->slow_gro;
4513 __nf_copy(dst, src, true);
4514 }
4515
4516 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4517 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4518 {
4519 to->secmark = from->secmark;
4520 }
4521
skb_init_secmark(struct sk_buff * skb)4522 static inline void skb_init_secmark(struct sk_buff *skb)
4523 {
4524 skb->secmark = 0;
4525 }
4526 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4527 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4528 { }
4529
skb_init_secmark(struct sk_buff * skb)4530 static inline void skb_init_secmark(struct sk_buff *skb)
4531 { }
4532 #endif
4533
secpath_exists(const struct sk_buff * skb)4534 static inline int secpath_exists(const struct sk_buff *skb)
4535 {
4536 #ifdef CONFIG_XFRM
4537 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4538 #else
4539 return 0;
4540 #endif
4541 }
4542
skb_irq_freeable(const struct sk_buff * skb)4543 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4544 {
4545 return !skb->destructor &&
4546 !secpath_exists(skb) &&
4547 !skb_nfct(skb) &&
4548 !skb->_skb_refdst &&
4549 !skb_has_frag_list(skb);
4550 }
4551
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)4552 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4553 {
4554 skb->queue_mapping = queue_mapping;
4555 }
4556
skb_get_queue_mapping(const struct sk_buff * skb)4557 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4558 {
4559 return skb->queue_mapping;
4560 }
4561
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)4562 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4563 {
4564 to->queue_mapping = from->queue_mapping;
4565 }
4566
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)4567 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4568 {
4569 skb->queue_mapping = rx_queue + 1;
4570 }
4571
skb_get_rx_queue(const struct sk_buff * skb)4572 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4573 {
4574 return skb->queue_mapping - 1;
4575 }
4576
skb_rx_queue_recorded(const struct sk_buff * skb)4577 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4578 {
4579 return skb->queue_mapping != 0;
4580 }
4581
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)4582 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4583 {
4584 skb->dst_pending_confirm = val;
4585 }
4586
skb_get_dst_pending_confirm(const struct sk_buff * skb)4587 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4588 {
4589 return skb->dst_pending_confirm != 0;
4590 }
4591
skb_sec_path(const struct sk_buff * skb)4592 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4593 {
4594 #ifdef CONFIG_XFRM
4595 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4596 #else
4597 return NULL;
4598 #endif
4599 }
4600
4601 /* Keeps track of mac header offset relative to skb->head.
4602 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4603 * For non-tunnel skb it points to skb_mac_header() and for
4604 * tunnel skb it points to outer mac header.
4605 * Keeps track of level of encapsulation of network headers.
4606 */
4607 struct skb_gso_cb {
4608 union {
4609 int mac_offset;
4610 int data_offset;
4611 };
4612 int encap_level;
4613 __wsum csum;
4614 __u16 csum_start;
4615 };
4616 #define SKB_GSO_CB_OFFSET 32
4617 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4618
skb_tnl_header_len(const struct sk_buff * inner_skb)4619 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4620 {
4621 return (skb_mac_header(inner_skb) - inner_skb->head) -
4622 SKB_GSO_CB(inner_skb)->mac_offset;
4623 }
4624
gso_pskb_expand_head(struct sk_buff * skb,int extra)4625 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4626 {
4627 int new_headroom, headroom;
4628 int ret;
4629
4630 headroom = skb_headroom(skb);
4631 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4632 if (ret)
4633 return ret;
4634
4635 new_headroom = skb_headroom(skb);
4636 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4637 return 0;
4638 }
4639
gso_reset_checksum(struct sk_buff * skb,__wsum res)4640 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4641 {
4642 /* Do not update partial checksums if remote checksum is enabled. */
4643 if (skb->remcsum_offload)
4644 return;
4645
4646 SKB_GSO_CB(skb)->csum = res;
4647 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4648 }
4649
4650 /* Compute the checksum for a gso segment. First compute the checksum value
4651 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4652 * then add in skb->csum (checksum from csum_start to end of packet).
4653 * skb->csum and csum_start are then updated to reflect the checksum of the
4654 * resultant packet starting from the transport header-- the resultant checksum
4655 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4656 * header.
4657 */
gso_make_checksum(struct sk_buff * skb,__wsum res)4658 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4659 {
4660 unsigned char *csum_start = skb_transport_header(skb);
4661 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4662 __wsum partial = SKB_GSO_CB(skb)->csum;
4663
4664 SKB_GSO_CB(skb)->csum = res;
4665 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4666
4667 return csum_fold(csum_partial(csum_start, plen, partial));
4668 }
4669
skb_is_gso(const struct sk_buff * skb)4670 static inline bool skb_is_gso(const struct sk_buff *skb)
4671 {
4672 return skb_shinfo(skb)->gso_size;
4673 }
4674
4675 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)4676 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4677 {
4678 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4679 }
4680
4681 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)4682 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4683 {
4684 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4685 }
4686
4687 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)4688 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4689 {
4690 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4691 }
4692
skb_gso_reset(struct sk_buff * skb)4693 static inline void skb_gso_reset(struct sk_buff *skb)
4694 {
4695 skb_shinfo(skb)->gso_size = 0;
4696 skb_shinfo(skb)->gso_segs = 0;
4697 skb_shinfo(skb)->gso_type = 0;
4698 }
4699
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)4700 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4701 u16 increment)
4702 {
4703 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4704 return;
4705 shinfo->gso_size += increment;
4706 }
4707
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)4708 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4709 u16 decrement)
4710 {
4711 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4712 return;
4713 shinfo->gso_size -= decrement;
4714 }
4715
4716 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4717
skb_warn_if_lro(const struct sk_buff * skb)4718 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4719 {
4720 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4721 * wanted then gso_type will be set. */
4722 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4723
4724 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4725 unlikely(shinfo->gso_type == 0)) {
4726 __skb_warn_lro_forwarding(skb);
4727 return true;
4728 }
4729 return false;
4730 }
4731
skb_forward_csum(struct sk_buff * skb)4732 static inline void skb_forward_csum(struct sk_buff *skb)
4733 {
4734 /* Unfortunately we don't support this one. Any brave souls? */
4735 if (skb->ip_summed == CHECKSUM_COMPLETE)
4736 skb->ip_summed = CHECKSUM_NONE;
4737 }
4738
4739 /**
4740 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4741 * @skb: skb to check
4742 *
4743 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4744 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4745 * use this helper, to document places where we make this assertion.
4746 */
skb_checksum_none_assert(const struct sk_buff * skb)4747 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4748 {
4749 #ifdef DEBUG
4750 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4751 #endif
4752 }
4753
4754 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4755
4756 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4757 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4758 unsigned int transport_len,
4759 __sum16(*skb_chkf)(struct sk_buff *skb));
4760
4761 /**
4762 * skb_head_is_locked - Determine if the skb->head is locked down
4763 * @skb: skb to check
4764 *
4765 * The head on skbs build around a head frag can be removed if they are
4766 * not cloned. This function returns true if the skb head is locked down
4767 * due to either being allocated via kmalloc, or by being a clone with
4768 * multiple references to the head.
4769 */
skb_head_is_locked(const struct sk_buff * skb)4770 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4771 {
4772 return !skb->head_frag || skb_cloned(skb);
4773 }
4774
4775 /* Local Checksum Offload.
4776 * Compute outer checksum based on the assumption that the
4777 * inner checksum will be offloaded later.
4778 * See Documentation/networking/checksum-offloads.rst for
4779 * explanation of how this works.
4780 * Fill in outer checksum adjustment (e.g. with sum of outer
4781 * pseudo-header) before calling.
4782 * Also ensure that inner checksum is in linear data area.
4783 */
lco_csum(struct sk_buff * skb)4784 static inline __wsum lco_csum(struct sk_buff *skb)
4785 {
4786 unsigned char *csum_start = skb_checksum_start(skb);
4787 unsigned char *l4_hdr = skb_transport_header(skb);
4788 __wsum partial;
4789
4790 /* Start with complement of inner checksum adjustment */
4791 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4792 skb->csum_offset));
4793
4794 /* Add in checksum of our headers (incl. outer checksum
4795 * adjustment filled in by caller) and return result.
4796 */
4797 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4798 }
4799
skb_is_redirected(const struct sk_buff * skb)4800 static inline bool skb_is_redirected(const struct sk_buff *skb)
4801 {
4802 return skb->redirected;
4803 }
4804
skb_set_redirected(struct sk_buff * skb,bool from_ingress)4805 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4806 {
4807 skb->redirected = 1;
4808 #ifdef CONFIG_NET_REDIRECT
4809 skb->from_ingress = from_ingress;
4810 if (skb->from_ingress)
4811 skb->tstamp = 0;
4812 #endif
4813 }
4814
skb_reset_redirect(struct sk_buff * skb)4815 static inline void skb_reset_redirect(struct sk_buff *skb)
4816 {
4817 skb->redirected = 0;
4818 }
4819
skb_csum_is_sctp(struct sk_buff * skb)4820 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4821 {
4822 return skb->csum_not_inet;
4823 }
4824
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)4825 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4826 const u64 kcov_handle)
4827 {
4828 #ifdef CONFIG_KCOV
4829 skb->kcov_handle = kcov_handle;
4830 #endif
4831 }
4832
skb_get_kcov_handle(struct sk_buff * skb)4833 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4834 {
4835 #ifdef CONFIG_KCOV
4836 return skb->kcov_handle;
4837 #else
4838 return 0;
4839 #endif
4840 }
4841
4842 #ifdef CONFIG_PAGE_POOL
skb_mark_for_recycle(struct sk_buff * skb)4843 static inline void skb_mark_for_recycle(struct sk_buff *skb)
4844 {
4845 skb->pp_recycle = 1;
4846 }
4847 #endif
4848
skb_pp_recycle(struct sk_buff * skb,void * data)4849 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
4850 {
4851 if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
4852 return false;
4853 return page_pool_return_skb_page(virt_to_page(data));
4854 }
4855
4856 #endif /* __KERNEL__ */
4857 #endif /* _LINUX_SKBUFF_H */
4858