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 <net/checksum.h>
27 #include <linux/rcupdate.h>
28 #include <linux/dma-mapping.h>
29 #include <linux/netdev_features.h>
30 #include <net/flow_dissector.h>
31 #include <linux/in6.h>
32 #include <linux/if_packet.h>
33 #include <linux/llist.h>
34 #include <net/flow.h>
35 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
36 #include <linux/netfilter/nf_conntrack_common.h>
37 #endif
38 #include <net/net_debug.h>
39 #include <net/dropreason-core.h>
40 #include <net/netmem.h>
41 #include <linux/android_vendor.h>
42 #include <linux/android_kabi.h>
43
44 /**
45 * DOC: skb checksums
46 *
47 * The interface for checksum offload between the stack and networking drivers
48 * is as follows...
49 *
50 * IP checksum related features
51 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
52 *
53 * Drivers advertise checksum offload capabilities in the features of a device.
54 * From the stack's point of view these are capabilities offered by the driver.
55 * A driver typically only advertises features that it is capable of offloading
56 * to its device.
57 *
58 * .. flat-table:: Checksum related device features
59 * :widths: 1 10
60 *
61 * * - %NETIF_F_HW_CSUM
62 * - The driver (or its device) is able to compute one
63 * IP (one's complement) checksum for any combination
64 * of protocols or protocol layering. The checksum is
65 * computed and set in a packet per the CHECKSUM_PARTIAL
66 * interface (see below).
67 *
68 * * - %NETIF_F_IP_CSUM
69 * - Driver (device) is only able to checksum plain
70 * TCP or UDP packets over IPv4. These are specifically
71 * unencapsulated packets of the form IPv4|TCP or
72 * IPv4|UDP where the Protocol field in the IPv4 header
73 * is TCP or UDP. The IPv4 header may contain IP options.
74 * This feature cannot be set in features for a device
75 * with NETIF_F_HW_CSUM also set. This feature is being
76 * DEPRECATED (see below).
77 *
78 * * - %NETIF_F_IPV6_CSUM
79 * - Driver (device) is only able to checksum plain
80 * TCP or UDP packets over IPv6. These are specifically
81 * unencapsulated packets of the form IPv6|TCP or
82 * IPv6|UDP where the Next Header field in the IPv6
83 * header is either TCP or UDP. IPv6 extension headers
84 * are not supported with this feature. This feature
85 * cannot be set in features for a device with
86 * NETIF_F_HW_CSUM also set. This feature is being
87 * DEPRECATED (see below).
88 *
89 * * - %NETIF_F_RXCSUM
90 * - Driver (device) performs receive checksum offload.
91 * This flag is only used to disable the RX checksum
92 * feature for a device. The stack will accept receive
93 * checksum indication in packets received on a device
94 * regardless of whether NETIF_F_RXCSUM is set.
95 *
96 * Checksumming of received packets by device
97 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
98 *
99 * Indication of checksum verification is set in &sk_buff.ip_summed.
100 * Possible values are:
101 *
102 * - %CHECKSUM_NONE
103 *
104 * Device did not checksum this packet e.g. due to lack of capabilities.
105 * The packet contains full (though not verified) checksum in packet but
106 * not in skb->csum. Thus, skb->csum is undefined in this case.
107 *
108 * - %CHECKSUM_UNNECESSARY
109 *
110 * The hardware you're dealing with doesn't calculate the full checksum
111 * (as in %CHECKSUM_COMPLETE), but it does parse headers and verify checksums
112 * for specific protocols. For such packets it will set %CHECKSUM_UNNECESSARY
113 * if their checksums are okay. &sk_buff.csum is still undefined in this case
114 * though. A driver or device must never modify the checksum field in the
115 * packet even if checksum is verified.
116 *
117 * %CHECKSUM_UNNECESSARY is applicable to following protocols:
118 *
119 * - TCP: IPv6 and IPv4.
120 * - UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
121 * zero UDP checksum for either IPv4 or IPv6, the networking stack
122 * may perform further validation in this case.
123 * - GRE: only if the checksum is present in the header.
124 * - SCTP: indicates the CRC in SCTP header has been validated.
125 * - FCOE: indicates the CRC in FC frame has been validated.
126 *
127 * &sk_buff.csum_level indicates the number of consecutive checksums found in
128 * the packet minus one that have been verified as %CHECKSUM_UNNECESSARY.
129 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
130 * and a device is able to verify the checksums for UDP (possibly zero),
131 * GRE (checksum flag is set) and TCP, &sk_buff.csum_level would be set to
132 * two. If the device were only able to verify the UDP checksum and not
133 * GRE, either because it doesn't support GRE checksum or because GRE
134 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
135 * not considered in this case).
136 *
137 * - %CHECKSUM_COMPLETE
138 *
139 * This is the most generic way. The device supplied checksum of the _whole_
140 * packet as seen by netif_rx() and fills in &sk_buff.csum. This means the
141 * hardware doesn't need to parse L3/L4 headers to implement this.
142 *
143 * Notes:
144 *
145 * - Even if device supports only some protocols, but is able to produce
146 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
147 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
148 *
149 * - %CHECKSUM_PARTIAL
150 *
151 * A checksum is set up to be offloaded to a device as described in the
152 * output description for CHECKSUM_PARTIAL. This may occur on a packet
153 * received directly from another Linux OS, e.g., a virtualized Linux kernel
154 * on the same host, or it may be set in the input path in GRO or remote
155 * checksum offload. For the purposes of checksum verification, the checksum
156 * referred to by skb->csum_start + skb->csum_offset and any preceding
157 * checksums in the packet are considered verified. Any checksums in the
158 * packet that are after the checksum being offloaded are not considered to
159 * be verified.
160 *
161 * Checksumming on transmit for non-GSO
162 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
163 *
164 * The stack requests checksum offload in the &sk_buff.ip_summed for a packet.
165 * Values are:
166 *
167 * - %CHECKSUM_PARTIAL
168 *
169 * The driver is required to checksum the packet as seen by hard_start_xmit()
170 * from &sk_buff.csum_start up to the end, and to record/write the checksum at
171 * offset &sk_buff.csum_start + &sk_buff.csum_offset.
172 * A driver may verify that the
173 * csum_start and csum_offset values are valid values given the length and
174 * offset of the packet, but it should not attempt to validate that the
175 * checksum refers to a legitimate transport layer checksum -- it is the
176 * purview of the stack to validate that csum_start and csum_offset are set
177 * correctly.
178 *
179 * When the stack requests checksum offload for a packet, the driver MUST
180 * ensure that the checksum is set correctly. A driver can either offload the
181 * checksum calculation to the device, or call skb_checksum_help (in the case
182 * that the device does not support offload for a particular checksum).
183 *
184 * %NETIF_F_IP_CSUM and %NETIF_F_IPV6_CSUM are being deprecated in favor of
185 * %NETIF_F_HW_CSUM. New devices should use %NETIF_F_HW_CSUM to indicate
186 * checksum offload capability.
187 * skb_csum_hwoffload_help() can be called to resolve %CHECKSUM_PARTIAL based
188 * on network device checksumming capabilities: if a packet does not match
189 * them, skb_checksum_help() or skb_crc32c_help() (depending on the value of
190 * &sk_buff.csum_not_inet, see :ref:`crc`)
191 * is called to resolve the checksum.
192 *
193 * - %CHECKSUM_NONE
194 *
195 * The skb was already checksummed by the protocol, or a checksum is not
196 * required.
197 *
198 * - %CHECKSUM_UNNECESSARY
199 *
200 * This has the same meaning as CHECKSUM_NONE for checksum offload on
201 * output.
202 *
203 * - %CHECKSUM_COMPLETE
204 *
205 * Not used in checksum output. If a driver observes a packet with this value
206 * set in skbuff, it should treat the packet as if %CHECKSUM_NONE were set.
207 *
208 * .. _crc:
209 *
210 * Non-IP checksum (CRC) offloads
211 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
212 *
213 * .. flat-table::
214 * :widths: 1 10
215 *
216 * * - %NETIF_F_SCTP_CRC
217 * - This feature indicates that a device is capable of
218 * offloading the SCTP CRC in a packet. To perform this offload the stack
219 * will set csum_start and csum_offset accordingly, set ip_summed to
220 * %CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication
221 * in the skbuff that the %CHECKSUM_PARTIAL refers to CRC32c.
222 * A driver that supports both IP checksum offload and SCTP CRC32c offload
223 * must verify which offload is configured for a packet by testing the
224 * value of &sk_buff.csum_not_inet; skb_crc32c_csum_help() is provided to
225 * resolve %CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
226 *
227 * * - %NETIF_F_FCOE_CRC
228 * - This feature indicates that a device is capable of offloading the FCOE
229 * CRC in a packet. To perform this offload the stack will set ip_summed
230 * to %CHECKSUM_PARTIAL and set csum_start and csum_offset
231 * accordingly. Note that there is no indication in the skbuff that the
232 * %CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
233 * both IP checksum offload and FCOE CRC offload must verify which offload
234 * is configured for a packet, presumably by inspecting packet headers.
235 *
236 * Checksumming on output with GSO
237 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
238 *
239 * In the case of a GSO packet (skb_is_gso() is true), checksum offload
240 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
241 * gso_type is %SKB_GSO_TCPV4 or %SKB_GSO_TCPV6, TCP checksum offload as
242 * part of the GSO operation is implied. If a checksum is being offloaded
243 * with GSO then ip_summed is %CHECKSUM_PARTIAL, and both csum_start and
244 * csum_offset are set to refer to the outermost checksum being offloaded
245 * (two offloaded checksums are possible with UDP encapsulation).
246 */
247
248 /* Don't change this without changing skb_csum_unnecessary! */
249 #define CHECKSUM_NONE 0
250 #define CHECKSUM_UNNECESSARY 1
251 #define CHECKSUM_COMPLETE 2
252 #define CHECKSUM_PARTIAL 3
253
254 /* Maximum value in skb->csum_level */
255 #define SKB_MAX_CSUM_LEVEL 3
256
257 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
258 #define SKB_WITH_OVERHEAD(X) \
259 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
260
261 /* For X bytes available in skb->head, what is the minimal
262 * allocation needed, knowing struct skb_shared_info needs
263 * to be aligned.
264 */
265 #define SKB_HEAD_ALIGN(X) (SKB_DATA_ALIGN(X) + \
266 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
267
268 #define SKB_MAX_ORDER(X, ORDER) \
269 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
270 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
271 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
272
273 /* return minimum truesize of one skb containing X bytes of data */
274 #define SKB_TRUESIZE(X) ((X) + \
275 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
276 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
277
278 struct ahash_request;
279 struct net_device;
280 struct scatterlist;
281 struct pipe_inode_info;
282 struct iov_iter;
283 struct napi_struct;
284 struct bpf_prog;
285 union bpf_attr;
286 struct skb_ext;
287 struct ts_config;
288
289 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
290 struct nf_bridge_info {
291 enum {
292 BRNF_PROTO_UNCHANGED,
293 BRNF_PROTO_8021Q,
294 BRNF_PROTO_PPPOE
295 } orig_proto:8;
296 u8 pkt_otherhost:1;
297 u8 in_prerouting:1;
298 u8 bridged_dnat:1;
299 u8 sabotage_in_done:1;
300 __u16 frag_max_size;
301 int physinif;
302
303 /* always valid & non-NULL from FORWARD on, for physdev match */
304 struct net_device *physoutdev;
305 union {
306 /* prerouting: detect dnat in orig/reply direction */
307 __be32 ipv4_daddr;
308 struct in6_addr ipv6_daddr;
309
310 /* after prerouting + nat detected: store original source
311 * mac since neigh resolution overwrites it, only used while
312 * skb is out in neigh layer.
313 */
314 char neigh_header[8];
315 };
316 };
317 #endif
318
319 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
320 /* Chain in tc_skb_ext will be used to share the tc chain with
321 * ovs recirc_id. It will be set to the current chain by tc
322 * and read by ovs to recirc_id.
323 */
324 struct tc_skb_ext {
325 union {
326 u64 act_miss_cookie;
327 __u32 chain;
328 };
329 __u16 mru;
330 __u16 zone;
331 u8 post_ct:1;
332 u8 post_ct_snat:1;
333 u8 post_ct_dnat:1;
334 u8 act_miss:1; /* Set if act_miss_cookie is used */
335 u8 l2_miss:1; /* Set by bridge upon FDB or MDB miss */
336 };
337 #endif
338
339 struct sk_buff_head {
340 /* These two members must be first to match sk_buff. */
341 struct_group_tagged(sk_buff_list, list,
342 struct sk_buff *next;
343 struct sk_buff *prev;
344 );
345
346 __u32 qlen;
347 spinlock_t lock;
348 };
349
350 struct sk_buff;
351
352 #ifndef CONFIG_MAX_SKB_FRAGS
353 # define CONFIG_MAX_SKB_FRAGS 17
354 #endif
355
356 #define MAX_SKB_FRAGS CONFIG_MAX_SKB_FRAGS
357
358 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
359 * segment using its current segmentation instead.
360 */
361 #define GSO_BY_FRAGS 0xFFFF
362
363 typedef struct skb_frag {
364 netmem_ref netmem;
365 unsigned int len;
366 unsigned int offset;
367 } skb_frag_t;
368
369 /**
370 * skb_frag_size() - Returns the size of a skb fragment
371 * @frag: skb fragment
372 */
skb_frag_size(const skb_frag_t * frag)373 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
374 {
375 return frag->len;
376 }
377
378 /**
379 * skb_frag_size_set() - Sets the size of a skb fragment
380 * @frag: skb fragment
381 * @size: size of fragment
382 */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)383 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
384 {
385 frag->len = size;
386 }
387
388 /**
389 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
390 * @frag: skb fragment
391 * @delta: value to add
392 */
skb_frag_size_add(skb_frag_t * frag,int delta)393 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
394 {
395 frag->len += delta;
396 }
397
398 /**
399 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
400 * @frag: skb fragment
401 * @delta: value to subtract
402 */
skb_frag_size_sub(skb_frag_t * frag,int delta)403 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
404 {
405 frag->len -= delta;
406 }
407
408 /**
409 * skb_frag_must_loop - Test if %p is a high memory page
410 * @p: fragment's page
411 */
skb_frag_must_loop(struct page * p)412 static inline bool skb_frag_must_loop(struct page *p)
413 {
414 #if defined(CONFIG_HIGHMEM)
415 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
416 return true;
417 #endif
418 return false;
419 }
420
421 /**
422 * skb_frag_foreach_page - loop over pages in a fragment
423 *
424 * @f: skb frag to operate on
425 * @f_off: offset from start of f->netmem
426 * @f_len: length from f_off to loop over
427 * @p: (temp var) current page
428 * @p_off: (temp var) offset from start of current page,
429 * non-zero only on first page.
430 * @p_len: (temp var) length in current page,
431 * < PAGE_SIZE only on first and last page.
432 * @copied: (temp var) length so far, excluding current p_len.
433 *
434 * A fragment can hold a compound page, in which case per-page
435 * operations, notably kmap_atomic, must be called for each
436 * regular page.
437 */
438 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
439 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
440 p_off = (f_off) & (PAGE_SIZE - 1), \
441 p_len = skb_frag_must_loop(p) ? \
442 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
443 copied = 0; \
444 copied < f_len; \
445 copied += p_len, p++, p_off = 0, \
446 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
447
448 /**
449 * struct skb_shared_hwtstamps - hardware time stamps
450 * @hwtstamp: hardware time stamp transformed into duration
451 * since arbitrary point in time
452 * @netdev_data: address/cookie of network device driver used as
453 * reference to actual hardware time stamp
454 *
455 * Software time stamps generated by ktime_get_real() are stored in
456 * skb->tstamp.
457 *
458 * hwtstamps can only be compared against other hwtstamps from
459 * the same device.
460 *
461 * This structure is attached to packets as part of the
462 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
463 */
464 struct skb_shared_hwtstamps {
465 union {
466 ktime_t hwtstamp;
467 void *netdev_data;
468 };
469 };
470
471 /* Definitions for tx_flags in struct skb_shared_info */
472 enum {
473 /* generate hardware time stamp */
474 SKBTX_HW_TSTAMP = 1 << 0,
475
476 /* generate software time stamp when queueing packet to NIC */
477 SKBTX_SW_TSTAMP = 1 << 1,
478
479 /* device driver is going to provide hardware time stamp */
480 SKBTX_IN_PROGRESS = 1 << 2,
481
482 /* generate hardware time stamp based on cycles if supported */
483 SKBTX_HW_TSTAMP_USE_CYCLES = 1 << 3,
484
485 /* generate wifi status information (where possible) */
486 SKBTX_WIFI_STATUS = 1 << 4,
487
488 /* determine hardware time stamp based on time or cycles */
489 SKBTX_HW_TSTAMP_NETDEV = 1 << 5,
490
491 /* generate software time stamp when entering packet scheduling */
492 SKBTX_SCHED_TSTAMP = 1 << 6,
493 };
494
495 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
496 SKBTX_SCHED_TSTAMP)
497 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | \
498 SKBTX_HW_TSTAMP_USE_CYCLES | \
499 SKBTX_ANY_SW_TSTAMP)
500
501 /* Definitions for flags in struct skb_shared_info */
502 enum {
503 /* use zcopy routines */
504 SKBFL_ZEROCOPY_ENABLE = BIT(0),
505
506 /* This indicates at least one fragment might be overwritten
507 * (as in vmsplice(), sendfile() ...)
508 * If we need to compute a TX checksum, we'll need to copy
509 * all frags to avoid possible bad checksum
510 */
511 SKBFL_SHARED_FRAG = BIT(1),
512
513 /* segment contains only zerocopy data and should not be
514 * charged to the kernel memory.
515 */
516 SKBFL_PURE_ZEROCOPY = BIT(2),
517
518 SKBFL_DONT_ORPHAN = BIT(3),
519
520 /* page references are managed by the ubuf_info, so it's safe to
521 * use frags only up until ubuf_info is released
522 */
523 SKBFL_MANAGED_FRAG_REFS = BIT(4),
524 };
525
526 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
527 #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \
528 SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS)
529
530 struct ubuf_info_ops {
531 void (*complete)(struct sk_buff *, struct ubuf_info *,
532 bool zerocopy_success);
533 /* has to be compatible with skb_zcopy_set() */
534 int (*link_skb)(struct sk_buff *skb, struct ubuf_info *uarg);
535 };
536
537 /*
538 * The callback notifies userspace to release buffers when skb DMA is done in
539 * lower device, the skb last reference should be 0 when calling this.
540 * The zerocopy_success argument is true if zero copy transmit occurred,
541 * false on data copy or out of memory error caused by data copy attempt.
542 * The ctx field is used to track device context.
543 * The desc field is used to track userspace buffer index.
544 */
545 struct ubuf_info {
546 const struct ubuf_info_ops *ops;
547 refcount_t refcnt;
548 u8 flags;
549 };
550
551 struct ubuf_info_msgzc {
552 struct ubuf_info ubuf;
553
554 union {
555 struct {
556 unsigned long desc;
557 void *ctx;
558 };
559 struct {
560 u32 id;
561 u16 len;
562 u16 zerocopy:1;
563 u32 bytelen;
564 };
565 };
566
567 struct mmpin {
568 struct user_struct *user;
569 unsigned int num_pg;
570 } mmp;
571 };
572
573 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
574 #define uarg_to_msgzc(ubuf_ptr) container_of((ubuf_ptr), struct ubuf_info_msgzc, \
575 ubuf)
576
577 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
578 void mm_unaccount_pinned_pages(struct mmpin *mmp);
579
580 /* Preserve some data across TX submission and completion.
581 *
582 * Note, this state is stored in the driver. Extending the layout
583 * might need some special care.
584 */
585 struct xsk_tx_metadata_compl {
586 __u64 *tx_timestamp;
587 };
588
589 /* This data is invariant across clones and lives at
590 * the end of the header data, ie. at skb->end.
591 */
592 struct skb_shared_info {
593 __u8 flags;
594 __u8 meta_len;
595 __u8 nr_frags;
596 __u8 tx_flags;
597 unsigned short gso_size;
598 /* Warning: this field is not always filled in (UFO)! */
599 unsigned short gso_segs;
600 struct sk_buff *frag_list;
601 union {
602 struct skb_shared_hwtstamps hwtstamps;
603 struct xsk_tx_metadata_compl xsk_meta;
604 };
605 unsigned int gso_type;
606 u32 tskey;
607
608 /*
609 * Warning : all fields before dataref are cleared in __alloc_skb()
610 */
611 atomic_t dataref;
612 unsigned int xdp_frags_size;
613
614 /* Intermediate layers must ensure that destructor_arg
615 * remains valid until skb destructor */
616 void * destructor_arg;
617
618 ANDROID_KABI_RESERVE(1);
619 ANDROID_KABI_RESERVE(2);
620 ANDROID_OEM_DATA_ARRAY(1, 3);
621
622 /* must be last field, see pskb_expand_head() */
623 skb_frag_t frags[MAX_SKB_FRAGS];
624 };
625
626 /**
627 * DOC: dataref and headerless skbs
628 *
629 * Transport layers send out clones of payload skbs they hold for
630 * retransmissions. To allow lower layers of the stack to prepend their headers
631 * we split &skb_shared_info.dataref into two halves.
632 * The lower 16 bits count the overall number of references.
633 * The higher 16 bits indicate how many of the references are payload-only.
634 * skb_header_cloned() checks if skb is allowed to add / write the headers.
635 *
636 * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr
637 * (via __skb_header_release()). Any clone created from marked skb will get
638 * &sk_buff.hdr_len populated with the available headroom.
639 * If there's the only clone in existence it's able to modify the headroom
640 * at will. The sequence of calls inside the transport layer is::
641 *
642 * <alloc skb>
643 * skb_reserve()
644 * __skb_header_release()
645 * skb_clone()
646 * // send the clone down the stack
647 *
648 * This is not a very generic construct and it depends on the transport layers
649 * doing the right thing. In practice there's usually only one payload-only skb.
650 * Having multiple payload-only skbs with different lengths of hdr_len is not
651 * possible. The payload-only skbs should never leave their owner.
652 */
653 #define SKB_DATAREF_SHIFT 16
654 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
655
656
657 enum {
658 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
659 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
660 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
661 };
662
663 enum {
664 SKB_GSO_TCPV4 = 1 << 0,
665
666 /* This indicates the skb is from an untrusted source. */
667 SKB_GSO_DODGY = 1 << 1,
668
669 /* This indicates the tcp segment has CWR set. */
670 SKB_GSO_TCP_ECN = 1 << 2,
671
672 SKB_GSO_TCP_FIXEDID = 1 << 3,
673
674 SKB_GSO_TCPV6 = 1 << 4,
675
676 SKB_GSO_FCOE = 1 << 5,
677
678 SKB_GSO_GRE = 1 << 6,
679
680 SKB_GSO_GRE_CSUM = 1 << 7,
681
682 SKB_GSO_IPXIP4 = 1 << 8,
683
684 SKB_GSO_IPXIP6 = 1 << 9,
685
686 SKB_GSO_UDP_TUNNEL = 1 << 10,
687
688 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
689
690 SKB_GSO_PARTIAL = 1 << 12,
691
692 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
693
694 SKB_GSO_SCTP = 1 << 14,
695
696 SKB_GSO_ESP = 1 << 15,
697
698 SKB_GSO_UDP = 1 << 16,
699
700 SKB_GSO_UDP_L4 = 1 << 17,
701
702 SKB_GSO_FRAGLIST = 1 << 18,
703 };
704
705 #if BITS_PER_LONG > 32
706 #define NET_SKBUFF_DATA_USES_OFFSET 1
707 #endif
708
709 #ifdef NET_SKBUFF_DATA_USES_OFFSET
710 typedef unsigned int sk_buff_data_t;
711 #else
712 typedef unsigned char *sk_buff_data_t;
713 #endif
714
715 enum skb_tstamp_type {
716 SKB_CLOCK_REALTIME,
717 SKB_CLOCK_MONOTONIC,
718 SKB_CLOCK_TAI,
719 __SKB_CLOCK_MAX = SKB_CLOCK_TAI,
720 };
721
722 /**
723 * DOC: Basic sk_buff geometry
724 *
725 * struct sk_buff itself is a metadata structure and does not hold any packet
726 * data. All the data is held in associated buffers.
727 *
728 * &sk_buff.head points to the main "head" buffer. The head buffer is divided
729 * into two parts:
730 *
731 * - data buffer, containing headers and sometimes payload;
732 * this is the part of the skb operated on by the common helpers
733 * such as skb_put() or skb_pull();
734 * - shared info (struct skb_shared_info) which holds an array of pointers
735 * to read-only data in the (page, offset, length) format.
736 *
737 * Optionally &skb_shared_info.frag_list may point to another skb.
738 *
739 * Basic diagram may look like this::
740 *
741 * ---------------
742 * | sk_buff |
743 * ---------------
744 * ,--------------------------- + head
745 * / ,----------------- + data
746 * / / ,----------- + tail
747 * | | | , + end
748 * | | | |
749 * v v v v
750 * -----------------------------------------------
751 * | headroom | data | tailroom | skb_shared_info |
752 * -----------------------------------------------
753 * + [page frag]
754 * + [page frag]
755 * + [page frag]
756 * + [page frag] ---------
757 * + frag_list --> | sk_buff |
758 * ---------
759 *
760 */
761
762 /**
763 * struct sk_buff - socket buffer
764 * @next: Next buffer in list
765 * @prev: Previous buffer in list
766 * @tstamp: Time we arrived/left
767 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
768 * for retransmit timer
769 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
770 * @list: queue head
771 * @ll_node: anchor in an llist (eg socket defer_list)
772 * @sk: Socket we are owned by
773 * @dev: Device we arrived on/are leaving by
774 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
775 * @cb: Control buffer. Free for use by every layer. Put private vars here
776 * @_skb_refdst: destination entry (with norefcount bit)
777 * @len: Length of actual data
778 * @data_len: Data length
779 * @mac_len: Length of link layer header
780 * @hdr_len: writable header length of cloned skb
781 * @csum: Checksum (must include start/offset pair)
782 * @csum_start: Offset from skb->head where checksumming should start
783 * @csum_offset: Offset from csum_start where checksum should be stored
784 * @priority: Packet queueing priority
785 * @ignore_df: allow local fragmentation
786 * @cloned: Head may be cloned (check refcnt to be sure)
787 * @ip_summed: Driver fed us an IP checksum
788 * @nohdr: Payload reference only, must not modify header
789 * @pkt_type: Packet class
790 * @fclone: skbuff clone status
791 * @ipvs_property: skbuff is owned by ipvs
792 * @inner_protocol_type: whether the inner protocol is
793 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
794 * @remcsum_offload: remote checksum offload is enabled
795 * @offload_fwd_mark: Packet was L2-forwarded in hardware
796 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
797 * @tc_skip_classify: do not classify packet. set by IFB device
798 * @tc_at_ingress: used within tc_classify to distinguish in/egress
799 * @redirected: packet was redirected by packet classifier
800 * @from_ingress: packet was redirected from the ingress path
801 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
802 * @peeked: this packet has been seen already, so stats have been
803 * done for it, don't do them again
804 * @nf_trace: netfilter packet trace flag
805 * @protocol: Packet protocol from driver
806 * @destructor: Destruct function
807 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
808 * @_sk_redir: socket redirection information for skmsg
809 * @_nfct: Associated connection, if any (with nfctinfo bits)
810 * @skb_iif: ifindex of device we arrived on
811 * @tc_index: Traffic control index
812 * @hash: the packet hash
813 * @queue_mapping: Queue mapping for multiqueue devices
814 * @head_frag: skb was allocated from page fragments,
815 * not allocated by kmalloc() or vmalloc().
816 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
817 * @pp_recycle: mark the packet for recycling instead of freeing (implies
818 * page_pool support on driver)
819 * @active_extensions: active extensions (skb_ext_id types)
820 * @ndisc_nodetype: router type (from link layer)
821 * @ooo_okay: allow the mapping of a socket to a queue to be changed
822 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
823 * ports.
824 * @sw_hash: indicates hash was computed in software stack
825 * @wifi_acked_valid: wifi_acked was set
826 * @wifi_acked: whether frame was acked on wifi or not
827 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
828 * @encapsulation: indicates the inner headers in the skbuff are valid
829 * @encap_hdr_csum: software checksum is needed
830 * @csum_valid: checksum is already valid
831 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
832 * @csum_complete_sw: checksum was completed by software
833 * @csum_level: indicates the number of consecutive checksums found in
834 * the packet minus one that have been verified as
835 * CHECKSUM_UNNECESSARY (max 3)
836 * @unreadable: indicates that at least 1 of the fragments in this skb is
837 * unreadable.
838 * @dst_pending_confirm: need to confirm neighbour
839 * @decrypted: Decrypted SKB
840 * @slow_gro: state present at GRO time, slower prepare step required
841 * @tstamp_type: When set, skb->tstamp has the
842 * delivery_time clock base of skb->tstamp.
843 * @napi_id: id of the NAPI struct this skb came from
844 * @sender_cpu: (aka @napi_id) source CPU in XPS
845 * @alloc_cpu: CPU which did the skb allocation.
846 * @secmark: security marking
847 * @mark: Generic packet mark
848 * @reserved_tailroom: (aka @mark) number of bytes of free space available
849 * at the tail of an sk_buff
850 * @vlan_all: vlan fields (proto & tci)
851 * @vlan_proto: vlan encapsulation protocol
852 * @vlan_tci: vlan tag control information
853 * @inner_protocol: Protocol (encapsulation)
854 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
855 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
856 * @inner_transport_header: Inner transport layer header (encapsulation)
857 * @inner_network_header: Network layer header (encapsulation)
858 * @inner_mac_header: Link layer header (encapsulation)
859 * @transport_header: Transport layer header
860 * @network_header: Network layer header
861 * @mac_header: Link layer header
862 * @kcov_handle: KCOV remote handle for remote coverage collection
863 * @tail: Tail pointer
864 * @end: End pointer
865 * @head: Head of buffer
866 * @data: Data head pointer
867 * @truesize: Buffer size
868 * @users: User count - see {datagram,tcp}.c
869 * @extensions: allocated extensions, valid if active_extensions is nonzero
870 */
871
872 struct sk_buff {
873 union {
874 struct {
875 /* These two members must be first to match sk_buff_head. */
876 struct sk_buff *next;
877 struct sk_buff *prev;
878
879 union {
880 struct net_device *dev;
881 /* Some protocols might use this space to store information,
882 * while device pointer would be NULL.
883 * UDP receive path is one user.
884 */
885 unsigned long dev_scratch;
886 };
887 };
888 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
889 struct list_head list;
890 struct llist_node ll_node;
891 };
892
893 struct sock *sk;
894
895 union {
896 ktime_t tstamp;
897 u64 skb_mstamp_ns; /* earliest departure time */
898 };
899 /*
900 * This is the control buffer. It is free to use for every
901 * layer. Please put your private variables there. If you
902 * want to keep them across layers you have to do a skb_clone()
903 * first. This is owned by whoever has the skb queued ATM.
904 */
905 char cb[48] __aligned(8);
906
907 union {
908 struct {
909 unsigned long _skb_refdst;
910 void (*destructor)(struct sk_buff *skb);
911 };
912 struct list_head tcp_tsorted_anchor;
913 #ifdef CONFIG_NET_SOCK_MSG
914 unsigned long _sk_redir;
915 #endif
916 };
917
918 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
919 unsigned long _nfct;
920 #endif
921 unsigned int len,
922 data_len;
923 __u16 mac_len,
924 hdr_len;
925
926 /* Following fields are _not_ copied in __copy_skb_header()
927 * Note that queue_mapping is here mostly to fill a hole.
928 */
929 __u16 queue_mapping;
930
931 /* if you move cloned around you also must adapt those constants */
932 #ifdef __BIG_ENDIAN_BITFIELD
933 #define CLONED_MASK (1 << 7)
934 #else
935 #define CLONED_MASK 1
936 #endif
937 #define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset)
938
939 /* private: */
940 __u8 __cloned_offset[0];
941 /* public: */
942 __u8 cloned:1,
943 nohdr:1,
944 fclone:2,
945 peeked:1,
946 head_frag:1,
947 pfmemalloc:1,
948 pp_recycle:1; /* page_pool recycle indicator */
949 #ifdef CONFIG_SKB_EXTENSIONS
950 __u8 active_extensions;
951 #endif
952
953 /* Fields enclosed in headers group are copied
954 * using a single memcpy() in __copy_skb_header()
955 */
956 struct_group(headers,
957
958 /* private: */
959 __u8 __pkt_type_offset[0];
960 /* public: */
961 __u8 pkt_type:3; /* see PKT_TYPE_MAX */
962 __u8 ignore_df:1;
963 __u8 dst_pending_confirm:1;
964 __u8 ip_summed:2;
965 __u8 ooo_okay:1;
966
967 /* private: */
968 __u8 __mono_tc_offset[0];
969 /* public: */
970 __u8 tstamp_type:2; /* See skb_tstamp_type */
971 #ifdef CONFIG_NET_XGRESS
972 __u8 tc_at_ingress:1; /* See TC_AT_INGRESS_MASK */
973 __u8 tc_skip_classify:1;
974 #endif
975 __u8 remcsum_offload:1;
976 __u8 csum_complete_sw:1;
977 __u8 csum_level:2;
978 __u8 inner_protocol_type:1;
979
980 __u8 l4_hash:1;
981 __u8 sw_hash:1;
982 #ifdef CONFIG_WIRELESS
983 __u8 wifi_acked_valid:1;
984 __u8 wifi_acked:1;
985 #endif
986 __u8 no_fcs:1;
987 /* Indicates the inner headers are valid in the skbuff. */
988 __u8 encapsulation:1;
989 __u8 encap_hdr_csum:1;
990 __u8 csum_valid:1;
991 #ifdef CONFIG_IPV6_NDISC_NODETYPE
992 __u8 ndisc_nodetype:2;
993 #endif
994
995 #if IS_ENABLED(CONFIG_IP_VS)
996 __u8 ipvs_property:1;
997 #endif
998 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
999 __u8 nf_trace:1;
1000 #endif
1001 #ifdef CONFIG_NET_SWITCHDEV
1002 __u8 offload_fwd_mark:1;
1003 __u8 offload_l3_fwd_mark:1;
1004 #endif
1005 __u8 redirected:1;
1006 #ifdef CONFIG_NET_REDIRECT
1007 __u8 from_ingress:1;
1008 #endif
1009 #ifdef CONFIG_NETFILTER_SKIP_EGRESS
1010 __u8 nf_skip_egress:1;
1011 #endif
1012 #ifdef CONFIG_SKB_DECRYPTED
1013 __u8 decrypted:1;
1014 #endif
1015 __u8 slow_gro:1;
1016 #if IS_ENABLED(CONFIG_IP_SCTP)
1017 __u8 csum_not_inet:1;
1018 #endif
1019 __u8 unreadable:1;
1020 #if defined(CONFIG_NET_SCHED) || defined(CONFIG_NET_XGRESS)
1021 __u16 tc_index; /* traffic control index */
1022 #endif
1023
1024 u16 alloc_cpu;
1025
1026 union {
1027 __wsum csum;
1028 struct {
1029 __u16 csum_start;
1030 __u16 csum_offset;
1031 };
1032 };
1033 __u32 priority;
1034 int skb_iif;
1035 __u32 hash;
1036 union {
1037 u32 vlan_all;
1038 struct {
1039 __be16 vlan_proto;
1040 __u16 vlan_tci;
1041 };
1042 };
1043 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
1044 union {
1045 unsigned int napi_id;
1046 unsigned int sender_cpu;
1047 };
1048 #endif
1049 #ifdef CONFIG_NETWORK_SECMARK
1050 __u32 secmark;
1051 #endif
1052
1053 union {
1054 __u32 mark;
1055 __u32 reserved_tailroom;
1056 };
1057
1058 union {
1059 __be16 inner_protocol;
1060 __u8 inner_ipproto;
1061 };
1062
1063 __u16 inner_transport_header;
1064 __u16 inner_network_header;
1065 __u16 inner_mac_header;
1066
1067 __be16 protocol;
1068 __u16 transport_header;
1069 __u16 network_header;
1070 __u16 mac_header;
1071
1072 #ifdef CONFIG_KCOV
1073 u64 kcov_handle;
1074 #endif
1075
1076 ANDROID_KABI_RESERVE(1);
1077 ANDROID_KABI_RESERVE(2);
1078
1079 ); /* end headers group */
1080
1081 /* These elements must be at the end, see alloc_skb() for details. */
1082 sk_buff_data_t tail;
1083 sk_buff_data_t end;
1084 unsigned char *head,
1085 *data;
1086 unsigned int truesize;
1087 refcount_t users;
1088
1089 #ifdef CONFIG_SKB_EXTENSIONS
1090 /* only usable after checking ->active_extensions != 0 */
1091 struct skb_ext *extensions;
1092 #endif
1093 };
1094
1095 /* if you move pkt_type around you also must adapt those constants */
1096 #ifdef __BIG_ENDIAN_BITFIELD
1097 #define PKT_TYPE_MAX (7 << 5)
1098 #else
1099 #define PKT_TYPE_MAX 7
1100 #endif
1101 #define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset)
1102
1103 /* if you move tc_at_ingress or tstamp_type
1104 * around, you also must adapt these constants.
1105 */
1106 #ifdef __BIG_ENDIAN_BITFIELD
1107 #define SKB_TSTAMP_TYPE_MASK (3 << 6)
1108 #define SKB_TSTAMP_TYPE_RSHIFT (6)
1109 #define TC_AT_INGRESS_MASK (1 << 5)
1110 #else
1111 #define SKB_TSTAMP_TYPE_MASK (3)
1112 #define TC_AT_INGRESS_MASK (1 << 2)
1113 #endif
1114 #define SKB_BF_MONO_TC_OFFSET offsetof(struct sk_buff, __mono_tc_offset)
1115
1116 #ifdef __KERNEL__
1117 /*
1118 * Handling routines are only of interest to the kernel
1119 */
1120
1121 #define SKB_ALLOC_FCLONE 0x01
1122 #define SKB_ALLOC_RX 0x02
1123 #define SKB_ALLOC_NAPI 0x04
1124
1125 /**
1126 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
1127 * @skb: buffer
1128 */
skb_pfmemalloc(const struct sk_buff * skb)1129 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
1130 {
1131 return unlikely(skb->pfmemalloc);
1132 }
1133
1134 /*
1135 * skb might have a dst pointer attached, refcounted or not.
1136 * _skb_refdst low order bit is set if refcount was _not_ taken
1137 */
1138 #define SKB_DST_NOREF 1UL
1139 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
1140
1141 /**
1142 * skb_dst - returns skb dst_entry
1143 * @skb: buffer
1144 *
1145 * Returns skb dst_entry, regardless of reference taken or not.
1146 */
skb_dst(const struct sk_buff * skb)1147 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1148 {
1149 /* If refdst was not refcounted, check we still are in a
1150 * rcu_read_lock section
1151 */
1152 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1153 !rcu_read_lock_held() &&
1154 !rcu_read_lock_bh_held());
1155 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
1156 }
1157
1158 /**
1159 * skb_dst_set - sets skb dst
1160 * @skb: buffer
1161 * @dst: dst entry
1162 *
1163 * Sets skb dst, assuming a reference was taken on dst and should
1164 * be released by skb_dst_drop()
1165 */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)1166 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1167 {
1168 skb->slow_gro |= !!dst;
1169 skb->_skb_refdst = (unsigned long)dst;
1170 }
1171
1172 /**
1173 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1174 * @skb: buffer
1175 * @dst: dst entry
1176 *
1177 * Sets skb dst, assuming a reference was not taken on dst.
1178 * If dst entry is cached, we do not take reference and dst_release
1179 * will be avoided by refdst_drop. If dst entry is not cached, we take
1180 * reference, so that last dst_release can destroy the dst immediately.
1181 */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)1182 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1183 {
1184 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1185 skb->slow_gro |= !!dst;
1186 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1187 }
1188
1189 /**
1190 * skb_dst_is_noref - Test if skb dst isn't refcounted
1191 * @skb: buffer
1192 */
skb_dst_is_noref(const struct sk_buff * skb)1193 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1194 {
1195 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1196 }
1197
1198 /* For mangling skb->pkt_type from user space side from applications
1199 * such as nft, tc, etc, we only allow a conservative subset of
1200 * possible pkt_types to be set.
1201 */
skb_pkt_type_ok(u32 ptype)1202 static inline bool skb_pkt_type_ok(u32 ptype)
1203 {
1204 return ptype <= PACKET_OTHERHOST;
1205 }
1206
1207 /**
1208 * skb_napi_id - Returns the skb's NAPI id
1209 * @skb: buffer
1210 */
skb_napi_id(const struct sk_buff * skb)1211 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1212 {
1213 #ifdef CONFIG_NET_RX_BUSY_POLL
1214 return skb->napi_id;
1215 #else
1216 return 0;
1217 #endif
1218 }
1219
skb_wifi_acked_valid(const struct sk_buff * skb)1220 static inline bool skb_wifi_acked_valid(const struct sk_buff *skb)
1221 {
1222 #ifdef CONFIG_WIRELESS
1223 return skb->wifi_acked_valid;
1224 #else
1225 return 0;
1226 #endif
1227 }
1228
1229 /**
1230 * skb_unref - decrement the skb's reference count
1231 * @skb: buffer
1232 *
1233 * Returns true if we can free the skb.
1234 */
skb_unref(struct sk_buff * skb)1235 static inline bool skb_unref(struct sk_buff *skb)
1236 {
1237 if (unlikely(!skb))
1238 return false;
1239 if (!IS_ENABLED(CONFIG_DEBUG_NET) && likely(refcount_read(&skb->users) == 1))
1240 smp_rmb();
1241 else if (likely(!refcount_dec_and_test(&skb->users)))
1242 return false;
1243
1244 return true;
1245 }
1246
skb_data_unref(const struct sk_buff * skb,struct skb_shared_info * shinfo)1247 static inline bool skb_data_unref(const struct sk_buff *skb,
1248 struct skb_shared_info *shinfo)
1249 {
1250 int bias;
1251
1252 if (!skb->cloned)
1253 return true;
1254
1255 bias = skb->nohdr ? (1 << SKB_DATAREF_SHIFT) + 1 : 1;
1256
1257 if (atomic_read(&shinfo->dataref) == bias)
1258 smp_rmb();
1259 else if (atomic_sub_return(bias, &shinfo->dataref))
1260 return false;
1261
1262 return true;
1263 }
1264
1265 void __fix_address sk_skb_reason_drop(struct sock *sk, struct sk_buff *skb,
1266 enum skb_drop_reason reason);
1267
1268 static inline void
kfree_skb_reason(struct sk_buff * skb,enum skb_drop_reason reason)1269 kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason)
1270 {
1271 sk_skb_reason_drop(NULL, skb, reason);
1272 }
1273
1274 /**
1275 * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1276 * @skb: buffer to free
1277 */
kfree_skb(struct sk_buff * skb)1278 static inline void kfree_skb(struct sk_buff *skb)
1279 {
1280 kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1281 }
1282
1283 void skb_release_head_state(struct sk_buff *skb);
1284 void kfree_skb_list_reason(struct sk_buff *segs,
1285 enum skb_drop_reason reason);
1286 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1287 void skb_tx_error(struct sk_buff *skb);
1288
kfree_skb_list(struct sk_buff * segs)1289 static inline void kfree_skb_list(struct sk_buff *segs)
1290 {
1291 kfree_skb_list_reason(segs, SKB_DROP_REASON_NOT_SPECIFIED);
1292 }
1293
1294 #ifdef CONFIG_TRACEPOINTS
1295 void consume_skb(struct sk_buff *skb);
1296 #else
consume_skb(struct sk_buff * skb)1297 static inline void consume_skb(struct sk_buff *skb)
1298 {
1299 return kfree_skb(skb);
1300 }
1301 #endif
1302
1303 void __consume_stateless_skb(struct sk_buff *skb);
1304 void __kfree_skb(struct sk_buff *skb);
1305
1306 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1307 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1308 bool *fragstolen, int *delta_truesize);
1309
1310 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1311 int node);
1312 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1313 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1314 struct sk_buff *build_skb_around(struct sk_buff *skb,
1315 void *data, unsigned int frag_size);
1316 void skb_attempt_defer_free(struct sk_buff *skb);
1317
1318 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1319 struct sk_buff *slab_build_skb(void *data);
1320
1321 /**
1322 * alloc_skb - allocate a network buffer
1323 * @size: size to allocate
1324 * @priority: allocation mask
1325 *
1326 * This function is a convenient wrapper around __alloc_skb().
1327 */
alloc_skb(unsigned int size,gfp_t priority)1328 static inline struct sk_buff *alloc_skb(unsigned int size,
1329 gfp_t priority)
1330 {
1331 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1332 }
1333
1334 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1335 unsigned long data_len,
1336 int max_page_order,
1337 int *errcode,
1338 gfp_t gfp_mask);
1339 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1340
1341 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1342 struct sk_buff_fclones {
1343 struct sk_buff skb1;
1344
1345 struct sk_buff skb2;
1346
1347 refcount_t fclone_ref;
1348 };
1349
1350 /**
1351 * skb_fclone_busy - check if fclone is busy
1352 * @sk: socket
1353 * @skb: buffer
1354 *
1355 * Returns true if skb is a fast clone, and its clone is not freed.
1356 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1357 * so we also check that didn't happen.
1358 */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1359 static inline bool skb_fclone_busy(const struct sock *sk,
1360 const struct sk_buff *skb)
1361 {
1362 const struct sk_buff_fclones *fclones;
1363
1364 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1365
1366 return skb->fclone == SKB_FCLONE_ORIG &&
1367 refcount_read(&fclones->fclone_ref) > 1 &&
1368 READ_ONCE(fclones->skb2.sk) == sk;
1369 }
1370
1371 /**
1372 * alloc_skb_fclone - allocate a network buffer from fclone cache
1373 * @size: size to allocate
1374 * @priority: allocation mask
1375 *
1376 * This function is a convenient wrapper around __alloc_skb().
1377 */
alloc_skb_fclone(unsigned int size,gfp_t priority)1378 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1379 gfp_t priority)
1380 {
1381 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1382 }
1383
1384 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1385 void skb_headers_offset_update(struct sk_buff *skb, int off);
1386 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1387 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1388 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1389 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1390 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1391 gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1392 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1393 gfp_t gfp_mask)
1394 {
1395 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1396 }
1397
1398 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1399 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1400 unsigned int headroom);
1401 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1402 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1403 int newtailroom, gfp_t priority);
1404 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1405 int offset, int len);
1406 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1407 int offset, int len);
1408 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1409 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1410
1411 /**
1412 * skb_pad - zero pad the tail of an skb
1413 * @skb: buffer to pad
1414 * @pad: space to pad
1415 *
1416 * Ensure that a buffer is followed by a padding area that is zero
1417 * filled. Used by network drivers which may DMA or transfer data
1418 * beyond the buffer end onto the wire.
1419 *
1420 * May return error in out of memory cases. The skb is freed on error.
1421 */
skb_pad(struct sk_buff * skb,int pad)1422 static inline int skb_pad(struct sk_buff *skb, int pad)
1423 {
1424 return __skb_pad(skb, pad, true);
1425 }
1426 #define dev_kfree_skb(a) consume_skb(a)
1427
1428 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1429 int offset, size_t size, size_t max_frags);
1430
1431 struct skb_seq_state {
1432 __u32 lower_offset;
1433 __u32 upper_offset;
1434 __u32 frag_idx;
1435 __u32 stepped_offset;
1436 struct sk_buff *root_skb;
1437 struct sk_buff *cur_skb;
1438 __u8 *frag_data;
1439 __u32 frag_off;
1440 };
1441
1442 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1443 unsigned int to, struct skb_seq_state *st);
1444 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1445 struct skb_seq_state *st);
1446 void skb_abort_seq_read(struct skb_seq_state *st);
1447 int skb_copy_seq_read(struct skb_seq_state *st, int offset, void *to, int len);
1448
1449 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1450 unsigned int to, struct ts_config *config);
1451
1452 /*
1453 * Packet hash types specify the type of hash in skb_set_hash.
1454 *
1455 * Hash types refer to the protocol layer addresses which are used to
1456 * construct a packet's hash. The hashes are used to differentiate or identify
1457 * flows of the protocol layer for the hash type. Hash types are either
1458 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1459 *
1460 * Properties of hashes:
1461 *
1462 * 1) Two packets in different flows have different hash values
1463 * 2) Two packets in the same flow should have the same hash value
1464 *
1465 * A hash at a higher layer is considered to be more specific. A driver should
1466 * set the most specific hash possible.
1467 *
1468 * A driver cannot indicate a more specific hash than the layer at which a hash
1469 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1470 *
1471 * A driver may indicate a hash level which is less specific than the
1472 * actual layer the hash was computed on. For instance, a hash computed
1473 * at L4 may be considered an L3 hash. This should only be done if the
1474 * driver can't unambiguously determine that the HW computed the hash at
1475 * the higher layer. Note that the "should" in the second property above
1476 * permits this.
1477 */
1478 enum pkt_hash_types {
1479 PKT_HASH_TYPE_NONE, /* Undefined type */
1480 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1481 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1482 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1483 };
1484
skb_clear_hash(struct sk_buff * skb)1485 static inline void skb_clear_hash(struct sk_buff *skb)
1486 {
1487 skb->hash = 0;
1488 skb->sw_hash = 0;
1489 skb->l4_hash = 0;
1490 }
1491
skb_clear_hash_if_not_l4(struct sk_buff * skb)1492 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1493 {
1494 if (!skb->l4_hash)
1495 skb_clear_hash(skb);
1496 }
1497
1498 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1499 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1500 {
1501 skb->l4_hash = is_l4;
1502 skb->sw_hash = is_sw;
1503 skb->hash = hash;
1504 }
1505
1506 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1507 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1508 {
1509 /* Used by drivers to set hash from HW */
1510 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1511 }
1512
1513 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1514 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1515 {
1516 __skb_set_hash(skb, hash, true, is_l4);
1517 }
1518
1519 u32 __skb_get_hash_symmetric_net(const struct net *net, const struct sk_buff *skb);
1520
__skb_get_hash_symmetric(const struct sk_buff * skb)1521 static inline u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1522 {
1523 return __skb_get_hash_symmetric_net(NULL, skb);
1524 }
1525
1526 void __skb_get_hash_net(const struct net *net, struct sk_buff *skb);
1527 u32 skb_get_poff(const struct sk_buff *skb);
1528 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1529 const struct flow_keys_basic *keys, int hlen);
1530 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1531 const void *data, int hlen_proto);
1532
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto)1533 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1534 int thoff, u8 ip_proto)
1535 {
1536 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1537 }
1538
1539 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1540 const struct flow_dissector_key *key,
1541 unsigned int key_count);
1542
1543 struct bpf_flow_dissector;
1544 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1545 __be16 proto, int nhoff, int hlen, unsigned int flags);
1546
1547 bool __skb_flow_dissect(const struct net *net,
1548 const struct sk_buff *skb,
1549 struct flow_dissector *flow_dissector,
1550 void *target_container, const void *data,
1551 __be16 proto, int nhoff, int hlen, unsigned int flags);
1552
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1553 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1554 struct flow_dissector *flow_dissector,
1555 void *target_container, unsigned int flags)
1556 {
1557 return __skb_flow_dissect(NULL, skb, flow_dissector,
1558 target_container, NULL, 0, 0, 0, flags);
1559 }
1560
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1561 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1562 struct flow_keys *flow,
1563 unsigned int flags)
1564 {
1565 memset(flow, 0, sizeof(*flow));
1566 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1567 flow, NULL, 0, 0, 0, flags);
1568 }
1569
1570 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)1571 skb_flow_dissect_flow_keys_basic(const struct net *net,
1572 const struct sk_buff *skb,
1573 struct flow_keys_basic *flow,
1574 const void *data, __be16 proto,
1575 int nhoff, int hlen, unsigned int flags)
1576 {
1577 memset(flow, 0, sizeof(*flow));
1578 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1579 data, proto, nhoff, hlen, flags);
1580 }
1581
1582 void skb_flow_dissect_meta(const struct sk_buff *skb,
1583 struct flow_dissector *flow_dissector,
1584 void *target_container);
1585
1586 /* Gets a skb connection tracking info, ctinfo map should be a
1587 * map of mapsize to translate enum ip_conntrack_info states
1588 * to user states.
1589 */
1590 void
1591 skb_flow_dissect_ct(const struct sk_buff *skb,
1592 struct flow_dissector *flow_dissector,
1593 void *target_container,
1594 u16 *ctinfo_map, size_t mapsize,
1595 bool post_ct, u16 zone);
1596 void
1597 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1598 struct flow_dissector *flow_dissector,
1599 void *target_container);
1600
1601 void skb_flow_dissect_hash(const struct sk_buff *skb,
1602 struct flow_dissector *flow_dissector,
1603 void *target_container);
1604
skb_get_hash_net(const struct net * net,struct sk_buff * skb)1605 static inline __u32 skb_get_hash_net(const struct net *net, struct sk_buff *skb)
1606 {
1607 if (!skb->l4_hash && !skb->sw_hash)
1608 __skb_get_hash_net(net, skb);
1609
1610 return skb->hash;
1611 }
1612
skb_get_hash(struct sk_buff * skb)1613 static inline __u32 skb_get_hash(struct sk_buff *skb)
1614 {
1615 if (!skb->l4_hash && !skb->sw_hash)
1616 __skb_get_hash_net(NULL, skb);
1617
1618 return skb->hash;
1619 }
1620
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1621 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1622 {
1623 if (!skb->l4_hash && !skb->sw_hash) {
1624 struct flow_keys keys;
1625 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1626
1627 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1628 }
1629
1630 return skb->hash;
1631 }
1632
1633 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1634 const siphash_key_t *perturb);
1635
skb_get_hash_raw(const struct sk_buff * skb)1636 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1637 {
1638 return skb->hash;
1639 }
1640
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1641 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1642 {
1643 to->hash = from->hash;
1644 to->sw_hash = from->sw_hash;
1645 to->l4_hash = from->l4_hash;
1646 };
1647
skb_cmp_decrypted(const struct sk_buff * skb1,const struct sk_buff * skb2)1648 static inline int skb_cmp_decrypted(const struct sk_buff *skb1,
1649 const struct sk_buff *skb2)
1650 {
1651 #ifdef CONFIG_SKB_DECRYPTED
1652 return skb2->decrypted - skb1->decrypted;
1653 #else
1654 return 0;
1655 #endif
1656 }
1657
skb_is_decrypted(const struct sk_buff * skb)1658 static inline bool skb_is_decrypted(const struct sk_buff *skb)
1659 {
1660 #ifdef CONFIG_SKB_DECRYPTED
1661 return skb->decrypted;
1662 #else
1663 return false;
1664 #endif
1665 }
1666
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1667 static inline void skb_copy_decrypted(struct sk_buff *to,
1668 const struct sk_buff *from)
1669 {
1670 #ifdef CONFIG_SKB_DECRYPTED
1671 to->decrypted = from->decrypted;
1672 #endif
1673 }
1674
1675 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1676 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1677 {
1678 return skb->head + skb->end;
1679 }
1680
skb_end_offset(const struct sk_buff * skb)1681 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1682 {
1683 return skb->end;
1684 }
1685
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1686 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1687 {
1688 skb->end = offset;
1689 }
1690 #else
skb_end_pointer(const struct sk_buff * skb)1691 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1692 {
1693 return skb->end;
1694 }
1695
skb_end_offset(const struct sk_buff * skb)1696 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1697 {
1698 return skb->end - skb->head;
1699 }
1700
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1701 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1702 {
1703 skb->end = skb->head + offset;
1704 }
1705 #endif
1706
1707 extern const struct ubuf_info_ops msg_zerocopy_ubuf_ops;
1708
1709 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
1710 struct ubuf_info *uarg);
1711
1712 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
1713
1714 int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk,
1715 struct sk_buff *skb, struct iov_iter *from,
1716 size_t length);
1717
1718 int zerocopy_fill_skb_from_iter(struct sk_buff *skb,
1719 struct iov_iter *from, size_t length);
1720
skb_zerocopy_iter_dgram(struct sk_buff * skb,struct msghdr * msg,int len)1721 static inline int skb_zerocopy_iter_dgram(struct sk_buff *skb,
1722 struct msghdr *msg, int len)
1723 {
1724 return __zerocopy_sg_from_iter(msg, skb->sk, skb, &msg->msg_iter, len);
1725 }
1726
1727 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
1728 struct msghdr *msg, int len,
1729 struct ubuf_info *uarg);
1730
1731 /* Internal */
1732 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1733
skb_hwtstamps(struct sk_buff * skb)1734 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1735 {
1736 return &skb_shinfo(skb)->hwtstamps;
1737 }
1738
skb_zcopy(struct sk_buff * skb)1739 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1740 {
1741 bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1742
1743 return is_zcopy ? skb_uarg(skb) : NULL;
1744 }
1745
skb_zcopy_pure(const struct sk_buff * skb)1746 static inline bool skb_zcopy_pure(const struct sk_buff *skb)
1747 {
1748 return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY;
1749 }
1750
skb_zcopy_managed(const struct sk_buff * skb)1751 static inline bool skb_zcopy_managed(const struct sk_buff *skb)
1752 {
1753 return skb_shinfo(skb)->flags & SKBFL_MANAGED_FRAG_REFS;
1754 }
1755
skb_pure_zcopy_same(const struct sk_buff * skb1,const struct sk_buff * skb2)1756 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1,
1757 const struct sk_buff *skb2)
1758 {
1759 return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2);
1760 }
1761
net_zcopy_get(struct ubuf_info * uarg)1762 static inline void net_zcopy_get(struct ubuf_info *uarg)
1763 {
1764 refcount_inc(&uarg->refcnt);
1765 }
1766
skb_zcopy_init(struct sk_buff * skb,struct ubuf_info * uarg)1767 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1768 {
1769 skb_shinfo(skb)->destructor_arg = uarg;
1770 skb_shinfo(skb)->flags |= uarg->flags;
1771 }
1772
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1773 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1774 bool *have_ref)
1775 {
1776 if (skb && uarg && !skb_zcopy(skb)) {
1777 if (unlikely(have_ref && *have_ref))
1778 *have_ref = false;
1779 else
1780 net_zcopy_get(uarg);
1781 skb_zcopy_init(skb, uarg);
1782 }
1783 }
1784
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1785 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1786 {
1787 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1788 skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1789 }
1790
skb_zcopy_is_nouarg(struct sk_buff * skb)1791 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1792 {
1793 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1794 }
1795
skb_zcopy_get_nouarg(struct sk_buff * skb)1796 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1797 {
1798 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1799 }
1800
net_zcopy_put(struct ubuf_info * uarg)1801 static inline void net_zcopy_put(struct ubuf_info *uarg)
1802 {
1803 if (uarg)
1804 uarg->ops->complete(NULL, uarg, true);
1805 }
1806
net_zcopy_put_abort(struct ubuf_info * uarg,bool have_uref)1807 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1808 {
1809 if (uarg) {
1810 if (uarg->ops == &msg_zerocopy_ubuf_ops)
1811 msg_zerocopy_put_abort(uarg, have_uref);
1812 else if (have_uref)
1813 net_zcopy_put(uarg);
1814 }
1815 }
1816
1817 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy_success)1818 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1819 {
1820 struct ubuf_info *uarg = skb_zcopy(skb);
1821
1822 if (uarg) {
1823 if (!skb_zcopy_is_nouarg(skb))
1824 uarg->ops->complete(skb, uarg, zerocopy_success);
1825
1826 skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY;
1827 }
1828 }
1829
1830 void __skb_zcopy_downgrade_managed(struct sk_buff *skb);
1831
skb_zcopy_downgrade_managed(struct sk_buff * skb)1832 static inline void skb_zcopy_downgrade_managed(struct sk_buff *skb)
1833 {
1834 if (unlikely(skb_zcopy_managed(skb)))
1835 __skb_zcopy_downgrade_managed(skb);
1836 }
1837
1838 /* Return true if frags in this skb are readable by the host. */
skb_frags_readable(const struct sk_buff * skb)1839 static inline bool skb_frags_readable(const struct sk_buff *skb)
1840 {
1841 return !skb->unreadable;
1842 }
1843
skb_mark_not_on_list(struct sk_buff * skb)1844 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1845 {
1846 skb->next = NULL;
1847 }
1848
skb_poison_list(struct sk_buff * skb)1849 static inline void skb_poison_list(struct sk_buff *skb)
1850 {
1851 #ifdef CONFIG_DEBUG_NET
1852 skb->next = SKB_LIST_POISON_NEXT;
1853 #endif
1854 }
1855
1856 /* Iterate through singly-linked GSO fragments of an skb. */
1857 #define skb_list_walk_safe(first, skb, next_skb) \
1858 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1859 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1860
skb_list_del_init(struct sk_buff * skb)1861 static inline void skb_list_del_init(struct sk_buff *skb)
1862 {
1863 __list_del_entry(&skb->list);
1864 skb_mark_not_on_list(skb);
1865 }
1866
1867 /**
1868 * skb_queue_empty - check if a queue is empty
1869 * @list: queue head
1870 *
1871 * Returns true if the queue is empty, false otherwise.
1872 */
skb_queue_empty(const struct sk_buff_head * list)1873 static inline int skb_queue_empty(const struct sk_buff_head *list)
1874 {
1875 return list->next == (const struct sk_buff *) list;
1876 }
1877
1878 /**
1879 * skb_queue_empty_lockless - check if a queue is empty
1880 * @list: queue head
1881 *
1882 * Returns true if the queue is empty, false otherwise.
1883 * This variant can be used in lockless contexts.
1884 */
skb_queue_empty_lockless(const struct sk_buff_head * list)1885 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1886 {
1887 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1888 }
1889
1890
1891 /**
1892 * skb_queue_is_last - check if skb is the last entry in the queue
1893 * @list: queue head
1894 * @skb: buffer
1895 *
1896 * Returns true if @skb is the last buffer on the list.
1897 */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1898 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1899 const struct sk_buff *skb)
1900 {
1901 return skb->next == (const struct sk_buff *) list;
1902 }
1903
1904 /**
1905 * skb_queue_is_first - check if skb is the first entry in the queue
1906 * @list: queue head
1907 * @skb: buffer
1908 *
1909 * Returns true if @skb is the first buffer on the list.
1910 */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1911 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1912 const struct sk_buff *skb)
1913 {
1914 return skb->prev == (const struct sk_buff *) list;
1915 }
1916
1917 /**
1918 * skb_queue_next - return the next packet in the queue
1919 * @list: queue head
1920 * @skb: current buffer
1921 *
1922 * Return the next packet in @list after @skb. It is only valid to
1923 * call this if skb_queue_is_last() evaluates to false.
1924 */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1925 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1926 const struct sk_buff *skb)
1927 {
1928 /* This BUG_ON may seem severe, but if we just return then we
1929 * are going to dereference garbage.
1930 */
1931 BUG_ON(skb_queue_is_last(list, skb));
1932 return skb->next;
1933 }
1934
1935 /**
1936 * skb_queue_prev - return the prev packet in the queue
1937 * @list: queue head
1938 * @skb: current buffer
1939 *
1940 * Return the prev packet in @list before @skb. It is only valid to
1941 * call this if skb_queue_is_first() evaluates to false.
1942 */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1943 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1944 const struct sk_buff *skb)
1945 {
1946 /* This BUG_ON may seem severe, but if we just return then we
1947 * are going to dereference garbage.
1948 */
1949 BUG_ON(skb_queue_is_first(list, skb));
1950 return skb->prev;
1951 }
1952
1953 /**
1954 * skb_get - reference buffer
1955 * @skb: buffer to reference
1956 *
1957 * Makes another reference to a socket buffer and returns a pointer
1958 * to the buffer.
1959 */
skb_get(struct sk_buff * skb)1960 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1961 {
1962 refcount_inc(&skb->users);
1963 return skb;
1964 }
1965
1966 /*
1967 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1968 */
1969
1970 /**
1971 * skb_cloned - is the buffer a clone
1972 * @skb: buffer to check
1973 *
1974 * Returns true if the buffer was generated with skb_clone() and is
1975 * one of multiple shared copies of the buffer. Cloned buffers are
1976 * shared data so must not be written to under normal circumstances.
1977 */
skb_cloned(const struct sk_buff * skb)1978 static inline int skb_cloned(const struct sk_buff *skb)
1979 {
1980 return skb->cloned &&
1981 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1982 }
1983
skb_unclone(struct sk_buff * skb,gfp_t pri)1984 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1985 {
1986 might_sleep_if(gfpflags_allow_blocking(pri));
1987
1988 if (skb_cloned(skb))
1989 return pskb_expand_head(skb, 0, 0, pri);
1990
1991 return 0;
1992 }
1993
1994 /* This variant of skb_unclone() makes sure skb->truesize
1995 * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1996 *
1997 * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
1998 * when various debugging features are in place.
1999 */
2000 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
skb_unclone_keeptruesize(struct sk_buff * skb,gfp_t pri)2001 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
2002 {
2003 might_sleep_if(gfpflags_allow_blocking(pri));
2004
2005 if (skb_cloned(skb))
2006 return __skb_unclone_keeptruesize(skb, pri);
2007 return 0;
2008 }
2009
2010 /**
2011 * skb_header_cloned - is the header a clone
2012 * @skb: buffer to check
2013 *
2014 * Returns true if modifying the header part of the buffer requires
2015 * the data to be copied.
2016 */
skb_header_cloned(const struct sk_buff * skb)2017 static inline int skb_header_cloned(const struct sk_buff *skb)
2018 {
2019 int dataref;
2020
2021 if (!skb->cloned)
2022 return 0;
2023
2024 dataref = atomic_read(&skb_shinfo(skb)->dataref);
2025 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
2026 return dataref != 1;
2027 }
2028
skb_header_unclone(struct sk_buff * skb,gfp_t pri)2029 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
2030 {
2031 might_sleep_if(gfpflags_allow_blocking(pri));
2032
2033 if (skb_header_cloned(skb))
2034 return pskb_expand_head(skb, 0, 0, pri);
2035
2036 return 0;
2037 }
2038
2039 /**
2040 * __skb_header_release() - allow clones to use the headroom
2041 * @skb: buffer to operate on
2042 *
2043 * See "DOC: dataref and headerless skbs".
2044 */
__skb_header_release(struct sk_buff * skb)2045 static inline void __skb_header_release(struct sk_buff *skb)
2046 {
2047 skb->nohdr = 1;
2048 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
2049 }
2050
2051
2052 /**
2053 * skb_shared - is the buffer shared
2054 * @skb: buffer to check
2055 *
2056 * Returns true if more than one person has a reference to this
2057 * buffer.
2058 */
skb_shared(const struct sk_buff * skb)2059 static inline int skb_shared(const struct sk_buff *skb)
2060 {
2061 return refcount_read(&skb->users) != 1;
2062 }
2063
2064 /**
2065 * skb_share_check - check if buffer is shared and if so clone it
2066 * @skb: buffer to check
2067 * @pri: priority for memory allocation
2068 *
2069 * If the buffer is shared the buffer is cloned and the old copy
2070 * drops a reference. A new clone with a single reference is returned.
2071 * If the buffer is not shared the original buffer is returned. When
2072 * being called from interrupt status or with spinlocks held pri must
2073 * be GFP_ATOMIC.
2074 *
2075 * NULL is returned on a memory allocation failure.
2076 */
skb_share_check(struct sk_buff * skb,gfp_t pri)2077 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
2078 {
2079 might_sleep_if(gfpflags_allow_blocking(pri));
2080 if (skb_shared(skb)) {
2081 struct sk_buff *nskb = skb_clone(skb, pri);
2082
2083 if (likely(nskb))
2084 consume_skb(skb);
2085 else
2086 kfree_skb(skb);
2087 skb = nskb;
2088 }
2089 return skb;
2090 }
2091
2092 /*
2093 * Copy shared buffers into a new sk_buff. We effectively do COW on
2094 * packets to handle cases where we have a local reader and forward
2095 * and a couple of other messy ones. The normal one is tcpdumping
2096 * a packet that's being forwarded.
2097 */
2098
2099 /**
2100 * skb_unshare - make a copy of a shared buffer
2101 * @skb: buffer to check
2102 * @pri: priority for memory allocation
2103 *
2104 * If the socket buffer is a clone then this function creates a new
2105 * copy of the data, drops a reference count on the old copy and returns
2106 * the new copy with the reference count at 1. If the buffer is not a clone
2107 * the original buffer is returned. When called with a spinlock held or
2108 * from interrupt state @pri must be %GFP_ATOMIC
2109 *
2110 * %NULL is returned on a memory allocation failure.
2111 */
skb_unshare(struct sk_buff * skb,gfp_t pri)2112 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
2113 gfp_t pri)
2114 {
2115 might_sleep_if(gfpflags_allow_blocking(pri));
2116 if (skb_cloned(skb)) {
2117 struct sk_buff *nskb = skb_copy(skb, pri);
2118
2119 /* Free our shared copy */
2120 if (likely(nskb))
2121 consume_skb(skb);
2122 else
2123 kfree_skb(skb);
2124 skb = nskb;
2125 }
2126 return skb;
2127 }
2128
2129 /**
2130 * skb_peek - peek at the head of an &sk_buff_head
2131 * @list_: list to peek at
2132 *
2133 * Peek an &sk_buff. Unlike most other operations you _MUST_
2134 * be careful with this one. A peek leaves the buffer on the
2135 * list and someone else may run off with it. You must hold
2136 * the appropriate locks or have a private queue to do this.
2137 *
2138 * Returns %NULL for an empty list or a pointer to the head element.
2139 * The reference count is not incremented and the reference is therefore
2140 * volatile. Use with caution.
2141 */
skb_peek(const struct sk_buff_head * list_)2142 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
2143 {
2144 struct sk_buff *skb = list_->next;
2145
2146 if (skb == (struct sk_buff *)list_)
2147 skb = NULL;
2148 return skb;
2149 }
2150
2151 /**
2152 * __skb_peek - peek at the head of a non-empty &sk_buff_head
2153 * @list_: list to peek at
2154 *
2155 * Like skb_peek(), but the caller knows that the list is not empty.
2156 */
__skb_peek(const struct sk_buff_head * list_)2157 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
2158 {
2159 return list_->next;
2160 }
2161
2162 /**
2163 * skb_peek_next - peek skb following the given one from a queue
2164 * @skb: skb to start from
2165 * @list_: list to peek at
2166 *
2167 * Returns %NULL when the end of the list is met or a pointer to the
2168 * next element. The reference count is not incremented and the
2169 * reference is therefore volatile. Use with caution.
2170 */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)2171 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
2172 const struct sk_buff_head *list_)
2173 {
2174 struct sk_buff *next = skb->next;
2175
2176 if (next == (struct sk_buff *)list_)
2177 next = NULL;
2178 return next;
2179 }
2180
2181 /**
2182 * skb_peek_tail - peek at the tail of an &sk_buff_head
2183 * @list_: list to peek at
2184 *
2185 * Peek an &sk_buff. Unlike most other operations you _MUST_
2186 * be careful with this one. A peek leaves the buffer on the
2187 * list and someone else may run off with it. You must hold
2188 * the appropriate locks or have a private queue to do this.
2189 *
2190 * Returns %NULL for an empty list or a pointer to the tail element.
2191 * The reference count is not incremented and the reference is therefore
2192 * volatile. Use with caution.
2193 */
skb_peek_tail(const struct sk_buff_head * list_)2194 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
2195 {
2196 struct sk_buff *skb = READ_ONCE(list_->prev);
2197
2198 if (skb == (struct sk_buff *)list_)
2199 skb = NULL;
2200 return skb;
2201
2202 }
2203
2204 /**
2205 * skb_queue_len - get queue length
2206 * @list_: list to measure
2207 *
2208 * Return the length of an &sk_buff queue.
2209 */
skb_queue_len(const struct sk_buff_head * list_)2210 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
2211 {
2212 return list_->qlen;
2213 }
2214
2215 /**
2216 * skb_queue_len_lockless - get queue length
2217 * @list_: list to measure
2218 *
2219 * Return the length of an &sk_buff queue.
2220 * This variant can be used in lockless contexts.
2221 */
skb_queue_len_lockless(const struct sk_buff_head * list_)2222 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
2223 {
2224 return READ_ONCE(list_->qlen);
2225 }
2226
2227 /**
2228 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
2229 * @list: queue to initialize
2230 *
2231 * This initializes only the list and queue length aspects of
2232 * an sk_buff_head object. This allows to initialize the list
2233 * aspects of an sk_buff_head without reinitializing things like
2234 * the spinlock. It can also be used for on-stack sk_buff_head
2235 * objects where the spinlock is known to not be used.
2236 */
__skb_queue_head_init(struct sk_buff_head * list)2237 static inline void __skb_queue_head_init(struct sk_buff_head *list)
2238 {
2239 list->prev = list->next = (struct sk_buff *)list;
2240 list->qlen = 0;
2241 }
2242
2243 /*
2244 * This function creates a split out lock class for each invocation;
2245 * this is needed for now since a whole lot of users of the skb-queue
2246 * infrastructure in drivers have different locking usage (in hardirq)
2247 * than the networking core (in softirq only). In the long run either the
2248 * network layer or drivers should need annotation to consolidate the
2249 * main types of usage into 3 classes.
2250 */
skb_queue_head_init(struct sk_buff_head * list)2251 static inline void skb_queue_head_init(struct sk_buff_head *list)
2252 {
2253 spin_lock_init(&list->lock);
2254 __skb_queue_head_init(list);
2255 }
2256
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)2257 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
2258 struct lock_class_key *class)
2259 {
2260 skb_queue_head_init(list);
2261 lockdep_set_class(&list->lock, class);
2262 }
2263
2264 /*
2265 * Insert an sk_buff on a list.
2266 *
2267 * The "__skb_xxxx()" functions are the non-atomic ones that
2268 * can only be called with interrupts disabled.
2269 */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)2270 static inline void __skb_insert(struct sk_buff *newsk,
2271 struct sk_buff *prev, struct sk_buff *next,
2272 struct sk_buff_head *list)
2273 {
2274 /* See skb_queue_empty_lockless() and skb_peek_tail()
2275 * for the opposite READ_ONCE()
2276 */
2277 WRITE_ONCE(newsk->next, next);
2278 WRITE_ONCE(newsk->prev, prev);
2279 WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk);
2280 WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk);
2281 WRITE_ONCE(list->qlen, list->qlen + 1);
2282 }
2283
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)2284 static inline void __skb_queue_splice(const struct sk_buff_head *list,
2285 struct sk_buff *prev,
2286 struct sk_buff *next)
2287 {
2288 struct sk_buff *first = list->next;
2289 struct sk_buff *last = list->prev;
2290
2291 WRITE_ONCE(first->prev, prev);
2292 WRITE_ONCE(prev->next, first);
2293
2294 WRITE_ONCE(last->next, next);
2295 WRITE_ONCE(next->prev, last);
2296 }
2297
2298 /**
2299 * skb_queue_splice - join two skb lists, this is designed for stacks
2300 * @list: the new list to add
2301 * @head: the place to add it in the first list
2302 */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)2303 static inline void skb_queue_splice(const struct sk_buff_head *list,
2304 struct sk_buff_head *head)
2305 {
2306 if (!skb_queue_empty(list)) {
2307 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2308 head->qlen += list->qlen;
2309 }
2310 }
2311
2312 /**
2313 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2314 * @list: the new list to add
2315 * @head: the place to add it in the first list
2316 *
2317 * The list at @list is reinitialised
2318 */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)2319 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2320 struct sk_buff_head *head)
2321 {
2322 if (!skb_queue_empty(list)) {
2323 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
2324 head->qlen += list->qlen;
2325 __skb_queue_head_init(list);
2326 }
2327 }
2328
2329 /**
2330 * skb_queue_splice_tail - join two skb lists, each list being a queue
2331 * @list: the new list to add
2332 * @head: the place to add it in the first list
2333 */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)2334 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2335 struct sk_buff_head *head)
2336 {
2337 if (!skb_queue_empty(list)) {
2338 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2339 head->qlen += list->qlen;
2340 }
2341 }
2342
2343 /**
2344 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2345 * @list: the new list to add
2346 * @head: the place to add it in the first list
2347 *
2348 * Each of the lists is a queue.
2349 * The list at @list is reinitialised
2350 */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)2351 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2352 struct sk_buff_head *head)
2353 {
2354 if (!skb_queue_empty(list)) {
2355 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2356 head->qlen += list->qlen;
2357 __skb_queue_head_init(list);
2358 }
2359 }
2360
2361 /**
2362 * __skb_queue_after - queue a buffer at the list head
2363 * @list: list to use
2364 * @prev: place after this buffer
2365 * @newsk: buffer to queue
2366 *
2367 * Queue a buffer int the middle of a list. This function takes no locks
2368 * and you must therefore hold required locks before calling it.
2369 *
2370 * A buffer cannot be placed on two lists at the same time.
2371 */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)2372 static inline void __skb_queue_after(struct sk_buff_head *list,
2373 struct sk_buff *prev,
2374 struct sk_buff *newsk)
2375 {
2376 __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list);
2377 }
2378
2379 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2380 struct sk_buff_head *list);
2381
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)2382 static inline void __skb_queue_before(struct sk_buff_head *list,
2383 struct sk_buff *next,
2384 struct sk_buff *newsk)
2385 {
2386 __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list);
2387 }
2388
2389 /**
2390 * __skb_queue_head - queue a buffer at the list head
2391 * @list: list to use
2392 * @newsk: buffer to queue
2393 *
2394 * Queue a buffer at the start of a list. This function takes no locks
2395 * and you must therefore hold required locks before calling it.
2396 *
2397 * A buffer cannot be placed on two lists at the same time.
2398 */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)2399 static inline void __skb_queue_head(struct sk_buff_head *list,
2400 struct sk_buff *newsk)
2401 {
2402 __skb_queue_after(list, (struct sk_buff *)list, newsk);
2403 }
2404 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2405
2406 /**
2407 * __skb_queue_tail - queue a buffer at the list tail
2408 * @list: list to use
2409 * @newsk: buffer to queue
2410 *
2411 * Queue a buffer at the end of a list. This function takes no locks
2412 * and you must therefore hold required locks before calling it.
2413 *
2414 * A buffer cannot be placed on two lists at the same time.
2415 */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2416 static inline void __skb_queue_tail(struct sk_buff_head *list,
2417 struct sk_buff *newsk)
2418 {
2419 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2420 }
2421 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2422
2423 /*
2424 * remove sk_buff from list. _Must_ be called atomically, and with
2425 * the list known..
2426 */
2427 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2428 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2429 {
2430 struct sk_buff *next, *prev;
2431
2432 WRITE_ONCE(list->qlen, list->qlen - 1);
2433 next = skb->next;
2434 prev = skb->prev;
2435 skb->next = skb->prev = NULL;
2436 WRITE_ONCE(next->prev, prev);
2437 WRITE_ONCE(prev->next, next);
2438 }
2439
2440 /**
2441 * __skb_dequeue - remove from the head of the queue
2442 * @list: list to dequeue from
2443 *
2444 * Remove the head of the list. This function does not take any locks
2445 * so must be used with appropriate locks held only. The head item is
2446 * returned or %NULL if the list is empty.
2447 */
__skb_dequeue(struct sk_buff_head * list)2448 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2449 {
2450 struct sk_buff *skb = skb_peek(list);
2451 if (skb)
2452 __skb_unlink(skb, list);
2453 return skb;
2454 }
2455 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2456
2457 /**
2458 * __skb_dequeue_tail - remove from the tail of the queue
2459 * @list: list to dequeue from
2460 *
2461 * Remove the tail of the list. This function does not take any locks
2462 * so must be used with appropriate locks held only. The tail item is
2463 * returned or %NULL if the list is empty.
2464 */
__skb_dequeue_tail(struct sk_buff_head * list)2465 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2466 {
2467 struct sk_buff *skb = skb_peek_tail(list);
2468 if (skb)
2469 __skb_unlink(skb, list);
2470 return skb;
2471 }
2472 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2473
2474
skb_is_nonlinear(const struct sk_buff * skb)2475 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2476 {
2477 return skb->data_len;
2478 }
2479
skb_headlen(const struct sk_buff * skb)2480 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2481 {
2482 return skb->len - skb->data_len;
2483 }
2484
__skb_pagelen(const struct sk_buff * skb)2485 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2486 {
2487 unsigned int i, len = 0;
2488
2489 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2490 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2491 return len;
2492 }
2493
skb_pagelen(const struct sk_buff * skb)2494 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2495 {
2496 return skb_headlen(skb) + __skb_pagelen(skb);
2497 }
2498
skb_frag_fill_netmem_desc(skb_frag_t * frag,netmem_ref netmem,int off,int size)2499 static inline void skb_frag_fill_netmem_desc(skb_frag_t *frag,
2500 netmem_ref netmem, int off,
2501 int size)
2502 {
2503 frag->netmem = netmem;
2504 frag->offset = off;
2505 skb_frag_size_set(frag, size);
2506 }
2507
skb_frag_fill_page_desc(skb_frag_t * frag,struct page * page,int off,int size)2508 static inline void skb_frag_fill_page_desc(skb_frag_t *frag,
2509 struct page *page,
2510 int off, int size)
2511 {
2512 skb_frag_fill_netmem_desc(frag, page_to_netmem(page), off, size);
2513 }
2514
__skb_fill_netmem_desc_noacc(struct skb_shared_info * shinfo,int i,netmem_ref netmem,int off,int size)2515 static inline void __skb_fill_netmem_desc_noacc(struct skb_shared_info *shinfo,
2516 int i, netmem_ref netmem,
2517 int off, int size)
2518 {
2519 skb_frag_t *frag = &shinfo->frags[i];
2520
2521 skb_frag_fill_netmem_desc(frag, netmem, off, size);
2522 }
2523
__skb_fill_page_desc_noacc(struct skb_shared_info * shinfo,int i,struct page * page,int off,int size)2524 static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo,
2525 int i, struct page *page,
2526 int off, int size)
2527 {
2528 __skb_fill_netmem_desc_noacc(shinfo, i, page_to_netmem(page), off,
2529 size);
2530 }
2531
2532 /**
2533 * skb_len_add - adds a number to len fields of skb
2534 * @skb: buffer to add len to
2535 * @delta: number of bytes to add
2536 */
skb_len_add(struct sk_buff * skb,int delta)2537 static inline void skb_len_add(struct sk_buff *skb, int delta)
2538 {
2539 skb->len += delta;
2540 skb->data_len += delta;
2541 skb->truesize += delta;
2542 }
2543
2544 /**
2545 * __skb_fill_netmem_desc - initialise a fragment in an skb
2546 * @skb: buffer containing fragment to be initialised
2547 * @i: fragment index to initialise
2548 * @netmem: the netmem to use for this fragment
2549 * @off: the offset to the data with @page
2550 * @size: the length of the data
2551 *
2552 * Initialises the @i'th fragment of @skb to point to &size bytes at
2553 * offset @off within @page.
2554 *
2555 * Does not take any additional reference on the fragment.
2556 */
__skb_fill_netmem_desc(struct sk_buff * skb,int i,netmem_ref netmem,int off,int size)2557 static inline void __skb_fill_netmem_desc(struct sk_buff *skb, int i,
2558 netmem_ref netmem, int off, int size)
2559 {
2560 struct page *page;
2561
2562 __skb_fill_netmem_desc_noacc(skb_shinfo(skb), i, netmem, off, size);
2563
2564 if (netmem_is_net_iov(netmem)) {
2565 skb->unreadable = true;
2566 return;
2567 }
2568
2569 page = netmem_to_page(netmem);
2570
2571 /* Propagate page pfmemalloc to the skb if we can. The problem is
2572 * that not all callers have unique ownership of the page but rely
2573 * on page_is_pfmemalloc doing the right thing(tm).
2574 */
2575 page = compound_head(page);
2576 if (page_is_pfmemalloc(page))
2577 skb->pfmemalloc = true;
2578 }
2579
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2580 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2581 struct page *page, int off, int size)
2582 {
2583 __skb_fill_netmem_desc(skb, i, page_to_netmem(page), off, size);
2584 }
2585
skb_fill_netmem_desc(struct sk_buff * skb,int i,netmem_ref netmem,int off,int size)2586 static inline void skb_fill_netmem_desc(struct sk_buff *skb, int i,
2587 netmem_ref netmem, int off, int size)
2588 {
2589 __skb_fill_netmem_desc(skb, i, netmem, off, size);
2590 skb_shinfo(skb)->nr_frags = i + 1;
2591 }
2592
2593 /**
2594 * skb_fill_page_desc - initialise a paged fragment in an skb
2595 * @skb: buffer containing fragment to be initialised
2596 * @i: paged fragment index to initialise
2597 * @page: the page to use for this fragment
2598 * @off: the offset to the data with @page
2599 * @size: the length of the data
2600 *
2601 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2602 * @skb to point to @size bytes at offset @off within @page. In
2603 * addition updates @skb such that @i is the last fragment.
2604 *
2605 * Does not take any additional reference on the fragment.
2606 */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2607 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2608 struct page *page, int off, int size)
2609 {
2610 skb_fill_netmem_desc(skb, i, page_to_netmem(page), off, size);
2611 }
2612
2613 /**
2614 * skb_fill_page_desc_noacc - initialise a paged fragment in an skb
2615 * @skb: buffer containing fragment to be initialised
2616 * @i: paged fragment index to initialise
2617 * @page: the page to use for this fragment
2618 * @off: the offset to the data with @page
2619 * @size: the length of the data
2620 *
2621 * Variant of skb_fill_page_desc() which does not deal with
2622 * pfmemalloc, if page is not owned by us.
2623 */
skb_fill_page_desc_noacc(struct sk_buff * skb,int i,struct page * page,int off,int size)2624 static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i,
2625 struct page *page, int off,
2626 int size)
2627 {
2628 struct skb_shared_info *shinfo = skb_shinfo(skb);
2629
2630 __skb_fill_page_desc_noacc(shinfo, i, page, off, size);
2631 shinfo->nr_frags = i + 1;
2632 }
2633
2634 void skb_add_rx_frag_netmem(struct sk_buff *skb, int i, netmem_ref netmem,
2635 int off, int size, unsigned int truesize);
2636
skb_add_rx_frag(struct sk_buff * skb,int i,struct page * page,int off,int size,unsigned int truesize)2637 static inline void skb_add_rx_frag(struct sk_buff *skb, int i,
2638 struct page *page, int off, int size,
2639 unsigned int truesize)
2640 {
2641 skb_add_rx_frag_netmem(skb, i, page_to_netmem(page), off, size,
2642 truesize);
2643 }
2644
2645 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2646 unsigned int truesize);
2647
2648 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2649
2650 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2651 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2652 {
2653 return skb->head + skb->tail;
2654 }
2655
skb_reset_tail_pointer(struct sk_buff * skb)2656 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2657 {
2658 skb->tail = skb->data - skb->head;
2659 }
2660
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2661 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2662 {
2663 skb_reset_tail_pointer(skb);
2664 skb->tail += offset;
2665 }
2666
2667 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2668 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2669 {
2670 return skb->tail;
2671 }
2672
skb_reset_tail_pointer(struct sk_buff * skb)2673 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2674 {
2675 skb->tail = skb->data;
2676 }
2677
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2678 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2679 {
2680 skb->tail = skb->data + offset;
2681 }
2682
2683 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2684
skb_assert_len(struct sk_buff * skb)2685 static inline void skb_assert_len(struct sk_buff *skb)
2686 {
2687 #ifdef CONFIG_DEBUG_NET
2688 if (WARN_ONCE(!skb->len, "%s\n", __func__))
2689 DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2690 #endif /* CONFIG_DEBUG_NET */
2691 }
2692
2693 /*
2694 * Add data to an sk_buff
2695 */
2696 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2697 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2698 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2699 {
2700 void *tmp = skb_tail_pointer(skb);
2701 SKB_LINEAR_ASSERT(skb);
2702 skb->tail += len;
2703 skb->len += len;
2704 return tmp;
2705 }
2706
__skb_put_zero(struct sk_buff * skb,unsigned int len)2707 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2708 {
2709 void *tmp = __skb_put(skb, len);
2710
2711 memset(tmp, 0, len);
2712 return tmp;
2713 }
2714
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2715 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2716 unsigned int len)
2717 {
2718 void *tmp = __skb_put(skb, len);
2719
2720 memcpy(tmp, data, len);
2721 return tmp;
2722 }
2723
__skb_put_u8(struct sk_buff * skb,u8 val)2724 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2725 {
2726 *(u8 *)__skb_put(skb, 1) = val;
2727 }
2728
skb_put_zero(struct sk_buff * skb,unsigned int len)2729 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2730 {
2731 void *tmp = skb_put(skb, len);
2732
2733 memset(tmp, 0, len);
2734
2735 return tmp;
2736 }
2737
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2738 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2739 unsigned int len)
2740 {
2741 void *tmp = skb_put(skb, len);
2742
2743 memcpy(tmp, data, len);
2744
2745 return tmp;
2746 }
2747
skb_put_u8(struct sk_buff * skb,u8 val)2748 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2749 {
2750 *(u8 *)skb_put(skb, 1) = val;
2751 }
2752
2753 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2754 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2755 {
2756 DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
2757
2758 skb->data -= len;
2759 skb->len += len;
2760 return skb->data;
2761 }
2762
2763 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2764 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2765 {
2766 DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
2767
2768 skb->len -= len;
2769 if (unlikely(skb->len < skb->data_len)) {
2770 #if defined(CONFIG_DEBUG_NET)
2771 skb->len += len;
2772 pr_err("__skb_pull(len=%u)\n", len);
2773 skb_dump(KERN_ERR, skb, false);
2774 #endif
2775 BUG();
2776 }
2777 return skb->data += len;
2778 }
2779
skb_pull_inline(struct sk_buff * skb,unsigned int len)2780 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2781 {
2782 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2783 }
2784
2785 void *skb_pull_data(struct sk_buff *skb, size_t len);
2786
2787 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2788
2789 static inline enum skb_drop_reason
pskb_may_pull_reason(struct sk_buff * skb,unsigned int len)2790 pskb_may_pull_reason(struct sk_buff *skb, unsigned int len)
2791 {
2792 DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
2793
2794 if (likely(len <= skb_headlen(skb)))
2795 return SKB_NOT_DROPPED_YET;
2796
2797 if (unlikely(len > skb->len))
2798 return SKB_DROP_REASON_PKT_TOO_SMALL;
2799
2800 if (unlikely(!__pskb_pull_tail(skb, len - skb_headlen(skb))))
2801 return SKB_DROP_REASON_NOMEM;
2802
2803 return SKB_NOT_DROPPED_YET;
2804 }
2805
pskb_may_pull(struct sk_buff * skb,unsigned int len)2806 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2807 {
2808 return pskb_may_pull_reason(skb, len) == SKB_NOT_DROPPED_YET;
2809 }
2810
pskb_pull(struct sk_buff * skb,unsigned int len)2811 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2812 {
2813 if (!pskb_may_pull(skb, len))
2814 return NULL;
2815
2816 skb->len -= len;
2817 return skb->data += len;
2818 }
2819
2820 void skb_condense(struct sk_buff *skb);
2821
2822 /**
2823 * skb_headroom - bytes at buffer head
2824 * @skb: buffer to check
2825 *
2826 * Return the number of bytes of free space at the head of an &sk_buff.
2827 */
skb_headroom(const struct sk_buff * skb)2828 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2829 {
2830 return skb->data - skb->head;
2831 }
2832
2833 /**
2834 * skb_tailroom - bytes at buffer end
2835 * @skb: buffer to check
2836 *
2837 * Return the number of bytes of free space at the tail of an sk_buff
2838 */
skb_tailroom(const struct sk_buff * skb)2839 static inline int skb_tailroom(const struct sk_buff *skb)
2840 {
2841 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2842 }
2843
2844 /**
2845 * skb_availroom - bytes at buffer end
2846 * @skb: buffer to check
2847 *
2848 * Return the number of bytes of free space at the tail of an sk_buff
2849 * allocated by sk_stream_alloc()
2850 */
skb_availroom(const struct sk_buff * skb)2851 static inline int skb_availroom(const struct sk_buff *skb)
2852 {
2853 if (skb_is_nonlinear(skb))
2854 return 0;
2855
2856 return skb->end - skb->tail - skb->reserved_tailroom;
2857 }
2858
2859 /**
2860 * skb_reserve - adjust headroom
2861 * @skb: buffer to alter
2862 * @len: bytes to move
2863 *
2864 * Increase the headroom of an empty &sk_buff by reducing the tail
2865 * room. This is only allowed for an empty buffer.
2866 */
skb_reserve(struct sk_buff * skb,int len)2867 static inline void skb_reserve(struct sk_buff *skb, int len)
2868 {
2869 skb->data += len;
2870 skb->tail += len;
2871 }
2872
2873 /**
2874 * skb_tailroom_reserve - adjust reserved_tailroom
2875 * @skb: buffer to alter
2876 * @mtu: maximum amount of headlen permitted
2877 * @needed_tailroom: minimum amount of reserved_tailroom
2878 *
2879 * Set reserved_tailroom so that headlen can be as large as possible but
2880 * not larger than mtu and tailroom cannot be smaller than
2881 * needed_tailroom.
2882 * The required headroom should already have been reserved before using
2883 * this function.
2884 */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2885 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2886 unsigned int needed_tailroom)
2887 {
2888 SKB_LINEAR_ASSERT(skb);
2889 if (mtu < skb_tailroom(skb) - needed_tailroom)
2890 /* use at most mtu */
2891 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2892 else
2893 /* use up to all available space */
2894 skb->reserved_tailroom = needed_tailroom;
2895 }
2896
2897 #define ENCAP_TYPE_ETHER 0
2898 #define ENCAP_TYPE_IPPROTO 1
2899
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2900 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2901 __be16 protocol)
2902 {
2903 skb->inner_protocol = protocol;
2904 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2905 }
2906
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2907 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2908 __u8 ipproto)
2909 {
2910 skb->inner_ipproto = ipproto;
2911 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2912 }
2913
skb_reset_inner_headers(struct sk_buff * skb)2914 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2915 {
2916 skb->inner_mac_header = skb->mac_header;
2917 skb->inner_network_header = skb->network_header;
2918 skb->inner_transport_header = skb->transport_header;
2919 }
2920
skb_reset_mac_len(struct sk_buff * skb)2921 static inline void skb_reset_mac_len(struct sk_buff *skb)
2922 {
2923 skb->mac_len = skb->network_header - skb->mac_header;
2924 }
2925
skb_inner_transport_header(const struct sk_buff * skb)2926 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2927 *skb)
2928 {
2929 return skb->head + skb->inner_transport_header;
2930 }
2931
skb_inner_transport_offset(const struct sk_buff * skb)2932 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2933 {
2934 return skb_inner_transport_header(skb) - skb->data;
2935 }
2936
skb_reset_inner_transport_header(struct sk_buff * skb)2937 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2938 {
2939 skb->inner_transport_header = skb->data - skb->head;
2940 }
2941
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2942 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2943 const int offset)
2944 {
2945 skb_reset_inner_transport_header(skb);
2946 skb->inner_transport_header += offset;
2947 }
2948
skb_inner_network_header(const struct sk_buff * skb)2949 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2950 {
2951 return skb->head + skb->inner_network_header;
2952 }
2953
skb_reset_inner_network_header(struct sk_buff * skb)2954 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2955 {
2956 skb->inner_network_header = skb->data - skb->head;
2957 }
2958
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2959 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2960 const int offset)
2961 {
2962 skb_reset_inner_network_header(skb);
2963 skb->inner_network_header += offset;
2964 }
2965
skb_inner_network_header_was_set(const struct sk_buff * skb)2966 static inline bool skb_inner_network_header_was_set(const struct sk_buff *skb)
2967 {
2968 return skb->inner_network_header > 0;
2969 }
2970
skb_inner_mac_header(const struct sk_buff * skb)2971 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2972 {
2973 return skb->head + skb->inner_mac_header;
2974 }
2975
skb_reset_inner_mac_header(struct sk_buff * skb)2976 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2977 {
2978 skb->inner_mac_header = skb->data - skb->head;
2979 }
2980
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)2981 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2982 const int offset)
2983 {
2984 skb_reset_inner_mac_header(skb);
2985 skb->inner_mac_header += offset;
2986 }
skb_transport_header_was_set(const struct sk_buff * skb)2987 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2988 {
2989 return skb->transport_header != (typeof(skb->transport_header))~0U;
2990 }
2991
skb_transport_header(const struct sk_buff * skb)2992 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2993 {
2994 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
2995 return skb->head + skb->transport_header;
2996 }
2997
skb_reset_transport_header(struct sk_buff * skb)2998 static inline void skb_reset_transport_header(struct sk_buff *skb)
2999 {
3000 skb->transport_header = skb->data - skb->head;
3001 }
3002
3003 /**
3004 * skb_reset_transport_header_careful - conditionally reset transport header
3005 * @skb: buffer to alter
3006 *
3007 * Hardened version of skb_reset_transport_header().
3008 *
3009 * Returns: true if the operation was a success.
3010 */
3011 static inline bool __must_check
skb_reset_transport_header_careful(struct sk_buff * skb)3012 skb_reset_transport_header_careful(struct sk_buff *skb)
3013 {
3014 long offset = skb->data - skb->head;
3015
3016 if (unlikely(offset != (typeof(skb->transport_header))offset))
3017 return false;
3018
3019 if (unlikely(offset == (typeof(skb->transport_header))~0U))
3020 return false;
3021
3022 skb->transport_header = offset;
3023 return true;
3024 }
3025
skb_set_transport_header(struct sk_buff * skb,const int offset)3026 static inline void skb_set_transport_header(struct sk_buff *skb,
3027 const int offset)
3028 {
3029 skb_reset_transport_header(skb);
3030 skb->transport_header += offset;
3031 }
3032
skb_network_header(const struct sk_buff * skb)3033 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
3034 {
3035 return skb->head + skb->network_header;
3036 }
3037
skb_reset_network_header(struct sk_buff * skb)3038 static inline void skb_reset_network_header(struct sk_buff *skb)
3039 {
3040 skb->network_header = skb->data - skb->head;
3041 }
3042
skb_set_network_header(struct sk_buff * skb,const int offset)3043 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
3044 {
3045 skb_reset_network_header(skb);
3046 skb->network_header += offset;
3047 }
3048
skb_mac_header_was_set(const struct sk_buff * skb)3049 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
3050 {
3051 return skb->mac_header != (typeof(skb->mac_header))~0U;
3052 }
3053
skb_mac_header(const struct sk_buff * skb)3054 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
3055 {
3056 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
3057 return skb->head + skb->mac_header;
3058 }
3059
skb_mac_offset(const struct sk_buff * skb)3060 static inline int skb_mac_offset(const struct sk_buff *skb)
3061 {
3062 return skb_mac_header(skb) - skb->data;
3063 }
3064
skb_mac_header_len(const struct sk_buff * skb)3065 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
3066 {
3067 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
3068 return skb->network_header - skb->mac_header;
3069 }
3070
skb_unset_mac_header(struct sk_buff * skb)3071 static inline void skb_unset_mac_header(struct sk_buff *skb)
3072 {
3073 skb->mac_header = (typeof(skb->mac_header))~0U;
3074 }
3075
skb_reset_mac_header(struct sk_buff * skb)3076 static inline void skb_reset_mac_header(struct sk_buff *skb)
3077 {
3078 skb->mac_header = skb->data - skb->head;
3079 }
3080
skb_set_mac_header(struct sk_buff * skb,const int offset)3081 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
3082 {
3083 skb_reset_mac_header(skb);
3084 skb->mac_header += offset;
3085 }
3086
skb_pop_mac_header(struct sk_buff * skb)3087 static inline void skb_pop_mac_header(struct sk_buff *skb)
3088 {
3089 skb->mac_header = skb->network_header;
3090 }
3091
skb_probe_transport_header(struct sk_buff * skb)3092 static inline void skb_probe_transport_header(struct sk_buff *skb)
3093 {
3094 struct flow_keys_basic keys;
3095
3096 if (skb_transport_header_was_set(skb))
3097 return;
3098
3099 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
3100 NULL, 0, 0, 0, 0))
3101 skb_set_transport_header(skb, keys.control.thoff);
3102 }
3103
skb_mac_header_rebuild(struct sk_buff * skb)3104 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
3105 {
3106 if (skb_mac_header_was_set(skb)) {
3107 const unsigned char *old_mac = skb_mac_header(skb);
3108
3109 skb_set_mac_header(skb, -skb->mac_len);
3110 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
3111 }
3112 }
3113
3114 /* Move the full mac header up to current network_header.
3115 * Leaves skb->data pointing at offset skb->mac_len into the mac_header.
3116 * Must be provided the complete mac header length.
3117 */
skb_mac_header_rebuild_full(struct sk_buff * skb,u32 full_mac_len)3118 static inline void skb_mac_header_rebuild_full(struct sk_buff *skb, u32 full_mac_len)
3119 {
3120 if (skb_mac_header_was_set(skb)) {
3121 const unsigned char *old_mac = skb_mac_header(skb);
3122
3123 skb_set_mac_header(skb, -full_mac_len);
3124 memmove(skb_mac_header(skb), old_mac, full_mac_len);
3125 __skb_push(skb, full_mac_len - skb->mac_len);
3126 }
3127 }
3128
skb_checksum_start_offset(const struct sk_buff * skb)3129 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
3130 {
3131 return skb->csum_start - skb_headroom(skb);
3132 }
3133
skb_checksum_start(const struct sk_buff * skb)3134 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
3135 {
3136 return skb->head + skb->csum_start;
3137 }
3138
skb_transport_offset(const struct sk_buff * skb)3139 static inline int skb_transport_offset(const struct sk_buff *skb)
3140 {
3141 return skb_transport_header(skb) - skb->data;
3142 }
3143
skb_network_header_len(const struct sk_buff * skb)3144 static inline u32 skb_network_header_len(const struct sk_buff *skb)
3145 {
3146 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
3147 return skb->transport_header - skb->network_header;
3148 }
3149
skb_inner_network_header_len(const struct sk_buff * skb)3150 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
3151 {
3152 return skb->inner_transport_header - skb->inner_network_header;
3153 }
3154
skb_network_offset(const struct sk_buff * skb)3155 static inline int skb_network_offset(const struct sk_buff *skb)
3156 {
3157 return skb_network_header(skb) - skb->data;
3158 }
3159
skb_inner_network_offset(const struct sk_buff * skb)3160 static inline int skb_inner_network_offset(const struct sk_buff *skb)
3161 {
3162 return skb_inner_network_header(skb) - skb->data;
3163 }
3164
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)3165 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
3166 {
3167 return pskb_may_pull(skb, skb_network_offset(skb) + len);
3168 }
3169
3170 /*
3171 * CPUs often take a performance hit when accessing unaligned memory
3172 * locations. The actual performance hit varies, it can be small if the
3173 * hardware handles it or large if we have to take an exception and fix it
3174 * in software.
3175 *
3176 * Since an ethernet header is 14 bytes network drivers often end up with
3177 * the IP header at an unaligned offset. The IP header can be aligned by
3178 * shifting the start of the packet by 2 bytes. Drivers should do this
3179 * with:
3180 *
3181 * skb_reserve(skb, NET_IP_ALIGN);
3182 *
3183 * The downside to this alignment of the IP header is that the DMA is now
3184 * unaligned. On some architectures the cost of an unaligned DMA is high
3185 * and this cost outweighs the gains made by aligning the IP header.
3186 *
3187 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
3188 * to be overridden.
3189 */
3190 #ifndef NET_IP_ALIGN
3191 #define NET_IP_ALIGN 2
3192 #endif
3193
3194 /*
3195 * The networking layer reserves some headroom in skb data (via
3196 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
3197 * the header has to grow. In the default case, if the header has to grow
3198 * 32 bytes or less we avoid the reallocation.
3199 *
3200 * Unfortunately this headroom changes the DMA alignment of the resulting
3201 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
3202 * on some architectures. An architecture can override this value,
3203 * perhaps setting it to a cacheline in size (since that will maintain
3204 * cacheline alignment of the DMA). It must be a power of 2.
3205 *
3206 * Various parts of the networking layer expect at least 32 bytes of
3207 * headroom, you should not reduce this.
3208 *
3209 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
3210 * to reduce average number of cache lines per packet.
3211 * get_rps_cpu() for example only access one 64 bytes aligned block :
3212 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
3213 */
3214 #ifndef NET_SKB_PAD
3215 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
3216 #endif
3217
3218 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
3219
__skb_set_length(struct sk_buff * skb,unsigned int len)3220 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
3221 {
3222 if (WARN_ON(skb_is_nonlinear(skb)))
3223 return;
3224 skb->len = len;
3225 skb_set_tail_pointer(skb, len);
3226 }
3227
__skb_trim(struct sk_buff * skb,unsigned int len)3228 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
3229 {
3230 __skb_set_length(skb, len);
3231 }
3232
3233 void skb_trim(struct sk_buff *skb, unsigned int len);
3234
__pskb_trim(struct sk_buff * skb,unsigned int len)3235 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
3236 {
3237 if (skb->data_len)
3238 return ___pskb_trim(skb, len);
3239 __skb_trim(skb, len);
3240 return 0;
3241 }
3242
pskb_trim(struct sk_buff * skb,unsigned int len)3243 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
3244 {
3245 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
3246 }
3247
3248 /**
3249 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
3250 * @skb: buffer to alter
3251 * @len: new length
3252 *
3253 * This is identical to pskb_trim except that the caller knows that
3254 * the skb is not cloned so we should never get an error due to out-
3255 * of-memory.
3256 */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)3257 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
3258 {
3259 int err = pskb_trim(skb, len);
3260 BUG_ON(err);
3261 }
3262
__skb_grow(struct sk_buff * skb,unsigned int len)3263 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
3264 {
3265 unsigned int diff = len - skb->len;
3266
3267 if (skb_tailroom(skb) < diff) {
3268 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
3269 GFP_ATOMIC);
3270 if (ret)
3271 return ret;
3272 }
3273 __skb_set_length(skb, len);
3274 return 0;
3275 }
3276
3277 /**
3278 * skb_orphan - orphan a buffer
3279 * @skb: buffer to orphan
3280 *
3281 * If a buffer currently has an owner then we call the owner's
3282 * destructor function and make the @skb unowned. The buffer continues
3283 * to exist but is no longer charged to its former owner.
3284 */
skb_orphan(struct sk_buff * skb)3285 static inline void skb_orphan(struct sk_buff *skb)
3286 {
3287 if (skb->destructor) {
3288 skb->destructor(skb);
3289 skb->destructor = NULL;
3290 skb->sk = NULL;
3291 } else {
3292 BUG_ON(skb->sk);
3293 }
3294 }
3295
3296 /**
3297 * skb_orphan_frags - orphan the frags contained in a buffer
3298 * @skb: buffer to orphan frags from
3299 * @gfp_mask: allocation mask for replacement pages
3300 *
3301 * For each frag in the SKB which needs a destructor (i.e. has an
3302 * owner) create a copy of that frag and release the original
3303 * page by calling the destructor.
3304 */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)3305 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
3306 {
3307 if (likely(!skb_zcopy(skb)))
3308 return 0;
3309 if (skb_shinfo(skb)->flags & SKBFL_DONT_ORPHAN)
3310 return 0;
3311 return skb_copy_ubufs(skb, gfp_mask);
3312 }
3313
3314 /* 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)3315 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
3316 {
3317 if (likely(!skb_zcopy(skb)))
3318 return 0;
3319 return skb_copy_ubufs(skb, gfp_mask);
3320 }
3321
3322 /**
3323 * __skb_queue_purge_reason - empty a list
3324 * @list: list to empty
3325 * @reason: drop reason
3326 *
3327 * Delete all buffers on an &sk_buff list. Each buffer is removed from
3328 * the list and one reference dropped. This function does not take the
3329 * list lock and the caller must hold the relevant locks to use it.
3330 */
__skb_queue_purge_reason(struct sk_buff_head * list,enum skb_drop_reason reason)3331 static inline void __skb_queue_purge_reason(struct sk_buff_head *list,
3332 enum skb_drop_reason reason)
3333 {
3334 struct sk_buff *skb;
3335
3336 while ((skb = __skb_dequeue(list)) != NULL)
3337 kfree_skb_reason(skb, reason);
3338 }
3339
__skb_queue_purge(struct sk_buff_head * list)3340 static inline void __skb_queue_purge(struct sk_buff_head *list)
3341 {
3342 __skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE);
3343 }
3344
3345 void skb_queue_purge_reason(struct sk_buff_head *list,
3346 enum skb_drop_reason reason);
3347
skb_queue_purge(struct sk_buff_head * list)3348 static inline void skb_queue_purge(struct sk_buff_head *list)
3349 {
3350 skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE);
3351 }
3352
3353 unsigned int skb_rbtree_purge(struct rb_root *root);
3354 void skb_errqueue_purge(struct sk_buff_head *list);
3355
3356 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3357
3358 /**
3359 * netdev_alloc_frag - allocate a page fragment
3360 * @fragsz: fragment size
3361 *
3362 * Allocates a frag from a page for receive buffer.
3363 * Uses GFP_ATOMIC allocations.
3364 */
netdev_alloc_frag(unsigned int fragsz)3365 static inline void *netdev_alloc_frag(unsigned int fragsz)
3366 {
3367 return __netdev_alloc_frag_align(fragsz, ~0u);
3368 }
3369
netdev_alloc_frag_align(unsigned int fragsz,unsigned int align)3370 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
3371 unsigned int align)
3372 {
3373 WARN_ON_ONCE(!is_power_of_2(align));
3374 return __netdev_alloc_frag_align(fragsz, -align);
3375 }
3376
3377 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
3378 gfp_t gfp_mask);
3379
3380 /**
3381 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
3382 * @dev: network device to receive on
3383 * @length: length to allocate
3384 *
3385 * Allocate a new &sk_buff and assign it a usage count of one. The
3386 * buffer has unspecified headroom built in. Users should allocate
3387 * the headroom they think they need without accounting for the
3388 * built in space. The built in space is used for optimisations.
3389 *
3390 * %NULL is returned if there is no free memory. Although this function
3391 * allocates memory it can be called from an interrupt.
3392 */
netdev_alloc_skb(struct net_device * dev,unsigned int length)3393 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
3394 unsigned int length)
3395 {
3396 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
3397 }
3398
3399 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)3400 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
3401 gfp_t gfp_mask)
3402 {
3403 return __netdev_alloc_skb(NULL, length, gfp_mask);
3404 }
3405
3406 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)3407 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
3408 {
3409 return netdev_alloc_skb(NULL, length);
3410 }
3411
3412
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)3413 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
3414 unsigned int length, gfp_t gfp)
3415 {
3416 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
3417
3418 if (NET_IP_ALIGN && skb)
3419 skb_reserve(skb, NET_IP_ALIGN);
3420 return skb;
3421 }
3422
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)3423 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
3424 unsigned int length)
3425 {
3426 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
3427 }
3428
skb_free_frag(void * addr)3429 static inline void skb_free_frag(void *addr)
3430 {
3431 page_frag_free(addr);
3432 }
3433
3434 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3435
napi_alloc_frag(unsigned int fragsz)3436 static inline void *napi_alloc_frag(unsigned int fragsz)
3437 {
3438 return __napi_alloc_frag_align(fragsz, ~0u);
3439 }
3440
napi_alloc_frag_align(unsigned int fragsz,unsigned int align)3441 static inline void *napi_alloc_frag_align(unsigned int fragsz,
3442 unsigned int align)
3443 {
3444 WARN_ON_ONCE(!is_power_of_2(align));
3445 return __napi_alloc_frag_align(fragsz, -align);
3446 }
3447
3448 struct sk_buff *napi_alloc_skb(struct napi_struct *napi, unsigned int length);
3449 void napi_consume_skb(struct sk_buff *skb, int budget);
3450
3451 void napi_skb_free_stolen_head(struct sk_buff *skb);
3452 void __napi_kfree_skb(struct sk_buff *skb, enum skb_drop_reason reason);
3453
3454 /**
3455 * __dev_alloc_pages - allocate page for network Rx
3456 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3457 * @order: size of the allocation
3458 *
3459 * Allocate a new page.
3460 *
3461 * %NULL is returned if there is no free memory.
3462 */
__dev_alloc_pages_noprof(gfp_t gfp_mask,unsigned int order)3463 static inline struct page *__dev_alloc_pages_noprof(gfp_t gfp_mask,
3464 unsigned int order)
3465 {
3466 /* This piece of code contains several assumptions.
3467 * 1. This is for device Rx, therefore a cold page is preferred.
3468 * 2. The expectation is the user wants a compound page.
3469 * 3. If requesting a order 0 page it will not be compound
3470 * due to the check to see if order has a value in prep_new_page
3471 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3472 * code in gfp_to_alloc_flags that should be enforcing this.
3473 */
3474 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3475
3476 return alloc_pages_node_noprof(NUMA_NO_NODE, gfp_mask, order);
3477 }
3478 #define __dev_alloc_pages(...) alloc_hooks(__dev_alloc_pages_noprof(__VA_ARGS__))
3479
3480 /*
3481 * This specialized allocator has to be a macro for its allocations to be
3482 * accounted separately (to have a separate alloc_tag).
3483 */
3484 #define dev_alloc_pages(_order) __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, _order)
3485
3486 /**
3487 * __dev_alloc_page - allocate a page for network Rx
3488 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3489 *
3490 * Allocate a new page.
3491 *
3492 * %NULL is returned if there is no free memory.
3493 */
__dev_alloc_page_noprof(gfp_t gfp_mask)3494 static inline struct page *__dev_alloc_page_noprof(gfp_t gfp_mask)
3495 {
3496 return __dev_alloc_pages_noprof(gfp_mask, 0);
3497 }
3498 #define __dev_alloc_page(...) alloc_hooks(__dev_alloc_page_noprof(__VA_ARGS__))
3499
3500 /*
3501 * This specialized allocator has to be a macro for its allocations to be
3502 * accounted separately (to have a separate alloc_tag).
3503 */
3504 #define dev_alloc_page() dev_alloc_pages(0)
3505
3506 /**
3507 * dev_page_is_reusable - check whether a page can be reused for network Rx
3508 * @page: the page to test
3509 *
3510 * A page shouldn't be considered for reusing/recycling if it was allocated
3511 * under memory pressure or at a distant memory node.
3512 *
3513 * Returns false if this page should be returned to page allocator, true
3514 * otherwise.
3515 */
dev_page_is_reusable(const struct page * page)3516 static inline bool dev_page_is_reusable(const struct page *page)
3517 {
3518 return likely(page_to_nid(page) == numa_mem_id() &&
3519 !page_is_pfmemalloc(page));
3520 }
3521
3522 /**
3523 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3524 * @page: The page that was allocated from skb_alloc_page
3525 * @skb: The skb that may need pfmemalloc set
3526 */
skb_propagate_pfmemalloc(const struct page * page,struct sk_buff * skb)3527 static inline void skb_propagate_pfmemalloc(const struct page *page,
3528 struct sk_buff *skb)
3529 {
3530 if (page_is_pfmemalloc(page))
3531 skb->pfmemalloc = true;
3532 }
3533
3534 /**
3535 * skb_frag_off() - Returns the offset of a skb fragment
3536 * @frag: the paged fragment
3537 */
skb_frag_off(const skb_frag_t * frag)3538 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3539 {
3540 return frag->offset;
3541 }
3542
3543 /**
3544 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3545 * @frag: skb fragment
3546 * @delta: value to add
3547 */
skb_frag_off_add(skb_frag_t * frag,int delta)3548 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3549 {
3550 frag->offset += delta;
3551 }
3552
3553 /**
3554 * skb_frag_off_set() - Sets the offset of a skb fragment
3555 * @frag: skb fragment
3556 * @offset: offset of fragment
3557 */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)3558 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3559 {
3560 frag->offset = offset;
3561 }
3562
3563 /**
3564 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3565 * @fragto: skb fragment where offset is set
3566 * @fragfrom: skb fragment offset is copied from
3567 */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3568 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3569 const skb_frag_t *fragfrom)
3570 {
3571 fragto->offset = fragfrom->offset;
3572 }
3573
3574 /* Return: true if the skb_frag contains a net_iov. */
skb_frag_is_net_iov(const skb_frag_t * frag)3575 static inline bool skb_frag_is_net_iov(const skb_frag_t *frag)
3576 {
3577 return netmem_is_net_iov(frag->netmem);
3578 }
3579
3580 /**
3581 * skb_frag_net_iov - retrieve the net_iov referred to by fragment
3582 * @frag: the fragment
3583 *
3584 * Return: the &struct net_iov associated with @frag. Returns NULL if this
3585 * frag has no associated net_iov.
3586 */
skb_frag_net_iov(const skb_frag_t * frag)3587 static inline struct net_iov *skb_frag_net_iov(const skb_frag_t *frag)
3588 {
3589 if (!skb_frag_is_net_iov(frag))
3590 return NULL;
3591
3592 return netmem_to_net_iov(frag->netmem);
3593 }
3594
3595 /**
3596 * skb_frag_page - retrieve the page referred to by a paged fragment
3597 * @frag: the paged fragment
3598 *
3599 * Return: the &struct page associated with @frag. Returns NULL if this frag
3600 * has no associated page.
3601 */
skb_frag_page(const skb_frag_t * frag)3602 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3603 {
3604 if (skb_frag_is_net_iov(frag))
3605 return NULL;
3606
3607 return netmem_to_page(frag->netmem);
3608 }
3609
3610 /**
3611 * skb_frag_netmem - retrieve the netmem referred to by a fragment
3612 * @frag: the fragment
3613 *
3614 * Return: the &netmem_ref associated with @frag.
3615 */
skb_frag_netmem(const skb_frag_t * frag)3616 static inline netmem_ref skb_frag_netmem(const skb_frag_t *frag)
3617 {
3618 return frag->netmem;
3619 }
3620
3621 int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb,
3622 unsigned int headroom);
3623 int skb_cow_data_for_xdp(struct page_pool *pool, struct sk_buff **pskb,
3624 struct bpf_prog *prog);
3625
3626 /**
3627 * skb_frag_address - gets the address of the data contained in a paged fragment
3628 * @frag: the paged fragment buffer
3629 *
3630 * Returns the address of the data within @frag. The page must already
3631 * be mapped.
3632 */
skb_frag_address(const skb_frag_t * frag)3633 static inline void *skb_frag_address(const skb_frag_t *frag)
3634 {
3635 if (!skb_frag_page(frag))
3636 return NULL;
3637
3638 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3639 }
3640
3641 /**
3642 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3643 * @frag: the paged fragment buffer
3644 *
3645 * Returns the address of the data within @frag. Checks that the page
3646 * is mapped and returns %NULL otherwise.
3647 */
skb_frag_address_safe(const skb_frag_t * frag)3648 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3649 {
3650 struct page *page = skb_frag_page(frag);
3651 void *ptr;
3652
3653 if (!page)
3654 return NULL;
3655
3656 ptr = page_address(page);
3657 if (unlikely(!ptr))
3658 return NULL;
3659
3660 return ptr + skb_frag_off(frag);
3661 }
3662
3663 /**
3664 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3665 * @fragto: skb fragment where page is set
3666 * @fragfrom: skb fragment page is copied from
3667 */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3668 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3669 const skb_frag_t *fragfrom)
3670 {
3671 fragto->netmem = fragfrom->netmem;
3672 }
3673
3674 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3675
3676 /**
3677 * skb_frag_dma_map - maps a paged fragment via the DMA API
3678 * @dev: the device to map the fragment to
3679 * @frag: the paged fragment to map
3680 * @offset: the offset within the fragment (starting at the
3681 * fragment's own offset)
3682 * @size: the number of bytes to map
3683 * @dir: the direction of the mapping (``PCI_DMA_*``)
3684 *
3685 * Maps the page associated with @frag to @device.
3686 */
skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3687 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3688 const skb_frag_t *frag,
3689 size_t offset, size_t size,
3690 enum dma_data_direction dir)
3691 {
3692 return dma_map_page(dev, skb_frag_page(frag),
3693 skb_frag_off(frag) + offset, size, dir);
3694 }
3695
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3696 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3697 gfp_t gfp_mask)
3698 {
3699 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3700 }
3701
3702
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3703 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3704 gfp_t gfp_mask)
3705 {
3706 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3707 }
3708
3709
3710 /**
3711 * skb_clone_writable - is the header of a clone writable
3712 * @skb: buffer to check
3713 * @len: length up to which to write
3714 *
3715 * Returns true if modifying the header part of the cloned buffer
3716 * does not requires the data to be copied.
3717 */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3718 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3719 {
3720 return !skb_header_cloned(skb) &&
3721 skb_headroom(skb) + len <= skb->hdr_len;
3722 }
3723
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3724 static inline int skb_try_make_writable(struct sk_buff *skb,
3725 unsigned int write_len)
3726 {
3727 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3728 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3729 }
3730
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3731 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3732 int cloned)
3733 {
3734 int delta = 0;
3735
3736 if (headroom > skb_headroom(skb))
3737 delta = headroom - skb_headroom(skb);
3738
3739 if (delta || cloned)
3740 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3741 GFP_ATOMIC);
3742 return 0;
3743 }
3744
3745 /**
3746 * skb_cow - copy header of skb when it is required
3747 * @skb: buffer to cow
3748 * @headroom: needed headroom
3749 *
3750 * If the skb passed lacks sufficient headroom or its data part
3751 * is shared, data is reallocated. If reallocation fails, an error
3752 * is returned and original skb is not changed.
3753 *
3754 * The result is skb with writable area skb->head...skb->tail
3755 * and at least @headroom of space at head.
3756 */
skb_cow(struct sk_buff * skb,unsigned int headroom)3757 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3758 {
3759 return __skb_cow(skb, headroom, skb_cloned(skb));
3760 }
3761
3762 /**
3763 * skb_cow_head - skb_cow but only making the head writable
3764 * @skb: buffer to cow
3765 * @headroom: needed headroom
3766 *
3767 * This function is identical to skb_cow except that we replace the
3768 * skb_cloned check by skb_header_cloned. It should be used when
3769 * you only need to push on some header and do not need to modify
3770 * the data.
3771 */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3772 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3773 {
3774 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3775 }
3776
3777 /**
3778 * skb_padto - pad an skbuff up to a minimal size
3779 * @skb: buffer to pad
3780 * @len: minimal length
3781 *
3782 * Pads up a buffer to ensure the trailing bytes exist and are
3783 * blanked. If the buffer already contains sufficient data it
3784 * is untouched. Otherwise it is extended. Returns zero on
3785 * success. The skb is freed on error.
3786 */
skb_padto(struct sk_buff * skb,unsigned int len)3787 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3788 {
3789 unsigned int size = skb->len;
3790 if (likely(size >= len))
3791 return 0;
3792 return skb_pad(skb, len - size);
3793 }
3794
3795 /**
3796 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3797 * @skb: buffer to pad
3798 * @len: minimal length
3799 * @free_on_error: free buffer on error
3800 *
3801 * Pads up a buffer to ensure the trailing bytes exist and are
3802 * blanked. If the buffer already contains sufficient data it
3803 * is untouched. Otherwise it is extended. Returns zero on
3804 * success. The skb is freed on error if @free_on_error is true.
3805 */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3806 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3807 unsigned int len,
3808 bool free_on_error)
3809 {
3810 unsigned int size = skb->len;
3811
3812 if (unlikely(size < len)) {
3813 len -= size;
3814 if (__skb_pad(skb, len, free_on_error))
3815 return -ENOMEM;
3816 __skb_put(skb, len);
3817 }
3818 return 0;
3819 }
3820
3821 /**
3822 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3823 * @skb: buffer to pad
3824 * @len: minimal length
3825 *
3826 * Pads up a buffer to ensure the trailing bytes exist and are
3827 * blanked. If the buffer already contains sufficient data it
3828 * is untouched. Otherwise it is extended. Returns zero on
3829 * success. The skb is freed on error.
3830 */
skb_put_padto(struct sk_buff * skb,unsigned int len)3831 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3832 {
3833 return __skb_put_padto(skb, len, true);
3834 }
3835
3836 bool csum_and_copy_from_iter_full(void *addr, size_t bytes, __wsum *csum, struct iov_iter *i)
3837 __must_check;
3838
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3839 static inline int skb_add_data(struct sk_buff *skb,
3840 struct iov_iter *from, int copy)
3841 {
3842 const int off = skb->len;
3843
3844 if (skb->ip_summed == CHECKSUM_NONE) {
3845 __wsum csum = 0;
3846 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3847 &csum, from)) {
3848 skb->csum = csum_block_add(skb->csum, csum, off);
3849 return 0;
3850 }
3851 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3852 return 0;
3853
3854 __skb_trim(skb, off);
3855 return -EFAULT;
3856 }
3857
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3858 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3859 const struct page *page, int off)
3860 {
3861 if (skb_zcopy(skb))
3862 return false;
3863 if (i) {
3864 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3865
3866 return page == skb_frag_page(frag) &&
3867 off == skb_frag_off(frag) + skb_frag_size(frag);
3868 }
3869 return false;
3870 }
3871
__skb_linearize(struct sk_buff * skb)3872 static inline int __skb_linearize(struct sk_buff *skb)
3873 {
3874 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3875 }
3876
3877 /**
3878 * skb_linearize - convert paged skb to linear one
3879 * @skb: buffer to linarize
3880 *
3881 * If there is no free memory -ENOMEM is returned, otherwise zero
3882 * is returned and the old skb data released.
3883 */
skb_linearize(struct sk_buff * skb)3884 static inline int skb_linearize(struct sk_buff *skb)
3885 {
3886 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3887 }
3888
3889 /**
3890 * skb_has_shared_frag - can any frag be overwritten
3891 * @skb: buffer to test
3892 *
3893 * Return true if the skb has at least one frag that might be modified
3894 * by an external entity (as in vmsplice()/sendfile())
3895 */
skb_has_shared_frag(const struct sk_buff * skb)3896 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3897 {
3898 return skb_is_nonlinear(skb) &&
3899 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3900 }
3901
3902 /**
3903 * skb_linearize_cow - make sure skb is linear and writable
3904 * @skb: buffer to process
3905 *
3906 * If there is no free memory -ENOMEM is returned, otherwise zero
3907 * is returned and the old skb data released.
3908 */
skb_linearize_cow(struct sk_buff * skb)3909 static inline int skb_linearize_cow(struct sk_buff *skb)
3910 {
3911 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3912 __skb_linearize(skb) : 0;
3913 }
3914
3915 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3916 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3917 unsigned int off)
3918 {
3919 if (skb->ip_summed == CHECKSUM_COMPLETE)
3920 skb->csum = csum_block_sub(skb->csum,
3921 csum_partial(start, len, 0), off);
3922 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3923 skb_checksum_start_offset(skb) < 0)
3924 skb->ip_summed = CHECKSUM_NONE;
3925 }
3926
3927 /**
3928 * skb_postpull_rcsum - update checksum for received skb after pull
3929 * @skb: buffer to update
3930 * @start: start of data before pull
3931 * @len: length of data pulled
3932 *
3933 * After doing a pull on a received packet, you need to call this to
3934 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3935 * CHECKSUM_NONE so that it can be recomputed from scratch.
3936 */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3937 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3938 const void *start, unsigned int len)
3939 {
3940 if (skb->ip_summed == CHECKSUM_COMPLETE)
3941 skb->csum = wsum_negate(csum_partial(start, len,
3942 wsum_negate(skb->csum)));
3943 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3944 skb_checksum_start_offset(skb) < 0)
3945 skb->ip_summed = CHECKSUM_NONE;
3946 }
3947
3948 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3949 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3950 unsigned int off)
3951 {
3952 if (skb->ip_summed == CHECKSUM_COMPLETE)
3953 skb->csum = csum_block_add(skb->csum,
3954 csum_partial(start, len, 0), off);
3955 }
3956
3957 /**
3958 * skb_postpush_rcsum - update checksum for received skb after push
3959 * @skb: buffer to update
3960 * @start: start of data after push
3961 * @len: length of data pushed
3962 *
3963 * After doing a push on a received packet, you need to call this to
3964 * update the CHECKSUM_COMPLETE checksum.
3965 */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3966 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3967 const void *start, unsigned int len)
3968 {
3969 __skb_postpush_rcsum(skb, start, len, 0);
3970 }
3971
3972 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3973
3974 /**
3975 * skb_push_rcsum - push skb and update receive checksum
3976 * @skb: buffer to update
3977 * @len: length of data pulled
3978 *
3979 * This function performs an skb_push on the packet and updates
3980 * the CHECKSUM_COMPLETE checksum. It should be used on
3981 * receive path processing instead of skb_push unless you know
3982 * that the checksum difference is zero (e.g., a valid IP header)
3983 * or you are setting ip_summed to CHECKSUM_NONE.
3984 */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3985 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3986 {
3987 skb_push(skb, len);
3988 skb_postpush_rcsum(skb, skb->data, len);
3989 return skb->data;
3990 }
3991
3992 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3993 /**
3994 * pskb_trim_rcsum - trim received skb and update checksum
3995 * @skb: buffer to trim
3996 * @len: new length
3997 *
3998 * This is exactly the same as pskb_trim except that it ensures the
3999 * checksum of received packets are still valid after the operation.
4000 * It can change skb pointers.
4001 */
4002
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)4003 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
4004 {
4005 if (likely(len >= skb->len))
4006 return 0;
4007 return pskb_trim_rcsum_slow(skb, len);
4008 }
4009
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)4010 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
4011 {
4012 if (skb->ip_summed == CHECKSUM_COMPLETE)
4013 skb->ip_summed = CHECKSUM_NONE;
4014 __skb_trim(skb, len);
4015 return 0;
4016 }
4017
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)4018 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
4019 {
4020 if (skb->ip_summed == CHECKSUM_COMPLETE)
4021 skb->ip_summed = CHECKSUM_NONE;
4022 return __skb_grow(skb, len);
4023 }
4024
4025 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
4026 #define skb_rb_first(root) rb_to_skb(rb_first(root))
4027 #define skb_rb_last(root) rb_to_skb(rb_last(root))
4028 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
4029 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
4030
4031 #define skb_queue_walk(queue, skb) \
4032 for (skb = (queue)->next; \
4033 skb != (struct sk_buff *)(queue); \
4034 skb = skb->next)
4035
4036 #define skb_queue_walk_safe(queue, skb, tmp) \
4037 for (skb = (queue)->next, tmp = skb->next; \
4038 skb != (struct sk_buff *)(queue); \
4039 skb = tmp, tmp = skb->next)
4040
4041 #define skb_queue_walk_from(queue, skb) \
4042 for (; skb != (struct sk_buff *)(queue); \
4043 skb = skb->next)
4044
4045 #define skb_rbtree_walk(skb, root) \
4046 for (skb = skb_rb_first(root); skb != NULL; \
4047 skb = skb_rb_next(skb))
4048
4049 #define skb_rbtree_walk_from(skb) \
4050 for (; skb != NULL; \
4051 skb = skb_rb_next(skb))
4052
4053 #define skb_rbtree_walk_from_safe(skb, tmp) \
4054 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
4055 skb = tmp)
4056
4057 #define skb_queue_walk_from_safe(queue, skb, tmp) \
4058 for (tmp = skb->next; \
4059 skb != (struct sk_buff *)(queue); \
4060 skb = tmp, tmp = skb->next)
4061
4062 #define skb_queue_reverse_walk(queue, skb) \
4063 for (skb = (queue)->prev; \
4064 skb != (struct sk_buff *)(queue); \
4065 skb = skb->prev)
4066
4067 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
4068 for (skb = (queue)->prev, tmp = skb->prev; \
4069 skb != (struct sk_buff *)(queue); \
4070 skb = tmp, tmp = skb->prev)
4071
4072 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
4073 for (tmp = skb->prev; \
4074 skb != (struct sk_buff *)(queue); \
4075 skb = tmp, tmp = skb->prev)
4076
skb_has_frag_list(const struct sk_buff * skb)4077 static inline bool skb_has_frag_list(const struct sk_buff *skb)
4078 {
4079 return skb_shinfo(skb)->frag_list != NULL;
4080 }
4081
skb_frag_list_init(struct sk_buff * skb)4082 static inline void skb_frag_list_init(struct sk_buff *skb)
4083 {
4084 skb_shinfo(skb)->frag_list = NULL;
4085 }
4086
4087 #define skb_walk_frags(skb, iter) \
4088 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
4089
4090
4091 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
4092 int *err, long *timeo_p,
4093 const struct sk_buff *skb);
4094 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
4095 struct sk_buff_head *queue,
4096 unsigned int flags,
4097 int *off, int *err,
4098 struct sk_buff **last);
4099 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
4100 struct sk_buff_head *queue,
4101 unsigned int flags, int *off, int *err,
4102 struct sk_buff **last);
4103 struct sk_buff *__skb_recv_datagram(struct sock *sk,
4104 struct sk_buff_head *sk_queue,
4105 unsigned int flags, int *off, int *err);
4106 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err);
4107 __poll_t datagram_poll(struct file *file, struct socket *sock,
4108 struct poll_table_struct *wait);
4109 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
4110 struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)4111 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
4112 struct msghdr *msg, int size)
4113 {
4114 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
4115 }
4116 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
4117 struct msghdr *msg);
4118 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
4119 struct iov_iter *to, int len,
4120 struct ahash_request *hash);
4121 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
4122 struct iov_iter *from, int len);
4123 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
4124 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
4125 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
4126 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
4127 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
4128 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
4129 int len);
4130 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
4131 struct pipe_inode_info *pipe, unsigned int len,
4132 unsigned int flags);
4133 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
4134 int len);
4135 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
4136 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
4137 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
4138 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
4139 int len, int hlen);
4140 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
4141 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
4142 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
4143 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
4144 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
4145 unsigned int offset);
4146 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
4147 int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len);
4148 int skb_ensure_writable_head_tail(struct sk_buff *skb, struct net_device *dev);
4149 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
4150 int skb_vlan_pop(struct sk_buff *skb);
4151 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
4152 int skb_eth_pop(struct sk_buff *skb);
4153 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
4154 const unsigned char *src);
4155 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
4156 int mac_len, bool ethernet);
4157 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
4158 bool ethernet);
4159 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
4160 int skb_mpls_dec_ttl(struct sk_buff *skb);
4161 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
4162 gfp_t gfp);
4163
memcpy_from_msg(void * data,struct msghdr * msg,int len)4164 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
4165 {
4166 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
4167 }
4168
memcpy_to_msg(struct msghdr * msg,void * data,int len)4169 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
4170 {
4171 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
4172 }
4173
4174 struct skb_checksum_ops {
4175 __wsum (*update)(const void *mem, int len, __wsum wsum);
4176 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
4177 };
4178
4179 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
4180
4181 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
4182 __wsum csum, const struct skb_checksum_ops *ops);
4183 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
4184 __wsum csum);
4185
4186 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,const void * data,int hlen,void * buffer)4187 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
4188 const void *data, int hlen, void *buffer)
4189 {
4190 if (likely(hlen - offset >= len))
4191 return (void *)data + offset;
4192
4193 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
4194 return NULL;
4195
4196 return buffer;
4197 }
4198
4199 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)4200 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
4201 {
4202 return __skb_header_pointer(skb, offset, len, skb->data,
4203 skb_headlen(skb), buffer);
4204 }
4205
4206 static inline void * __must_check
skb_pointer_if_linear(const struct sk_buff * skb,int offset,int len)4207 skb_pointer_if_linear(const struct sk_buff *skb, int offset, int len)
4208 {
4209 if (likely(skb_headlen(skb) - offset >= len))
4210 return skb->data + offset;
4211 return NULL;
4212 }
4213
4214 /**
4215 * skb_needs_linearize - check if we need to linearize a given skb
4216 * depending on the given device features.
4217 * @skb: socket buffer to check
4218 * @features: net device features
4219 *
4220 * Returns true if either:
4221 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
4222 * 2. skb is fragmented and the device does not support SG.
4223 */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)4224 static inline bool skb_needs_linearize(struct sk_buff *skb,
4225 netdev_features_t features)
4226 {
4227 return skb_is_nonlinear(skb) &&
4228 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
4229 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
4230 }
4231
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)4232 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
4233 void *to,
4234 const unsigned int len)
4235 {
4236 memcpy(to, skb->data, len);
4237 }
4238
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)4239 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
4240 const int offset, void *to,
4241 const unsigned int len)
4242 {
4243 memcpy(to, skb->data + offset, len);
4244 }
4245
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)4246 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
4247 const void *from,
4248 const unsigned int len)
4249 {
4250 memcpy(skb->data, from, len);
4251 }
4252
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)4253 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
4254 const int offset,
4255 const void *from,
4256 const unsigned int len)
4257 {
4258 memcpy(skb->data + offset, from, len);
4259 }
4260
4261 void skb_init(void);
4262
skb_get_ktime(const struct sk_buff * skb)4263 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
4264 {
4265 return skb->tstamp;
4266 }
4267
4268 /**
4269 * skb_get_timestamp - get timestamp from a skb
4270 * @skb: skb to get stamp from
4271 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
4272 *
4273 * Timestamps are stored in the skb as offsets to a base timestamp.
4274 * This function converts the offset back to a struct timeval and stores
4275 * it in stamp.
4276 */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)4277 static inline void skb_get_timestamp(const struct sk_buff *skb,
4278 struct __kernel_old_timeval *stamp)
4279 {
4280 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
4281 }
4282
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)4283 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
4284 struct __kernel_sock_timeval *stamp)
4285 {
4286 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4287
4288 stamp->tv_sec = ts.tv_sec;
4289 stamp->tv_usec = ts.tv_nsec / 1000;
4290 }
4291
skb_get_timestampns(const struct sk_buff * skb,struct __kernel_old_timespec * stamp)4292 static inline void skb_get_timestampns(const struct sk_buff *skb,
4293 struct __kernel_old_timespec *stamp)
4294 {
4295 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4296
4297 stamp->tv_sec = ts.tv_sec;
4298 stamp->tv_nsec = ts.tv_nsec;
4299 }
4300
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)4301 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
4302 struct __kernel_timespec *stamp)
4303 {
4304 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4305
4306 stamp->tv_sec = ts.tv_sec;
4307 stamp->tv_nsec = ts.tv_nsec;
4308 }
4309
__net_timestamp(struct sk_buff * skb)4310 static inline void __net_timestamp(struct sk_buff *skb)
4311 {
4312 skb->tstamp = ktime_get_real();
4313 skb->tstamp_type = SKB_CLOCK_REALTIME;
4314 }
4315
net_timedelta(ktime_t t)4316 static inline ktime_t net_timedelta(ktime_t t)
4317 {
4318 return ktime_sub(ktime_get_real(), t);
4319 }
4320
skb_set_delivery_time(struct sk_buff * skb,ktime_t kt,u8 tstamp_type)4321 static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt,
4322 u8 tstamp_type)
4323 {
4324 skb->tstamp = kt;
4325
4326 if (kt)
4327 skb->tstamp_type = tstamp_type;
4328 else
4329 skb->tstamp_type = SKB_CLOCK_REALTIME;
4330 }
4331
skb_set_delivery_type_by_clockid(struct sk_buff * skb,ktime_t kt,clockid_t clockid)4332 static inline void skb_set_delivery_type_by_clockid(struct sk_buff *skb,
4333 ktime_t kt, clockid_t clockid)
4334 {
4335 u8 tstamp_type = SKB_CLOCK_REALTIME;
4336
4337 switch (clockid) {
4338 case CLOCK_REALTIME:
4339 break;
4340 case CLOCK_MONOTONIC:
4341 tstamp_type = SKB_CLOCK_MONOTONIC;
4342 break;
4343 case CLOCK_TAI:
4344 tstamp_type = SKB_CLOCK_TAI;
4345 break;
4346 default:
4347 WARN_ON_ONCE(1);
4348 kt = 0;
4349 }
4350
4351 skb_set_delivery_time(skb, kt, tstamp_type);
4352 }
4353
4354 DECLARE_STATIC_KEY_FALSE(netstamp_needed_key);
4355
4356 /* It is used in the ingress path to clear the delivery_time.
4357 * If needed, set the skb->tstamp to the (rcv) timestamp.
4358 */
skb_clear_delivery_time(struct sk_buff * skb)4359 static inline void skb_clear_delivery_time(struct sk_buff *skb)
4360 {
4361 if (skb->tstamp_type) {
4362 skb->tstamp_type = SKB_CLOCK_REALTIME;
4363 if (static_branch_unlikely(&netstamp_needed_key))
4364 skb->tstamp = ktime_get_real();
4365 else
4366 skb->tstamp = 0;
4367 }
4368 }
4369
skb_clear_tstamp(struct sk_buff * skb)4370 static inline void skb_clear_tstamp(struct sk_buff *skb)
4371 {
4372 if (skb->tstamp_type)
4373 return;
4374
4375 skb->tstamp = 0;
4376 }
4377
skb_tstamp(const struct sk_buff * skb)4378 static inline ktime_t skb_tstamp(const struct sk_buff *skb)
4379 {
4380 if (skb->tstamp_type)
4381 return 0;
4382
4383 return skb->tstamp;
4384 }
4385
skb_tstamp_cond(const struct sk_buff * skb,bool cond)4386 static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond)
4387 {
4388 if (skb->tstamp_type != SKB_CLOCK_MONOTONIC && skb->tstamp)
4389 return skb->tstamp;
4390
4391 if (static_branch_unlikely(&netstamp_needed_key) || cond)
4392 return ktime_get_real();
4393
4394 return 0;
4395 }
4396
skb_metadata_len(const struct sk_buff * skb)4397 static inline u8 skb_metadata_len(const struct sk_buff *skb)
4398 {
4399 return skb_shinfo(skb)->meta_len;
4400 }
4401
skb_metadata_end(const struct sk_buff * skb)4402 static inline void *skb_metadata_end(const struct sk_buff *skb)
4403 {
4404 return skb_mac_header(skb);
4405 }
4406
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)4407 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
4408 const struct sk_buff *skb_b,
4409 u8 meta_len)
4410 {
4411 const void *a = skb_metadata_end(skb_a);
4412 const void *b = skb_metadata_end(skb_b);
4413 u64 diffs = 0;
4414
4415 if (!IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) ||
4416 BITS_PER_LONG != 64)
4417 goto slow;
4418
4419 /* Using more efficient variant than plain call to memcmp(). */
4420 switch (meta_len) {
4421 #define __it(x, op) (x -= sizeof(u##op))
4422 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
4423 case 32: diffs |= __it_diff(a, b, 64);
4424 fallthrough;
4425 case 24: diffs |= __it_diff(a, b, 64);
4426 fallthrough;
4427 case 16: diffs |= __it_diff(a, b, 64);
4428 fallthrough;
4429 case 8: diffs |= __it_diff(a, b, 64);
4430 break;
4431 case 28: diffs |= __it_diff(a, b, 64);
4432 fallthrough;
4433 case 20: diffs |= __it_diff(a, b, 64);
4434 fallthrough;
4435 case 12: diffs |= __it_diff(a, b, 64);
4436 fallthrough;
4437 case 4: diffs |= __it_diff(a, b, 32);
4438 break;
4439 default:
4440 slow:
4441 return memcmp(a - meta_len, b - meta_len, meta_len);
4442 }
4443 return diffs;
4444 }
4445
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)4446 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
4447 const struct sk_buff *skb_b)
4448 {
4449 u8 len_a = skb_metadata_len(skb_a);
4450 u8 len_b = skb_metadata_len(skb_b);
4451
4452 if (!(len_a | len_b))
4453 return false;
4454
4455 return len_a != len_b ?
4456 true : __skb_metadata_differs(skb_a, skb_b, len_a);
4457 }
4458
skb_metadata_set(struct sk_buff * skb,u8 meta_len)4459 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
4460 {
4461 skb_shinfo(skb)->meta_len = meta_len;
4462 }
4463
skb_metadata_clear(struct sk_buff * skb)4464 static inline void skb_metadata_clear(struct sk_buff *skb)
4465 {
4466 skb_metadata_set(skb, 0);
4467 }
4468
4469 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
4470
4471 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
4472
4473 void skb_clone_tx_timestamp(struct sk_buff *skb);
4474 bool skb_defer_rx_timestamp(struct sk_buff *skb);
4475
4476 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
4477
skb_clone_tx_timestamp(struct sk_buff * skb)4478 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
4479 {
4480 }
4481
skb_defer_rx_timestamp(struct sk_buff * skb)4482 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
4483 {
4484 return false;
4485 }
4486
4487 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
4488
4489 /**
4490 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
4491 *
4492 * PHY drivers may accept clones of transmitted packets for
4493 * timestamping via their phy_driver.txtstamp method. These drivers
4494 * must call this function to return the skb back to the stack with a
4495 * timestamp.
4496 *
4497 * @skb: clone of the original outgoing packet
4498 * @hwtstamps: hardware time stamps
4499 *
4500 */
4501 void skb_complete_tx_timestamp(struct sk_buff *skb,
4502 struct skb_shared_hwtstamps *hwtstamps);
4503
4504 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4505 struct skb_shared_hwtstamps *hwtstamps,
4506 struct sock *sk, int tstype);
4507
4508 /**
4509 * skb_tstamp_tx - queue clone of skb with send time stamps
4510 * @orig_skb: the original outgoing packet
4511 * @hwtstamps: hardware time stamps, may be NULL if not available
4512 *
4513 * If the skb has a socket associated, then this function clones the
4514 * skb (thus sharing the actual data and optional structures), stores
4515 * the optional hardware time stamping information (if non NULL) or
4516 * generates a software time stamp (otherwise), then queues the clone
4517 * to the error queue of the socket. Errors are silently ignored.
4518 */
4519 void skb_tstamp_tx(struct sk_buff *orig_skb,
4520 struct skb_shared_hwtstamps *hwtstamps);
4521
4522 /**
4523 * skb_tx_timestamp() - Driver hook for transmit timestamping
4524 *
4525 * Ethernet MAC Drivers should call this function in their hard_xmit()
4526 * function immediately before giving the sk_buff to the MAC hardware.
4527 *
4528 * Specifically, one should make absolutely sure that this function is
4529 * called before TX completion of this packet can trigger. Otherwise
4530 * the packet could potentially already be freed.
4531 *
4532 * @skb: A socket buffer.
4533 */
skb_tx_timestamp(struct sk_buff * skb)4534 static inline void skb_tx_timestamp(struct sk_buff *skb)
4535 {
4536 skb_clone_tx_timestamp(skb);
4537 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4538 skb_tstamp_tx(skb, NULL);
4539 }
4540
4541 /**
4542 * skb_complete_wifi_ack - deliver skb with wifi status
4543 *
4544 * @skb: the original outgoing packet
4545 * @acked: ack status
4546 *
4547 */
4548 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4549
4550 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4551 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4552
skb_csum_unnecessary(const struct sk_buff * skb)4553 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4554 {
4555 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4556 skb->csum_valid ||
4557 (skb->ip_summed == CHECKSUM_PARTIAL &&
4558 skb_checksum_start_offset(skb) >= 0));
4559 }
4560
4561 /**
4562 * skb_checksum_complete - Calculate checksum of an entire packet
4563 * @skb: packet to process
4564 *
4565 * This function calculates the checksum over the entire packet plus
4566 * the value of skb->csum. The latter can be used to supply the
4567 * checksum of a pseudo header as used by TCP/UDP. It returns the
4568 * checksum.
4569 *
4570 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4571 * this function can be used to verify that checksum on received
4572 * packets. In that case the function should return zero if the
4573 * checksum is correct. In particular, this function will return zero
4574 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4575 * hardware has already verified the correctness of the checksum.
4576 */
skb_checksum_complete(struct sk_buff * skb)4577 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4578 {
4579 return skb_csum_unnecessary(skb) ?
4580 0 : __skb_checksum_complete(skb);
4581 }
4582
__skb_decr_checksum_unnecessary(struct sk_buff * skb)4583 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4584 {
4585 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4586 if (skb->csum_level == 0)
4587 skb->ip_summed = CHECKSUM_NONE;
4588 else
4589 skb->csum_level--;
4590 }
4591 }
4592
__skb_incr_checksum_unnecessary(struct sk_buff * skb)4593 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4594 {
4595 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4596 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4597 skb->csum_level++;
4598 } else if (skb->ip_summed == CHECKSUM_NONE) {
4599 skb->ip_summed = CHECKSUM_UNNECESSARY;
4600 skb->csum_level = 0;
4601 }
4602 }
4603
__skb_reset_checksum_unnecessary(struct sk_buff * skb)4604 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4605 {
4606 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4607 skb->ip_summed = CHECKSUM_NONE;
4608 skb->csum_level = 0;
4609 }
4610 }
4611
4612 /* Check if we need to perform checksum complete validation.
4613 *
4614 * Returns true if checksum complete is needed, false otherwise
4615 * (either checksum is unnecessary or zero checksum is allowed).
4616 */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)4617 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4618 bool zero_okay,
4619 __sum16 check)
4620 {
4621 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4622 skb->csum_valid = 1;
4623 __skb_decr_checksum_unnecessary(skb);
4624 return false;
4625 }
4626
4627 return true;
4628 }
4629
4630 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4631 * in checksum_init.
4632 */
4633 #define CHECKSUM_BREAK 76
4634
4635 /* Unset checksum-complete
4636 *
4637 * Unset checksum complete can be done when packet is being modified
4638 * (uncompressed for instance) and checksum-complete value is
4639 * invalidated.
4640 */
skb_checksum_complete_unset(struct sk_buff * skb)4641 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4642 {
4643 if (skb->ip_summed == CHECKSUM_COMPLETE)
4644 skb->ip_summed = CHECKSUM_NONE;
4645 }
4646
4647 /* Validate (init) checksum based on checksum complete.
4648 *
4649 * Return values:
4650 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4651 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4652 * checksum is stored in skb->csum for use in __skb_checksum_complete
4653 * non-zero: value of invalid checksum
4654 *
4655 */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)4656 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4657 bool complete,
4658 __wsum psum)
4659 {
4660 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4661 if (!csum_fold(csum_add(psum, skb->csum))) {
4662 skb->csum_valid = 1;
4663 return 0;
4664 }
4665 }
4666
4667 skb->csum = psum;
4668
4669 if (complete || skb->len <= CHECKSUM_BREAK) {
4670 __sum16 csum;
4671
4672 csum = __skb_checksum_complete(skb);
4673 skb->csum_valid = !csum;
4674 return csum;
4675 }
4676
4677 return 0;
4678 }
4679
null_compute_pseudo(struct sk_buff * skb,int proto)4680 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4681 {
4682 return 0;
4683 }
4684
4685 /* Perform checksum validate (init). Note that this is a macro since we only
4686 * want to calculate the pseudo header which is an input function if necessary.
4687 * First we try to validate without any computation (checksum unnecessary) and
4688 * then calculate based on checksum complete calling the function to compute
4689 * pseudo header.
4690 *
4691 * Return values:
4692 * 0: checksum is validated or try to in skb_checksum_complete
4693 * non-zero: value of invalid checksum
4694 */
4695 #define __skb_checksum_validate(skb, proto, complete, \
4696 zero_okay, check, compute_pseudo) \
4697 ({ \
4698 __sum16 __ret = 0; \
4699 skb->csum_valid = 0; \
4700 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4701 __ret = __skb_checksum_validate_complete(skb, \
4702 complete, compute_pseudo(skb, proto)); \
4703 __ret; \
4704 })
4705
4706 #define skb_checksum_init(skb, proto, compute_pseudo) \
4707 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4708
4709 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4710 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4711
4712 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4713 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4714
4715 #define skb_checksum_validate_zero_check(skb, proto, check, \
4716 compute_pseudo) \
4717 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4718
4719 #define skb_checksum_simple_validate(skb) \
4720 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4721
__skb_checksum_convert_check(struct sk_buff * skb)4722 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4723 {
4724 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4725 }
4726
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4727 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4728 {
4729 skb->csum = ~pseudo;
4730 skb->ip_summed = CHECKSUM_COMPLETE;
4731 }
4732
4733 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4734 do { \
4735 if (__skb_checksum_convert_check(skb)) \
4736 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4737 } while (0)
4738
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4739 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4740 u16 start, u16 offset)
4741 {
4742 skb->ip_summed = CHECKSUM_PARTIAL;
4743 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4744 skb->csum_offset = offset - start;
4745 }
4746
4747 /* Update skbuf and packet to reflect the remote checksum offload operation.
4748 * When called, ptr indicates the starting point for skb->csum when
4749 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4750 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4751 */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4752 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4753 int start, int offset, bool nopartial)
4754 {
4755 __wsum delta;
4756
4757 if (!nopartial) {
4758 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4759 return;
4760 }
4761
4762 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4763 __skb_checksum_complete(skb);
4764 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4765 }
4766
4767 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4768
4769 /* Adjust skb->csum since we changed the packet */
4770 skb->csum = csum_add(skb->csum, delta);
4771 }
4772
skb_nfct(const struct sk_buff * skb)4773 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4774 {
4775 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4776 return (void *)(skb->_nfct & NFCT_PTRMASK);
4777 #else
4778 return NULL;
4779 #endif
4780 }
4781
skb_get_nfct(const struct sk_buff * skb)4782 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4783 {
4784 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4785 return skb->_nfct;
4786 #else
4787 return 0UL;
4788 #endif
4789 }
4790
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4791 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4792 {
4793 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4794 skb->slow_gro |= !!nfct;
4795 skb->_nfct = nfct;
4796 #endif
4797 }
4798
4799 #ifdef CONFIG_SKB_EXTENSIONS
4800 enum skb_ext_id {
4801 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4802 SKB_EXT_BRIDGE_NF,
4803 #endif
4804 #ifdef CONFIG_XFRM
4805 SKB_EXT_SEC_PATH,
4806 #endif
4807 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4808 TC_SKB_EXT,
4809 #endif
4810 #if IS_ENABLED(CONFIG_MPTCP)
4811 SKB_EXT_MPTCP,
4812 #endif
4813 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
4814 SKB_EXT_MCTP,
4815 #endif
4816 SKB_EXT_NUM, /* must be last */
4817 };
4818
4819 /**
4820 * struct skb_ext - sk_buff extensions
4821 * @refcnt: 1 on allocation, deallocated on 0
4822 * @offset: offset to add to @data to obtain extension address
4823 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4824 * @data: start of extension data, variable sized
4825 *
4826 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4827 * to use 'u8' types while allowing up to 2kb worth of extension data.
4828 */
4829 struct skb_ext {
4830 refcount_t refcnt;
4831 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4832 u8 chunks; /* same */
4833 char data[] __aligned(8);
4834 };
4835
4836 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4837 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4838 struct skb_ext *ext);
4839 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4840 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4841 void __skb_ext_put(struct skb_ext *ext);
4842
skb_ext_put(struct sk_buff * skb)4843 static inline void skb_ext_put(struct sk_buff *skb)
4844 {
4845 if (skb->active_extensions)
4846 __skb_ext_put(skb->extensions);
4847 }
4848
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4849 static inline void __skb_ext_copy(struct sk_buff *dst,
4850 const struct sk_buff *src)
4851 {
4852 dst->active_extensions = src->active_extensions;
4853
4854 if (src->active_extensions) {
4855 struct skb_ext *ext = src->extensions;
4856
4857 refcount_inc(&ext->refcnt);
4858 dst->extensions = ext;
4859 }
4860 }
4861
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4862 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4863 {
4864 skb_ext_put(dst);
4865 __skb_ext_copy(dst, src);
4866 }
4867
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4868 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4869 {
4870 return !!ext->offset[i];
4871 }
4872
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4873 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4874 {
4875 return skb->active_extensions & (1 << id);
4876 }
4877
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4878 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4879 {
4880 if (skb_ext_exist(skb, id))
4881 __skb_ext_del(skb, id);
4882 }
4883
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4884 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4885 {
4886 if (skb_ext_exist(skb, id)) {
4887 struct skb_ext *ext = skb->extensions;
4888
4889 return (void *)ext + (ext->offset[id] << 3);
4890 }
4891
4892 return NULL;
4893 }
4894
skb_ext_reset(struct sk_buff * skb)4895 static inline void skb_ext_reset(struct sk_buff *skb)
4896 {
4897 if (unlikely(skb->active_extensions)) {
4898 __skb_ext_put(skb->extensions);
4899 skb->active_extensions = 0;
4900 }
4901 }
4902
skb_has_extensions(struct sk_buff * skb)4903 static inline bool skb_has_extensions(struct sk_buff *skb)
4904 {
4905 return unlikely(skb->active_extensions);
4906 }
4907 #else
skb_ext_put(struct sk_buff * skb)4908 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4909 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4910 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4911 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)4912 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4913 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4914 #endif /* CONFIG_SKB_EXTENSIONS */
4915
nf_reset_ct(struct sk_buff * skb)4916 static inline void nf_reset_ct(struct sk_buff *skb)
4917 {
4918 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4919 nf_conntrack_put(skb_nfct(skb));
4920 skb->_nfct = 0;
4921 #endif
4922 }
4923
nf_reset_trace(struct sk_buff * skb)4924 static inline void nf_reset_trace(struct sk_buff *skb)
4925 {
4926 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4927 skb->nf_trace = 0;
4928 #endif
4929 }
4930
ipvs_reset(struct sk_buff * skb)4931 static inline void ipvs_reset(struct sk_buff *skb)
4932 {
4933 #if IS_ENABLED(CONFIG_IP_VS)
4934 skb->ipvs_property = 0;
4935 #endif
4936 }
4937
4938 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4939 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4940 bool copy)
4941 {
4942 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4943 dst->_nfct = src->_nfct;
4944 nf_conntrack_get(skb_nfct(src));
4945 #endif
4946 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4947 if (copy)
4948 dst->nf_trace = src->nf_trace;
4949 #endif
4950 }
4951
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4952 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4953 {
4954 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4955 nf_conntrack_put(skb_nfct(dst));
4956 #endif
4957 dst->slow_gro = src->slow_gro;
4958 __nf_copy(dst, src, true);
4959 }
4960
4961 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4962 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4963 {
4964 to->secmark = from->secmark;
4965 }
4966
skb_init_secmark(struct sk_buff * skb)4967 static inline void skb_init_secmark(struct sk_buff *skb)
4968 {
4969 skb->secmark = 0;
4970 }
4971 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4972 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4973 { }
4974
skb_init_secmark(struct sk_buff * skb)4975 static inline void skb_init_secmark(struct sk_buff *skb)
4976 { }
4977 #endif
4978
secpath_exists(const struct sk_buff * skb)4979 static inline int secpath_exists(const struct sk_buff *skb)
4980 {
4981 #ifdef CONFIG_XFRM
4982 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4983 #else
4984 return 0;
4985 #endif
4986 }
4987
skb_irq_freeable(const struct sk_buff * skb)4988 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4989 {
4990 return !skb->destructor &&
4991 !secpath_exists(skb) &&
4992 !skb_nfct(skb) &&
4993 !skb->_skb_refdst &&
4994 !skb_has_frag_list(skb);
4995 }
4996
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)4997 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4998 {
4999 skb->queue_mapping = queue_mapping;
5000 }
5001
skb_get_queue_mapping(const struct sk_buff * skb)5002 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
5003 {
5004 return skb->queue_mapping;
5005 }
5006
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)5007 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
5008 {
5009 to->queue_mapping = from->queue_mapping;
5010 }
5011
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)5012 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
5013 {
5014 skb->queue_mapping = rx_queue + 1;
5015 }
5016
skb_get_rx_queue(const struct sk_buff * skb)5017 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
5018 {
5019 return skb->queue_mapping - 1;
5020 }
5021
skb_rx_queue_recorded(const struct sk_buff * skb)5022 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
5023 {
5024 return skb->queue_mapping != 0;
5025 }
5026
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)5027 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
5028 {
5029 skb->dst_pending_confirm = val;
5030 }
5031
skb_get_dst_pending_confirm(const struct sk_buff * skb)5032 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
5033 {
5034 return skb->dst_pending_confirm != 0;
5035 }
5036
skb_sec_path(const struct sk_buff * skb)5037 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
5038 {
5039 #ifdef CONFIG_XFRM
5040 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
5041 #else
5042 return NULL;
5043 #endif
5044 }
5045
skb_is_gso(const struct sk_buff * skb)5046 static inline bool skb_is_gso(const struct sk_buff *skb)
5047 {
5048 return skb_shinfo(skb)->gso_size;
5049 }
5050
5051 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)5052 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
5053 {
5054 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
5055 }
5056
5057 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)5058 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
5059 {
5060 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
5061 }
5062
5063 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)5064 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
5065 {
5066 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
5067 }
5068
skb_gso_reset(struct sk_buff * skb)5069 static inline void skb_gso_reset(struct sk_buff *skb)
5070 {
5071 skb_shinfo(skb)->gso_size = 0;
5072 skb_shinfo(skb)->gso_segs = 0;
5073 skb_shinfo(skb)->gso_type = 0;
5074 }
5075
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)5076 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
5077 u16 increment)
5078 {
5079 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
5080 return;
5081 shinfo->gso_size += increment;
5082 }
5083
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)5084 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
5085 u16 decrement)
5086 {
5087 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
5088 return;
5089 shinfo->gso_size -= decrement;
5090 }
5091
5092 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
5093
skb_warn_if_lro(const struct sk_buff * skb)5094 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
5095 {
5096 /* LRO sets gso_size but not gso_type, whereas if GSO is really
5097 * wanted then gso_type will be set. */
5098 const struct skb_shared_info *shinfo = skb_shinfo(skb);
5099
5100 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
5101 unlikely(shinfo->gso_type == 0)) {
5102 __skb_warn_lro_forwarding(skb);
5103 return true;
5104 }
5105 return false;
5106 }
5107
skb_forward_csum(struct sk_buff * skb)5108 static inline void skb_forward_csum(struct sk_buff *skb)
5109 {
5110 /* Unfortunately we don't support this one. Any brave souls? */
5111 if (skb->ip_summed == CHECKSUM_COMPLETE)
5112 skb->ip_summed = CHECKSUM_NONE;
5113 }
5114
5115 /**
5116 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
5117 * @skb: skb to check
5118 *
5119 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
5120 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
5121 * use this helper, to document places where we make this assertion.
5122 */
skb_checksum_none_assert(const struct sk_buff * skb)5123 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
5124 {
5125 DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE);
5126 }
5127
5128 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
5129
5130 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
5131 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
5132 unsigned int transport_len,
5133 __sum16(*skb_chkf)(struct sk_buff *skb));
5134
5135 /**
5136 * skb_head_is_locked - Determine if the skb->head is locked down
5137 * @skb: skb to check
5138 *
5139 * The head on skbs build around a head frag can be removed if they are
5140 * not cloned. This function returns true if the skb head is locked down
5141 * due to either being allocated via kmalloc, or by being a clone with
5142 * multiple references to the head.
5143 */
skb_head_is_locked(const struct sk_buff * skb)5144 static inline bool skb_head_is_locked(const struct sk_buff *skb)
5145 {
5146 return !skb->head_frag || skb_cloned(skb);
5147 }
5148
5149 /* Local Checksum Offload.
5150 * Compute outer checksum based on the assumption that the
5151 * inner checksum will be offloaded later.
5152 * See Documentation/networking/checksum-offloads.rst for
5153 * explanation of how this works.
5154 * Fill in outer checksum adjustment (e.g. with sum of outer
5155 * pseudo-header) before calling.
5156 * Also ensure that inner checksum is in linear data area.
5157 */
lco_csum(struct sk_buff * skb)5158 static inline __wsum lco_csum(struct sk_buff *skb)
5159 {
5160 unsigned char *csum_start = skb_checksum_start(skb);
5161 unsigned char *l4_hdr = skb_transport_header(skb);
5162 __wsum partial;
5163
5164 /* Start with complement of inner checksum adjustment */
5165 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
5166 skb->csum_offset));
5167
5168 /* Add in checksum of our headers (incl. outer checksum
5169 * adjustment filled in by caller) and return result.
5170 */
5171 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
5172 }
5173
skb_is_redirected(const struct sk_buff * skb)5174 static inline bool skb_is_redirected(const struct sk_buff *skb)
5175 {
5176 return skb->redirected;
5177 }
5178
skb_set_redirected(struct sk_buff * skb,bool from_ingress)5179 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
5180 {
5181 skb->redirected = 1;
5182 #ifdef CONFIG_NET_REDIRECT
5183 skb->from_ingress = from_ingress;
5184 if (skb->from_ingress)
5185 skb_clear_tstamp(skb);
5186 #endif
5187 }
5188
skb_reset_redirect(struct sk_buff * skb)5189 static inline void skb_reset_redirect(struct sk_buff *skb)
5190 {
5191 skb->redirected = 0;
5192 }
5193
skb_set_redirected_noclear(struct sk_buff * skb,bool from_ingress)5194 static inline void skb_set_redirected_noclear(struct sk_buff *skb,
5195 bool from_ingress)
5196 {
5197 skb->redirected = 1;
5198 #ifdef CONFIG_NET_REDIRECT
5199 skb->from_ingress = from_ingress;
5200 #endif
5201 }
5202
skb_csum_is_sctp(struct sk_buff * skb)5203 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
5204 {
5205 #if IS_ENABLED(CONFIG_IP_SCTP)
5206 return skb->csum_not_inet;
5207 #else
5208 return 0;
5209 #endif
5210 }
5211
skb_reset_csum_not_inet(struct sk_buff * skb)5212 static inline void skb_reset_csum_not_inet(struct sk_buff *skb)
5213 {
5214 skb->ip_summed = CHECKSUM_NONE;
5215 #if IS_ENABLED(CONFIG_IP_SCTP)
5216 skb->csum_not_inet = 0;
5217 #endif
5218 }
5219
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)5220 static inline void skb_set_kcov_handle(struct sk_buff *skb,
5221 const u64 kcov_handle)
5222 {
5223 #ifdef CONFIG_KCOV
5224 skb->kcov_handle = kcov_handle;
5225 #endif
5226 }
5227
skb_get_kcov_handle(struct sk_buff * skb)5228 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
5229 {
5230 #ifdef CONFIG_KCOV
5231 return skb->kcov_handle;
5232 #else
5233 return 0;
5234 #endif
5235 }
5236
skb_mark_for_recycle(struct sk_buff * skb)5237 static inline void skb_mark_for_recycle(struct sk_buff *skb)
5238 {
5239 #ifdef CONFIG_PAGE_POOL
5240 skb->pp_recycle = 1;
5241 #endif
5242 }
5243
5244 ssize_t skb_splice_from_iter(struct sk_buff *skb, struct iov_iter *iter,
5245 ssize_t maxsize, gfp_t gfp);
5246
5247 #endif /* __KERNEL__ */
5248 #endif /* _LINUX_SKBUFF_H */
5249