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, <[email protected]>
7 * Florian La Roche, <[email protected]>
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 <linux/page_frag_cache.h>
35 #include <net/flow.h>
36 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
37 #include <linux/netfilter/nf_conntrack_common.h>
38 #endif
39 #include <net/net_debug.h>
40 #include <net/dropreason-core.h>
41 #include <net/netmem.h>
42
43 /**
44 * DOC: skb checksums
45 *
46 * The interface for checksum offload between the stack and networking drivers
47 * is as follows...
48 *
49 * IP checksum related features
50 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
51 *
52 * Drivers advertise checksum offload capabilities in the features of a device.
53 * From the stack's point of view these are capabilities offered by the driver.
54 * A driver typically only advertises features that it is capable of offloading
55 * to its device.
56 *
57 * .. flat-table:: Checksum related device features
58 * :widths: 1 10
59 *
60 * * - %NETIF_F_HW_CSUM
61 * - The driver (or its device) is able to compute one
62 * IP (one's complement) checksum for any combination
63 * of protocols or protocol layering. The checksum is
64 * computed and set in a packet per the CHECKSUM_PARTIAL
65 * interface (see below).
66 *
67 * * - %NETIF_F_IP_CSUM
68 * - Driver (device) is only able to checksum plain
69 * TCP or UDP packets over IPv4. These are specifically
70 * unencapsulated packets of the form IPv4|TCP or
71 * IPv4|UDP where the Protocol field in the IPv4 header
72 * is TCP or UDP. The IPv4 header may contain IP options.
73 * This feature cannot be set in features for a device
74 * with NETIF_F_HW_CSUM also set. This feature is being
75 * DEPRECATED (see below).
76 *
77 * * - %NETIF_F_IPV6_CSUM
78 * - Driver (device) is only able to checksum plain
79 * TCP or UDP packets over IPv6. These are specifically
80 * unencapsulated packets of the form IPv6|TCP or
81 * IPv6|UDP where the Next Header field in the IPv6
82 * header is either TCP or UDP. IPv6 extension headers
83 * are not supported with this feature. This feature
84 * cannot be set in features for a device with
85 * NETIF_F_HW_CSUM also set. This feature is being
86 * DEPRECATED (see below).
87 *
88 * * - %NETIF_F_RXCSUM
89 * - Driver (device) performs receive checksum offload.
90 * This flag is only used to disable the RX checksum
91 * feature for a device. The stack will accept receive
92 * checksum indication in packets received on a device
93 * regardless of whether NETIF_F_RXCSUM is set.
94 *
95 * Checksumming of received packets by device
96 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
97 *
98 * Indication of checksum verification is set in &sk_buff.ip_summed.
99 * Possible values are:
100 *
101 * - %CHECKSUM_NONE
102 *
103 * Device did not checksum this packet e.g. due to lack of capabilities.
104 * The packet contains full (though not verified) checksum in packet but
105 * not in skb->csum. Thus, skb->csum is undefined in this case.
106 *
107 * - %CHECKSUM_UNNECESSARY
108 *
109 * The hardware you're dealing with doesn't calculate the full checksum
110 * (as in %CHECKSUM_COMPLETE), but it does parse headers and verify checksums
111 * for specific protocols. For such packets it will set %CHECKSUM_UNNECESSARY
112 * if their checksums are okay. &sk_buff.csum is still undefined in this case
113 * though. A driver or device must never modify the checksum field in the
114 * packet even if checksum is verified.
115 *
116 * %CHECKSUM_UNNECESSARY is applicable to following protocols:
117 *
118 * - TCP: IPv6 and IPv4.
119 * - UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
120 * zero UDP checksum for either IPv4 or IPv6, the networking stack
121 * may perform further validation in this case.
122 * - GRE: only if the checksum is present in the header.
123 * - SCTP: indicates the CRC in SCTP header has been validated.
124 * - FCOE: indicates the CRC in FC frame has been validated.
125 *
126 * &sk_buff.csum_level indicates the number of consecutive checksums found in
127 * the packet minus one that have been verified as %CHECKSUM_UNNECESSARY.
128 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
129 * and a device is able to verify the checksums for UDP (possibly zero),
130 * GRE (checksum flag is set) and TCP, &sk_buff.csum_level would be set to
131 * two. If the device were only able to verify the UDP checksum and not
132 * GRE, either because it doesn't support GRE checksum or because GRE
133 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
134 * not considered in this case).
135 *
136 * - %CHECKSUM_COMPLETE
137 *
138 * This is the most generic way. The device supplied checksum of the _whole_
139 * packet as seen by netif_rx() and fills in &sk_buff.csum. This means the
140 * hardware doesn't need to parse L3/L4 headers to implement this.
141 *
142 * Notes:
143 *
144 * - Even if device supports only some protocols, but is able to produce
145 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
146 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
147 *
148 * - %CHECKSUM_PARTIAL
149 *
150 * A checksum is set up to be offloaded to a device as described in the
151 * output description for CHECKSUM_PARTIAL. This may occur on a packet
152 * received directly from another Linux OS, e.g., a virtualized Linux kernel
153 * on the same host, or it may be set in the input path in GRO or remote
154 * checksum offload. For the purposes of checksum verification, the checksum
155 * referred to by skb->csum_start + skb->csum_offset and any preceding
156 * checksums in the packet are considered verified. Any checksums in the
157 * packet that are after the checksum being offloaded are not considered to
158 * be verified.
159 *
160 * Checksumming on transmit for non-GSO
161 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
162 *
163 * The stack requests checksum offload in the &sk_buff.ip_summed for a packet.
164 * Values are:
165 *
166 * - %CHECKSUM_PARTIAL
167 *
168 * The driver is required to checksum the packet as seen by hard_start_xmit()
169 * from &sk_buff.csum_start up to the end, and to record/write the checksum at
170 * offset &sk_buff.csum_start + &sk_buff.csum_offset.
171 * A driver may verify that the
172 * csum_start and csum_offset values are valid values given the length and
173 * offset of the packet, but it should not attempt to validate that the
174 * checksum refers to a legitimate transport layer checksum -- it is the
175 * purview of the stack to validate that csum_start and csum_offset are set
176 * correctly.
177 *
178 * When the stack requests checksum offload for a packet, the driver MUST
179 * ensure that the checksum is set correctly. A driver can either offload the
180 * checksum calculation to the device, or call skb_checksum_help (in the case
181 * that the device does not support offload for a particular checksum).
182 *
183 * %NETIF_F_IP_CSUM and %NETIF_F_IPV6_CSUM are being deprecated in favor of
184 * %NETIF_F_HW_CSUM. New devices should use %NETIF_F_HW_CSUM to indicate
185 * checksum offload capability.
186 * skb_csum_hwoffload_help() can be called to resolve %CHECKSUM_PARTIAL based
187 * on network device checksumming capabilities: if a packet does not match
188 * them, skb_checksum_help() or skb_crc32c_help() (depending on the value of
189 * &sk_buff.csum_not_inet, see :ref:`crc`)
190 * is called to resolve the checksum.
191 *
192 * - %CHECKSUM_NONE
193 *
194 * The skb was already checksummed by the protocol, or a checksum is not
195 * required.
196 *
197 * - %CHECKSUM_UNNECESSARY
198 *
199 * This has the same meaning as CHECKSUM_NONE for checksum offload on
200 * output.
201 *
202 * - %CHECKSUM_COMPLETE
203 *
204 * Not used in checksum output. If a driver observes a packet with this value
205 * set in skbuff, it should treat the packet as if %CHECKSUM_NONE were set.
206 *
207 * .. _crc:
208 *
209 * Non-IP checksum (CRC) offloads
210 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
211 *
212 * .. flat-table::
213 * :widths: 1 10
214 *
215 * * - %NETIF_F_SCTP_CRC
216 * - This feature indicates that a device is capable of
217 * offloading the SCTP CRC in a packet. To perform this offload the stack
218 * will set csum_start and csum_offset accordingly, set ip_summed to
219 * %CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication
220 * in the skbuff that the %CHECKSUM_PARTIAL refers to CRC32c.
221 * A driver that supports both IP checksum offload and SCTP CRC32c offload
222 * must verify which offload is configured for a packet by testing the
223 * value of &sk_buff.csum_not_inet; skb_crc32c_csum_help() is provided to
224 * resolve %CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
225 *
226 * * - %NETIF_F_FCOE_CRC
227 * - This feature indicates that a device is capable of offloading the FCOE
228 * CRC in a packet. To perform this offload the stack will set ip_summed
229 * to %CHECKSUM_PARTIAL and set csum_start and csum_offset
230 * accordingly. Note that there is no indication in the skbuff that the
231 * %CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
232 * both IP checksum offload and FCOE CRC offload must verify which offload
233 * is configured for a packet, presumably by inspecting packet headers.
234 *
235 * Checksumming on output with GSO
236 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
237 *
238 * In the case of a GSO packet (skb_is_gso() is true), checksum offload
239 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
240 * gso_type is %SKB_GSO_TCPV4 or %SKB_GSO_TCPV6, TCP checksum offload as
241 * part of the GSO operation is implied. If a checksum is being offloaded
242 * with GSO then ip_summed is %CHECKSUM_PARTIAL, and both csum_start and
243 * csum_offset are set to refer to the outermost checksum being offloaded
244 * (two offloaded checksums are possible with UDP encapsulation).
245 */
246
247 /* Don't change this without changing skb_csum_unnecessary! */
248 #define CHECKSUM_NONE 0
249 #define CHECKSUM_UNNECESSARY 1
250 #define CHECKSUM_COMPLETE 2
251 #define CHECKSUM_PARTIAL 3
252
253 /* Maximum value in skb->csum_level */
254 #define SKB_MAX_CSUM_LEVEL 3
255
256 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
257 #define SKB_WITH_OVERHEAD(X) \
258 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
259
260 /* For X bytes available in skb->head, what is the minimal
261 * allocation needed, knowing struct skb_shared_info needs
262 * to be aligned.
263 */
264 #define SKB_HEAD_ALIGN(X) (SKB_DATA_ALIGN(X) + \
265 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
266
267 #define SKB_MAX_ORDER(X, ORDER) \
268 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
269 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
270 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
271
272 /* return minimum truesize of one skb containing X bytes of data */
273 #define SKB_TRUESIZE(X) ((X) + \
274 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
275 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
276
277 struct ahash_request;
278 struct net_device;
279 struct scatterlist;
280 struct pipe_inode_info;
281 struct iov_iter;
282 struct napi_struct;
283 struct bpf_prog;
284 union bpf_attr;
285 struct skb_ext;
286 struct ts_config;
287
288 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
289 struct nf_bridge_info {
290 enum {
291 BRNF_PROTO_UNCHANGED,
292 BRNF_PROTO_8021Q,
293 BRNF_PROTO_PPPOE
294 } orig_proto:8;
295 u8 pkt_otherhost:1;
296 u8 in_prerouting:1;
297 u8 bridged_dnat:1;
298 u8 sabotage_in_done:1;
299 __u16 frag_max_size;
300 int physinif;
301
302 /* always valid & non-NULL from FORWARD on, for physdev match */
303 struct net_device *physoutdev;
304 union {
305 /* prerouting: detect dnat in orig/reply direction */
306 __be32 ipv4_daddr;
307 struct in6_addr ipv6_daddr;
308
309 /* after prerouting + nat detected: store original source
310 * mac since neigh resolution overwrites it, only used while
311 * skb is out in neigh layer.
312 */
313 char neigh_header[8];
314 };
315 };
316 #endif
317
318 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
319 /* Chain in tc_skb_ext will be used to share the tc chain with
320 * ovs recirc_id. It will be set to the current chain by tc
321 * and read by ovs to recirc_id.
322 */
323 struct tc_skb_ext {
324 union {
325 u64 act_miss_cookie;
326 __u32 chain;
327 };
328 __u16 mru;
329 __u16 zone;
330 u8 post_ct:1;
331 u8 post_ct_snat:1;
332 u8 post_ct_dnat:1;
333 u8 act_miss:1; /* Set if act_miss_cookie is used */
334 u8 l2_miss:1; /* Set by bridge upon FDB or MDB miss */
335 };
336 #endif
337
338 struct sk_buff_head {
339 /* These two members must be first to match sk_buff. */
340 struct_group_tagged(sk_buff_list, list,
341 struct sk_buff *next;
342 struct sk_buff *prev;
343 );
344
345 __u32 qlen;
346 spinlock_t lock;
347 };
348
349 struct sk_buff;
350
351 #ifndef CONFIG_MAX_SKB_FRAGS
352 # define CONFIG_MAX_SKB_FRAGS 17
353 #endif
354
355 #define MAX_SKB_FRAGS CONFIG_MAX_SKB_FRAGS
356
357 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
358 * segment using its current segmentation instead.
359 */
360 #define GSO_BY_FRAGS 0xFFFF
361
362 typedef struct skb_frag {
363 netmem_ref netmem;
364 unsigned int len;
365 unsigned int offset;
366 } skb_frag_t;
367
368 /**
369 * skb_frag_size() - Returns the size of a skb fragment
370 * @frag: skb fragment
371 */
skb_frag_size(const skb_frag_t * frag)372 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
373 {
374 return frag->len;
375 }
376
377 /**
378 * skb_frag_size_set() - Sets the size of a skb fragment
379 * @frag: skb fragment
380 * @size: size of fragment
381 */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)382 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
383 {
384 frag->len = size;
385 }
386
387 /**
388 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
389 * @frag: skb fragment
390 * @delta: value to add
391 */
skb_frag_size_add(skb_frag_t * frag,int delta)392 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
393 {
394 frag->len += delta;
395 }
396
397 /**
398 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
399 * @frag: skb fragment
400 * @delta: value to subtract
401 */
skb_frag_size_sub(skb_frag_t * frag,int delta)402 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
403 {
404 frag->len -= delta;
405 }
406
407 /**
408 * skb_frag_must_loop - Test if %p is a high memory page
409 * @p: fragment's page
410 */
skb_frag_must_loop(struct page * p)411 static inline bool skb_frag_must_loop(struct page *p)
412 {
413 #if defined(CONFIG_HIGHMEM)
414 if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
415 return true;
416 #endif
417 return false;
418 }
419
420 /**
421 * skb_frag_foreach_page - loop over pages in a fragment
422 *
423 * @f: skb frag to operate on
424 * @f_off: offset from start of f->netmem
425 * @f_len: length from f_off to loop over
426 * @p: (temp var) current page
427 * @p_off: (temp var) offset from start of current page,
428 * non-zero only on first page.
429 * @p_len: (temp var) length in current page,
430 * < PAGE_SIZE only on first and last page.
431 * @copied: (temp var) length so far, excluding current p_len.
432 *
433 * A fragment can hold a compound page, in which case per-page
434 * operations, notably kmap_atomic, must be called for each
435 * regular page.
436 */
437 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
438 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
439 p_off = (f_off) & (PAGE_SIZE - 1), \
440 p_len = skb_frag_must_loop(p) ? \
441 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
442 copied = 0; \
443 copied < f_len; \
444 copied += p_len, p++, p_off = 0, \
445 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
446
447 /**
448 * struct skb_shared_hwtstamps - hardware time stamps
449 * @hwtstamp: hardware time stamp transformed into duration
450 * since arbitrary point in time
451 * @netdev_data: address/cookie of network device driver used as
452 * reference to actual hardware time stamp
453 *
454 * Software time stamps generated by ktime_get_real() are stored in
455 * skb->tstamp.
456 *
457 * hwtstamps can only be compared against other hwtstamps from
458 * the same device.
459 *
460 * This structure is attached to packets as part of the
461 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
462 */
463 struct skb_shared_hwtstamps {
464 union {
465 ktime_t hwtstamp;
466 void *netdev_data;
467 };
468 };
469
470 /* Definitions for tx_flags in struct skb_shared_info */
471 enum {
472 /* generate hardware time stamp */
473 SKBTX_HW_TSTAMP = 1 << 0,
474
475 /* generate software time stamp when queueing packet to NIC */
476 SKBTX_SW_TSTAMP = 1 << 1,
477
478 /* device driver is going to provide hardware time stamp */
479 SKBTX_IN_PROGRESS = 1 << 2,
480
481 /* generate hardware time stamp based on cycles if supported */
482 SKBTX_HW_TSTAMP_USE_CYCLES = 1 << 3,
483
484 /* generate wifi status information (where possible) */
485 SKBTX_WIFI_STATUS = 1 << 4,
486
487 /* determine hardware time stamp based on time or cycles */
488 SKBTX_HW_TSTAMP_NETDEV = 1 << 5,
489
490 /* generate software time stamp when entering packet scheduling */
491 SKBTX_SCHED_TSTAMP = 1 << 6,
492 };
493
494 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
495 SKBTX_SCHED_TSTAMP)
496 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | \
497 SKBTX_HW_TSTAMP_USE_CYCLES | \
498 SKBTX_ANY_SW_TSTAMP)
499
500 /* Definitions for flags in struct skb_shared_info */
501 enum {
502 /* use zcopy routines */
503 SKBFL_ZEROCOPY_ENABLE = BIT(0),
504
505 /* This indicates at least one fragment might be overwritten
506 * (as in vmsplice(), sendfile() ...)
507 * If we need to compute a TX checksum, we'll need to copy
508 * all frags to avoid possible bad checksum
509 */
510 SKBFL_SHARED_FRAG = BIT(1),
511
512 /* segment contains only zerocopy data and should not be
513 * charged to the kernel memory.
514 */
515 SKBFL_PURE_ZEROCOPY = BIT(2),
516
517 SKBFL_DONT_ORPHAN = BIT(3),
518
519 /* page references are managed by the ubuf_info, so it's safe to
520 * use frags only up until ubuf_info is released
521 */
522 SKBFL_MANAGED_FRAG_REFS = BIT(4),
523 };
524
525 #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
526 #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \
527 SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS)
528
529 struct ubuf_info_ops {
530 void (*complete)(struct sk_buff *, struct ubuf_info *,
531 bool zerocopy_success);
532 /* has to be compatible with skb_zcopy_set() */
533 int (*link_skb)(struct sk_buff *skb, struct ubuf_info *uarg);
534 };
535
536 /*
537 * The callback notifies userspace to release buffers when skb DMA is done in
538 * lower device, the skb last reference should be 0 when calling this.
539 * The zerocopy_success argument is true if zero copy transmit occurred,
540 * false on data copy or out of memory error caused by data copy attempt.
541 * The ctx field is used to track device context.
542 * The desc field is used to track userspace buffer index.
543 */
544 struct ubuf_info {
545 const struct ubuf_info_ops *ops;
546 refcount_t refcnt;
547 u8 flags;
548 };
549
550 struct ubuf_info_msgzc {
551 struct ubuf_info ubuf;
552
553 union {
554 struct {
555 unsigned long desc;
556 void *ctx;
557 };
558 struct {
559 u32 id;
560 u16 len;
561 u16 zerocopy:1;
562 u32 bytelen;
563 };
564 };
565
566 struct mmpin {
567 struct user_struct *user;
568 unsigned int num_pg;
569 } mmp;
570 };
571
572 #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
573 #define uarg_to_msgzc(ubuf_ptr) container_of((ubuf_ptr), struct ubuf_info_msgzc, \
574 ubuf)
575
576 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
577 void mm_unaccount_pinned_pages(struct mmpin *mmp);
578
579 /* Preserve some data across TX submission and completion.
580 *
581 * Note, this state is stored in the driver. Extending the layout
582 * might need some special care.
583 */
584 struct xsk_tx_metadata_compl {
585 __u64 *tx_timestamp;
586 };
587
588 /* This data is invariant across clones and lives at
589 * the end of the header data, ie. at skb->end.
590 */
591 struct skb_shared_info {
592 __u8 flags;
593 __u8 meta_len;
594 __u8 nr_frags;
595 __u8 tx_flags;
596 unsigned short gso_size;
597 /* Warning: this field is not always filled in (UFO)! */
598 unsigned short gso_segs;
599 struct sk_buff *frag_list;
600 union {
601 struct skb_shared_hwtstamps hwtstamps;
602 struct xsk_tx_metadata_compl xsk_meta;
603 };
604 unsigned int gso_type;
605 u32 tskey;
606
607 /*
608 * Warning : all fields before dataref are cleared in __alloc_skb()
609 */
610 atomic_t dataref;
611
612 union {
613 struct {
614 u32 xdp_frags_size;
615 u32 xdp_frags_truesize;
616 };
617
618 /*
619 * Intermediate layers must ensure that destructor_arg
620 * remains valid until skb destructor.
621 */
622 void *destructor_arg;
623 };
624
625 /* must be last field, see pskb_expand_head() */
626 skb_frag_t frags[MAX_SKB_FRAGS];
627 };
628
629 /**
630 * DOC: dataref and headerless skbs
631 *
632 * Transport layers send out clones of payload skbs they hold for
633 * retransmissions. To allow lower layers of the stack to prepend their headers
634 * we split &skb_shared_info.dataref into two halves.
635 * The lower 16 bits count the overall number of references.
636 * The higher 16 bits indicate how many of the references are payload-only.
637 * skb_header_cloned() checks if skb is allowed to add / write the headers.
638 *
639 * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr
640 * (via __skb_header_release()). Any clone created from marked skb will get
641 * &sk_buff.hdr_len populated with the available headroom.
642 * If there's the only clone in existence it's able to modify the headroom
643 * at will. The sequence of calls inside the transport layer is::
644 *
645 * <alloc skb>
646 * skb_reserve()
647 * __skb_header_release()
648 * skb_clone()
649 * // send the clone down the stack
650 *
651 * This is not a very generic construct and it depends on the transport layers
652 * doing the right thing. In practice there's usually only one payload-only skb.
653 * Having multiple payload-only skbs with different lengths of hdr_len is not
654 * possible. The payload-only skbs should never leave their owner.
655 */
656 #define SKB_DATAREF_SHIFT 16
657 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
658
659
660 enum {
661 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
662 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
663 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
664 };
665
666 enum {
667 SKB_GSO_TCPV4 = 1 << 0,
668
669 /* This indicates the skb is from an untrusted source. */
670 SKB_GSO_DODGY = 1 << 1,
671
672 /* This indicates the tcp segment has CWR set. */
673 SKB_GSO_TCP_ECN = 1 << 2,
674
675 SKB_GSO_TCP_FIXEDID = 1 << 3,
676
677 SKB_GSO_TCPV6 = 1 << 4,
678
679 SKB_GSO_FCOE = 1 << 5,
680
681 SKB_GSO_GRE = 1 << 6,
682
683 SKB_GSO_GRE_CSUM = 1 << 7,
684
685 SKB_GSO_IPXIP4 = 1 << 8,
686
687 SKB_GSO_IPXIP6 = 1 << 9,
688
689 SKB_GSO_UDP_TUNNEL = 1 << 10,
690
691 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
692
693 SKB_GSO_PARTIAL = 1 << 12,
694
695 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
696
697 SKB_GSO_SCTP = 1 << 14,
698
699 SKB_GSO_ESP = 1 << 15,
700
701 SKB_GSO_UDP = 1 << 16,
702
703 SKB_GSO_UDP_L4 = 1 << 17,
704
705 SKB_GSO_FRAGLIST = 1 << 18,
706 };
707
708 #if BITS_PER_LONG > 32
709 #define NET_SKBUFF_DATA_USES_OFFSET 1
710 #endif
711
712 #ifdef NET_SKBUFF_DATA_USES_OFFSET
713 typedef unsigned int sk_buff_data_t;
714 #else
715 typedef unsigned char *sk_buff_data_t;
716 #endif
717
718 enum skb_tstamp_type {
719 SKB_CLOCK_REALTIME,
720 SKB_CLOCK_MONOTONIC,
721 SKB_CLOCK_TAI,
722 __SKB_CLOCK_MAX = SKB_CLOCK_TAI,
723 };
724
725 /**
726 * DOC: Basic sk_buff geometry
727 *
728 * struct sk_buff itself is a metadata structure and does not hold any packet
729 * data. All the data is held in associated buffers.
730 *
731 * &sk_buff.head points to the main "head" buffer. The head buffer is divided
732 * into two parts:
733 *
734 * - data buffer, containing headers and sometimes payload;
735 * this is the part of the skb operated on by the common helpers
736 * such as skb_put() or skb_pull();
737 * - shared info (struct skb_shared_info) which holds an array of pointers
738 * to read-only data in the (page, offset, length) format.
739 *
740 * Optionally &skb_shared_info.frag_list may point to another skb.
741 *
742 * Basic diagram may look like this::
743 *
744 * ---------------
745 * | sk_buff |
746 * ---------------
747 * ,--------------------------- + head
748 * / ,----------------- + data
749 * / / ,----------- + tail
750 * | | | , + end
751 * | | | |
752 * v v v v
753 * -----------------------------------------------
754 * | headroom | data | tailroom | skb_shared_info |
755 * -----------------------------------------------
756 * + [page frag]
757 * + [page frag]
758 * + [page frag]
759 * + [page frag] ---------
760 * + frag_list --> | sk_buff |
761 * ---------
762 *
763 */
764
765 /**
766 * struct sk_buff - socket buffer
767 * @next: Next buffer in list
768 * @prev: Previous buffer in list
769 * @tstamp: Time we arrived/left
770 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
771 * for retransmit timer
772 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
773 * @list: queue head
774 * @ll_node: anchor in an llist (eg socket defer_list)
775 * @sk: Socket we are owned by
776 * @dev: Device we arrived on/are leaving by
777 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
778 * @cb: Control buffer. Free for use by every layer. Put private vars here
779 * @_skb_refdst: destination entry (with norefcount bit)
780 * @len: Length of actual data
781 * @data_len: Data length
782 * @mac_len: Length of link layer header
783 * @hdr_len: writable header length of cloned skb
784 * @csum: Checksum (must include start/offset pair)
785 * @csum_start: Offset from skb->head where checksumming should start
786 * @csum_offset: Offset from csum_start where checksum should be stored
787 * @priority: Packet queueing priority
788 * @ignore_df: allow local fragmentation
789 * @cloned: Head may be cloned (check refcnt to be sure)
790 * @ip_summed: Driver fed us an IP checksum
791 * @nohdr: Payload reference only, must not modify header
792 * @pkt_type: Packet class
793 * @fclone: skbuff clone status
794 * @ipvs_property: skbuff is owned by ipvs
795 * @inner_protocol_type: whether the inner protocol is
796 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
797 * @remcsum_offload: remote checksum offload is enabled
798 * @offload_fwd_mark: Packet was L2-forwarded in hardware
799 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
800 * @tc_skip_classify: do not classify packet. set by IFB device
801 * @tc_at_ingress: used within tc_classify to distinguish in/egress
802 * @redirected: packet was redirected by packet classifier
803 * @from_ingress: packet was redirected from the ingress path
804 * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
805 * @peeked: this packet has been seen already, so stats have been
806 * done for it, don't do them again
807 * @nf_trace: netfilter packet trace flag
808 * @protocol: Packet protocol from driver
809 * @destructor: Destruct function
810 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
811 * @_sk_redir: socket redirection information for skmsg
812 * @_nfct: Associated connection, if any (with nfctinfo bits)
813 * @skb_iif: ifindex of device we arrived on
814 * @tc_index: Traffic control index
815 * @hash: the packet hash
816 * @queue_mapping: Queue mapping for multiqueue devices
817 * @head_frag: skb was allocated from page fragments,
818 * not allocated by kmalloc() or vmalloc().
819 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
820 * @pp_recycle: mark the packet for recycling instead of freeing (implies
821 * page_pool support on driver)
822 * @active_extensions: active extensions (skb_ext_id types)
823 * @ndisc_nodetype: router type (from link layer)
824 * @ooo_okay: allow the mapping of a socket to a queue to be changed
825 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
826 * ports.
827 * @sw_hash: indicates hash was computed in software stack
828 * @wifi_acked_valid: wifi_acked was set
829 * @wifi_acked: whether frame was acked on wifi or not
830 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
831 * @encapsulation: indicates the inner headers in the skbuff are valid
832 * @encap_hdr_csum: software checksum is needed
833 * @csum_valid: checksum is already valid
834 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
835 * @csum_complete_sw: checksum was completed by software
836 * @csum_level: indicates the number of consecutive checksums found in
837 * the packet minus one that have been verified as
838 * CHECKSUM_UNNECESSARY (max 3)
839 * @unreadable: indicates that at least 1 of the fragments in this skb is
840 * unreadable.
841 * @dst_pending_confirm: need to confirm neighbour
842 * @decrypted: Decrypted SKB
843 * @slow_gro: state present at GRO time, slower prepare step required
844 * @tstamp_type: When set, skb->tstamp has the
845 * delivery_time clock base of skb->tstamp.
846 * @napi_id: id of the NAPI struct this skb came from
847 * @sender_cpu: (aka @napi_id) source CPU in XPS
848 * @alloc_cpu: CPU which did the skb allocation.
849 * @secmark: security marking
850 * @mark: Generic packet mark
851 * @reserved_tailroom: (aka @mark) number of bytes of free space available
852 * at the tail of an sk_buff
853 * @vlan_all: vlan fields (proto & tci)
854 * @vlan_proto: vlan encapsulation protocol
855 * @vlan_tci: vlan tag control information
856 * @inner_protocol: Protocol (encapsulation)
857 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
858 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
859 * @inner_transport_header: Inner transport layer header (encapsulation)
860 * @inner_network_header: Network layer header (encapsulation)
861 * @inner_mac_header: Link layer header (encapsulation)
862 * @transport_header: Transport layer header
863 * @network_header: Network layer header
864 * @mac_header: Link layer header
865 * @kcov_handle: KCOV remote handle for remote coverage collection
866 * @tail: Tail pointer
867 * @end: End pointer
868 * @head: Head of buffer
869 * @data: Data head pointer
870 * @truesize: Buffer size
871 * @users: User count - see {datagram,tcp}.c
872 * @extensions: allocated extensions, valid if active_extensions is nonzero
873 */
874
875 struct sk_buff {
876 union {
877 struct {
878 /* These two members must be first to match sk_buff_head. */
879 struct sk_buff *next;
880 struct sk_buff *prev;
881
882 union {
883 struct net_device *dev;
884 /* Some protocols might use this space to store information,
885 * while device pointer would be NULL.
886 * UDP receive path is one user.
887 */
888 unsigned long dev_scratch;
889 };
890 };
891 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
892 struct list_head list;
893 struct llist_node ll_node;
894 };
895
896 struct sock *sk;
897
898 union {
899 ktime_t tstamp;
900 u64 skb_mstamp_ns; /* earliest departure time */
901 };
902 /*
903 * This is the control buffer. It is free to use for every
904 * layer. Please put your private variables there. If you
905 * want to keep them across layers you have to do a skb_clone()
906 * first. This is owned by whoever has the skb queued ATM.
907 */
908 char cb[48] __aligned(8);
909
910 union {
911 struct {
912 unsigned long _skb_refdst;
913 void (*destructor)(struct sk_buff *skb);
914 };
915 struct list_head tcp_tsorted_anchor;
916 #ifdef CONFIG_NET_SOCK_MSG
917 unsigned long _sk_redir;
918 #endif
919 };
920
921 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
922 unsigned long _nfct;
923 #endif
924 unsigned int len,
925 data_len;
926 __u16 mac_len,
927 hdr_len;
928
929 /* Following fields are _not_ copied in __copy_skb_header()
930 * Note that queue_mapping is here mostly to fill a hole.
931 */
932 __u16 queue_mapping;
933
934 /* if you move cloned around you also must adapt those constants */
935 #ifdef __BIG_ENDIAN_BITFIELD
936 #define CLONED_MASK (1 << 7)
937 #else
938 #define CLONED_MASK 1
939 #endif
940 #define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset)
941
942 /* private: */
943 __u8 __cloned_offset[0];
944 /* public: */
945 __u8 cloned:1,
946 nohdr:1,
947 fclone:2,
948 peeked:1,
949 head_frag:1,
950 pfmemalloc:1,
951 pp_recycle:1; /* page_pool recycle indicator */
952 #ifdef CONFIG_SKB_EXTENSIONS
953 __u8 active_extensions;
954 #endif
955
956 /* Fields enclosed in headers group are copied
957 * using a single memcpy() in __copy_skb_header()
958 */
959 struct_group(headers,
960
961 /* private: */
962 __u8 __pkt_type_offset[0];
963 /* public: */
964 __u8 pkt_type:3; /* see PKT_TYPE_MAX */
965 __u8 ignore_df:1;
966 __u8 dst_pending_confirm:1;
967 __u8 ip_summed:2;
968 __u8 ooo_okay:1;
969
970 /* private: */
971 __u8 __mono_tc_offset[0];
972 /* public: */
973 __u8 tstamp_type:2; /* See skb_tstamp_type */
974 #ifdef CONFIG_NET_XGRESS
975 __u8 tc_at_ingress:1; /* See TC_AT_INGRESS_MASK */
976 __u8 tc_skip_classify:1;
977 #endif
978 __u8 remcsum_offload:1;
979 __u8 csum_complete_sw:1;
980 __u8 csum_level:2;
981 __u8 inner_protocol_type:1;
982
983 __u8 l4_hash:1;
984 __u8 sw_hash:1;
985 #ifdef CONFIG_WIRELESS
986 __u8 wifi_acked_valid:1;
987 __u8 wifi_acked:1;
988 #endif
989 __u8 no_fcs:1;
990 /* Indicates the inner headers are valid in the skbuff. */
991 __u8 encapsulation:1;
992 __u8 encap_hdr_csum:1;
993 __u8 csum_valid:1;
994 #ifdef CONFIG_IPV6_NDISC_NODETYPE
995 __u8 ndisc_nodetype:2;
996 #endif
997
998 #if IS_ENABLED(CONFIG_IP_VS)
999 __u8 ipvs_property:1;
1000 #endif
1001 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
1002 __u8 nf_trace:1;
1003 #endif
1004 #ifdef CONFIG_NET_SWITCHDEV
1005 __u8 offload_fwd_mark:1;
1006 __u8 offload_l3_fwd_mark:1;
1007 #endif
1008 __u8 redirected:1;
1009 #ifdef CONFIG_NET_REDIRECT
1010 __u8 from_ingress:1;
1011 #endif
1012 #ifdef CONFIG_NETFILTER_SKIP_EGRESS
1013 __u8 nf_skip_egress:1;
1014 #endif
1015 #ifdef CONFIG_SKB_DECRYPTED
1016 __u8 decrypted:1;
1017 #endif
1018 __u8 slow_gro:1;
1019 #if IS_ENABLED(CONFIG_IP_SCTP)
1020 __u8 csum_not_inet:1;
1021 #endif
1022 __u8 unreadable:1;
1023 #if defined(CONFIG_NET_SCHED) || defined(CONFIG_NET_XGRESS)
1024 __u16 tc_index; /* traffic control index */
1025 #endif
1026
1027 u16 alloc_cpu;
1028
1029 union {
1030 __wsum csum;
1031 struct {
1032 __u16 csum_start;
1033 __u16 csum_offset;
1034 };
1035 };
1036 __u32 priority;
1037 int skb_iif;
1038 __u32 hash;
1039 union {
1040 u32 vlan_all;
1041 struct {
1042 __be16 vlan_proto;
1043 __u16 vlan_tci;
1044 };
1045 };
1046 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
1047 union {
1048 unsigned int napi_id;
1049 unsigned int sender_cpu;
1050 };
1051 #endif
1052 #ifdef CONFIG_NETWORK_SECMARK
1053 __u32 secmark;
1054 #endif
1055
1056 union {
1057 __u32 mark;
1058 __u32 reserved_tailroom;
1059 };
1060
1061 union {
1062 __be16 inner_protocol;
1063 __u8 inner_ipproto;
1064 };
1065
1066 __u16 inner_transport_header;
1067 __u16 inner_network_header;
1068 __u16 inner_mac_header;
1069
1070 __be16 protocol;
1071 __u16 transport_header;
1072 __u16 network_header;
1073 __u16 mac_header;
1074
1075 #ifdef CONFIG_KCOV
1076 u64 kcov_handle;
1077 #endif
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 #if defined(CONFIG_FAIL_SKB_REALLOC)
2694 void skb_might_realloc(struct sk_buff *skb);
2695 #else
skb_might_realloc(struct sk_buff * skb)2696 static inline void skb_might_realloc(struct sk_buff *skb) {}
2697 #endif
2698
2699 /*
2700 * Add data to an sk_buff
2701 */
2702 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2703 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2704 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2705 {
2706 void *tmp = skb_tail_pointer(skb);
2707 SKB_LINEAR_ASSERT(skb);
2708 skb->tail += len;
2709 skb->len += len;
2710 return tmp;
2711 }
2712
__skb_put_zero(struct sk_buff * skb,unsigned int len)2713 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2714 {
2715 void *tmp = __skb_put(skb, len);
2716
2717 memset(tmp, 0, len);
2718 return tmp;
2719 }
2720
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2721 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2722 unsigned int len)
2723 {
2724 void *tmp = __skb_put(skb, len);
2725
2726 memcpy(tmp, data, len);
2727 return tmp;
2728 }
2729
__skb_put_u8(struct sk_buff * skb,u8 val)2730 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2731 {
2732 *(u8 *)__skb_put(skb, 1) = val;
2733 }
2734
skb_put_zero(struct sk_buff * skb,unsigned int len)2735 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2736 {
2737 void *tmp = skb_put(skb, len);
2738
2739 memset(tmp, 0, len);
2740
2741 return tmp;
2742 }
2743
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2744 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2745 unsigned int len)
2746 {
2747 void *tmp = skb_put(skb, len);
2748
2749 memcpy(tmp, data, len);
2750
2751 return tmp;
2752 }
2753
skb_put_u8(struct sk_buff * skb,u8 val)2754 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2755 {
2756 *(u8 *)skb_put(skb, 1) = val;
2757 }
2758
2759 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2760 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2761 {
2762 DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
2763
2764 skb->data -= len;
2765 skb->len += len;
2766 return skb->data;
2767 }
2768
2769 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2770 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2771 {
2772 DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
2773
2774 skb->len -= len;
2775 if (unlikely(skb->len < skb->data_len)) {
2776 #if defined(CONFIG_DEBUG_NET)
2777 skb->len += len;
2778 pr_err("__skb_pull(len=%u)\n", len);
2779 skb_dump(KERN_ERR, skb, false);
2780 #endif
2781 BUG();
2782 }
2783 return skb->data += len;
2784 }
2785
skb_pull_inline(struct sk_buff * skb,unsigned int len)2786 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2787 {
2788 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2789 }
2790
2791 void *skb_pull_data(struct sk_buff *skb, size_t len);
2792
2793 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2794
2795 static inline enum skb_drop_reason
pskb_may_pull_reason(struct sk_buff * skb,unsigned int len)2796 pskb_may_pull_reason(struct sk_buff *skb, unsigned int len)
2797 {
2798 DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
2799 skb_might_realloc(skb);
2800
2801 if (likely(len <= skb_headlen(skb)))
2802 return SKB_NOT_DROPPED_YET;
2803
2804 if (unlikely(len > skb->len))
2805 return SKB_DROP_REASON_PKT_TOO_SMALL;
2806
2807 if (unlikely(!__pskb_pull_tail(skb, len - skb_headlen(skb))))
2808 return SKB_DROP_REASON_NOMEM;
2809
2810 return SKB_NOT_DROPPED_YET;
2811 }
2812
pskb_may_pull(struct sk_buff * skb,unsigned int len)2813 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2814 {
2815 return pskb_may_pull_reason(skb, len) == SKB_NOT_DROPPED_YET;
2816 }
2817
pskb_pull(struct sk_buff * skb,unsigned int len)2818 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2819 {
2820 if (!pskb_may_pull(skb, len))
2821 return NULL;
2822
2823 skb->len -= len;
2824 return skb->data += len;
2825 }
2826
2827 void skb_condense(struct sk_buff *skb);
2828
2829 /**
2830 * skb_headroom - bytes at buffer head
2831 * @skb: buffer to check
2832 *
2833 * Return the number of bytes of free space at the head of an &sk_buff.
2834 */
skb_headroom(const struct sk_buff * skb)2835 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2836 {
2837 return skb->data - skb->head;
2838 }
2839
2840 /**
2841 * skb_tailroom - bytes at buffer end
2842 * @skb: buffer to check
2843 *
2844 * Return the number of bytes of free space at the tail of an sk_buff
2845 */
skb_tailroom(const struct sk_buff * skb)2846 static inline int skb_tailroom(const struct sk_buff *skb)
2847 {
2848 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2849 }
2850
2851 /**
2852 * skb_availroom - bytes at buffer end
2853 * @skb: buffer to check
2854 *
2855 * Return the number of bytes of free space at the tail of an sk_buff
2856 * allocated by sk_stream_alloc()
2857 */
skb_availroom(const struct sk_buff * skb)2858 static inline int skb_availroom(const struct sk_buff *skb)
2859 {
2860 if (skb_is_nonlinear(skb))
2861 return 0;
2862
2863 return skb->end - skb->tail - skb->reserved_tailroom;
2864 }
2865
2866 /**
2867 * skb_reserve - adjust headroom
2868 * @skb: buffer to alter
2869 * @len: bytes to move
2870 *
2871 * Increase the headroom of an empty &sk_buff by reducing the tail
2872 * room. This is only allowed for an empty buffer.
2873 */
skb_reserve(struct sk_buff * skb,int len)2874 static inline void skb_reserve(struct sk_buff *skb, int len)
2875 {
2876 skb->data += len;
2877 skb->tail += len;
2878 }
2879
2880 /**
2881 * skb_tailroom_reserve - adjust reserved_tailroom
2882 * @skb: buffer to alter
2883 * @mtu: maximum amount of headlen permitted
2884 * @needed_tailroom: minimum amount of reserved_tailroom
2885 *
2886 * Set reserved_tailroom so that headlen can be as large as possible but
2887 * not larger than mtu and tailroom cannot be smaller than
2888 * needed_tailroom.
2889 * The required headroom should already have been reserved before using
2890 * this function.
2891 */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2892 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2893 unsigned int needed_tailroom)
2894 {
2895 SKB_LINEAR_ASSERT(skb);
2896 if (mtu < skb_tailroom(skb) - needed_tailroom)
2897 /* use at most mtu */
2898 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2899 else
2900 /* use up to all available space */
2901 skb->reserved_tailroom = needed_tailroom;
2902 }
2903
2904 #define ENCAP_TYPE_ETHER 0
2905 #define ENCAP_TYPE_IPPROTO 1
2906
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2907 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2908 __be16 protocol)
2909 {
2910 skb->inner_protocol = protocol;
2911 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2912 }
2913
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2914 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2915 __u8 ipproto)
2916 {
2917 skb->inner_ipproto = ipproto;
2918 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2919 }
2920
skb_reset_inner_headers(struct sk_buff * skb)2921 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2922 {
2923 skb->inner_mac_header = skb->mac_header;
2924 skb->inner_network_header = skb->network_header;
2925 skb->inner_transport_header = skb->transport_header;
2926 }
2927
skb_mac_header_was_set(const struct sk_buff * skb)2928 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2929 {
2930 return skb->mac_header != (typeof(skb->mac_header))~0U;
2931 }
2932
skb_reset_mac_len(struct sk_buff * skb)2933 static inline void skb_reset_mac_len(struct sk_buff *skb)
2934 {
2935 if (!skb_mac_header_was_set(skb)) {
2936 DEBUG_NET_WARN_ON_ONCE(1);
2937 skb->mac_len = 0;
2938 } else {
2939 skb->mac_len = skb->network_header - skb->mac_header;
2940 }
2941 }
2942
skb_inner_transport_header(const struct sk_buff * skb)2943 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2944 *skb)
2945 {
2946 return skb->head + skb->inner_transport_header;
2947 }
2948
skb_inner_transport_offset(const struct sk_buff * skb)2949 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2950 {
2951 return skb_inner_transport_header(skb) - skb->data;
2952 }
2953
skb_reset_inner_transport_header(struct sk_buff * skb)2954 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2955 {
2956 long offset = skb->data - skb->head;
2957
2958 DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->inner_transport_header))offset);
2959 skb->inner_transport_header = offset;
2960 }
2961
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2962 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2963 const int offset)
2964 {
2965 skb_reset_inner_transport_header(skb);
2966 skb->inner_transport_header += offset;
2967 }
2968
skb_inner_network_header(const struct sk_buff * skb)2969 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2970 {
2971 return skb->head + skb->inner_network_header;
2972 }
2973
skb_reset_inner_network_header(struct sk_buff * skb)2974 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2975 {
2976 long offset = skb->data - skb->head;
2977
2978 DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->inner_network_header))offset);
2979 skb->inner_network_header = offset;
2980 }
2981
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2982 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2983 const int offset)
2984 {
2985 skb_reset_inner_network_header(skb);
2986 skb->inner_network_header += offset;
2987 }
2988
skb_inner_network_header_was_set(const struct sk_buff * skb)2989 static inline bool skb_inner_network_header_was_set(const struct sk_buff *skb)
2990 {
2991 return skb->inner_network_header > 0;
2992 }
2993
skb_inner_mac_header(const struct sk_buff * skb)2994 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2995 {
2996 return skb->head + skb->inner_mac_header;
2997 }
2998
skb_reset_inner_mac_header(struct sk_buff * skb)2999 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
3000 {
3001 long offset = skb->data - skb->head;
3002
3003 DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->inner_mac_header))offset);
3004 skb->inner_mac_header = offset;
3005 }
3006
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)3007 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
3008 const int offset)
3009 {
3010 skb_reset_inner_mac_header(skb);
3011 skb->inner_mac_header += offset;
3012 }
skb_transport_header_was_set(const struct sk_buff * skb)3013 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
3014 {
3015 return skb->transport_header != (typeof(skb->transport_header))~0U;
3016 }
3017
skb_transport_header(const struct sk_buff * skb)3018 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
3019 {
3020 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
3021 return skb->head + skb->transport_header;
3022 }
3023
skb_reset_transport_header(struct sk_buff * skb)3024 static inline void skb_reset_transport_header(struct sk_buff *skb)
3025 {
3026 long offset = skb->data - skb->head;
3027
3028 DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->transport_header))offset);
3029 skb->transport_header = offset;
3030 }
3031
skb_set_transport_header(struct sk_buff * skb,const int offset)3032 static inline void skb_set_transport_header(struct sk_buff *skb,
3033 const int offset)
3034 {
3035 skb_reset_transport_header(skb);
3036 skb->transport_header += offset;
3037 }
3038
skb_network_header(const struct sk_buff * skb)3039 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
3040 {
3041 return skb->head + skb->network_header;
3042 }
3043
skb_reset_network_header(struct sk_buff * skb)3044 static inline void skb_reset_network_header(struct sk_buff *skb)
3045 {
3046 long offset = skb->data - skb->head;
3047
3048 DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->network_header))offset);
3049 skb->network_header = offset;
3050 }
3051
skb_set_network_header(struct sk_buff * skb,const int offset)3052 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
3053 {
3054 skb_reset_network_header(skb);
3055 skb->network_header += offset;
3056 }
3057
skb_mac_header(const struct sk_buff * skb)3058 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
3059 {
3060 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
3061 return skb->head + skb->mac_header;
3062 }
3063
skb_mac_offset(const struct sk_buff * skb)3064 static inline int skb_mac_offset(const struct sk_buff *skb)
3065 {
3066 return skb_mac_header(skb) - skb->data;
3067 }
3068
skb_mac_header_len(const struct sk_buff * skb)3069 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
3070 {
3071 DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
3072 return skb->network_header - skb->mac_header;
3073 }
3074
skb_unset_mac_header(struct sk_buff * skb)3075 static inline void skb_unset_mac_header(struct sk_buff *skb)
3076 {
3077 skb->mac_header = (typeof(skb->mac_header))~0U;
3078 }
3079
skb_reset_mac_header(struct sk_buff * skb)3080 static inline void skb_reset_mac_header(struct sk_buff *skb)
3081 {
3082 long offset = skb->data - skb->head;
3083
3084 DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->mac_header))offset);
3085 skb->mac_header = offset;
3086 }
3087
skb_set_mac_header(struct sk_buff * skb,const int offset)3088 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
3089 {
3090 skb_reset_mac_header(skb);
3091 skb->mac_header += offset;
3092 }
3093
skb_pop_mac_header(struct sk_buff * skb)3094 static inline void skb_pop_mac_header(struct sk_buff *skb)
3095 {
3096 skb->mac_header = skb->network_header;
3097 }
3098
skb_probe_transport_header(struct sk_buff * skb)3099 static inline void skb_probe_transport_header(struct sk_buff *skb)
3100 {
3101 struct flow_keys_basic keys;
3102
3103 if (skb_transport_header_was_set(skb))
3104 return;
3105
3106 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
3107 NULL, 0, 0, 0, 0))
3108 skb_set_transport_header(skb, keys.control.thoff);
3109 }
3110
skb_mac_header_rebuild(struct sk_buff * skb)3111 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
3112 {
3113 if (skb_mac_header_was_set(skb)) {
3114 const unsigned char *old_mac = skb_mac_header(skb);
3115
3116 skb_set_mac_header(skb, -skb->mac_len);
3117 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
3118 }
3119 }
3120
3121 /* Move the full mac header up to current network_header.
3122 * Leaves skb->data pointing at offset skb->mac_len into the mac_header.
3123 * Must be provided the complete mac header length.
3124 */
skb_mac_header_rebuild_full(struct sk_buff * skb,u32 full_mac_len)3125 static inline void skb_mac_header_rebuild_full(struct sk_buff *skb, u32 full_mac_len)
3126 {
3127 if (skb_mac_header_was_set(skb)) {
3128 const unsigned char *old_mac = skb_mac_header(skb);
3129
3130 skb_set_mac_header(skb, -full_mac_len);
3131 memmove(skb_mac_header(skb), old_mac, full_mac_len);
3132 __skb_push(skb, full_mac_len - skb->mac_len);
3133 }
3134 }
3135
skb_checksum_start_offset(const struct sk_buff * skb)3136 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
3137 {
3138 return skb->csum_start - skb_headroom(skb);
3139 }
3140
skb_checksum_start(const struct sk_buff * skb)3141 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
3142 {
3143 return skb->head + skb->csum_start;
3144 }
3145
skb_transport_offset(const struct sk_buff * skb)3146 static inline int skb_transport_offset(const struct sk_buff *skb)
3147 {
3148 return skb_transport_header(skb) - skb->data;
3149 }
3150
skb_network_header_len(const struct sk_buff * skb)3151 static inline u32 skb_network_header_len(const struct sk_buff *skb)
3152 {
3153 DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
3154 return skb->transport_header - skb->network_header;
3155 }
3156
skb_inner_network_header_len(const struct sk_buff * skb)3157 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
3158 {
3159 return skb->inner_transport_header - skb->inner_network_header;
3160 }
3161
skb_network_offset(const struct sk_buff * skb)3162 static inline int skb_network_offset(const struct sk_buff *skb)
3163 {
3164 return skb_network_header(skb) - skb->data;
3165 }
3166
skb_inner_network_offset(const struct sk_buff * skb)3167 static inline int skb_inner_network_offset(const struct sk_buff *skb)
3168 {
3169 return skb_inner_network_header(skb) - skb->data;
3170 }
3171
3172 static inline enum skb_drop_reason
pskb_network_may_pull_reason(struct sk_buff * skb,unsigned int len)3173 pskb_network_may_pull_reason(struct sk_buff *skb, unsigned int len)
3174 {
3175 return pskb_may_pull_reason(skb, skb_network_offset(skb) + len);
3176 }
3177
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)3178 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
3179 {
3180 return pskb_network_may_pull_reason(skb, len) == SKB_NOT_DROPPED_YET;
3181 }
3182
3183 /*
3184 * CPUs often take a performance hit when accessing unaligned memory
3185 * locations. The actual performance hit varies, it can be small if the
3186 * hardware handles it or large if we have to take an exception and fix it
3187 * in software.
3188 *
3189 * Since an ethernet header is 14 bytes network drivers often end up with
3190 * the IP header at an unaligned offset. The IP header can be aligned by
3191 * shifting the start of the packet by 2 bytes. Drivers should do this
3192 * with:
3193 *
3194 * skb_reserve(skb, NET_IP_ALIGN);
3195 *
3196 * The downside to this alignment of the IP header is that the DMA is now
3197 * unaligned. On some architectures the cost of an unaligned DMA is high
3198 * and this cost outweighs the gains made by aligning the IP header.
3199 *
3200 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
3201 * to be overridden.
3202 */
3203 #ifndef NET_IP_ALIGN
3204 #define NET_IP_ALIGN 2
3205 #endif
3206
3207 /*
3208 * The networking layer reserves some headroom in skb data (via
3209 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
3210 * the header has to grow. In the default case, if the header has to grow
3211 * 32 bytes or less we avoid the reallocation.
3212 *
3213 * Unfortunately this headroom changes the DMA alignment of the resulting
3214 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
3215 * on some architectures. An architecture can override this value,
3216 * perhaps setting it to a cacheline in size (since that will maintain
3217 * cacheline alignment of the DMA). It must be a power of 2.
3218 *
3219 * Various parts of the networking layer expect at least 32 bytes of
3220 * headroom, you should not reduce this.
3221 *
3222 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
3223 * to reduce average number of cache lines per packet.
3224 * get_rps_cpu() for example only access one 64 bytes aligned block :
3225 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
3226 */
3227 #ifndef NET_SKB_PAD
3228 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
3229 #endif
3230
3231 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
3232
__skb_set_length(struct sk_buff * skb,unsigned int len)3233 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
3234 {
3235 if (WARN_ON(skb_is_nonlinear(skb)))
3236 return;
3237 skb->len = len;
3238 skb_set_tail_pointer(skb, len);
3239 }
3240
__skb_trim(struct sk_buff * skb,unsigned int len)3241 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
3242 {
3243 __skb_set_length(skb, len);
3244 }
3245
3246 void skb_trim(struct sk_buff *skb, unsigned int len);
3247
__pskb_trim(struct sk_buff * skb,unsigned int len)3248 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
3249 {
3250 if (skb->data_len)
3251 return ___pskb_trim(skb, len);
3252 __skb_trim(skb, len);
3253 return 0;
3254 }
3255
pskb_trim(struct sk_buff * skb,unsigned int len)3256 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
3257 {
3258 skb_might_realloc(skb);
3259 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
3260 }
3261
3262 /**
3263 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
3264 * @skb: buffer to alter
3265 * @len: new length
3266 *
3267 * This is identical to pskb_trim except that the caller knows that
3268 * the skb is not cloned so we should never get an error due to out-
3269 * of-memory.
3270 */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)3271 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
3272 {
3273 int err = pskb_trim(skb, len);
3274 BUG_ON(err);
3275 }
3276
__skb_grow(struct sk_buff * skb,unsigned int len)3277 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
3278 {
3279 unsigned int diff = len - skb->len;
3280
3281 if (skb_tailroom(skb) < diff) {
3282 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
3283 GFP_ATOMIC);
3284 if (ret)
3285 return ret;
3286 }
3287 __skb_set_length(skb, len);
3288 return 0;
3289 }
3290
3291 /**
3292 * skb_orphan - orphan a buffer
3293 * @skb: buffer to orphan
3294 *
3295 * If a buffer currently has an owner then we call the owner's
3296 * destructor function and make the @skb unowned. The buffer continues
3297 * to exist but is no longer charged to its former owner.
3298 */
skb_orphan(struct sk_buff * skb)3299 static inline void skb_orphan(struct sk_buff *skb)
3300 {
3301 if (skb->destructor) {
3302 skb->destructor(skb);
3303 skb->destructor = NULL;
3304 skb->sk = NULL;
3305 } else {
3306 BUG_ON(skb->sk);
3307 }
3308 }
3309
3310 /**
3311 * skb_orphan_frags - orphan the frags contained in a buffer
3312 * @skb: buffer to orphan frags from
3313 * @gfp_mask: allocation mask for replacement pages
3314 *
3315 * For each frag in the SKB which needs a destructor (i.e. has an
3316 * owner) create a copy of that frag and release the original
3317 * page by calling the destructor.
3318 */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)3319 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
3320 {
3321 if (likely(!skb_zcopy(skb)))
3322 return 0;
3323 if (skb_shinfo(skb)->flags & SKBFL_DONT_ORPHAN)
3324 return 0;
3325 return skb_copy_ubufs(skb, gfp_mask);
3326 }
3327
3328 /* 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)3329 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
3330 {
3331 if (likely(!skb_zcopy(skb)))
3332 return 0;
3333 return skb_copy_ubufs(skb, gfp_mask);
3334 }
3335
3336 /**
3337 * __skb_queue_purge_reason - empty a list
3338 * @list: list to empty
3339 * @reason: drop reason
3340 *
3341 * Delete all buffers on an &sk_buff list. Each buffer is removed from
3342 * the list and one reference dropped. This function does not take the
3343 * list lock and the caller must hold the relevant locks to use it.
3344 */
__skb_queue_purge_reason(struct sk_buff_head * list,enum skb_drop_reason reason)3345 static inline void __skb_queue_purge_reason(struct sk_buff_head *list,
3346 enum skb_drop_reason reason)
3347 {
3348 struct sk_buff *skb;
3349
3350 while ((skb = __skb_dequeue(list)) != NULL)
3351 kfree_skb_reason(skb, reason);
3352 }
3353
__skb_queue_purge(struct sk_buff_head * list)3354 static inline void __skb_queue_purge(struct sk_buff_head *list)
3355 {
3356 __skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE);
3357 }
3358
3359 void skb_queue_purge_reason(struct sk_buff_head *list,
3360 enum skb_drop_reason reason);
3361
skb_queue_purge(struct sk_buff_head * list)3362 static inline void skb_queue_purge(struct sk_buff_head *list)
3363 {
3364 skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE);
3365 }
3366
3367 unsigned int skb_rbtree_purge(struct rb_root *root);
3368 void skb_errqueue_purge(struct sk_buff_head *list);
3369
3370 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3371
3372 /**
3373 * netdev_alloc_frag - allocate a page fragment
3374 * @fragsz: fragment size
3375 *
3376 * Allocates a frag from a page for receive buffer.
3377 * Uses GFP_ATOMIC allocations.
3378 */
netdev_alloc_frag(unsigned int fragsz)3379 static inline void *netdev_alloc_frag(unsigned int fragsz)
3380 {
3381 return __netdev_alloc_frag_align(fragsz, ~0u);
3382 }
3383
netdev_alloc_frag_align(unsigned int fragsz,unsigned int align)3384 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
3385 unsigned int align)
3386 {
3387 WARN_ON_ONCE(!is_power_of_2(align));
3388 return __netdev_alloc_frag_align(fragsz, -align);
3389 }
3390
3391 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
3392 gfp_t gfp_mask);
3393
3394 /**
3395 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
3396 * @dev: network device to receive on
3397 * @length: length to allocate
3398 *
3399 * Allocate a new &sk_buff and assign it a usage count of one. The
3400 * buffer has unspecified headroom built in. Users should allocate
3401 * the headroom they think they need without accounting for the
3402 * built in space. The built in space is used for optimisations.
3403 *
3404 * %NULL is returned if there is no free memory. Although this function
3405 * allocates memory it can be called from an interrupt.
3406 */
netdev_alloc_skb(struct net_device * dev,unsigned int length)3407 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
3408 unsigned int length)
3409 {
3410 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
3411 }
3412
3413 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)3414 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
3415 gfp_t gfp_mask)
3416 {
3417 return __netdev_alloc_skb(NULL, length, gfp_mask);
3418 }
3419
3420 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)3421 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
3422 {
3423 return netdev_alloc_skb(NULL, length);
3424 }
3425
3426
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)3427 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
3428 unsigned int length, gfp_t gfp)
3429 {
3430 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
3431
3432 if (NET_IP_ALIGN && skb)
3433 skb_reserve(skb, NET_IP_ALIGN);
3434 return skb;
3435 }
3436
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)3437 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
3438 unsigned int length)
3439 {
3440 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
3441 }
3442
skb_free_frag(void * addr)3443 static inline void skb_free_frag(void *addr)
3444 {
3445 page_frag_free(addr);
3446 }
3447
3448 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3449
napi_alloc_frag(unsigned int fragsz)3450 static inline void *napi_alloc_frag(unsigned int fragsz)
3451 {
3452 return __napi_alloc_frag_align(fragsz, ~0u);
3453 }
3454
napi_alloc_frag_align(unsigned int fragsz,unsigned int align)3455 static inline void *napi_alloc_frag_align(unsigned int fragsz,
3456 unsigned int align)
3457 {
3458 WARN_ON_ONCE(!is_power_of_2(align));
3459 return __napi_alloc_frag_align(fragsz, -align);
3460 }
3461
3462 struct sk_buff *napi_alloc_skb(struct napi_struct *napi, unsigned int length);
3463 void napi_consume_skb(struct sk_buff *skb, int budget);
3464
3465 void napi_skb_free_stolen_head(struct sk_buff *skb);
3466 void __napi_kfree_skb(struct sk_buff *skb, enum skb_drop_reason reason);
3467
3468 /**
3469 * __dev_alloc_pages - allocate page for network Rx
3470 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3471 * @order: size of the allocation
3472 *
3473 * Allocate a new page.
3474 *
3475 * %NULL is returned if there is no free memory.
3476 */
__dev_alloc_pages_noprof(gfp_t gfp_mask,unsigned int order)3477 static inline struct page *__dev_alloc_pages_noprof(gfp_t gfp_mask,
3478 unsigned int order)
3479 {
3480 /* This piece of code contains several assumptions.
3481 * 1. This is for device Rx, therefore a cold page is preferred.
3482 * 2. The expectation is the user wants a compound page.
3483 * 3. If requesting a order 0 page it will not be compound
3484 * due to the check to see if order has a value in prep_new_page
3485 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3486 * code in gfp_to_alloc_flags that should be enforcing this.
3487 */
3488 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3489
3490 return alloc_pages_node_noprof(NUMA_NO_NODE, gfp_mask, order);
3491 }
3492 #define __dev_alloc_pages(...) alloc_hooks(__dev_alloc_pages_noprof(__VA_ARGS__))
3493
3494 /*
3495 * This specialized allocator has to be a macro for its allocations to be
3496 * accounted separately (to have a separate alloc_tag).
3497 */
3498 #define dev_alloc_pages(_order) __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, _order)
3499
3500 /**
3501 * __dev_alloc_page - allocate a page for network Rx
3502 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3503 *
3504 * Allocate a new page.
3505 *
3506 * %NULL is returned if there is no free memory.
3507 */
__dev_alloc_page_noprof(gfp_t gfp_mask)3508 static inline struct page *__dev_alloc_page_noprof(gfp_t gfp_mask)
3509 {
3510 return __dev_alloc_pages_noprof(gfp_mask, 0);
3511 }
3512 #define __dev_alloc_page(...) alloc_hooks(__dev_alloc_page_noprof(__VA_ARGS__))
3513
3514 /*
3515 * This specialized allocator has to be a macro for its allocations to be
3516 * accounted separately (to have a separate alloc_tag).
3517 */
3518 #define dev_alloc_page() dev_alloc_pages(0)
3519
3520 /**
3521 * dev_page_is_reusable - check whether a page can be reused for network Rx
3522 * @page: the page to test
3523 *
3524 * A page shouldn't be considered for reusing/recycling if it was allocated
3525 * under memory pressure or at a distant memory node.
3526 *
3527 * Returns: false if this page should be returned to page allocator, true
3528 * otherwise.
3529 */
dev_page_is_reusable(const struct page * page)3530 static inline bool dev_page_is_reusable(const struct page *page)
3531 {
3532 return likely(page_to_nid(page) == numa_mem_id() &&
3533 !page_is_pfmemalloc(page));
3534 }
3535
3536 /**
3537 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3538 * @page: The page that was allocated from skb_alloc_page
3539 * @skb: The skb that may need pfmemalloc set
3540 */
skb_propagate_pfmemalloc(const struct page * page,struct sk_buff * skb)3541 static inline void skb_propagate_pfmemalloc(const struct page *page,
3542 struct sk_buff *skb)
3543 {
3544 if (page_is_pfmemalloc(page))
3545 skb->pfmemalloc = true;
3546 }
3547
3548 /**
3549 * skb_frag_off() - Returns the offset of a skb fragment
3550 * @frag: the paged fragment
3551 */
skb_frag_off(const skb_frag_t * frag)3552 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3553 {
3554 return frag->offset;
3555 }
3556
3557 /**
3558 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3559 * @frag: skb fragment
3560 * @delta: value to add
3561 */
skb_frag_off_add(skb_frag_t * frag,int delta)3562 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3563 {
3564 frag->offset += delta;
3565 }
3566
3567 /**
3568 * skb_frag_off_set() - Sets the offset of a skb fragment
3569 * @frag: skb fragment
3570 * @offset: offset of fragment
3571 */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)3572 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3573 {
3574 frag->offset = offset;
3575 }
3576
3577 /**
3578 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3579 * @fragto: skb fragment where offset is set
3580 * @fragfrom: skb fragment offset is copied from
3581 */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3582 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3583 const skb_frag_t *fragfrom)
3584 {
3585 fragto->offset = fragfrom->offset;
3586 }
3587
3588 /* Return: true if the skb_frag contains a net_iov. */
skb_frag_is_net_iov(const skb_frag_t * frag)3589 static inline bool skb_frag_is_net_iov(const skb_frag_t *frag)
3590 {
3591 return netmem_is_net_iov(frag->netmem);
3592 }
3593
3594 /**
3595 * skb_frag_net_iov - retrieve the net_iov referred to by fragment
3596 * @frag: the fragment
3597 *
3598 * Return: the &struct net_iov associated with @frag. Returns NULL if this
3599 * frag has no associated net_iov.
3600 */
skb_frag_net_iov(const skb_frag_t * frag)3601 static inline struct net_iov *skb_frag_net_iov(const skb_frag_t *frag)
3602 {
3603 if (!skb_frag_is_net_iov(frag))
3604 return NULL;
3605
3606 return netmem_to_net_iov(frag->netmem);
3607 }
3608
3609 /**
3610 * skb_frag_page - retrieve the page referred to by a paged fragment
3611 * @frag: the paged fragment
3612 *
3613 * Return: the &struct page associated with @frag. Returns NULL if this frag
3614 * has no associated page.
3615 */
skb_frag_page(const skb_frag_t * frag)3616 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3617 {
3618 if (skb_frag_is_net_iov(frag))
3619 return NULL;
3620
3621 return netmem_to_page(frag->netmem);
3622 }
3623
3624 /**
3625 * skb_frag_netmem - retrieve the netmem referred to by a fragment
3626 * @frag: the fragment
3627 *
3628 * Return: the &netmem_ref associated with @frag.
3629 */
skb_frag_netmem(const skb_frag_t * frag)3630 static inline netmem_ref skb_frag_netmem(const skb_frag_t *frag)
3631 {
3632 return frag->netmem;
3633 }
3634
3635 int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb,
3636 unsigned int headroom);
3637 int skb_cow_data_for_xdp(struct page_pool *pool, struct sk_buff **pskb,
3638 const struct bpf_prog *prog);
3639
3640 /**
3641 * skb_frag_address - gets the address of the data contained in a paged fragment
3642 * @frag: the paged fragment buffer
3643 *
3644 * Returns: the address of the data within @frag. The page must already
3645 * be mapped.
3646 */
skb_frag_address(const skb_frag_t * frag)3647 static inline void *skb_frag_address(const skb_frag_t *frag)
3648 {
3649 if (!skb_frag_page(frag))
3650 return NULL;
3651
3652 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3653 }
3654
3655 /**
3656 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3657 * @frag: the paged fragment buffer
3658 *
3659 * Returns: the address of the data within @frag. Checks that the page
3660 * is mapped and returns %NULL otherwise.
3661 */
skb_frag_address_safe(const skb_frag_t * frag)3662 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3663 {
3664 void *ptr = page_address(skb_frag_page(frag));
3665 if (unlikely(!ptr))
3666 return NULL;
3667
3668 return ptr + skb_frag_off(frag);
3669 }
3670
3671 /**
3672 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3673 * @fragto: skb fragment where page is set
3674 * @fragfrom: skb fragment page is copied from
3675 */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3676 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3677 const skb_frag_t *fragfrom)
3678 {
3679 fragto->netmem = fragfrom->netmem;
3680 }
3681
3682 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3683
3684 /**
3685 * __skb_frag_dma_map - maps a paged fragment via the DMA API
3686 * @dev: the device to map the fragment to
3687 * @frag: the paged fragment to map
3688 * @offset: the offset within the fragment (starting at the
3689 * fragment's own offset)
3690 * @size: the number of bytes to map
3691 * @dir: the direction of the mapping (``PCI_DMA_*``)
3692 *
3693 * Maps the page associated with @frag to @device.
3694 */
__skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3695 static inline dma_addr_t __skb_frag_dma_map(struct device *dev,
3696 const skb_frag_t *frag,
3697 size_t offset, size_t size,
3698 enum dma_data_direction dir)
3699 {
3700 return dma_map_page(dev, skb_frag_page(frag),
3701 skb_frag_off(frag) + offset, size, dir);
3702 }
3703
3704 #define skb_frag_dma_map(dev, frag, ...) \
3705 CONCATENATE(_skb_frag_dma_map, \
3706 COUNT_ARGS(__VA_ARGS__))(dev, frag, ##__VA_ARGS__)
3707
3708 #define __skb_frag_dma_map1(dev, frag, offset, uf, uo) ({ \
3709 const skb_frag_t *uf = (frag); \
3710 size_t uo = (offset); \
3711 \
3712 __skb_frag_dma_map(dev, uf, uo, skb_frag_size(uf) - uo, \
3713 DMA_TO_DEVICE); \
3714 })
3715 #define _skb_frag_dma_map1(dev, frag, offset) \
3716 __skb_frag_dma_map1(dev, frag, offset, __UNIQUE_ID(frag_), \
3717 __UNIQUE_ID(offset_))
3718 #define _skb_frag_dma_map0(dev, frag) \
3719 _skb_frag_dma_map1(dev, frag, 0)
3720 #define _skb_frag_dma_map2(dev, frag, offset, size) \
3721 __skb_frag_dma_map(dev, frag, offset, size, DMA_TO_DEVICE)
3722 #define _skb_frag_dma_map3(dev, frag, offset, size, dir) \
3723 __skb_frag_dma_map(dev, frag, offset, size, dir)
3724
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3725 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3726 gfp_t gfp_mask)
3727 {
3728 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3729 }
3730
3731
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3732 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3733 gfp_t gfp_mask)
3734 {
3735 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3736 }
3737
3738
3739 /**
3740 * skb_clone_writable - is the header of a clone writable
3741 * @skb: buffer to check
3742 * @len: length up to which to write
3743 *
3744 * Returns true if modifying the header part of the cloned buffer
3745 * does not requires the data to be copied.
3746 */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3747 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3748 {
3749 return !skb_header_cloned(skb) &&
3750 skb_headroom(skb) + len <= skb->hdr_len;
3751 }
3752
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3753 static inline int skb_try_make_writable(struct sk_buff *skb,
3754 unsigned int write_len)
3755 {
3756 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3757 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3758 }
3759
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3760 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3761 int cloned)
3762 {
3763 int delta = 0;
3764
3765 if (headroom > skb_headroom(skb))
3766 delta = headroom - skb_headroom(skb);
3767
3768 if (delta || cloned)
3769 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3770 GFP_ATOMIC);
3771 return 0;
3772 }
3773
3774 /**
3775 * skb_cow - copy header of skb when it is required
3776 * @skb: buffer to cow
3777 * @headroom: needed headroom
3778 *
3779 * If the skb passed lacks sufficient headroom or its data part
3780 * is shared, data is reallocated. If reallocation fails, an error
3781 * is returned and original skb is not changed.
3782 *
3783 * The result is skb with writable area skb->head...skb->tail
3784 * and at least @headroom of space at head.
3785 */
skb_cow(struct sk_buff * skb,unsigned int headroom)3786 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3787 {
3788 return __skb_cow(skb, headroom, skb_cloned(skb));
3789 }
3790
3791 /**
3792 * skb_cow_head - skb_cow but only making the head writable
3793 * @skb: buffer to cow
3794 * @headroom: needed headroom
3795 *
3796 * This function is identical to skb_cow except that we replace the
3797 * skb_cloned check by skb_header_cloned. It should be used when
3798 * you only need to push on some header and do not need to modify
3799 * the data.
3800 */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3801 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3802 {
3803 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3804 }
3805
3806 /**
3807 * skb_padto - pad an skbuff up to a minimal size
3808 * @skb: buffer to pad
3809 * @len: minimal length
3810 *
3811 * Pads up a buffer to ensure the trailing bytes exist and are
3812 * blanked. If the buffer already contains sufficient data it
3813 * is untouched. Otherwise it is extended. Returns zero on
3814 * success. The skb is freed on error.
3815 */
skb_padto(struct sk_buff * skb,unsigned int len)3816 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3817 {
3818 unsigned int size = skb->len;
3819 if (likely(size >= len))
3820 return 0;
3821 return skb_pad(skb, len - size);
3822 }
3823
3824 /**
3825 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
3826 * @skb: buffer to pad
3827 * @len: minimal length
3828 * @free_on_error: free buffer on error
3829 *
3830 * Pads up a buffer to ensure the trailing bytes exist and are
3831 * blanked. If the buffer already contains sufficient data it
3832 * is untouched. Otherwise it is extended. Returns zero on
3833 * success. The skb is freed on error if @free_on_error is true.
3834 */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3835 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3836 unsigned int len,
3837 bool free_on_error)
3838 {
3839 unsigned int size = skb->len;
3840
3841 if (unlikely(size < len)) {
3842 len -= size;
3843 if (__skb_pad(skb, len, free_on_error))
3844 return -ENOMEM;
3845 __skb_put(skb, len);
3846 }
3847 return 0;
3848 }
3849
3850 /**
3851 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3852 * @skb: buffer to pad
3853 * @len: minimal length
3854 *
3855 * Pads up a buffer to ensure the trailing bytes exist and are
3856 * blanked. If the buffer already contains sufficient data it
3857 * is untouched. Otherwise it is extended. Returns zero on
3858 * success. The skb is freed on error.
3859 */
skb_put_padto(struct sk_buff * skb,unsigned int len)3860 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3861 {
3862 return __skb_put_padto(skb, len, true);
3863 }
3864
3865 bool csum_and_copy_from_iter_full(void *addr, size_t bytes, __wsum *csum, struct iov_iter *i)
3866 __must_check;
3867
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3868 static inline int skb_add_data(struct sk_buff *skb,
3869 struct iov_iter *from, int copy)
3870 {
3871 const int off = skb->len;
3872
3873 if (skb->ip_summed == CHECKSUM_NONE) {
3874 __wsum csum = 0;
3875 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3876 &csum, from)) {
3877 skb->csum = csum_block_add(skb->csum, csum, off);
3878 return 0;
3879 }
3880 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3881 return 0;
3882
3883 __skb_trim(skb, off);
3884 return -EFAULT;
3885 }
3886
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3887 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3888 const struct page *page, int off)
3889 {
3890 if (skb_zcopy(skb))
3891 return false;
3892 if (i) {
3893 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3894
3895 return page == skb_frag_page(frag) &&
3896 off == skb_frag_off(frag) + skb_frag_size(frag);
3897 }
3898 return false;
3899 }
3900
__skb_linearize(struct sk_buff * skb)3901 static inline int __skb_linearize(struct sk_buff *skb)
3902 {
3903 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3904 }
3905
3906 /**
3907 * skb_linearize - convert paged skb to linear one
3908 * @skb: buffer to linarize
3909 *
3910 * If there is no free memory -ENOMEM is returned, otherwise zero
3911 * is returned and the old skb data released.
3912 */
skb_linearize(struct sk_buff * skb)3913 static inline int skb_linearize(struct sk_buff *skb)
3914 {
3915 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3916 }
3917
3918 /**
3919 * skb_has_shared_frag - can any frag be overwritten
3920 * @skb: buffer to test
3921 *
3922 * Return: true if the skb has at least one frag that might be modified
3923 * by an external entity (as in vmsplice()/sendfile())
3924 */
skb_has_shared_frag(const struct sk_buff * skb)3925 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3926 {
3927 return skb_is_nonlinear(skb) &&
3928 skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3929 }
3930
3931 /**
3932 * skb_linearize_cow - make sure skb is linear and writable
3933 * @skb: buffer to process
3934 *
3935 * If there is no free memory -ENOMEM is returned, otherwise zero
3936 * is returned and the old skb data released.
3937 */
skb_linearize_cow(struct sk_buff * skb)3938 static inline int skb_linearize_cow(struct sk_buff *skb)
3939 {
3940 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3941 __skb_linearize(skb) : 0;
3942 }
3943
3944 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3945 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3946 unsigned int off)
3947 {
3948 if (skb->ip_summed == CHECKSUM_COMPLETE)
3949 skb->csum = csum_block_sub(skb->csum,
3950 csum_partial(start, len, 0), off);
3951 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3952 skb_checksum_start_offset(skb) < 0)
3953 skb->ip_summed = CHECKSUM_NONE;
3954 }
3955
3956 /**
3957 * skb_postpull_rcsum - update checksum for received skb after pull
3958 * @skb: buffer to update
3959 * @start: start of data before pull
3960 * @len: length of data pulled
3961 *
3962 * After doing a pull on a received packet, you need to call this to
3963 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3964 * CHECKSUM_NONE so that it can be recomputed from scratch.
3965 */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3966 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3967 const void *start, unsigned int len)
3968 {
3969 if (skb->ip_summed == CHECKSUM_COMPLETE)
3970 skb->csum = wsum_negate(csum_partial(start, len,
3971 wsum_negate(skb->csum)));
3972 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3973 skb_checksum_start_offset(skb) < 0)
3974 skb->ip_summed = CHECKSUM_NONE;
3975 }
3976
3977 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3978 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3979 unsigned int off)
3980 {
3981 if (skb->ip_summed == CHECKSUM_COMPLETE)
3982 skb->csum = csum_block_add(skb->csum,
3983 csum_partial(start, len, 0), off);
3984 }
3985
3986 /**
3987 * skb_postpush_rcsum - update checksum for received skb after push
3988 * @skb: buffer to update
3989 * @start: start of data after push
3990 * @len: length of data pushed
3991 *
3992 * After doing a push on a received packet, you need to call this to
3993 * update the CHECKSUM_COMPLETE checksum.
3994 */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3995 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3996 const void *start, unsigned int len)
3997 {
3998 __skb_postpush_rcsum(skb, start, len, 0);
3999 }
4000
4001 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
4002
4003 /**
4004 * skb_push_rcsum - push skb and update receive checksum
4005 * @skb: buffer to update
4006 * @len: length of data pulled
4007 *
4008 * This function performs an skb_push on the packet and updates
4009 * the CHECKSUM_COMPLETE checksum. It should be used on
4010 * receive path processing instead of skb_push unless you know
4011 * that the checksum difference is zero (e.g., a valid IP header)
4012 * or you are setting ip_summed to CHECKSUM_NONE.
4013 */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)4014 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
4015 {
4016 skb_push(skb, len);
4017 skb_postpush_rcsum(skb, skb->data, len);
4018 return skb->data;
4019 }
4020
4021 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
4022 /**
4023 * pskb_trim_rcsum - trim received skb and update checksum
4024 * @skb: buffer to trim
4025 * @len: new length
4026 *
4027 * This is exactly the same as pskb_trim except that it ensures the
4028 * checksum of received packets are still valid after the operation.
4029 * It can change skb pointers.
4030 */
4031
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)4032 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
4033 {
4034 skb_might_realloc(skb);
4035 if (likely(len >= skb->len))
4036 return 0;
4037 return pskb_trim_rcsum_slow(skb, len);
4038 }
4039
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)4040 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
4041 {
4042 if (skb->ip_summed == CHECKSUM_COMPLETE)
4043 skb->ip_summed = CHECKSUM_NONE;
4044 __skb_trim(skb, len);
4045 return 0;
4046 }
4047
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)4048 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
4049 {
4050 if (skb->ip_summed == CHECKSUM_COMPLETE)
4051 skb->ip_summed = CHECKSUM_NONE;
4052 return __skb_grow(skb, len);
4053 }
4054
4055 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
4056 #define skb_rb_first(root) rb_to_skb(rb_first(root))
4057 #define skb_rb_last(root) rb_to_skb(rb_last(root))
4058 #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
4059 #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
4060
4061 #define skb_queue_walk(queue, skb) \
4062 for (skb = (queue)->next; \
4063 skb != (struct sk_buff *)(queue); \
4064 skb = skb->next)
4065
4066 #define skb_queue_walk_safe(queue, skb, tmp) \
4067 for (skb = (queue)->next, tmp = skb->next; \
4068 skb != (struct sk_buff *)(queue); \
4069 skb = tmp, tmp = skb->next)
4070
4071 #define skb_queue_walk_from(queue, skb) \
4072 for (; skb != (struct sk_buff *)(queue); \
4073 skb = skb->next)
4074
4075 #define skb_rbtree_walk(skb, root) \
4076 for (skb = skb_rb_first(root); skb != NULL; \
4077 skb = skb_rb_next(skb))
4078
4079 #define skb_rbtree_walk_from(skb) \
4080 for (; skb != NULL; \
4081 skb = skb_rb_next(skb))
4082
4083 #define skb_rbtree_walk_from_safe(skb, tmp) \
4084 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
4085 skb = tmp)
4086
4087 #define skb_queue_walk_from_safe(queue, skb, tmp) \
4088 for (tmp = skb->next; \
4089 skb != (struct sk_buff *)(queue); \
4090 skb = tmp, tmp = skb->next)
4091
4092 #define skb_queue_reverse_walk(queue, skb) \
4093 for (skb = (queue)->prev; \
4094 skb != (struct sk_buff *)(queue); \
4095 skb = skb->prev)
4096
4097 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
4098 for (skb = (queue)->prev, tmp = skb->prev; \
4099 skb != (struct sk_buff *)(queue); \
4100 skb = tmp, tmp = skb->prev)
4101
4102 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
4103 for (tmp = skb->prev; \
4104 skb != (struct sk_buff *)(queue); \
4105 skb = tmp, tmp = skb->prev)
4106
skb_has_frag_list(const struct sk_buff * skb)4107 static inline bool skb_has_frag_list(const struct sk_buff *skb)
4108 {
4109 return skb_shinfo(skb)->frag_list != NULL;
4110 }
4111
skb_frag_list_init(struct sk_buff * skb)4112 static inline void skb_frag_list_init(struct sk_buff *skb)
4113 {
4114 skb_shinfo(skb)->frag_list = NULL;
4115 }
4116
4117 #define skb_walk_frags(skb, iter) \
4118 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
4119
4120
4121 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
4122 int *err, long *timeo_p,
4123 const struct sk_buff *skb);
4124 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
4125 struct sk_buff_head *queue,
4126 unsigned int flags,
4127 int *off, int *err,
4128 struct sk_buff **last);
4129 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
4130 struct sk_buff_head *queue,
4131 unsigned int flags, int *off, int *err,
4132 struct sk_buff **last);
4133 struct sk_buff *__skb_recv_datagram(struct sock *sk,
4134 struct sk_buff_head *sk_queue,
4135 unsigned int flags, int *off, int *err);
4136 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err);
4137 __poll_t datagram_poll(struct file *file, struct socket *sock,
4138 struct poll_table_struct *wait);
4139 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
4140 struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)4141 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
4142 struct msghdr *msg, int size)
4143 {
4144 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
4145 }
4146 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
4147 struct msghdr *msg);
4148 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
4149 struct iov_iter *to, int len,
4150 struct ahash_request *hash);
4151 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
4152 struct iov_iter *from, int len);
4153 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
4154 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
4155 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
4156 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
4157 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
4158 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
4159 int len);
4160 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
4161 struct pipe_inode_info *pipe, unsigned int len,
4162 unsigned int flags);
4163 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
4164 int len);
4165 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
4166 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
4167 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
4168 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
4169 int len, int hlen);
4170 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
4171 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
4172 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
4173 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
4174 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
4175 unsigned int offset);
4176 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
4177 int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len);
4178 int skb_ensure_writable_head_tail(struct sk_buff *skb, struct net_device *dev);
4179 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
4180 int skb_vlan_pop(struct sk_buff *skb);
4181 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
4182 int skb_eth_pop(struct sk_buff *skb);
4183 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
4184 const unsigned char *src);
4185 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
4186 int mac_len, bool ethernet);
4187 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
4188 bool ethernet);
4189 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
4190 int skb_mpls_dec_ttl(struct sk_buff *skb);
4191 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
4192 gfp_t gfp);
4193
memcpy_from_msg(void * data,struct msghdr * msg,int len)4194 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
4195 {
4196 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
4197 }
4198
memcpy_to_msg(struct msghdr * msg,void * data,int len)4199 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
4200 {
4201 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
4202 }
4203
4204 struct skb_checksum_ops {
4205 __wsum (*update)(const void *mem, int len, __wsum wsum);
4206 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
4207 };
4208
4209 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
4210
4211 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
4212 __wsum csum, const struct skb_checksum_ops *ops);
4213 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
4214 __wsum csum);
4215
4216 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,const void * data,int hlen,void * buffer)4217 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
4218 const void *data, int hlen, void *buffer)
4219 {
4220 if (likely(hlen - offset >= len))
4221 return (void *)data + offset;
4222
4223 if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
4224 return NULL;
4225
4226 return buffer;
4227 }
4228
4229 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)4230 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
4231 {
4232 return __skb_header_pointer(skb, offset, len, skb->data,
4233 skb_headlen(skb), buffer);
4234 }
4235
4236 static inline void * __must_check
skb_pointer_if_linear(const struct sk_buff * skb,int offset,int len)4237 skb_pointer_if_linear(const struct sk_buff *skb, int offset, int len)
4238 {
4239 if (likely(skb_headlen(skb) - offset >= len))
4240 return skb->data + offset;
4241 return NULL;
4242 }
4243
4244 /**
4245 * skb_needs_linearize - check if we need to linearize a given skb
4246 * depending on the given device features.
4247 * @skb: socket buffer to check
4248 * @features: net device features
4249 *
4250 * Returns true if either:
4251 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
4252 * 2. skb is fragmented and the device does not support SG.
4253 */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)4254 static inline bool skb_needs_linearize(struct sk_buff *skb,
4255 netdev_features_t features)
4256 {
4257 return skb_is_nonlinear(skb) &&
4258 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
4259 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
4260 }
4261
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)4262 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
4263 void *to,
4264 const unsigned int len)
4265 {
4266 memcpy(to, skb->data, len);
4267 }
4268
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)4269 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
4270 const int offset, void *to,
4271 const unsigned int len)
4272 {
4273 memcpy(to, skb->data + offset, len);
4274 }
4275
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)4276 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
4277 const void *from,
4278 const unsigned int len)
4279 {
4280 memcpy(skb->data, from, len);
4281 }
4282
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)4283 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
4284 const int offset,
4285 const void *from,
4286 const unsigned int len)
4287 {
4288 memcpy(skb->data + offset, from, len);
4289 }
4290
4291 void skb_init(void);
4292
skb_get_ktime(const struct sk_buff * skb)4293 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
4294 {
4295 return skb->tstamp;
4296 }
4297
4298 /**
4299 * skb_get_timestamp - get timestamp from a skb
4300 * @skb: skb to get stamp from
4301 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
4302 *
4303 * Timestamps are stored in the skb as offsets to a base timestamp.
4304 * This function converts the offset back to a struct timeval and stores
4305 * it in stamp.
4306 */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)4307 static inline void skb_get_timestamp(const struct sk_buff *skb,
4308 struct __kernel_old_timeval *stamp)
4309 {
4310 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
4311 }
4312
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)4313 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
4314 struct __kernel_sock_timeval *stamp)
4315 {
4316 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4317
4318 stamp->tv_sec = ts.tv_sec;
4319 stamp->tv_usec = ts.tv_nsec / 1000;
4320 }
4321
skb_get_timestampns(const struct sk_buff * skb,struct __kernel_old_timespec * stamp)4322 static inline void skb_get_timestampns(const struct sk_buff *skb,
4323 struct __kernel_old_timespec *stamp)
4324 {
4325 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4326
4327 stamp->tv_sec = ts.tv_sec;
4328 stamp->tv_nsec = ts.tv_nsec;
4329 }
4330
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)4331 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
4332 struct __kernel_timespec *stamp)
4333 {
4334 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4335
4336 stamp->tv_sec = ts.tv_sec;
4337 stamp->tv_nsec = ts.tv_nsec;
4338 }
4339
__net_timestamp(struct sk_buff * skb)4340 static inline void __net_timestamp(struct sk_buff *skb)
4341 {
4342 skb->tstamp = ktime_get_real();
4343 skb->tstamp_type = SKB_CLOCK_REALTIME;
4344 }
4345
net_timedelta(ktime_t t)4346 static inline ktime_t net_timedelta(ktime_t t)
4347 {
4348 return ktime_sub(ktime_get_real(), t);
4349 }
4350
skb_set_delivery_time(struct sk_buff * skb,ktime_t kt,u8 tstamp_type)4351 static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt,
4352 u8 tstamp_type)
4353 {
4354 skb->tstamp = kt;
4355
4356 if (kt)
4357 skb->tstamp_type = tstamp_type;
4358 else
4359 skb->tstamp_type = SKB_CLOCK_REALTIME;
4360 }
4361
skb_set_delivery_type_by_clockid(struct sk_buff * skb,ktime_t kt,clockid_t clockid)4362 static inline void skb_set_delivery_type_by_clockid(struct sk_buff *skb,
4363 ktime_t kt, clockid_t clockid)
4364 {
4365 u8 tstamp_type = SKB_CLOCK_REALTIME;
4366
4367 switch (clockid) {
4368 case CLOCK_REALTIME:
4369 break;
4370 case CLOCK_MONOTONIC:
4371 tstamp_type = SKB_CLOCK_MONOTONIC;
4372 break;
4373 case CLOCK_TAI:
4374 tstamp_type = SKB_CLOCK_TAI;
4375 break;
4376 default:
4377 WARN_ON_ONCE(1);
4378 kt = 0;
4379 }
4380
4381 skb_set_delivery_time(skb, kt, tstamp_type);
4382 }
4383
4384 DECLARE_STATIC_KEY_FALSE(netstamp_needed_key);
4385
4386 /* It is used in the ingress path to clear the delivery_time.
4387 * If needed, set the skb->tstamp to the (rcv) timestamp.
4388 */
skb_clear_delivery_time(struct sk_buff * skb)4389 static inline void skb_clear_delivery_time(struct sk_buff *skb)
4390 {
4391 if (skb->tstamp_type) {
4392 skb->tstamp_type = SKB_CLOCK_REALTIME;
4393 if (static_branch_unlikely(&netstamp_needed_key))
4394 skb->tstamp = ktime_get_real();
4395 else
4396 skb->tstamp = 0;
4397 }
4398 }
4399
skb_clear_tstamp(struct sk_buff * skb)4400 static inline void skb_clear_tstamp(struct sk_buff *skb)
4401 {
4402 if (skb->tstamp_type)
4403 return;
4404
4405 skb->tstamp = 0;
4406 }
4407
skb_tstamp(const struct sk_buff * skb)4408 static inline ktime_t skb_tstamp(const struct sk_buff *skb)
4409 {
4410 if (skb->tstamp_type)
4411 return 0;
4412
4413 return skb->tstamp;
4414 }
4415
skb_tstamp_cond(const struct sk_buff * skb,bool cond)4416 static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond)
4417 {
4418 if (skb->tstamp_type != SKB_CLOCK_MONOTONIC && skb->tstamp)
4419 return skb->tstamp;
4420
4421 if (static_branch_unlikely(&netstamp_needed_key) || cond)
4422 return ktime_get_real();
4423
4424 return 0;
4425 }
4426
skb_metadata_len(const struct sk_buff * skb)4427 static inline u8 skb_metadata_len(const struct sk_buff *skb)
4428 {
4429 return skb_shinfo(skb)->meta_len;
4430 }
4431
skb_metadata_end(const struct sk_buff * skb)4432 static inline void *skb_metadata_end(const struct sk_buff *skb)
4433 {
4434 return skb_mac_header(skb);
4435 }
4436
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)4437 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
4438 const struct sk_buff *skb_b,
4439 u8 meta_len)
4440 {
4441 const void *a = skb_metadata_end(skb_a);
4442 const void *b = skb_metadata_end(skb_b);
4443 u64 diffs = 0;
4444
4445 if (!IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) ||
4446 BITS_PER_LONG != 64)
4447 goto slow;
4448
4449 /* Using more efficient variant than plain call to memcmp(). */
4450 switch (meta_len) {
4451 #define __it(x, op) (x -= sizeof(u##op))
4452 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
4453 case 32: diffs |= __it_diff(a, b, 64);
4454 fallthrough;
4455 case 24: diffs |= __it_diff(a, b, 64);
4456 fallthrough;
4457 case 16: diffs |= __it_diff(a, b, 64);
4458 fallthrough;
4459 case 8: diffs |= __it_diff(a, b, 64);
4460 break;
4461 case 28: diffs |= __it_diff(a, b, 64);
4462 fallthrough;
4463 case 20: diffs |= __it_diff(a, b, 64);
4464 fallthrough;
4465 case 12: diffs |= __it_diff(a, b, 64);
4466 fallthrough;
4467 case 4: diffs |= __it_diff(a, b, 32);
4468 break;
4469 default:
4470 slow:
4471 return memcmp(a - meta_len, b - meta_len, meta_len);
4472 }
4473 return diffs;
4474 }
4475
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)4476 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
4477 const struct sk_buff *skb_b)
4478 {
4479 u8 len_a = skb_metadata_len(skb_a);
4480 u8 len_b = skb_metadata_len(skb_b);
4481
4482 if (!(len_a | len_b))
4483 return false;
4484
4485 return len_a != len_b ?
4486 true : __skb_metadata_differs(skb_a, skb_b, len_a);
4487 }
4488
skb_metadata_set(struct sk_buff * skb,u8 meta_len)4489 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
4490 {
4491 skb_shinfo(skb)->meta_len = meta_len;
4492 }
4493
skb_metadata_clear(struct sk_buff * skb)4494 static inline void skb_metadata_clear(struct sk_buff *skb)
4495 {
4496 skb_metadata_set(skb, 0);
4497 }
4498
4499 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
4500
4501 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
4502
4503 void skb_clone_tx_timestamp(struct sk_buff *skb);
4504 bool skb_defer_rx_timestamp(struct sk_buff *skb);
4505
4506 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
4507
skb_clone_tx_timestamp(struct sk_buff * skb)4508 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
4509 {
4510 }
4511
skb_defer_rx_timestamp(struct sk_buff * skb)4512 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
4513 {
4514 return false;
4515 }
4516
4517 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
4518
4519 /**
4520 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
4521 *
4522 * PHY drivers may accept clones of transmitted packets for
4523 * timestamping via their phy_driver.txtstamp method. These drivers
4524 * must call this function to return the skb back to the stack with a
4525 * timestamp.
4526 *
4527 * @skb: clone of the original outgoing packet
4528 * @hwtstamps: hardware time stamps
4529 *
4530 */
4531 void skb_complete_tx_timestamp(struct sk_buff *skb,
4532 struct skb_shared_hwtstamps *hwtstamps);
4533
4534 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4535 struct skb_shared_hwtstamps *hwtstamps,
4536 struct sock *sk, int tstype);
4537
4538 /**
4539 * skb_tstamp_tx - queue clone of skb with send time stamps
4540 * @orig_skb: the original outgoing packet
4541 * @hwtstamps: hardware time stamps, may be NULL if not available
4542 *
4543 * If the skb has a socket associated, then this function clones the
4544 * skb (thus sharing the actual data and optional structures), stores
4545 * the optional hardware time stamping information (if non NULL) or
4546 * generates a software time stamp (otherwise), then queues the clone
4547 * to the error queue of the socket. Errors are silently ignored.
4548 */
4549 void skb_tstamp_tx(struct sk_buff *orig_skb,
4550 struct skb_shared_hwtstamps *hwtstamps);
4551
4552 /**
4553 * skb_tx_timestamp() - Driver hook for transmit timestamping
4554 *
4555 * Ethernet MAC Drivers should call this function in their hard_xmit()
4556 * function immediately before giving the sk_buff to the MAC hardware.
4557 *
4558 * Specifically, one should make absolutely sure that this function is
4559 * called before TX completion of this packet can trigger. Otherwise
4560 * the packet could potentially already be freed.
4561 *
4562 * @skb: A socket buffer.
4563 */
skb_tx_timestamp(struct sk_buff * skb)4564 static inline void skb_tx_timestamp(struct sk_buff *skb)
4565 {
4566 skb_clone_tx_timestamp(skb);
4567 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4568 skb_tstamp_tx(skb, NULL);
4569 }
4570
4571 /**
4572 * skb_complete_wifi_ack - deliver skb with wifi status
4573 *
4574 * @skb: the original outgoing packet
4575 * @acked: ack status
4576 *
4577 */
4578 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4579
4580 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4581 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4582
skb_csum_unnecessary(const struct sk_buff * skb)4583 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4584 {
4585 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4586 skb->csum_valid ||
4587 (skb->ip_summed == CHECKSUM_PARTIAL &&
4588 skb_checksum_start_offset(skb) >= 0));
4589 }
4590
4591 /**
4592 * skb_checksum_complete - Calculate checksum of an entire packet
4593 * @skb: packet to process
4594 *
4595 * This function calculates the checksum over the entire packet plus
4596 * the value of skb->csum. The latter can be used to supply the
4597 * checksum of a pseudo header as used by TCP/UDP. It returns the
4598 * checksum.
4599 *
4600 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
4601 * this function can be used to verify that checksum on received
4602 * packets. In that case the function should return zero if the
4603 * checksum is correct. In particular, this function will return zero
4604 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4605 * hardware has already verified the correctness of the checksum.
4606 */
skb_checksum_complete(struct sk_buff * skb)4607 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4608 {
4609 return skb_csum_unnecessary(skb) ?
4610 0 : __skb_checksum_complete(skb);
4611 }
4612
__skb_decr_checksum_unnecessary(struct sk_buff * skb)4613 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4614 {
4615 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4616 if (skb->csum_level == 0)
4617 skb->ip_summed = CHECKSUM_NONE;
4618 else
4619 skb->csum_level--;
4620 }
4621 }
4622
__skb_incr_checksum_unnecessary(struct sk_buff * skb)4623 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4624 {
4625 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4626 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4627 skb->csum_level++;
4628 } else if (skb->ip_summed == CHECKSUM_NONE) {
4629 skb->ip_summed = CHECKSUM_UNNECESSARY;
4630 skb->csum_level = 0;
4631 }
4632 }
4633
__skb_reset_checksum_unnecessary(struct sk_buff * skb)4634 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4635 {
4636 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4637 skb->ip_summed = CHECKSUM_NONE;
4638 skb->csum_level = 0;
4639 }
4640 }
4641
4642 /* Check if we need to perform checksum complete validation.
4643 *
4644 * Returns: true if checksum complete is needed, false otherwise
4645 * (either checksum is unnecessary or zero checksum is allowed).
4646 */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)4647 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4648 bool zero_okay,
4649 __sum16 check)
4650 {
4651 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4652 skb->csum_valid = 1;
4653 __skb_decr_checksum_unnecessary(skb);
4654 return false;
4655 }
4656
4657 return true;
4658 }
4659
4660 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4661 * in checksum_init.
4662 */
4663 #define CHECKSUM_BREAK 76
4664
4665 /* Unset checksum-complete
4666 *
4667 * Unset checksum complete can be done when packet is being modified
4668 * (uncompressed for instance) and checksum-complete value is
4669 * invalidated.
4670 */
skb_checksum_complete_unset(struct sk_buff * skb)4671 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4672 {
4673 if (skb->ip_summed == CHECKSUM_COMPLETE)
4674 skb->ip_summed = CHECKSUM_NONE;
4675 }
4676
4677 /* Validate (init) checksum based on checksum complete.
4678 *
4679 * Return values:
4680 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4681 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4682 * checksum is stored in skb->csum for use in __skb_checksum_complete
4683 * non-zero: value of invalid checksum
4684 *
4685 */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)4686 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4687 bool complete,
4688 __wsum psum)
4689 {
4690 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4691 if (!csum_fold(csum_add(psum, skb->csum))) {
4692 skb->csum_valid = 1;
4693 return 0;
4694 }
4695 }
4696
4697 skb->csum = psum;
4698
4699 if (complete || skb->len <= CHECKSUM_BREAK) {
4700 __sum16 csum;
4701
4702 csum = __skb_checksum_complete(skb);
4703 skb->csum_valid = !csum;
4704 return csum;
4705 }
4706
4707 return 0;
4708 }
4709
null_compute_pseudo(struct sk_buff * skb,int proto)4710 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4711 {
4712 return 0;
4713 }
4714
4715 /* Perform checksum validate (init). Note that this is a macro since we only
4716 * want to calculate the pseudo header which is an input function if necessary.
4717 * First we try to validate without any computation (checksum unnecessary) and
4718 * then calculate based on checksum complete calling the function to compute
4719 * pseudo header.
4720 *
4721 * Return values:
4722 * 0: checksum is validated or try to in skb_checksum_complete
4723 * non-zero: value of invalid checksum
4724 */
4725 #define __skb_checksum_validate(skb, proto, complete, \
4726 zero_okay, check, compute_pseudo) \
4727 ({ \
4728 __sum16 __ret = 0; \
4729 skb->csum_valid = 0; \
4730 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4731 __ret = __skb_checksum_validate_complete(skb, \
4732 complete, compute_pseudo(skb, proto)); \
4733 __ret; \
4734 })
4735
4736 #define skb_checksum_init(skb, proto, compute_pseudo) \
4737 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4738
4739 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4740 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4741
4742 #define skb_checksum_validate(skb, proto, compute_pseudo) \
4743 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4744
4745 #define skb_checksum_validate_zero_check(skb, proto, check, \
4746 compute_pseudo) \
4747 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4748
4749 #define skb_checksum_simple_validate(skb) \
4750 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4751
__skb_checksum_convert_check(struct sk_buff * skb)4752 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4753 {
4754 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4755 }
4756
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4757 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4758 {
4759 skb->csum = ~pseudo;
4760 skb->ip_summed = CHECKSUM_COMPLETE;
4761 }
4762
4763 #define skb_checksum_try_convert(skb, proto, compute_pseudo) \
4764 do { \
4765 if (__skb_checksum_convert_check(skb)) \
4766 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4767 } while (0)
4768
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4769 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4770 u16 start, u16 offset)
4771 {
4772 skb->ip_summed = CHECKSUM_PARTIAL;
4773 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4774 skb->csum_offset = offset - start;
4775 }
4776
4777 /* Update skbuf and packet to reflect the remote checksum offload operation.
4778 * When called, ptr indicates the starting point for skb->csum when
4779 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4780 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4781 */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4782 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4783 int start, int offset, bool nopartial)
4784 {
4785 __wsum delta;
4786
4787 if (!nopartial) {
4788 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4789 return;
4790 }
4791
4792 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4793 __skb_checksum_complete(skb);
4794 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4795 }
4796
4797 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4798
4799 /* Adjust skb->csum since we changed the packet */
4800 skb->csum = csum_add(skb->csum, delta);
4801 }
4802
skb_nfct(const struct sk_buff * skb)4803 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4804 {
4805 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4806 return (void *)(skb->_nfct & NFCT_PTRMASK);
4807 #else
4808 return NULL;
4809 #endif
4810 }
4811
skb_get_nfct(const struct sk_buff * skb)4812 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4813 {
4814 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4815 return skb->_nfct;
4816 #else
4817 return 0UL;
4818 #endif
4819 }
4820
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4821 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4822 {
4823 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4824 skb->slow_gro |= !!nfct;
4825 skb->_nfct = nfct;
4826 #endif
4827 }
4828
4829 #ifdef CONFIG_SKB_EXTENSIONS
4830 enum skb_ext_id {
4831 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4832 SKB_EXT_BRIDGE_NF,
4833 #endif
4834 #ifdef CONFIG_XFRM
4835 SKB_EXT_SEC_PATH,
4836 #endif
4837 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4838 TC_SKB_EXT,
4839 #endif
4840 #if IS_ENABLED(CONFIG_MPTCP)
4841 SKB_EXT_MPTCP,
4842 #endif
4843 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
4844 SKB_EXT_MCTP,
4845 #endif
4846 SKB_EXT_NUM, /* must be last */
4847 };
4848
4849 /**
4850 * struct skb_ext - sk_buff extensions
4851 * @refcnt: 1 on allocation, deallocated on 0
4852 * @offset: offset to add to @data to obtain extension address
4853 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4854 * @data: start of extension data, variable sized
4855 *
4856 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4857 * to use 'u8' types while allowing up to 2kb worth of extension data.
4858 */
4859 struct skb_ext {
4860 refcount_t refcnt;
4861 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4862 u8 chunks; /* same */
4863 char data[] __aligned(8);
4864 };
4865
4866 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4867 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4868 struct skb_ext *ext);
4869 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4870 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4871 void __skb_ext_put(struct skb_ext *ext);
4872
skb_ext_put(struct sk_buff * skb)4873 static inline void skb_ext_put(struct sk_buff *skb)
4874 {
4875 if (skb->active_extensions)
4876 __skb_ext_put(skb->extensions);
4877 }
4878
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4879 static inline void __skb_ext_copy(struct sk_buff *dst,
4880 const struct sk_buff *src)
4881 {
4882 dst->active_extensions = src->active_extensions;
4883
4884 if (src->active_extensions) {
4885 struct skb_ext *ext = src->extensions;
4886
4887 refcount_inc(&ext->refcnt);
4888 dst->extensions = ext;
4889 }
4890 }
4891
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4892 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4893 {
4894 skb_ext_put(dst);
4895 __skb_ext_copy(dst, src);
4896 }
4897
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4898 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4899 {
4900 return !!ext->offset[i];
4901 }
4902
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4903 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4904 {
4905 return skb->active_extensions & (1 << id);
4906 }
4907
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4908 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4909 {
4910 if (skb_ext_exist(skb, id))
4911 __skb_ext_del(skb, id);
4912 }
4913
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4914 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4915 {
4916 if (skb_ext_exist(skb, id)) {
4917 struct skb_ext *ext = skb->extensions;
4918
4919 return (void *)ext + (ext->offset[id] << 3);
4920 }
4921
4922 return NULL;
4923 }
4924
skb_ext_reset(struct sk_buff * skb)4925 static inline void skb_ext_reset(struct sk_buff *skb)
4926 {
4927 if (unlikely(skb->active_extensions)) {
4928 __skb_ext_put(skb->extensions);
4929 skb->active_extensions = 0;
4930 }
4931 }
4932
skb_has_extensions(struct sk_buff * skb)4933 static inline bool skb_has_extensions(struct sk_buff *skb)
4934 {
4935 return unlikely(skb->active_extensions);
4936 }
4937 #else
skb_ext_put(struct sk_buff * skb)4938 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4939 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4940 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4941 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)4942 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4943 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4944 #endif /* CONFIG_SKB_EXTENSIONS */
4945
nf_reset_ct(struct sk_buff * skb)4946 static inline void nf_reset_ct(struct sk_buff *skb)
4947 {
4948 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4949 nf_conntrack_put(skb_nfct(skb));
4950 skb->_nfct = 0;
4951 #endif
4952 }
4953
nf_reset_trace(struct sk_buff * skb)4954 static inline void nf_reset_trace(struct sk_buff *skb)
4955 {
4956 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4957 skb->nf_trace = 0;
4958 #endif
4959 }
4960
ipvs_reset(struct sk_buff * skb)4961 static inline void ipvs_reset(struct sk_buff *skb)
4962 {
4963 #if IS_ENABLED(CONFIG_IP_VS)
4964 skb->ipvs_property = 0;
4965 #endif
4966 }
4967
4968 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4969 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4970 bool copy)
4971 {
4972 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4973 dst->_nfct = src->_nfct;
4974 nf_conntrack_get(skb_nfct(src));
4975 #endif
4976 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4977 if (copy)
4978 dst->nf_trace = src->nf_trace;
4979 #endif
4980 }
4981
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4982 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4983 {
4984 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4985 nf_conntrack_put(skb_nfct(dst));
4986 #endif
4987 dst->slow_gro = src->slow_gro;
4988 __nf_copy(dst, src, true);
4989 }
4990
4991 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4992 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4993 {
4994 to->secmark = from->secmark;
4995 }
4996
skb_init_secmark(struct sk_buff * skb)4997 static inline void skb_init_secmark(struct sk_buff *skb)
4998 {
4999 skb->secmark = 0;
5000 }
5001 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)5002 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
5003 { }
5004
skb_init_secmark(struct sk_buff * skb)5005 static inline void skb_init_secmark(struct sk_buff *skb)
5006 { }
5007 #endif
5008
secpath_exists(const struct sk_buff * skb)5009 static inline int secpath_exists(const struct sk_buff *skb)
5010 {
5011 #ifdef CONFIG_XFRM
5012 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
5013 #else
5014 return 0;
5015 #endif
5016 }
5017
skb_irq_freeable(const struct sk_buff * skb)5018 static inline bool skb_irq_freeable(const struct sk_buff *skb)
5019 {
5020 return !skb->destructor &&
5021 !secpath_exists(skb) &&
5022 !skb_nfct(skb) &&
5023 !skb->_skb_refdst &&
5024 !skb_has_frag_list(skb);
5025 }
5026
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)5027 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
5028 {
5029 skb->queue_mapping = queue_mapping;
5030 }
5031
skb_get_queue_mapping(const struct sk_buff * skb)5032 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
5033 {
5034 return skb->queue_mapping;
5035 }
5036
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)5037 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
5038 {
5039 to->queue_mapping = from->queue_mapping;
5040 }
5041
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)5042 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
5043 {
5044 skb->queue_mapping = rx_queue + 1;
5045 }
5046
skb_get_rx_queue(const struct sk_buff * skb)5047 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
5048 {
5049 return skb->queue_mapping - 1;
5050 }
5051
skb_rx_queue_recorded(const struct sk_buff * skb)5052 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
5053 {
5054 return skb->queue_mapping != 0;
5055 }
5056
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)5057 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
5058 {
5059 skb->dst_pending_confirm = val;
5060 }
5061
skb_get_dst_pending_confirm(const struct sk_buff * skb)5062 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
5063 {
5064 return skb->dst_pending_confirm != 0;
5065 }
5066
skb_sec_path(const struct sk_buff * skb)5067 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
5068 {
5069 #ifdef CONFIG_XFRM
5070 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
5071 #else
5072 return NULL;
5073 #endif
5074 }
5075
skb_is_gso(const struct sk_buff * skb)5076 static inline bool skb_is_gso(const struct sk_buff *skb)
5077 {
5078 return skb_shinfo(skb)->gso_size;
5079 }
5080
5081 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)5082 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
5083 {
5084 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
5085 }
5086
5087 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)5088 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
5089 {
5090 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
5091 }
5092
5093 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)5094 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
5095 {
5096 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
5097 }
5098
skb_gso_reset(struct sk_buff * skb)5099 static inline void skb_gso_reset(struct sk_buff *skb)
5100 {
5101 skb_shinfo(skb)->gso_size = 0;
5102 skb_shinfo(skb)->gso_segs = 0;
5103 skb_shinfo(skb)->gso_type = 0;
5104 }
5105
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)5106 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
5107 u16 increment)
5108 {
5109 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
5110 return;
5111 shinfo->gso_size += increment;
5112 }
5113
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)5114 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
5115 u16 decrement)
5116 {
5117 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
5118 return;
5119 shinfo->gso_size -= decrement;
5120 }
5121
5122 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
5123
skb_warn_if_lro(const struct sk_buff * skb)5124 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
5125 {
5126 /* LRO sets gso_size but not gso_type, whereas if GSO is really
5127 * wanted then gso_type will be set. */
5128 const struct skb_shared_info *shinfo = skb_shinfo(skb);
5129
5130 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
5131 unlikely(shinfo->gso_type == 0)) {
5132 __skb_warn_lro_forwarding(skb);
5133 return true;
5134 }
5135 return false;
5136 }
5137
skb_forward_csum(struct sk_buff * skb)5138 static inline void skb_forward_csum(struct sk_buff *skb)
5139 {
5140 /* Unfortunately we don't support this one. Any brave souls? */
5141 if (skb->ip_summed == CHECKSUM_COMPLETE)
5142 skb->ip_summed = CHECKSUM_NONE;
5143 }
5144
5145 /**
5146 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
5147 * @skb: skb to check
5148 *
5149 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
5150 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
5151 * use this helper, to document places where we make this assertion.
5152 */
skb_checksum_none_assert(const struct sk_buff * skb)5153 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
5154 {
5155 DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE);
5156 }
5157
5158 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
5159
5160 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
5161 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
5162 unsigned int transport_len,
5163 __sum16(*skb_chkf)(struct sk_buff *skb));
5164
5165 /**
5166 * skb_head_is_locked - Determine if the skb->head is locked down
5167 * @skb: skb to check
5168 *
5169 * The head on skbs build around a head frag can be removed if they are
5170 * not cloned. This function returns true if the skb head is locked down
5171 * due to either being allocated via kmalloc, or by being a clone with
5172 * multiple references to the head.
5173 */
skb_head_is_locked(const struct sk_buff * skb)5174 static inline bool skb_head_is_locked(const struct sk_buff *skb)
5175 {
5176 return !skb->head_frag || skb_cloned(skb);
5177 }
5178
5179 /* Local Checksum Offload.
5180 * Compute outer checksum based on the assumption that the
5181 * inner checksum will be offloaded later.
5182 * See Documentation/networking/checksum-offloads.rst for
5183 * explanation of how this works.
5184 * Fill in outer checksum adjustment (e.g. with sum of outer
5185 * pseudo-header) before calling.
5186 * Also ensure that inner checksum is in linear data area.
5187 */
lco_csum(struct sk_buff * skb)5188 static inline __wsum lco_csum(struct sk_buff *skb)
5189 {
5190 unsigned char *csum_start = skb_checksum_start(skb);
5191 unsigned char *l4_hdr = skb_transport_header(skb);
5192 __wsum partial;
5193
5194 /* Start with complement of inner checksum adjustment */
5195 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
5196 skb->csum_offset));
5197
5198 /* Add in checksum of our headers (incl. outer checksum
5199 * adjustment filled in by caller) and return result.
5200 */
5201 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
5202 }
5203
skb_is_redirected(const struct sk_buff * skb)5204 static inline bool skb_is_redirected(const struct sk_buff *skb)
5205 {
5206 return skb->redirected;
5207 }
5208
skb_set_redirected(struct sk_buff * skb,bool from_ingress)5209 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
5210 {
5211 skb->redirected = 1;
5212 #ifdef CONFIG_NET_REDIRECT
5213 skb->from_ingress = from_ingress;
5214 if (skb->from_ingress)
5215 skb_clear_tstamp(skb);
5216 #endif
5217 }
5218
skb_reset_redirect(struct sk_buff * skb)5219 static inline void skb_reset_redirect(struct sk_buff *skb)
5220 {
5221 skb->redirected = 0;
5222 }
5223
skb_set_redirected_noclear(struct sk_buff * skb,bool from_ingress)5224 static inline void skb_set_redirected_noclear(struct sk_buff *skb,
5225 bool from_ingress)
5226 {
5227 skb->redirected = 1;
5228 #ifdef CONFIG_NET_REDIRECT
5229 skb->from_ingress = from_ingress;
5230 #endif
5231 }
5232
skb_csum_is_sctp(struct sk_buff * skb)5233 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
5234 {
5235 #if IS_ENABLED(CONFIG_IP_SCTP)
5236 return skb->csum_not_inet;
5237 #else
5238 return 0;
5239 #endif
5240 }
5241
skb_reset_csum_not_inet(struct sk_buff * skb)5242 static inline void skb_reset_csum_not_inet(struct sk_buff *skb)
5243 {
5244 skb->ip_summed = CHECKSUM_NONE;
5245 #if IS_ENABLED(CONFIG_IP_SCTP)
5246 skb->csum_not_inet = 0;
5247 #endif
5248 }
5249
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)5250 static inline void skb_set_kcov_handle(struct sk_buff *skb,
5251 const u64 kcov_handle)
5252 {
5253 #ifdef CONFIG_KCOV
5254 skb->kcov_handle = kcov_handle;
5255 #endif
5256 }
5257
skb_get_kcov_handle(struct sk_buff * skb)5258 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
5259 {
5260 #ifdef CONFIG_KCOV
5261 return skb->kcov_handle;
5262 #else
5263 return 0;
5264 #endif
5265 }
5266
skb_mark_for_recycle(struct sk_buff * skb)5267 static inline void skb_mark_for_recycle(struct sk_buff *skb)
5268 {
5269 #ifdef CONFIG_PAGE_POOL
5270 skb->pp_recycle = 1;
5271 #endif
5272 }
5273
5274 ssize_t skb_splice_from_iter(struct sk_buff *skb, struct iov_iter *iter,
5275 ssize_t maxsize, gfp_t gfp);
5276
5277 #endif /* __KERNEL__ */
5278 #endif /* _LINUX_SKBUFF_H */
5279