padata: use %*pb[l] to print bitmaps including cpumasks and nodemasks
[deliverable/linux.git] / include / linux / skbuff.h
1 /*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14 #ifndef _LINUX_SKBUFF_H
15 #define _LINUX_SKBUFF_H
16
17 #include <linux/kernel.h>
18 #include <linux/kmemcheck.h>
19 #include <linux/compiler.h>
20 #include <linux/time.h>
21 #include <linux/bug.h>
22 #include <linux/cache.h>
23 #include <linux/rbtree.h>
24 #include <linux/socket.h>
25
26 #include <linux/atomic.h>
27 #include <asm/types.h>
28 #include <linux/spinlock.h>
29 #include <linux/net.h>
30 #include <linux/textsearch.h>
31 #include <net/checksum.h>
32 #include <linux/rcupdate.h>
33 #include <linux/hrtimer.h>
34 #include <linux/dma-mapping.h>
35 #include <linux/netdev_features.h>
36 #include <linux/sched.h>
37 #include <net/flow_keys.h>
38
39 /* A. Checksumming of received packets by device.
40 *
41 * CHECKSUM_NONE:
42 *
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
46 *
47 * CHECKSUM_UNNECESSARY:
48 *
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
51 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
56 *
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
58 * TCP: IPv6 and IPv4.
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
64 *
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
74 *
75 * CHECKSUM_COMPLETE:
76 *
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
80 *
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
83 *
84 * CHECKSUM_PARTIAL:
85 *
86 * This is identical to the case for output below. This may occur on a packet
87 * received directly from another Linux OS, e.g., a virtualized Linux kernel
88 * on the same host. The packet can be treated in the same way as
89 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
90 * checksum must be filled in by the OS or the hardware.
91 *
92 * B. Checksumming on output.
93 *
94 * CHECKSUM_NONE:
95 *
96 * The skb was already checksummed by the protocol, or a checksum is not
97 * required.
98 *
99 * CHECKSUM_PARTIAL:
100 *
101 * The device is required to checksum the packet as seen by hard_start_xmit()
102 * from skb->csum_start up to the end, and to record/write the checksum at
103 * offset skb->csum_start + skb->csum_offset.
104 *
105 * The device must show its capabilities in dev->features, set up at device
106 * setup time, e.g. netdev_features.h:
107 *
108 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
109 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
110 * IPv4. Sigh. Vendors like this way for an unknown reason.
111 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
112 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
113 * NETIF_F_... - Well, you get the picture.
114 *
115 * CHECKSUM_UNNECESSARY:
116 *
117 * Normally, the device will do per protocol specific checksumming. Protocol
118 * implementations that do not want the NIC to perform the checksum
119 * calculation should use this flag in their outgoing skbs.
120 *
121 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
122 * offload. Correspondingly, the FCoE protocol driver
123 * stack should use CHECKSUM_UNNECESSARY.
124 *
125 * Any questions? No questions, good. --ANK
126 */
127
128 /* Don't change this without changing skb_csum_unnecessary! */
129 #define CHECKSUM_NONE 0
130 #define CHECKSUM_UNNECESSARY 1
131 #define CHECKSUM_COMPLETE 2
132 #define CHECKSUM_PARTIAL 3
133
134 /* Maximum value in skb->csum_level */
135 #define SKB_MAX_CSUM_LEVEL 3
136
137 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
138 #define SKB_WITH_OVERHEAD(X) \
139 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
140 #define SKB_MAX_ORDER(X, ORDER) \
141 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
142 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
143 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
144
145 /* return minimum truesize of one skb containing X bytes of data */
146 #define SKB_TRUESIZE(X) ((X) + \
147 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
148 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
149
150 struct net_device;
151 struct scatterlist;
152 struct pipe_inode_info;
153 struct iov_iter;
154 struct napi_struct;
155
156 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
157 struct nf_conntrack {
158 atomic_t use;
159 };
160 #endif
161
162 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
163 struct nf_bridge_info {
164 atomic_t use;
165 unsigned int mask;
166 struct net_device *physindev;
167 struct net_device *physoutdev;
168 unsigned long data[32 / sizeof(unsigned long)];
169 };
170 #endif
171
172 struct sk_buff_head {
173 /* These two members must be first. */
174 struct sk_buff *next;
175 struct sk_buff *prev;
176
177 __u32 qlen;
178 spinlock_t lock;
179 };
180
181 struct sk_buff;
182
183 /* To allow 64K frame to be packed as single skb without frag_list we
184 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
185 * buffers which do not start on a page boundary.
186 *
187 * Since GRO uses frags we allocate at least 16 regardless of page
188 * size.
189 */
190 #if (65536/PAGE_SIZE + 1) < 16
191 #define MAX_SKB_FRAGS 16UL
192 #else
193 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
194 #endif
195
196 typedef struct skb_frag_struct skb_frag_t;
197
198 struct skb_frag_struct {
199 struct {
200 struct page *p;
201 } page;
202 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
203 __u32 page_offset;
204 __u32 size;
205 #else
206 __u16 page_offset;
207 __u16 size;
208 #endif
209 };
210
211 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
212 {
213 return frag->size;
214 }
215
216 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
217 {
218 frag->size = size;
219 }
220
221 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
222 {
223 frag->size += delta;
224 }
225
226 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
227 {
228 frag->size -= delta;
229 }
230
231 #define HAVE_HW_TIME_STAMP
232
233 /**
234 * struct skb_shared_hwtstamps - hardware time stamps
235 * @hwtstamp: hardware time stamp transformed into duration
236 * since arbitrary point in time
237 *
238 * Software time stamps generated by ktime_get_real() are stored in
239 * skb->tstamp.
240 *
241 * hwtstamps can only be compared against other hwtstamps from
242 * the same device.
243 *
244 * This structure is attached to packets as part of the
245 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
246 */
247 struct skb_shared_hwtstamps {
248 ktime_t hwtstamp;
249 };
250
251 /* Definitions for tx_flags in struct skb_shared_info */
252 enum {
253 /* generate hardware time stamp */
254 SKBTX_HW_TSTAMP = 1 << 0,
255
256 /* generate software time stamp when queueing packet to NIC */
257 SKBTX_SW_TSTAMP = 1 << 1,
258
259 /* device driver is going to provide hardware time stamp */
260 SKBTX_IN_PROGRESS = 1 << 2,
261
262 /* device driver supports TX zero-copy buffers */
263 SKBTX_DEV_ZEROCOPY = 1 << 3,
264
265 /* generate wifi status information (where possible) */
266 SKBTX_WIFI_STATUS = 1 << 4,
267
268 /* This indicates at least one fragment might be overwritten
269 * (as in vmsplice(), sendfile() ...)
270 * If we need to compute a TX checksum, we'll need to copy
271 * all frags to avoid possible bad checksum
272 */
273 SKBTX_SHARED_FRAG = 1 << 5,
274
275 /* generate software time stamp when entering packet scheduling */
276 SKBTX_SCHED_TSTAMP = 1 << 6,
277
278 /* generate software timestamp on peer data acknowledgment */
279 SKBTX_ACK_TSTAMP = 1 << 7,
280 };
281
282 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
283 SKBTX_SCHED_TSTAMP | \
284 SKBTX_ACK_TSTAMP)
285 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
286
287 /*
288 * The callback notifies userspace to release buffers when skb DMA is done in
289 * lower device, the skb last reference should be 0 when calling this.
290 * The zerocopy_success argument is true if zero copy transmit occurred,
291 * false on data copy or out of memory error caused by data copy attempt.
292 * The ctx field is used to track device context.
293 * The desc field is used to track userspace buffer index.
294 */
295 struct ubuf_info {
296 void (*callback)(struct ubuf_info *, bool zerocopy_success);
297 void *ctx;
298 unsigned long desc;
299 };
300
301 /* This data is invariant across clones and lives at
302 * the end of the header data, ie. at skb->end.
303 */
304 struct skb_shared_info {
305 unsigned char nr_frags;
306 __u8 tx_flags;
307 unsigned short gso_size;
308 /* Warning: this field is not always filled in (UFO)! */
309 unsigned short gso_segs;
310 unsigned short gso_type;
311 struct sk_buff *frag_list;
312 struct skb_shared_hwtstamps hwtstamps;
313 u32 tskey;
314 __be32 ip6_frag_id;
315
316 /*
317 * Warning : all fields before dataref are cleared in __alloc_skb()
318 */
319 atomic_t dataref;
320
321 /* Intermediate layers must ensure that destructor_arg
322 * remains valid until skb destructor */
323 void * destructor_arg;
324
325 /* must be last field, see pskb_expand_head() */
326 skb_frag_t frags[MAX_SKB_FRAGS];
327 };
328
329 /* We divide dataref into two halves. The higher 16 bits hold references
330 * to the payload part of skb->data. The lower 16 bits hold references to
331 * the entire skb->data. A clone of a headerless skb holds the length of
332 * the header in skb->hdr_len.
333 *
334 * All users must obey the rule that the skb->data reference count must be
335 * greater than or equal to the payload reference count.
336 *
337 * Holding a reference to the payload part means that the user does not
338 * care about modifications to the header part of skb->data.
339 */
340 #define SKB_DATAREF_SHIFT 16
341 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
342
343
344 enum {
345 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
346 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
347 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
348 };
349
350 enum {
351 SKB_GSO_TCPV4 = 1 << 0,
352 SKB_GSO_UDP = 1 << 1,
353
354 /* This indicates the skb is from an untrusted source. */
355 SKB_GSO_DODGY = 1 << 2,
356
357 /* This indicates the tcp segment has CWR set. */
358 SKB_GSO_TCP_ECN = 1 << 3,
359
360 SKB_GSO_TCPV6 = 1 << 4,
361
362 SKB_GSO_FCOE = 1 << 5,
363
364 SKB_GSO_GRE = 1 << 6,
365
366 SKB_GSO_GRE_CSUM = 1 << 7,
367
368 SKB_GSO_IPIP = 1 << 8,
369
370 SKB_GSO_SIT = 1 << 9,
371
372 SKB_GSO_UDP_TUNNEL = 1 << 10,
373
374 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
375
376 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
377 };
378
379 #if BITS_PER_LONG > 32
380 #define NET_SKBUFF_DATA_USES_OFFSET 1
381 #endif
382
383 #ifdef NET_SKBUFF_DATA_USES_OFFSET
384 typedef unsigned int sk_buff_data_t;
385 #else
386 typedef unsigned char *sk_buff_data_t;
387 #endif
388
389 /**
390 * struct skb_mstamp - multi resolution time stamps
391 * @stamp_us: timestamp in us resolution
392 * @stamp_jiffies: timestamp in jiffies
393 */
394 struct skb_mstamp {
395 union {
396 u64 v64;
397 struct {
398 u32 stamp_us;
399 u32 stamp_jiffies;
400 };
401 };
402 };
403
404 /**
405 * skb_mstamp_get - get current timestamp
406 * @cl: place to store timestamps
407 */
408 static inline void skb_mstamp_get(struct skb_mstamp *cl)
409 {
410 u64 val = local_clock();
411
412 do_div(val, NSEC_PER_USEC);
413 cl->stamp_us = (u32)val;
414 cl->stamp_jiffies = (u32)jiffies;
415 }
416
417 /**
418 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
419 * @t1: pointer to newest sample
420 * @t0: pointer to oldest sample
421 */
422 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
423 const struct skb_mstamp *t0)
424 {
425 s32 delta_us = t1->stamp_us - t0->stamp_us;
426 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
427
428 /* If delta_us is negative, this might be because interval is too big,
429 * or local_clock() drift is too big : fallback using jiffies.
430 */
431 if (delta_us <= 0 ||
432 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
433
434 delta_us = jiffies_to_usecs(delta_jiffies);
435
436 return delta_us;
437 }
438
439
440 /**
441 * struct sk_buff - socket buffer
442 * @next: Next buffer in list
443 * @prev: Previous buffer in list
444 * @tstamp: Time we arrived/left
445 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
446 * @sk: Socket we are owned by
447 * @dev: Device we arrived on/are leaving by
448 * @cb: Control buffer. Free for use by every layer. Put private vars here
449 * @_skb_refdst: destination entry (with norefcount bit)
450 * @sp: the security path, used for xfrm
451 * @len: Length of actual data
452 * @data_len: Data length
453 * @mac_len: Length of link layer header
454 * @hdr_len: writable header length of cloned skb
455 * @csum: Checksum (must include start/offset pair)
456 * @csum_start: Offset from skb->head where checksumming should start
457 * @csum_offset: Offset from csum_start where checksum should be stored
458 * @priority: Packet queueing priority
459 * @ignore_df: allow local fragmentation
460 * @cloned: Head may be cloned (check refcnt to be sure)
461 * @ip_summed: Driver fed us an IP checksum
462 * @nohdr: Payload reference only, must not modify header
463 * @nfctinfo: Relationship of this skb to the connection
464 * @pkt_type: Packet class
465 * @fclone: skbuff clone status
466 * @ipvs_property: skbuff is owned by ipvs
467 * @peeked: this packet has been seen already, so stats have been
468 * done for it, don't do them again
469 * @nf_trace: netfilter packet trace flag
470 * @protocol: Packet protocol from driver
471 * @destructor: Destruct function
472 * @nfct: Associated connection, if any
473 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
474 * @skb_iif: ifindex of device we arrived on
475 * @tc_index: Traffic control index
476 * @tc_verd: traffic control verdict
477 * @hash: the packet hash
478 * @queue_mapping: Queue mapping for multiqueue devices
479 * @xmit_more: More SKBs are pending for this queue
480 * @ndisc_nodetype: router type (from link layer)
481 * @ooo_okay: allow the mapping of a socket to a queue to be changed
482 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
483 * ports.
484 * @sw_hash: indicates hash was computed in software stack
485 * @wifi_acked_valid: wifi_acked was set
486 * @wifi_acked: whether frame was acked on wifi or not
487 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
488 * @napi_id: id of the NAPI struct this skb came from
489 * @secmark: security marking
490 * @mark: Generic packet mark
491 * @dropcount: total number of sk_receive_queue overflows
492 * @vlan_proto: vlan encapsulation protocol
493 * @vlan_tci: vlan tag control information
494 * @inner_protocol: Protocol (encapsulation)
495 * @inner_transport_header: Inner transport layer header (encapsulation)
496 * @inner_network_header: Network layer header (encapsulation)
497 * @inner_mac_header: Link layer header (encapsulation)
498 * @transport_header: Transport layer header
499 * @network_header: Network layer header
500 * @mac_header: Link layer header
501 * @tail: Tail pointer
502 * @end: End pointer
503 * @head: Head of buffer
504 * @data: Data head pointer
505 * @truesize: Buffer size
506 * @users: User count - see {datagram,tcp}.c
507 */
508
509 struct sk_buff {
510 union {
511 struct {
512 /* These two members must be first. */
513 struct sk_buff *next;
514 struct sk_buff *prev;
515
516 union {
517 ktime_t tstamp;
518 struct skb_mstamp skb_mstamp;
519 };
520 };
521 struct rb_node rbnode; /* used in netem & tcp stack */
522 };
523 struct sock *sk;
524 struct net_device *dev;
525
526 /*
527 * This is the control buffer. It is free to use for every
528 * layer. Please put your private variables there. If you
529 * want to keep them across layers you have to do a skb_clone()
530 * first. This is owned by whoever has the skb queued ATM.
531 */
532 char cb[48] __aligned(8);
533
534 unsigned long _skb_refdst;
535 void (*destructor)(struct sk_buff *skb);
536 #ifdef CONFIG_XFRM
537 struct sec_path *sp;
538 #endif
539 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
540 struct nf_conntrack *nfct;
541 #endif
542 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
543 struct nf_bridge_info *nf_bridge;
544 #endif
545 unsigned int len,
546 data_len;
547 __u16 mac_len,
548 hdr_len;
549
550 /* Following fields are _not_ copied in __copy_skb_header()
551 * Note that queue_mapping is here mostly to fill a hole.
552 */
553 kmemcheck_bitfield_begin(flags1);
554 __u16 queue_mapping;
555 __u8 cloned:1,
556 nohdr:1,
557 fclone:2,
558 peeked:1,
559 head_frag:1,
560 xmit_more:1;
561 /* one bit hole */
562 kmemcheck_bitfield_end(flags1);
563
564 /* fields enclosed in headers_start/headers_end are copied
565 * using a single memcpy() in __copy_skb_header()
566 */
567 /* private: */
568 __u32 headers_start[0];
569 /* public: */
570
571 /* if you move pkt_type around you also must adapt those constants */
572 #ifdef __BIG_ENDIAN_BITFIELD
573 #define PKT_TYPE_MAX (7 << 5)
574 #else
575 #define PKT_TYPE_MAX 7
576 #endif
577 #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
578
579 __u8 __pkt_type_offset[0];
580 __u8 pkt_type:3;
581 __u8 pfmemalloc:1;
582 __u8 ignore_df:1;
583 __u8 nfctinfo:3;
584
585 __u8 nf_trace:1;
586 __u8 ip_summed:2;
587 __u8 ooo_okay:1;
588 __u8 l4_hash:1;
589 __u8 sw_hash:1;
590 __u8 wifi_acked_valid:1;
591 __u8 wifi_acked:1;
592
593 __u8 no_fcs:1;
594 /* Indicates the inner headers are valid in the skbuff. */
595 __u8 encapsulation:1;
596 __u8 encap_hdr_csum:1;
597 __u8 csum_valid:1;
598 __u8 csum_complete_sw:1;
599 __u8 csum_level:2;
600 __u8 csum_bad:1;
601
602 #ifdef CONFIG_IPV6_NDISC_NODETYPE
603 __u8 ndisc_nodetype:2;
604 #endif
605 __u8 ipvs_property:1;
606 __u8 inner_protocol_type:1;
607 __u8 remcsum_offload:1;
608 /* 3 or 5 bit hole */
609
610 #ifdef CONFIG_NET_SCHED
611 __u16 tc_index; /* traffic control index */
612 #ifdef CONFIG_NET_CLS_ACT
613 __u16 tc_verd; /* traffic control verdict */
614 #endif
615 #endif
616
617 union {
618 __wsum csum;
619 struct {
620 __u16 csum_start;
621 __u16 csum_offset;
622 };
623 };
624 __u32 priority;
625 int skb_iif;
626 __u32 hash;
627 __be16 vlan_proto;
628 __u16 vlan_tci;
629 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
630 union {
631 unsigned int napi_id;
632 unsigned int sender_cpu;
633 };
634 #endif
635 #ifdef CONFIG_NETWORK_SECMARK
636 __u32 secmark;
637 #endif
638 union {
639 __u32 mark;
640 __u32 dropcount;
641 __u32 reserved_tailroom;
642 };
643
644 union {
645 __be16 inner_protocol;
646 __u8 inner_ipproto;
647 };
648
649 __u16 inner_transport_header;
650 __u16 inner_network_header;
651 __u16 inner_mac_header;
652
653 __be16 protocol;
654 __u16 transport_header;
655 __u16 network_header;
656 __u16 mac_header;
657
658 /* private: */
659 __u32 headers_end[0];
660 /* public: */
661
662 /* These elements must be at the end, see alloc_skb() for details. */
663 sk_buff_data_t tail;
664 sk_buff_data_t end;
665 unsigned char *head,
666 *data;
667 unsigned int truesize;
668 atomic_t users;
669 };
670
671 #ifdef __KERNEL__
672 /*
673 * Handling routines are only of interest to the kernel
674 */
675 #include <linux/slab.h>
676
677
678 #define SKB_ALLOC_FCLONE 0x01
679 #define SKB_ALLOC_RX 0x02
680 #define SKB_ALLOC_NAPI 0x04
681
682 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
683 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
684 {
685 return unlikely(skb->pfmemalloc);
686 }
687
688 /*
689 * skb might have a dst pointer attached, refcounted or not.
690 * _skb_refdst low order bit is set if refcount was _not_ taken
691 */
692 #define SKB_DST_NOREF 1UL
693 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
694
695 /**
696 * skb_dst - returns skb dst_entry
697 * @skb: buffer
698 *
699 * Returns skb dst_entry, regardless of reference taken or not.
700 */
701 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
702 {
703 /* If refdst was not refcounted, check we still are in a
704 * rcu_read_lock section
705 */
706 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
707 !rcu_read_lock_held() &&
708 !rcu_read_lock_bh_held());
709 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
710 }
711
712 /**
713 * skb_dst_set - sets skb dst
714 * @skb: buffer
715 * @dst: dst entry
716 *
717 * Sets skb dst, assuming a reference was taken on dst and should
718 * be released by skb_dst_drop()
719 */
720 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
721 {
722 skb->_skb_refdst = (unsigned long)dst;
723 }
724
725 /**
726 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
727 * @skb: buffer
728 * @dst: dst entry
729 *
730 * Sets skb dst, assuming a reference was not taken on dst.
731 * If dst entry is cached, we do not take reference and dst_release
732 * will be avoided by refdst_drop. If dst entry is not cached, we take
733 * reference, so that last dst_release can destroy the dst immediately.
734 */
735 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
736 {
737 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
738 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
739 }
740
741 /**
742 * skb_dst_is_noref - Test if skb dst isn't refcounted
743 * @skb: buffer
744 */
745 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
746 {
747 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
748 }
749
750 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
751 {
752 return (struct rtable *)skb_dst(skb);
753 }
754
755 void kfree_skb(struct sk_buff *skb);
756 void kfree_skb_list(struct sk_buff *segs);
757 void skb_tx_error(struct sk_buff *skb);
758 void consume_skb(struct sk_buff *skb);
759 void __kfree_skb(struct sk_buff *skb);
760 extern struct kmem_cache *skbuff_head_cache;
761
762 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
763 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
764 bool *fragstolen, int *delta_truesize);
765
766 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
767 int node);
768 struct sk_buff *build_skb(void *data, unsigned int frag_size);
769 static inline struct sk_buff *alloc_skb(unsigned int size,
770 gfp_t priority)
771 {
772 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
773 }
774
775 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
776 unsigned long data_len,
777 int max_page_order,
778 int *errcode,
779 gfp_t gfp_mask);
780
781 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
782 struct sk_buff_fclones {
783 struct sk_buff skb1;
784
785 struct sk_buff skb2;
786
787 atomic_t fclone_ref;
788 };
789
790 /**
791 * skb_fclone_busy - check if fclone is busy
792 * @skb: buffer
793 *
794 * Returns true is skb is a fast clone, and its clone is not freed.
795 * Some drivers call skb_orphan() in their ndo_start_xmit(),
796 * so we also check that this didnt happen.
797 */
798 static inline bool skb_fclone_busy(const struct sock *sk,
799 const struct sk_buff *skb)
800 {
801 const struct sk_buff_fclones *fclones;
802
803 fclones = container_of(skb, struct sk_buff_fclones, skb1);
804
805 return skb->fclone == SKB_FCLONE_ORIG &&
806 atomic_read(&fclones->fclone_ref) > 1 &&
807 fclones->skb2.sk == sk;
808 }
809
810 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
811 gfp_t priority)
812 {
813 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
814 }
815
816 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
817 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
818 {
819 return __alloc_skb_head(priority, -1);
820 }
821
822 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
823 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
824 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
825 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
826 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
827 gfp_t gfp_mask, bool fclone);
828 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
829 gfp_t gfp_mask)
830 {
831 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
832 }
833
834 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
835 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
836 unsigned int headroom);
837 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
838 int newtailroom, gfp_t priority);
839 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
840 int offset, int len);
841 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
842 int len);
843 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
844 int skb_pad(struct sk_buff *skb, int pad);
845 #define dev_kfree_skb(a) consume_skb(a)
846
847 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
848 int getfrag(void *from, char *to, int offset,
849 int len, int odd, struct sk_buff *skb),
850 void *from, int length);
851
852 struct skb_seq_state {
853 __u32 lower_offset;
854 __u32 upper_offset;
855 __u32 frag_idx;
856 __u32 stepped_offset;
857 struct sk_buff *root_skb;
858 struct sk_buff *cur_skb;
859 __u8 *frag_data;
860 };
861
862 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
863 unsigned int to, struct skb_seq_state *st);
864 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
865 struct skb_seq_state *st);
866 void skb_abort_seq_read(struct skb_seq_state *st);
867
868 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
869 unsigned int to, struct ts_config *config,
870 struct ts_state *state);
871
872 /*
873 * Packet hash types specify the type of hash in skb_set_hash.
874 *
875 * Hash types refer to the protocol layer addresses which are used to
876 * construct a packet's hash. The hashes are used to differentiate or identify
877 * flows of the protocol layer for the hash type. Hash types are either
878 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
879 *
880 * Properties of hashes:
881 *
882 * 1) Two packets in different flows have different hash values
883 * 2) Two packets in the same flow should have the same hash value
884 *
885 * A hash at a higher layer is considered to be more specific. A driver should
886 * set the most specific hash possible.
887 *
888 * A driver cannot indicate a more specific hash than the layer at which a hash
889 * was computed. For instance an L3 hash cannot be set as an L4 hash.
890 *
891 * A driver may indicate a hash level which is less specific than the
892 * actual layer the hash was computed on. For instance, a hash computed
893 * at L4 may be considered an L3 hash. This should only be done if the
894 * driver can't unambiguously determine that the HW computed the hash at
895 * the higher layer. Note that the "should" in the second property above
896 * permits this.
897 */
898 enum pkt_hash_types {
899 PKT_HASH_TYPE_NONE, /* Undefined type */
900 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
901 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
902 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
903 };
904
905 static inline void
906 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
907 {
908 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
909 skb->sw_hash = 0;
910 skb->hash = hash;
911 }
912
913 void __skb_get_hash(struct sk_buff *skb);
914 static inline __u32 skb_get_hash(struct sk_buff *skb)
915 {
916 if (!skb->l4_hash && !skb->sw_hash)
917 __skb_get_hash(skb);
918
919 return skb->hash;
920 }
921
922 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
923 {
924 return skb->hash;
925 }
926
927 static inline void skb_clear_hash(struct sk_buff *skb)
928 {
929 skb->hash = 0;
930 skb->sw_hash = 0;
931 skb->l4_hash = 0;
932 }
933
934 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
935 {
936 if (!skb->l4_hash)
937 skb_clear_hash(skb);
938 }
939
940 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
941 {
942 to->hash = from->hash;
943 to->sw_hash = from->sw_hash;
944 to->l4_hash = from->l4_hash;
945 };
946
947 #ifdef NET_SKBUFF_DATA_USES_OFFSET
948 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
949 {
950 return skb->head + skb->end;
951 }
952
953 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
954 {
955 return skb->end;
956 }
957 #else
958 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
959 {
960 return skb->end;
961 }
962
963 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
964 {
965 return skb->end - skb->head;
966 }
967 #endif
968
969 /* Internal */
970 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
971
972 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
973 {
974 return &skb_shinfo(skb)->hwtstamps;
975 }
976
977 /**
978 * skb_queue_empty - check if a queue is empty
979 * @list: queue head
980 *
981 * Returns true if the queue is empty, false otherwise.
982 */
983 static inline int skb_queue_empty(const struct sk_buff_head *list)
984 {
985 return list->next == (const struct sk_buff *) list;
986 }
987
988 /**
989 * skb_queue_is_last - check if skb is the last entry in the queue
990 * @list: queue head
991 * @skb: buffer
992 *
993 * Returns true if @skb is the last buffer on the list.
994 */
995 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
996 const struct sk_buff *skb)
997 {
998 return skb->next == (const struct sk_buff *) list;
999 }
1000
1001 /**
1002 * skb_queue_is_first - check if skb is the first entry in the queue
1003 * @list: queue head
1004 * @skb: buffer
1005 *
1006 * Returns true if @skb is the first buffer on the list.
1007 */
1008 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1009 const struct sk_buff *skb)
1010 {
1011 return skb->prev == (const struct sk_buff *) list;
1012 }
1013
1014 /**
1015 * skb_queue_next - return the next packet in the queue
1016 * @list: queue head
1017 * @skb: current buffer
1018 *
1019 * Return the next packet in @list after @skb. It is only valid to
1020 * call this if skb_queue_is_last() evaluates to false.
1021 */
1022 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1023 const struct sk_buff *skb)
1024 {
1025 /* This BUG_ON may seem severe, but if we just return then we
1026 * are going to dereference garbage.
1027 */
1028 BUG_ON(skb_queue_is_last(list, skb));
1029 return skb->next;
1030 }
1031
1032 /**
1033 * skb_queue_prev - return the prev packet in the queue
1034 * @list: queue head
1035 * @skb: current buffer
1036 *
1037 * Return the prev packet in @list before @skb. It is only valid to
1038 * call this if skb_queue_is_first() evaluates to false.
1039 */
1040 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1041 const struct sk_buff *skb)
1042 {
1043 /* This BUG_ON may seem severe, but if we just return then we
1044 * are going to dereference garbage.
1045 */
1046 BUG_ON(skb_queue_is_first(list, skb));
1047 return skb->prev;
1048 }
1049
1050 /**
1051 * skb_get - reference buffer
1052 * @skb: buffer to reference
1053 *
1054 * Makes another reference to a socket buffer and returns a pointer
1055 * to the buffer.
1056 */
1057 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1058 {
1059 atomic_inc(&skb->users);
1060 return skb;
1061 }
1062
1063 /*
1064 * If users == 1, we are the only owner and are can avoid redundant
1065 * atomic change.
1066 */
1067
1068 /**
1069 * skb_cloned - is the buffer a clone
1070 * @skb: buffer to check
1071 *
1072 * Returns true if the buffer was generated with skb_clone() and is
1073 * one of multiple shared copies of the buffer. Cloned buffers are
1074 * shared data so must not be written to under normal circumstances.
1075 */
1076 static inline int skb_cloned(const struct sk_buff *skb)
1077 {
1078 return skb->cloned &&
1079 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1080 }
1081
1082 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1083 {
1084 might_sleep_if(pri & __GFP_WAIT);
1085
1086 if (skb_cloned(skb))
1087 return pskb_expand_head(skb, 0, 0, pri);
1088
1089 return 0;
1090 }
1091
1092 /**
1093 * skb_header_cloned - is the header a clone
1094 * @skb: buffer to check
1095 *
1096 * Returns true if modifying the header part of the buffer requires
1097 * the data to be copied.
1098 */
1099 static inline int skb_header_cloned(const struct sk_buff *skb)
1100 {
1101 int dataref;
1102
1103 if (!skb->cloned)
1104 return 0;
1105
1106 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1107 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1108 return dataref != 1;
1109 }
1110
1111 /**
1112 * skb_header_release - release reference to header
1113 * @skb: buffer to operate on
1114 *
1115 * Drop a reference to the header part of the buffer. This is done
1116 * by acquiring a payload reference. You must not read from the header
1117 * part of skb->data after this.
1118 * Note : Check if you can use __skb_header_release() instead.
1119 */
1120 static inline void skb_header_release(struct sk_buff *skb)
1121 {
1122 BUG_ON(skb->nohdr);
1123 skb->nohdr = 1;
1124 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1125 }
1126
1127 /**
1128 * __skb_header_release - release reference to header
1129 * @skb: buffer to operate on
1130 *
1131 * Variant of skb_header_release() assuming skb is private to caller.
1132 * We can avoid one atomic operation.
1133 */
1134 static inline void __skb_header_release(struct sk_buff *skb)
1135 {
1136 skb->nohdr = 1;
1137 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1138 }
1139
1140
1141 /**
1142 * skb_shared - is the buffer shared
1143 * @skb: buffer to check
1144 *
1145 * Returns true if more than one person has a reference to this
1146 * buffer.
1147 */
1148 static inline int skb_shared(const struct sk_buff *skb)
1149 {
1150 return atomic_read(&skb->users) != 1;
1151 }
1152
1153 /**
1154 * skb_share_check - check if buffer is shared and if so clone it
1155 * @skb: buffer to check
1156 * @pri: priority for memory allocation
1157 *
1158 * If the buffer is shared the buffer is cloned and the old copy
1159 * drops a reference. A new clone with a single reference is returned.
1160 * If the buffer is not shared the original buffer is returned. When
1161 * being called from interrupt status or with spinlocks held pri must
1162 * be GFP_ATOMIC.
1163 *
1164 * NULL is returned on a memory allocation failure.
1165 */
1166 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1167 {
1168 might_sleep_if(pri & __GFP_WAIT);
1169 if (skb_shared(skb)) {
1170 struct sk_buff *nskb = skb_clone(skb, pri);
1171
1172 if (likely(nskb))
1173 consume_skb(skb);
1174 else
1175 kfree_skb(skb);
1176 skb = nskb;
1177 }
1178 return skb;
1179 }
1180
1181 /*
1182 * Copy shared buffers into a new sk_buff. We effectively do COW on
1183 * packets to handle cases where we have a local reader and forward
1184 * and a couple of other messy ones. The normal one is tcpdumping
1185 * a packet thats being forwarded.
1186 */
1187
1188 /**
1189 * skb_unshare - make a copy of a shared buffer
1190 * @skb: buffer to check
1191 * @pri: priority for memory allocation
1192 *
1193 * If the socket buffer is a clone then this function creates a new
1194 * copy of the data, drops a reference count on the old copy and returns
1195 * the new copy with the reference count at 1. If the buffer is not a clone
1196 * the original buffer is returned. When called with a spinlock held or
1197 * from interrupt state @pri must be %GFP_ATOMIC
1198 *
1199 * %NULL is returned on a memory allocation failure.
1200 */
1201 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1202 gfp_t pri)
1203 {
1204 might_sleep_if(pri & __GFP_WAIT);
1205 if (skb_cloned(skb)) {
1206 struct sk_buff *nskb = skb_copy(skb, pri);
1207
1208 /* Free our shared copy */
1209 if (likely(nskb))
1210 consume_skb(skb);
1211 else
1212 kfree_skb(skb);
1213 skb = nskb;
1214 }
1215 return skb;
1216 }
1217
1218 /**
1219 * skb_peek - peek at the head of an &sk_buff_head
1220 * @list_: list to peek at
1221 *
1222 * Peek an &sk_buff. Unlike most other operations you _MUST_
1223 * be careful with this one. A peek leaves the buffer on the
1224 * list and someone else may run off with it. You must hold
1225 * the appropriate locks or have a private queue to do this.
1226 *
1227 * Returns %NULL for an empty list or a pointer to the head element.
1228 * The reference count is not incremented and the reference is therefore
1229 * volatile. Use with caution.
1230 */
1231 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1232 {
1233 struct sk_buff *skb = list_->next;
1234
1235 if (skb == (struct sk_buff *)list_)
1236 skb = NULL;
1237 return skb;
1238 }
1239
1240 /**
1241 * skb_peek_next - peek skb following the given one from a queue
1242 * @skb: skb to start from
1243 * @list_: list to peek at
1244 *
1245 * Returns %NULL when the end of the list is met or a pointer to the
1246 * next element. The reference count is not incremented and the
1247 * reference is therefore volatile. Use with caution.
1248 */
1249 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1250 const struct sk_buff_head *list_)
1251 {
1252 struct sk_buff *next = skb->next;
1253
1254 if (next == (struct sk_buff *)list_)
1255 next = NULL;
1256 return next;
1257 }
1258
1259 /**
1260 * skb_peek_tail - peek at the tail of an &sk_buff_head
1261 * @list_: list to peek at
1262 *
1263 * Peek an &sk_buff. Unlike most other operations you _MUST_
1264 * be careful with this one. A peek leaves the buffer on the
1265 * list and someone else may run off with it. You must hold
1266 * the appropriate locks or have a private queue to do this.
1267 *
1268 * Returns %NULL for an empty list or a pointer to the tail element.
1269 * The reference count is not incremented and the reference is therefore
1270 * volatile. Use with caution.
1271 */
1272 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1273 {
1274 struct sk_buff *skb = list_->prev;
1275
1276 if (skb == (struct sk_buff *)list_)
1277 skb = NULL;
1278 return skb;
1279
1280 }
1281
1282 /**
1283 * skb_queue_len - get queue length
1284 * @list_: list to measure
1285 *
1286 * Return the length of an &sk_buff queue.
1287 */
1288 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1289 {
1290 return list_->qlen;
1291 }
1292
1293 /**
1294 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1295 * @list: queue to initialize
1296 *
1297 * This initializes only the list and queue length aspects of
1298 * an sk_buff_head object. This allows to initialize the list
1299 * aspects of an sk_buff_head without reinitializing things like
1300 * the spinlock. It can also be used for on-stack sk_buff_head
1301 * objects where the spinlock is known to not be used.
1302 */
1303 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1304 {
1305 list->prev = list->next = (struct sk_buff *)list;
1306 list->qlen = 0;
1307 }
1308
1309 /*
1310 * This function creates a split out lock class for each invocation;
1311 * this is needed for now since a whole lot of users of the skb-queue
1312 * infrastructure in drivers have different locking usage (in hardirq)
1313 * than the networking core (in softirq only). In the long run either the
1314 * network layer or drivers should need annotation to consolidate the
1315 * main types of usage into 3 classes.
1316 */
1317 static inline void skb_queue_head_init(struct sk_buff_head *list)
1318 {
1319 spin_lock_init(&list->lock);
1320 __skb_queue_head_init(list);
1321 }
1322
1323 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1324 struct lock_class_key *class)
1325 {
1326 skb_queue_head_init(list);
1327 lockdep_set_class(&list->lock, class);
1328 }
1329
1330 /*
1331 * Insert an sk_buff on a list.
1332 *
1333 * The "__skb_xxxx()" functions are the non-atomic ones that
1334 * can only be called with interrupts disabled.
1335 */
1336 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1337 struct sk_buff_head *list);
1338 static inline void __skb_insert(struct sk_buff *newsk,
1339 struct sk_buff *prev, struct sk_buff *next,
1340 struct sk_buff_head *list)
1341 {
1342 newsk->next = next;
1343 newsk->prev = prev;
1344 next->prev = prev->next = newsk;
1345 list->qlen++;
1346 }
1347
1348 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1349 struct sk_buff *prev,
1350 struct sk_buff *next)
1351 {
1352 struct sk_buff *first = list->next;
1353 struct sk_buff *last = list->prev;
1354
1355 first->prev = prev;
1356 prev->next = first;
1357
1358 last->next = next;
1359 next->prev = last;
1360 }
1361
1362 /**
1363 * skb_queue_splice - join two skb lists, this is designed for stacks
1364 * @list: the new list to add
1365 * @head: the place to add it in the first list
1366 */
1367 static inline void skb_queue_splice(const struct sk_buff_head *list,
1368 struct sk_buff_head *head)
1369 {
1370 if (!skb_queue_empty(list)) {
1371 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1372 head->qlen += list->qlen;
1373 }
1374 }
1375
1376 /**
1377 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1378 * @list: the new list to add
1379 * @head: the place to add it in the first list
1380 *
1381 * The list at @list is reinitialised
1382 */
1383 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1384 struct sk_buff_head *head)
1385 {
1386 if (!skb_queue_empty(list)) {
1387 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1388 head->qlen += list->qlen;
1389 __skb_queue_head_init(list);
1390 }
1391 }
1392
1393 /**
1394 * skb_queue_splice_tail - join two skb lists, each list being a queue
1395 * @list: the new list to add
1396 * @head: the place to add it in the first list
1397 */
1398 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1399 struct sk_buff_head *head)
1400 {
1401 if (!skb_queue_empty(list)) {
1402 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1403 head->qlen += list->qlen;
1404 }
1405 }
1406
1407 /**
1408 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1409 * @list: the new list to add
1410 * @head: the place to add it in the first list
1411 *
1412 * Each of the lists is a queue.
1413 * The list at @list is reinitialised
1414 */
1415 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1416 struct sk_buff_head *head)
1417 {
1418 if (!skb_queue_empty(list)) {
1419 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1420 head->qlen += list->qlen;
1421 __skb_queue_head_init(list);
1422 }
1423 }
1424
1425 /**
1426 * __skb_queue_after - queue a buffer at the list head
1427 * @list: list to use
1428 * @prev: place after this buffer
1429 * @newsk: buffer to queue
1430 *
1431 * Queue a buffer int the middle of a list. This function takes no locks
1432 * and you must therefore hold required locks before calling it.
1433 *
1434 * A buffer cannot be placed on two lists at the same time.
1435 */
1436 static inline void __skb_queue_after(struct sk_buff_head *list,
1437 struct sk_buff *prev,
1438 struct sk_buff *newsk)
1439 {
1440 __skb_insert(newsk, prev, prev->next, list);
1441 }
1442
1443 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1444 struct sk_buff_head *list);
1445
1446 static inline void __skb_queue_before(struct sk_buff_head *list,
1447 struct sk_buff *next,
1448 struct sk_buff *newsk)
1449 {
1450 __skb_insert(newsk, next->prev, next, list);
1451 }
1452
1453 /**
1454 * __skb_queue_head - queue a buffer at the list head
1455 * @list: list to use
1456 * @newsk: buffer to queue
1457 *
1458 * Queue a buffer at the start of a list. This function takes no locks
1459 * and you must therefore hold required locks before calling it.
1460 *
1461 * A buffer cannot be placed on two lists at the same time.
1462 */
1463 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1464 static inline void __skb_queue_head(struct sk_buff_head *list,
1465 struct sk_buff *newsk)
1466 {
1467 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1468 }
1469
1470 /**
1471 * __skb_queue_tail - queue a buffer at the list tail
1472 * @list: list to use
1473 * @newsk: buffer to queue
1474 *
1475 * Queue a buffer at the end of a list. This function takes no locks
1476 * and you must therefore hold required locks before calling it.
1477 *
1478 * A buffer cannot be placed on two lists at the same time.
1479 */
1480 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1481 static inline void __skb_queue_tail(struct sk_buff_head *list,
1482 struct sk_buff *newsk)
1483 {
1484 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1485 }
1486
1487 /*
1488 * remove sk_buff from list. _Must_ be called atomically, and with
1489 * the list known..
1490 */
1491 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1492 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1493 {
1494 struct sk_buff *next, *prev;
1495
1496 list->qlen--;
1497 next = skb->next;
1498 prev = skb->prev;
1499 skb->next = skb->prev = NULL;
1500 next->prev = prev;
1501 prev->next = next;
1502 }
1503
1504 /**
1505 * __skb_dequeue - remove from the head of the queue
1506 * @list: list to dequeue from
1507 *
1508 * Remove the head of the list. This function does not take any locks
1509 * so must be used with appropriate locks held only. The head item is
1510 * returned or %NULL if the list is empty.
1511 */
1512 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1513 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1514 {
1515 struct sk_buff *skb = skb_peek(list);
1516 if (skb)
1517 __skb_unlink(skb, list);
1518 return skb;
1519 }
1520
1521 /**
1522 * __skb_dequeue_tail - remove from the tail of the queue
1523 * @list: list to dequeue from
1524 *
1525 * Remove the tail of the list. This function does not take any locks
1526 * so must be used with appropriate locks held only. The tail item is
1527 * returned or %NULL if the list is empty.
1528 */
1529 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1530 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1531 {
1532 struct sk_buff *skb = skb_peek_tail(list);
1533 if (skb)
1534 __skb_unlink(skb, list);
1535 return skb;
1536 }
1537
1538
1539 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1540 {
1541 return skb->data_len;
1542 }
1543
1544 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1545 {
1546 return skb->len - skb->data_len;
1547 }
1548
1549 static inline int skb_pagelen(const struct sk_buff *skb)
1550 {
1551 int i, len = 0;
1552
1553 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1554 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1555 return len + skb_headlen(skb);
1556 }
1557
1558 /**
1559 * __skb_fill_page_desc - initialise a paged fragment in an skb
1560 * @skb: buffer containing fragment to be initialised
1561 * @i: paged fragment index to initialise
1562 * @page: the page to use for this fragment
1563 * @off: the offset to the data with @page
1564 * @size: the length of the data
1565 *
1566 * Initialises the @i'th fragment of @skb to point to &size bytes at
1567 * offset @off within @page.
1568 *
1569 * Does not take any additional reference on the fragment.
1570 */
1571 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1572 struct page *page, int off, int size)
1573 {
1574 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1575
1576 /*
1577 * Propagate page->pfmemalloc to the skb if we can. The problem is
1578 * that not all callers have unique ownership of the page. If
1579 * pfmemalloc is set, we check the mapping as a mapping implies
1580 * page->index is set (index and pfmemalloc share space).
1581 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1582 * do not lose pfmemalloc information as the pages would not be
1583 * allocated using __GFP_MEMALLOC.
1584 */
1585 frag->page.p = page;
1586 frag->page_offset = off;
1587 skb_frag_size_set(frag, size);
1588
1589 page = compound_head(page);
1590 if (page->pfmemalloc && !page->mapping)
1591 skb->pfmemalloc = true;
1592 }
1593
1594 /**
1595 * skb_fill_page_desc - initialise a paged fragment in an skb
1596 * @skb: buffer containing fragment to be initialised
1597 * @i: paged fragment index to initialise
1598 * @page: the page to use for this fragment
1599 * @off: the offset to the data with @page
1600 * @size: the length of the data
1601 *
1602 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1603 * @skb to point to @size bytes at offset @off within @page. In
1604 * addition updates @skb such that @i is the last fragment.
1605 *
1606 * Does not take any additional reference on the fragment.
1607 */
1608 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1609 struct page *page, int off, int size)
1610 {
1611 __skb_fill_page_desc(skb, i, page, off, size);
1612 skb_shinfo(skb)->nr_frags = i + 1;
1613 }
1614
1615 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1616 int size, unsigned int truesize);
1617
1618 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1619 unsigned int truesize);
1620
1621 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1622 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1623 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1624
1625 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1626 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1627 {
1628 return skb->head + skb->tail;
1629 }
1630
1631 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1632 {
1633 skb->tail = skb->data - skb->head;
1634 }
1635
1636 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1637 {
1638 skb_reset_tail_pointer(skb);
1639 skb->tail += offset;
1640 }
1641
1642 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1643 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1644 {
1645 return skb->tail;
1646 }
1647
1648 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1649 {
1650 skb->tail = skb->data;
1651 }
1652
1653 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1654 {
1655 skb->tail = skb->data + offset;
1656 }
1657
1658 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1659
1660 /*
1661 * Add data to an sk_buff
1662 */
1663 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1664 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1665 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1666 {
1667 unsigned char *tmp = skb_tail_pointer(skb);
1668 SKB_LINEAR_ASSERT(skb);
1669 skb->tail += len;
1670 skb->len += len;
1671 return tmp;
1672 }
1673
1674 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1675 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1676 {
1677 skb->data -= len;
1678 skb->len += len;
1679 return skb->data;
1680 }
1681
1682 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1683 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1684 {
1685 skb->len -= len;
1686 BUG_ON(skb->len < skb->data_len);
1687 return skb->data += len;
1688 }
1689
1690 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1691 {
1692 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1693 }
1694
1695 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1696
1697 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1698 {
1699 if (len > skb_headlen(skb) &&
1700 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1701 return NULL;
1702 skb->len -= len;
1703 return skb->data += len;
1704 }
1705
1706 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1707 {
1708 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1709 }
1710
1711 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1712 {
1713 if (likely(len <= skb_headlen(skb)))
1714 return 1;
1715 if (unlikely(len > skb->len))
1716 return 0;
1717 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1718 }
1719
1720 /**
1721 * skb_headroom - bytes at buffer head
1722 * @skb: buffer to check
1723 *
1724 * Return the number of bytes of free space at the head of an &sk_buff.
1725 */
1726 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1727 {
1728 return skb->data - skb->head;
1729 }
1730
1731 /**
1732 * skb_tailroom - bytes at buffer end
1733 * @skb: buffer to check
1734 *
1735 * Return the number of bytes of free space at the tail of an sk_buff
1736 */
1737 static inline int skb_tailroom(const struct sk_buff *skb)
1738 {
1739 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1740 }
1741
1742 /**
1743 * skb_availroom - bytes at buffer end
1744 * @skb: buffer to check
1745 *
1746 * Return the number of bytes of free space at the tail of an sk_buff
1747 * allocated by sk_stream_alloc()
1748 */
1749 static inline int skb_availroom(const struct sk_buff *skb)
1750 {
1751 if (skb_is_nonlinear(skb))
1752 return 0;
1753
1754 return skb->end - skb->tail - skb->reserved_tailroom;
1755 }
1756
1757 /**
1758 * skb_reserve - adjust headroom
1759 * @skb: buffer to alter
1760 * @len: bytes to move
1761 *
1762 * Increase the headroom of an empty &sk_buff by reducing the tail
1763 * room. This is only allowed for an empty buffer.
1764 */
1765 static inline void skb_reserve(struct sk_buff *skb, int len)
1766 {
1767 skb->data += len;
1768 skb->tail += len;
1769 }
1770
1771 #define ENCAP_TYPE_ETHER 0
1772 #define ENCAP_TYPE_IPPROTO 1
1773
1774 static inline void skb_set_inner_protocol(struct sk_buff *skb,
1775 __be16 protocol)
1776 {
1777 skb->inner_protocol = protocol;
1778 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1779 }
1780
1781 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1782 __u8 ipproto)
1783 {
1784 skb->inner_ipproto = ipproto;
1785 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1786 }
1787
1788 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1789 {
1790 skb->inner_mac_header = skb->mac_header;
1791 skb->inner_network_header = skb->network_header;
1792 skb->inner_transport_header = skb->transport_header;
1793 }
1794
1795 static inline void skb_reset_mac_len(struct sk_buff *skb)
1796 {
1797 skb->mac_len = skb->network_header - skb->mac_header;
1798 }
1799
1800 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1801 *skb)
1802 {
1803 return skb->head + skb->inner_transport_header;
1804 }
1805
1806 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1807 {
1808 skb->inner_transport_header = skb->data - skb->head;
1809 }
1810
1811 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1812 const int offset)
1813 {
1814 skb_reset_inner_transport_header(skb);
1815 skb->inner_transport_header += offset;
1816 }
1817
1818 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1819 {
1820 return skb->head + skb->inner_network_header;
1821 }
1822
1823 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1824 {
1825 skb->inner_network_header = skb->data - skb->head;
1826 }
1827
1828 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1829 const int offset)
1830 {
1831 skb_reset_inner_network_header(skb);
1832 skb->inner_network_header += offset;
1833 }
1834
1835 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1836 {
1837 return skb->head + skb->inner_mac_header;
1838 }
1839
1840 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1841 {
1842 skb->inner_mac_header = skb->data - skb->head;
1843 }
1844
1845 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1846 const int offset)
1847 {
1848 skb_reset_inner_mac_header(skb);
1849 skb->inner_mac_header += offset;
1850 }
1851 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1852 {
1853 return skb->transport_header != (typeof(skb->transport_header))~0U;
1854 }
1855
1856 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1857 {
1858 return skb->head + skb->transport_header;
1859 }
1860
1861 static inline void skb_reset_transport_header(struct sk_buff *skb)
1862 {
1863 skb->transport_header = skb->data - skb->head;
1864 }
1865
1866 static inline void skb_set_transport_header(struct sk_buff *skb,
1867 const int offset)
1868 {
1869 skb_reset_transport_header(skb);
1870 skb->transport_header += offset;
1871 }
1872
1873 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1874 {
1875 return skb->head + skb->network_header;
1876 }
1877
1878 static inline void skb_reset_network_header(struct sk_buff *skb)
1879 {
1880 skb->network_header = skb->data - skb->head;
1881 }
1882
1883 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1884 {
1885 skb_reset_network_header(skb);
1886 skb->network_header += offset;
1887 }
1888
1889 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1890 {
1891 return skb->head + skb->mac_header;
1892 }
1893
1894 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1895 {
1896 return skb->mac_header != (typeof(skb->mac_header))~0U;
1897 }
1898
1899 static inline void skb_reset_mac_header(struct sk_buff *skb)
1900 {
1901 skb->mac_header = skb->data - skb->head;
1902 }
1903
1904 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1905 {
1906 skb_reset_mac_header(skb);
1907 skb->mac_header += offset;
1908 }
1909
1910 static inline void skb_pop_mac_header(struct sk_buff *skb)
1911 {
1912 skb->mac_header = skb->network_header;
1913 }
1914
1915 static inline void skb_probe_transport_header(struct sk_buff *skb,
1916 const int offset_hint)
1917 {
1918 struct flow_keys keys;
1919
1920 if (skb_transport_header_was_set(skb))
1921 return;
1922 else if (skb_flow_dissect(skb, &keys))
1923 skb_set_transport_header(skb, keys.thoff);
1924 else
1925 skb_set_transport_header(skb, offset_hint);
1926 }
1927
1928 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1929 {
1930 if (skb_mac_header_was_set(skb)) {
1931 const unsigned char *old_mac = skb_mac_header(skb);
1932
1933 skb_set_mac_header(skb, -skb->mac_len);
1934 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1935 }
1936 }
1937
1938 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1939 {
1940 return skb->csum_start - skb_headroom(skb);
1941 }
1942
1943 static inline int skb_transport_offset(const struct sk_buff *skb)
1944 {
1945 return skb_transport_header(skb) - skb->data;
1946 }
1947
1948 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1949 {
1950 return skb->transport_header - skb->network_header;
1951 }
1952
1953 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1954 {
1955 return skb->inner_transport_header - skb->inner_network_header;
1956 }
1957
1958 static inline int skb_network_offset(const struct sk_buff *skb)
1959 {
1960 return skb_network_header(skb) - skb->data;
1961 }
1962
1963 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1964 {
1965 return skb_inner_network_header(skb) - skb->data;
1966 }
1967
1968 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1969 {
1970 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1971 }
1972
1973 /*
1974 * CPUs often take a performance hit when accessing unaligned memory
1975 * locations. The actual performance hit varies, it can be small if the
1976 * hardware handles it or large if we have to take an exception and fix it
1977 * in software.
1978 *
1979 * Since an ethernet header is 14 bytes network drivers often end up with
1980 * the IP header at an unaligned offset. The IP header can be aligned by
1981 * shifting the start of the packet by 2 bytes. Drivers should do this
1982 * with:
1983 *
1984 * skb_reserve(skb, NET_IP_ALIGN);
1985 *
1986 * The downside to this alignment of the IP header is that the DMA is now
1987 * unaligned. On some architectures the cost of an unaligned DMA is high
1988 * and this cost outweighs the gains made by aligning the IP header.
1989 *
1990 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1991 * to be overridden.
1992 */
1993 #ifndef NET_IP_ALIGN
1994 #define NET_IP_ALIGN 2
1995 #endif
1996
1997 /*
1998 * The networking layer reserves some headroom in skb data (via
1999 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2000 * the header has to grow. In the default case, if the header has to grow
2001 * 32 bytes or less we avoid the reallocation.
2002 *
2003 * Unfortunately this headroom changes the DMA alignment of the resulting
2004 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2005 * on some architectures. An architecture can override this value,
2006 * perhaps setting it to a cacheline in size (since that will maintain
2007 * cacheline alignment of the DMA). It must be a power of 2.
2008 *
2009 * Various parts of the networking layer expect at least 32 bytes of
2010 * headroom, you should not reduce this.
2011 *
2012 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2013 * to reduce average number of cache lines per packet.
2014 * get_rps_cpus() for example only access one 64 bytes aligned block :
2015 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2016 */
2017 #ifndef NET_SKB_PAD
2018 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2019 #endif
2020
2021 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2022
2023 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2024 {
2025 if (unlikely(skb_is_nonlinear(skb))) {
2026 WARN_ON(1);
2027 return;
2028 }
2029 skb->len = len;
2030 skb_set_tail_pointer(skb, len);
2031 }
2032
2033 void skb_trim(struct sk_buff *skb, unsigned int len);
2034
2035 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2036 {
2037 if (skb->data_len)
2038 return ___pskb_trim(skb, len);
2039 __skb_trim(skb, len);
2040 return 0;
2041 }
2042
2043 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2044 {
2045 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2046 }
2047
2048 /**
2049 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2050 * @skb: buffer to alter
2051 * @len: new length
2052 *
2053 * This is identical to pskb_trim except that the caller knows that
2054 * the skb is not cloned so we should never get an error due to out-
2055 * of-memory.
2056 */
2057 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2058 {
2059 int err = pskb_trim(skb, len);
2060 BUG_ON(err);
2061 }
2062
2063 /**
2064 * skb_orphan - orphan a buffer
2065 * @skb: buffer to orphan
2066 *
2067 * If a buffer currently has an owner then we call the owner's
2068 * destructor function and make the @skb unowned. The buffer continues
2069 * to exist but is no longer charged to its former owner.
2070 */
2071 static inline void skb_orphan(struct sk_buff *skb)
2072 {
2073 if (skb->destructor) {
2074 skb->destructor(skb);
2075 skb->destructor = NULL;
2076 skb->sk = NULL;
2077 } else {
2078 BUG_ON(skb->sk);
2079 }
2080 }
2081
2082 /**
2083 * skb_orphan_frags - orphan the frags contained in a buffer
2084 * @skb: buffer to orphan frags from
2085 * @gfp_mask: allocation mask for replacement pages
2086 *
2087 * For each frag in the SKB which needs a destructor (i.e. has an
2088 * owner) create a copy of that frag and release the original
2089 * page by calling the destructor.
2090 */
2091 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2092 {
2093 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2094 return 0;
2095 return skb_copy_ubufs(skb, gfp_mask);
2096 }
2097
2098 /**
2099 * __skb_queue_purge - empty a list
2100 * @list: list to empty
2101 *
2102 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2103 * the list and one reference dropped. This function does not take the
2104 * list lock and the caller must hold the relevant locks to use it.
2105 */
2106 void skb_queue_purge(struct sk_buff_head *list);
2107 static inline void __skb_queue_purge(struct sk_buff_head *list)
2108 {
2109 struct sk_buff *skb;
2110 while ((skb = __skb_dequeue(list)) != NULL)
2111 kfree_skb(skb);
2112 }
2113
2114 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2115 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2116 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2117
2118 void *netdev_alloc_frag(unsigned int fragsz);
2119
2120 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2121 gfp_t gfp_mask);
2122
2123 /**
2124 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2125 * @dev: network device to receive on
2126 * @length: length to allocate
2127 *
2128 * Allocate a new &sk_buff and assign it a usage count of one. The
2129 * buffer has unspecified headroom built in. Users should allocate
2130 * the headroom they think they need without accounting for the
2131 * built in space. The built in space is used for optimisations.
2132 *
2133 * %NULL is returned if there is no free memory. Although this function
2134 * allocates memory it can be called from an interrupt.
2135 */
2136 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2137 unsigned int length)
2138 {
2139 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2140 }
2141
2142 /* legacy helper around __netdev_alloc_skb() */
2143 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2144 gfp_t gfp_mask)
2145 {
2146 return __netdev_alloc_skb(NULL, length, gfp_mask);
2147 }
2148
2149 /* legacy helper around netdev_alloc_skb() */
2150 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2151 {
2152 return netdev_alloc_skb(NULL, length);
2153 }
2154
2155
2156 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2157 unsigned int length, gfp_t gfp)
2158 {
2159 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2160
2161 if (NET_IP_ALIGN && skb)
2162 skb_reserve(skb, NET_IP_ALIGN);
2163 return skb;
2164 }
2165
2166 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2167 unsigned int length)
2168 {
2169 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2170 }
2171
2172 void *napi_alloc_frag(unsigned int fragsz);
2173 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2174 unsigned int length, gfp_t gfp_mask);
2175 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2176 unsigned int length)
2177 {
2178 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2179 }
2180
2181 /**
2182 * __dev_alloc_pages - allocate page for network Rx
2183 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2184 * @order: size of the allocation
2185 *
2186 * Allocate a new page.
2187 *
2188 * %NULL is returned if there is no free memory.
2189 */
2190 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2191 unsigned int order)
2192 {
2193 /* This piece of code contains several assumptions.
2194 * 1. This is for device Rx, therefor a cold page is preferred.
2195 * 2. The expectation is the user wants a compound page.
2196 * 3. If requesting a order 0 page it will not be compound
2197 * due to the check to see if order has a value in prep_new_page
2198 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2199 * code in gfp_to_alloc_flags that should be enforcing this.
2200 */
2201 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2202
2203 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2204 }
2205
2206 static inline struct page *dev_alloc_pages(unsigned int order)
2207 {
2208 return __dev_alloc_pages(GFP_ATOMIC, order);
2209 }
2210
2211 /**
2212 * __dev_alloc_page - allocate a page for network Rx
2213 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2214 *
2215 * Allocate a new page.
2216 *
2217 * %NULL is returned if there is no free memory.
2218 */
2219 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2220 {
2221 return __dev_alloc_pages(gfp_mask, 0);
2222 }
2223
2224 static inline struct page *dev_alloc_page(void)
2225 {
2226 return __dev_alloc_page(GFP_ATOMIC);
2227 }
2228
2229 /**
2230 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2231 * @page: The page that was allocated from skb_alloc_page
2232 * @skb: The skb that may need pfmemalloc set
2233 */
2234 static inline void skb_propagate_pfmemalloc(struct page *page,
2235 struct sk_buff *skb)
2236 {
2237 if (page && page->pfmemalloc)
2238 skb->pfmemalloc = true;
2239 }
2240
2241 /**
2242 * skb_frag_page - retrieve the page referred to by a paged fragment
2243 * @frag: the paged fragment
2244 *
2245 * Returns the &struct page associated with @frag.
2246 */
2247 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2248 {
2249 return frag->page.p;
2250 }
2251
2252 /**
2253 * __skb_frag_ref - take an addition reference on a paged fragment.
2254 * @frag: the paged fragment
2255 *
2256 * Takes an additional reference on the paged fragment @frag.
2257 */
2258 static inline void __skb_frag_ref(skb_frag_t *frag)
2259 {
2260 get_page(skb_frag_page(frag));
2261 }
2262
2263 /**
2264 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2265 * @skb: the buffer
2266 * @f: the fragment offset.
2267 *
2268 * Takes an additional reference on the @f'th paged fragment of @skb.
2269 */
2270 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2271 {
2272 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2273 }
2274
2275 /**
2276 * __skb_frag_unref - release a reference on a paged fragment.
2277 * @frag: the paged fragment
2278 *
2279 * Releases a reference on the paged fragment @frag.
2280 */
2281 static inline void __skb_frag_unref(skb_frag_t *frag)
2282 {
2283 put_page(skb_frag_page(frag));
2284 }
2285
2286 /**
2287 * skb_frag_unref - release a reference on a paged fragment of an skb.
2288 * @skb: the buffer
2289 * @f: the fragment offset
2290 *
2291 * Releases a reference on the @f'th paged fragment of @skb.
2292 */
2293 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2294 {
2295 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2296 }
2297
2298 /**
2299 * skb_frag_address - gets the address of the data contained in a paged fragment
2300 * @frag: the paged fragment buffer
2301 *
2302 * Returns the address of the data within @frag. The page must already
2303 * be mapped.
2304 */
2305 static inline void *skb_frag_address(const skb_frag_t *frag)
2306 {
2307 return page_address(skb_frag_page(frag)) + frag->page_offset;
2308 }
2309
2310 /**
2311 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2312 * @frag: the paged fragment buffer
2313 *
2314 * Returns the address of the data within @frag. Checks that the page
2315 * is mapped and returns %NULL otherwise.
2316 */
2317 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2318 {
2319 void *ptr = page_address(skb_frag_page(frag));
2320 if (unlikely(!ptr))
2321 return NULL;
2322
2323 return ptr + frag->page_offset;
2324 }
2325
2326 /**
2327 * __skb_frag_set_page - sets the page contained in a paged fragment
2328 * @frag: the paged fragment
2329 * @page: the page to set
2330 *
2331 * Sets the fragment @frag to contain @page.
2332 */
2333 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2334 {
2335 frag->page.p = page;
2336 }
2337
2338 /**
2339 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2340 * @skb: the buffer
2341 * @f: the fragment offset
2342 * @page: the page to set
2343 *
2344 * Sets the @f'th fragment of @skb to contain @page.
2345 */
2346 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2347 struct page *page)
2348 {
2349 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2350 }
2351
2352 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2353
2354 /**
2355 * skb_frag_dma_map - maps a paged fragment via the DMA API
2356 * @dev: the device to map the fragment to
2357 * @frag: the paged fragment to map
2358 * @offset: the offset within the fragment (starting at the
2359 * fragment's own offset)
2360 * @size: the number of bytes to map
2361 * @dir: the direction of the mapping (%PCI_DMA_*)
2362 *
2363 * Maps the page associated with @frag to @device.
2364 */
2365 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2366 const skb_frag_t *frag,
2367 size_t offset, size_t size,
2368 enum dma_data_direction dir)
2369 {
2370 return dma_map_page(dev, skb_frag_page(frag),
2371 frag->page_offset + offset, size, dir);
2372 }
2373
2374 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2375 gfp_t gfp_mask)
2376 {
2377 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2378 }
2379
2380
2381 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2382 gfp_t gfp_mask)
2383 {
2384 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2385 }
2386
2387
2388 /**
2389 * skb_clone_writable - is the header of a clone writable
2390 * @skb: buffer to check
2391 * @len: length up to which to write
2392 *
2393 * Returns true if modifying the header part of the cloned buffer
2394 * does not requires the data to be copied.
2395 */
2396 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2397 {
2398 return !skb_header_cloned(skb) &&
2399 skb_headroom(skb) + len <= skb->hdr_len;
2400 }
2401
2402 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2403 int cloned)
2404 {
2405 int delta = 0;
2406
2407 if (headroom > skb_headroom(skb))
2408 delta = headroom - skb_headroom(skb);
2409
2410 if (delta || cloned)
2411 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2412 GFP_ATOMIC);
2413 return 0;
2414 }
2415
2416 /**
2417 * skb_cow - copy header of skb when it is required
2418 * @skb: buffer to cow
2419 * @headroom: needed headroom
2420 *
2421 * If the skb passed lacks sufficient headroom or its data part
2422 * is shared, data is reallocated. If reallocation fails, an error
2423 * is returned and original skb is not changed.
2424 *
2425 * The result is skb with writable area skb->head...skb->tail
2426 * and at least @headroom of space at head.
2427 */
2428 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2429 {
2430 return __skb_cow(skb, headroom, skb_cloned(skb));
2431 }
2432
2433 /**
2434 * skb_cow_head - skb_cow but only making the head writable
2435 * @skb: buffer to cow
2436 * @headroom: needed headroom
2437 *
2438 * This function is identical to skb_cow except that we replace the
2439 * skb_cloned check by skb_header_cloned. It should be used when
2440 * you only need to push on some header and do not need to modify
2441 * the data.
2442 */
2443 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2444 {
2445 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2446 }
2447
2448 /**
2449 * skb_padto - pad an skbuff up to a minimal size
2450 * @skb: buffer to pad
2451 * @len: minimal length
2452 *
2453 * Pads up a buffer to ensure the trailing bytes exist and are
2454 * blanked. If the buffer already contains sufficient data it
2455 * is untouched. Otherwise it is extended. Returns zero on
2456 * success. The skb is freed on error.
2457 */
2458 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2459 {
2460 unsigned int size = skb->len;
2461 if (likely(size >= len))
2462 return 0;
2463 return skb_pad(skb, len - size);
2464 }
2465
2466 /**
2467 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2468 * @skb: buffer to pad
2469 * @len: minimal length
2470 *
2471 * Pads up a buffer to ensure the trailing bytes exist and are
2472 * blanked. If the buffer already contains sufficient data it
2473 * is untouched. Otherwise it is extended. Returns zero on
2474 * success. The skb is freed on error.
2475 */
2476 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2477 {
2478 unsigned int size = skb->len;
2479
2480 if (unlikely(size < len)) {
2481 len -= size;
2482 if (skb_pad(skb, len))
2483 return -ENOMEM;
2484 __skb_put(skb, len);
2485 }
2486 return 0;
2487 }
2488
2489 static inline int skb_add_data(struct sk_buff *skb,
2490 struct iov_iter *from, int copy)
2491 {
2492 const int off = skb->len;
2493
2494 if (skb->ip_summed == CHECKSUM_NONE) {
2495 __wsum csum = 0;
2496 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2497 &csum, from) == copy) {
2498 skb->csum = csum_block_add(skb->csum, csum, off);
2499 return 0;
2500 }
2501 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
2502 return 0;
2503
2504 __skb_trim(skb, off);
2505 return -EFAULT;
2506 }
2507
2508 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2509 const struct page *page, int off)
2510 {
2511 if (i) {
2512 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2513
2514 return page == skb_frag_page(frag) &&
2515 off == frag->page_offset + skb_frag_size(frag);
2516 }
2517 return false;
2518 }
2519
2520 static inline int __skb_linearize(struct sk_buff *skb)
2521 {
2522 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2523 }
2524
2525 /**
2526 * skb_linearize - convert paged skb to linear one
2527 * @skb: buffer to linarize
2528 *
2529 * If there is no free memory -ENOMEM is returned, otherwise zero
2530 * is returned and the old skb data released.
2531 */
2532 static inline int skb_linearize(struct sk_buff *skb)
2533 {
2534 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2535 }
2536
2537 /**
2538 * skb_has_shared_frag - can any frag be overwritten
2539 * @skb: buffer to test
2540 *
2541 * Return true if the skb has at least one frag that might be modified
2542 * by an external entity (as in vmsplice()/sendfile())
2543 */
2544 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2545 {
2546 return skb_is_nonlinear(skb) &&
2547 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2548 }
2549
2550 /**
2551 * skb_linearize_cow - make sure skb is linear and writable
2552 * @skb: buffer to process
2553 *
2554 * If there is no free memory -ENOMEM is returned, otherwise zero
2555 * is returned and the old skb data released.
2556 */
2557 static inline int skb_linearize_cow(struct sk_buff *skb)
2558 {
2559 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2560 __skb_linearize(skb) : 0;
2561 }
2562
2563 /**
2564 * skb_postpull_rcsum - update checksum for received skb after pull
2565 * @skb: buffer to update
2566 * @start: start of data before pull
2567 * @len: length of data pulled
2568 *
2569 * After doing a pull on a received packet, you need to call this to
2570 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2571 * CHECKSUM_NONE so that it can be recomputed from scratch.
2572 */
2573
2574 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2575 const void *start, unsigned int len)
2576 {
2577 if (skb->ip_summed == CHECKSUM_COMPLETE)
2578 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2579 }
2580
2581 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2582
2583 /**
2584 * pskb_trim_rcsum - trim received skb and update checksum
2585 * @skb: buffer to trim
2586 * @len: new length
2587 *
2588 * This is exactly the same as pskb_trim except that it ensures the
2589 * checksum of received packets are still valid after the operation.
2590 */
2591
2592 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2593 {
2594 if (likely(len >= skb->len))
2595 return 0;
2596 if (skb->ip_summed == CHECKSUM_COMPLETE)
2597 skb->ip_summed = CHECKSUM_NONE;
2598 return __pskb_trim(skb, len);
2599 }
2600
2601 #define skb_queue_walk(queue, skb) \
2602 for (skb = (queue)->next; \
2603 skb != (struct sk_buff *)(queue); \
2604 skb = skb->next)
2605
2606 #define skb_queue_walk_safe(queue, skb, tmp) \
2607 for (skb = (queue)->next, tmp = skb->next; \
2608 skb != (struct sk_buff *)(queue); \
2609 skb = tmp, tmp = skb->next)
2610
2611 #define skb_queue_walk_from(queue, skb) \
2612 for (; skb != (struct sk_buff *)(queue); \
2613 skb = skb->next)
2614
2615 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2616 for (tmp = skb->next; \
2617 skb != (struct sk_buff *)(queue); \
2618 skb = tmp, tmp = skb->next)
2619
2620 #define skb_queue_reverse_walk(queue, skb) \
2621 for (skb = (queue)->prev; \
2622 skb != (struct sk_buff *)(queue); \
2623 skb = skb->prev)
2624
2625 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2626 for (skb = (queue)->prev, tmp = skb->prev; \
2627 skb != (struct sk_buff *)(queue); \
2628 skb = tmp, tmp = skb->prev)
2629
2630 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2631 for (tmp = skb->prev; \
2632 skb != (struct sk_buff *)(queue); \
2633 skb = tmp, tmp = skb->prev)
2634
2635 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2636 {
2637 return skb_shinfo(skb)->frag_list != NULL;
2638 }
2639
2640 static inline void skb_frag_list_init(struct sk_buff *skb)
2641 {
2642 skb_shinfo(skb)->frag_list = NULL;
2643 }
2644
2645 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2646 {
2647 frag->next = skb_shinfo(skb)->frag_list;
2648 skb_shinfo(skb)->frag_list = frag;
2649 }
2650
2651 #define skb_walk_frags(skb, iter) \
2652 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2653
2654 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2655 int *peeked, int *off, int *err);
2656 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2657 int *err);
2658 unsigned int datagram_poll(struct file *file, struct socket *sock,
2659 struct poll_table_struct *wait);
2660 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2661 struct iov_iter *to, int size);
2662 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2663 struct msghdr *msg, int size)
2664 {
2665 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
2666 }
2667 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2668 struct msghdr *msg);
2669 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2670 struct iov_iter *from, int len);
2671 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
2672 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2673 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2674 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2675 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2676 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2677 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2678 int len, __wsum csum);
2679 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2680 struct pipe_inode_info *pipe, unsigned int len,
2681 unsigned int flags);
2682 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2683 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2684 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2685 int len, int hlen);
2686 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2687 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2688 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2689 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2690 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2691 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
2692 int skb_ensure_writable(struct sk_buff *skb, int write_len);
2693 int skb_vlan_pop(struct sk_buff *skb);
2694 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
2695
2696 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2697 {
2698 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2699 }
2700
2701 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2702 {
2703 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2704 }
2705
2706 struct skb_checksum_ops {
2707 __wsum (*update)(const void *mem, int len, __wsum wsum);
2708 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2709 };
2710
2711 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2712 __wsum csum, const struct skb_checksum_ops *ops);
2713 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2714 __wsum csum);
2715
2716 static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2717 int len, void *data, int hlen, void *buffer)
2718 {
2719 if (hlen - offset >= len)
2720 return data + offset;
2721
2722 if (!skb ||
2723 skb_copy_bits(skb, offset, buffer, len) < 0)
2724 return NULL;
2725
2726 return buffer;
2727 }
2728
2729 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2730 int len, void *buffer)
2731 {
2732 return __skb_header_pointer(skb, offset, len, skb->data,
2733 skb_headlen(skb), buffer);
2734 }
2735
2736 /**
2737 * skb_needs_linearize - check if we need to linearize a given skb
2738 * depending on the given device features.
2739 * @skb: socket buffer to check
2740 * @features: net device features
2741 *
2742 * Returns true if either:
2743 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2744 * 2. skb is fragmented and the device does not support SG.
2745 */
2746 static inline bool skb_needs_linearize(struct sk_buff *skb,
2747 netdev_features_t features)
2748 {
2749 return skb_is_nonlinear(skb) &&
2750 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2751 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2752 }
2753
2754 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2755 void *to,
2756 const unsigned int len)
2757 {
2758 memcpy(to, skb->data, len);
2759 }
2760
2761 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2762 const int offset, void *to,
2763 const unsigned int len)
2764 {
2765 memcpy(to, skb->data + offset, len);
2766 }
2767
2768 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2769 const void *from,
2770 const unsigned int len)
2771 {
2772 memcpy(skb->data, from, len);
2773 }
2774
2775 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2776 const int offset,
2777 const void *from,
2778 const unsigned int len)
2779 {
2780 memcpy(skb->data + offset, from, len);
2781 }
2782
2783 void skb_init(void);
2784
2785 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2786 {
2787 return skb->tstamp;
2788 }
2789
2790 /**
2791 * skb_get_timestamp - get timestamp from a skb
2792 * @skb: skb to get stamp from
2793 * @stamp: pointer to struct timeval to store stamp in
2794 *
2795 * Timestamps are stored in the skb as offsets to a base timestamp.
2796 * This function converts the offset back to a struct timeval and stores
2797 * it in stamp.
2798 */
2799 static inline void skb_get_timestamp(const struct sk_buff *skb,
2800 struct timeval *stamp)
2801 {
2802 *stamp = ktime_to_timeval(skb->tstamp);
2803 }
2804
2805 static inline void skb_get_timestampns(const struct sk_buff *skb,
2806 struct timespec *stamp)
2807 {
2808 *stamp = ktime_to_timespec(skb->tstamp);
2809 }
2810
2811 static inline void __net_timestamp(struct sk_buff *skb)
2812 {
2813 skb->tstamp = ktime_get_real();
2814 }
2815
2816 static inline ktime_t net_timedelta(ktime_t t)
2817 {
2818 return ktime_sub(ktime_get_real(), t);
2819 }
2820
2821 static inline ktime_t net_invalid_timestamp(void)
2822 {
2823 return ktime_set(0, 0);
2824 }
2825
2826 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2827
2828 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2829
2830 void skb_clone_tx_timestamp(struct sk_buff *skb);
2831 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2832
2833 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2834
2835 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2836 {
2837 }
2838
2839 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2840 {
2841 return false;
2842 }
2843
2844 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2845
2846 /**
2847 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2848 *
2849 * PHY drivers may accept clones of transmitted packets for
2850 * timestamping via their phy_driver.txtstamp method. These drivers
2851 * must call this function to return the skb back to the stack, with
2852 * or without a timestamp.
2853 *
2854 * @skb: clone of the the original outgoing packet
2855 * @hwtstamps: hardware time stamps, may be NULL if not available
2856 *
2857 */
2858 void skb_complete_tx_timestamp(struct sk_buff *skb,
2859 struct skb_shared_hwtstamps *hwtstamps);
2860
2861 void __skb_tstamp_tx(struct sk_buff *orig_skb,
2862 struct skb_shared_hwtstamps *hwtstamps,
2863 struct sock *sk, int tstype);
2864
2865 /**
2866 * skb_tstamp_tx - queue clone of skb with send time stamps
2867 * @orig_skb: the original outgoing packet
2868 * @hwtstamps: hardware time stamps, may be NULL if not available
2869 *
2870 * If the skb has a socket associated, then this function clones the
2871 * skb (thus sharing the actual data and optional structures), stores
2872 * the optional hardware time stamping information (if non NULL) or
2873 * generates a software time stamp (otherwise), then queues the clone
2874 * to the error queue of the socket. Errors are silently ignored.
2875 */
2876 void skb_tstamp_tx(struct sk_buff *orig_skb,
2877 struct skb_shared_hwtstamps *hwtstamps);
2878
2879 static inline void sw_tx_timestamp(struct sk_buff *skb)
2880 {
2881 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2882 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2883 skb_tstamp_tx(skb, NULL);
2884 }
2885
2886 /**
2887 * skb_tx_timestamp() - Driver hook for transmit timestamping
2888 *
2889 * Ethernet MAC Drivers should call this function in their hard_xmit()
2890 * function immediately before giving the sk_buff to the MAC hardware.
2891 *
2892 * Specifically, one should make absolutely sure that this function is
2893 * called before TX completion of this packet can trigger. Otherwise
2894 * the packet could potentially already be freed.
2895 *
2896 * @skb: A socket buffer.
2897 */
2898 static inline void skb_tx_timestamp(struct sk_buff *skb)
2899 {
2900 skb_clone_tx_timestamp(skb);
2901 sw_tx_timestamp(skb);
2902 }
2903
2904 /**
2905 * skb_complete_wifi_ack - deliver skb with wifi status
2906 *
2907 * @skb: the original outgoing packet
2908 * @acked: ack status
2909 *
2910 */
2911 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2912
2913 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2914 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2915
2916 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2917 {
2918 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
2919 }
2920
2921 /**
2922 * skb_checksum_complete - Calculate checksum of an entire packet
2923 * @skb: packet to process
2924 *
2925 * This function calculates the checksum over the entire packet plus
2926 * the value of skb->csum. The latter can be used to supply the
2927 * checksum of a pseudo header as used by TCP/UDP. It returns the
2928 * checksum.
2929 *
2930 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2931 * this function can be used to verify that checksum on received
2932 * packets. In that case the function should return zero if the
2933 * checksum is correct. In particular, this function will return zero
2934 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2935 * hardware has already verified the correctness of the checksum.
2936 */
2937 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2938 {
2939 return skb_csum_unnecessary(skb) ?
2940 0 : __skb_checksum_complete(skb);
2941 }
2942
2943 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2944 {
2945 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2946 if (skb->csum_level == 0)
2947 skb->ip_summed = CHECKSUM_NONE;
2948 else
2949 skb->csum_level--;
2950 }
2951 }
2952
2953 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2954 {
2955 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2956 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2957 skb->csum_level++;
2958 } else if (skb->ip_summed == CHECKSUM_NONE) {
2959 skb->ip_summed = CHECKSUM_UNNECESSARY;
2960 skb->csum_level = 0;
2961 }
2962 }
2963
2964 static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2965 {
2966 /* Mark current checksum as bad (typically called from GRO
2967 * path). In the case that ip_summed is CHECKSUM_NONE
2968 * this must be the first checksum encountered in the packet.
2969 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2970 * checksum after the last one validated. For UDP, a zero
2971 * checksum can not be marked as bad.
2972 */
2973
2974 if (skb->ip_summed == CHECKSUM_NONE ||
2975 skb->ip_summed == CHECKSUM_UNNECESSARY)
2976 skb->csum_bad = 1;
2977 }
2978
2979 /* Check if we need to perform checksum complete validation.
2980 *
2981 * Returns true if checksum complete is needed, false otherwise
2982 * (either checksum is unnecessary or zero checksum is allowed).
2983 */
2984 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2985 bool zero_okay,
2986 __sum16 check)
2987 {
2988 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2989 skb->csum_valid = 1;
2990 __skb_decr_checksum_unnecessary(skb);
2991 return false;
2992 }
2993
2994 return true;
2995 }
2996
2997 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2998 * in checksum_init.
2999 */
3000 #define CHECKSUM_BREAK 76
3001
3002 /* Validate (init) checksum based on checksum complete.
3003 *
3004 * Return values:
3005 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3006 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3007 * checksum is stored in skb->csum for use in __skb_checksum_complete
3008 * non-zero: value of invalid checksum
3009 *
3010 */
3011 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3012 bool complete,
3013 __wsum psum)
3014 {
3015 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3016 if (!csum_fold(csum_add(psum, skb->csum))) {
3017 skb->csum_valid = 1;
3018 return 0;
3019 }
3020 } else if (skb->csum_bad) {
3021 /* ip_summed == CHECKSUM_NONE in this case */
3022 return 1;
3023 }
3024
3025 skb->csum = psum;
3026
3027 if (complete || skb->len <= CHECKSUM_BREAK) {
3028 __sum16 csum;
3029
3030 csum = __skb_checksum_complete(skb);
3031 skb->csum_valid = !csum;
3032 return csum;
3033 }
3034
3035 return 0;
3036 }
3037
3038 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3039 {
3040 return 0;
3041 }
3042
3043 /* Perform checksum validate (init). Note that this is a macro since we only
3044 * want to calculate the pseudo header which is an input function if necessary.
3045 * First we try to validate without any computation (checksum unnecessary) and
3046 * then calculate based on checksum complete calling the function to compute
3047 * pseudo header.
3048 *
3049 * Return values:
3050 * 0: checksum is validated or try to in skb_checksum_complete
3051 * non-zero: value of invalid checksum
3052 */
3053 #define __skb_checksum_validate(skb, proto, complete, \
3054 zero_okay, check, compute_pseudo) \
3055 ({ \
3056 __sum16 __ret = 0; \
3057 skb->csum_valid = 0; \
3058 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3059 __ret = __skb_checksum_validate_complete(skb, \
3060 complete, compute_pseudo(skb, proto)); \
3061 __ret; \
3062 })
3063
3064 #define skb_checksum_init(skb, proto, compute_pseudo) \
3065 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3066
3067 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3068 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3069
3070 #define skb_checksum_validate(skb, proto, compute_pseudo) \
3071 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3072
3073 #define skb_checksum_validate_zero_check(skb, proto, check, \
3074 compute_pseudo) \
3075 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3076
3077 #define skb_checksum_simple_validate(skb) \
3078 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3079
3080 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3081 {
3082 return (skb->ip_summed == CHECKSUM_NONE &&
3083 skb->csum_valid && !skb->csum_bad);
3084 }
3085
3086 static inline void __skb_checksum_convert(struct sk_buff *skb,
3087 __sum16 check, __wsum pseudo)
3088 {
3089 skb->csum = ~pseudo;
3090 skb->ip_summed = CHECKSUM_COMPLETE;
3091 }
3092
3093 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3094 do { \
3095 if (__skb_checksum_convert_check(skb)) \
3096 __skb_checksum_convert(skb, check, \
3097 compute_pseudo(skb, proto)); \
3098 } while (0)
3099
3100 /* Update skbuf and packet to reflect the remote checksum offload operation.
3101 * When called, ptr indicates the starting point for skb->csum when
3102 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3103 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3104 */
3105 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
3106 int start, int offset)
3107 {
3108 __wsum delta;
3109
3110 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3111 __skb_checksum_complete(skb);
3112 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3113 }
3114
3115 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3116
3117 /* Adjust skb->csum since we changed the packet */
3118 skb->csum = csum_add(skb->csum, delta);
3119 }
3120
3121 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3122 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3123 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3124 {
3125 if (nfct && atomic_dec_and_test(&nfct->use))
3126 nf_conntrack_destroy(nfct);
3127 }
3128 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3129 {
3130 if (nfct)
3131 atomic_inc(&nfct->use);
3132 }
3133 #endif
3134 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3135 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3136 {
3137 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3138 kfree(nf_bridge);
3139 }
3140 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3141 {
3142 if (nf_bridge)
3143 atomic_inc(&nf_bridge->use);
3144 }
3145 #endif /* CONFIG_BRIDGE_NETFILTER */
3146 static inline void nf_reset(struct sk_buff *skb)
3147 {
3148 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3149 nf_conntrack_put(skb->nfct);
3150 skb->nfct = NULL;
3151 #endif
3152 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3153 nf_bridge_put(skb->nf_bridge);
3154 skb->nf_bridge = NULL;
3155 #endif
3156 }
3157
3158 static inline void nf_reset_trace(struct sk_buff *skb)
3159 {
3160 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3161 skb->nf_trace = 0;
3162 #endif
3163 }
3164
3165 /* Note: This doesn't put any conntrack and bridge info in dst. */
3166 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3167 bool copy)
3168 {
3169 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3170 dst->nfct = src->nfct;
3171 nf_conntrack_get(src->nfct);
3172 if (copy)
3173 dst->nfctinfo = src->nfctinfo;
3174 #endif
3175 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3176 dst->nf_bridge = src->nf_bridge;
3177 nf_bridge_get(src->nf_bridge);
3178 #endif
3179 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3180 if (copy)
3181 dst->nf_trace = src->nf_trace;
3182 #endif
3183 }
3184
3185 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3186 {
3187 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3188 nf_conntrack_put(dst->nfct);
3189 #endif
3190 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3191 nf_bridge_put(dst->nf_bridge);
3192 #endif
3193 __nf_copy(dst, src, true);
3194 }
3195
3196 #ifdef CONFIG_NETWORK_SECMARK
3197 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3198 {
3199 to->secmark = from->secmark;
3200 }
3201
3202 static inline void skb_init_secmark(struct sk_buff *skb)
3203 {
3204 skb->secmark = 0;
3205 }
3206 #else
3207 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3208 { }
3209
3210 static inline void skb_init_secmark(struct sk_buff *skb)
3211 { }
3212 #endif
3213
3214 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3215 {
3216 return !skb->destructor &&
3217 #if IS_ENABLED(CONFIG_XFRM)
3218 !skb->sp &&
3219 #endif
3220 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3221 !skb->nfct &&
3222 #endif
3223 !skb->_skb_refdst &&
3224 !skb_has_frag_list(skb);
3225 }
3226
3227 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3228 {
3229 skb->queue_mapping = queue_mapping;
3230 }
3231
3232 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3233 {
3234 return skb->queue_mapping;
3235 }
3236
3237 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3238 {
3239 to->queue_mapping = from->queue_mapping;
3240 }
3241
3242 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3243 {
3244 skb->queue_mapping = rx_queue + 1;
3245 }
3246
3247 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3248 {
3249 return skb->queue_mapping - 1;
3250 }
3251
3252 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3253 {
3254 return skb->queue_mapping != 0;
3255 }
3256
3257 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3258 unsigned int num_tx_queues);
3259
3260 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3261 {
3262 #ifdef CONFIG_XFRM
3263 return skb->sp;
3264 #else
3265 return NULL;
3266 #endif
3267 }
3268
3269 /* Keeps track of mac header offset relative to skb->head.
3270 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3271 * For non-tunnel skb it points to skb_mac_header() and for
3272 * tunnel skb it points to outer mac header.
3273 * Keeps track of level of encapsulation of network headers.
3274 */
3275 struct skb_gso_cb {
3276 int mac_offset;
3277 int encap_level;
3278 __u16 csum_start;
3279 };
3280 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3281
3282 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3283 {
3284 return (skb_mac_header(inner_skb) - inner_skb->head) -
3285 SKB_GSO_CB(inner_skb)->mac_offset;
3286 }
3287
3288 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3289 {
3290 int new_headroom, headroom;
3291 int ret;
3292
3293 headroom = skb_headroom(skb);
3294 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3295 if (ret)
3296 return ret;
3297
3298 new_headroom = skb_headroom(skb);
3299 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3300 return 0;
3301 }
3302
3303 /* Compute the checksum for a gso segment. First compute the checksum value
3304 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3305 * then add in skb->csum (checksum from csum_start to end of packet).
3306 * skb->csum and csum_start are then updated to reflect the checksum of the
3307 * resultant packet starting from the transport header-- the resultant checksum
3308 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3309 * header.
3310 */
3311 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3312 {
3313 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3314 skb_transport_offset(skb);
3315 __u16 csum;
3316
3317 csum = csum_fold(csum_partial(skb_transport_header(skb),
3318 plen, skb->csum));
3319 skb->csum = res;
3320 SKB_GSO_CB(skb)->csum_start -= plen;
3321
3322 return csum;
3323 }
3324
3325 static inline bool skb_is_gso(const struct sk_buff *skb)
3326 {
3327 return skb_shinfo(skb)->gso_size;
3328 }
3329
3330 /* Note: Should be called only if skb_is_gso(skb) is true */
3331 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
3332 {
3333 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3334 }
3335
3336 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
3337
3338 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3339 {
3340 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3341 * wanted then gso_type will be set. */
3342 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3343
3344 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3345 unlikely(shinfo->gso_type == 0)) {
3346 __skb_warn_lro_forwarding(skb);
3347 return true;
3348 }
3349 return false;
3350 }
3351
3352 static inline void skb_forward_csum(struct sk_buff *skb)
3353 {
3354 /* Unfortunately we don't support this one. Any brave souls? */
3355 if (skb->ip_summed == CHECKSUM_COMPLETE)
3356 skb->ip_summed = CHECKSUM_NONE;
3357 }
3358
3359 /**
3360 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3361 * @skb: skb to check
3362 *
3363 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3364 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3365 * use this helper, to document places where we make this assertion.
3366 */
3367 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
3368 {
3369 #ifdef DEBUG
3370 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3371 #endif
3372 }
3373
3374 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
3375
3376 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3377
3378 u32 skb_get_poff(const struct sk_buff *skb);
3379 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3380 const struct flow_keys *keys, int hlen);
3381
3382 /**
3383 * skb_head_is_locked - Determine if the skb->head is locked down
3384 * @skb: skb to check
3385 *
3386 * The head on skbs build around a head frag can be removed if they are
3387 * not cloned. This function returns true if the skb head is locked down
3388 * due to either being allocated via kmalloc, or by being a clone with
3389 * multiple references to the head.
3390 */
3391 static inline bool skb_head_is_locked(const struct sk_buff *skb)
3392 {
3393 return !skb->head_frag || skb_cloned(skb);
3394 }
3395
3396 /**
3397 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3398 *
3399 * @skb: GSO skb
3400 *
3401 * skb_gso_network_seglen is used to determine the real size of the
3402 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3403 *
3404 * The MAC/L2 header is not accounted for.
3405 */
3406 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3407 {
3408 unsigned int hdr_len = skb_transport_header(skb) -
3409 skb_network_header(skb);
3410 return hdr_len + skb_gso_transport_seglen(skb);
3411 }
3412 #endif /* __KERNEL__ */
3413 #endif /* _LINUX_SKBUFF_H */
This page took 0.101266 seconds and 5 git commands to generate.