net: remove deprecated syststamp timestamp
[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
24 #include <linux/atomic.h>
25 #include <asm/types.h>
26 #include <linux/spinlock.h>
27 #include <linux/net.h>
28 #include <linux/textsearch.h>
29 #include <net/checksum.h>
30 #include <linux/rcupdate.h>
31 #include <linux/dmaengine.h>
32 #include <linux/hrtimer.h>
33 #include <linux/dma-mapping.h>
34 #include <linux/netdev_features.h>
35 #include <linux/sched.h>
36 #include <net/flow_keys.h>
37
38 /* A. Checksumming of received packets by device.
39 *
40 * CHECKSUM_NONE:
41 *
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 *
46 * CHECKSUM_UNNECESSARY:
47 *
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
50 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
51 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
52 * undefined in this case though. It is a bad option, but, unfortunately,
53 * nowadays most vendors do this. Apparently with the secret goal to sell
54 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_COMPLETE:
57 *
58 * This is the most generic way. The device supplied checksum of the _whole_
59 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
60 * hardware doesn't need to parse L3/L4 headers to implement this.
61 *
62 * Note: Even if device supports only some protocols, but is able to produce
63 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
64 *
65 * CHECKSUM_PARTIAL:
66 *
67 * This is identical to the case for output below. This may occur on a packet
68 * received directly from another Linux OS, e.g., a virtualized Linux kernel
69 * on the same host. The packet can be treated in the same way as
70 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
71 * checksum must be filled in by the OS or the hardware.
72 *
73 * B. Checksumming on output.
74 *
75 * CHECKSUM_NONE:
76 *
77 * The skb was already checksummed by the protocol, or a checksum is not
78 * required.
79 *
80 * CHECKSUM_PARTIAL:
81 *
82 * The device is required to checksum the packet as seen by hard_start_xmit()
83 * from skb->csum_start up to the end, and to record/write the checksum at
84 * offset skb->csum_start + skb->csum_offset.
85 *
86 * The device must show its capabilities in dev->features, set up at device
87 * setup time, e.g. netdev_features.h:
88 *
89 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
90 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
91 * IPv4. Sigh. Vendors like this way for an unknown reason.
92 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
93 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
94 * NETIF_F_... - Well, you get the picture.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * Normally, the device will do per protocol specific checksumming. Protocol
99 * implementations that do not want the NIC to perform the checksum
100 * calculation should use this flag in their outgoing skbs.
101 *
102 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
106 * Any questions? No questions, good. --ANK
107 */
108
109 /* Don't change this without changing skb_csum_unnecessary! */
110 #define CHECKSUM_NONE 0
111 #define CHECKSUM_UNNECESSARY 1
112 #define CHECKSUM_COMPLETE 2
113 #define CHECKSUM_PARTIAL 3
114
115 #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
116 #define SKB_WITH_OVERHEAD(X) \
117 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
118 #define SKB_MAX_ORDER(X, ORDER) \
119 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
120 #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
121 #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
122
123 /* return minimum truesize of one skb containing X bytes of data */
124 #define SKB_TRUESIZE(X) ((X) + \
125 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
126 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
127
128 struct net_device;
129 struct scatterlist;
130 struct pipe_inode_info;
131
132 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
133 struct nf_conntrack {
134 atomic_t use;
135 };
136 #endif
137
138 #ifdef CONFIG_BRIDGE_NETFILTER
139 struct nf_bridge_info {
140 atomic_t use;
141 unsigned int mask;
142 struct net_device *physindev;
143 struct net_device *physoutdev;
144 unsigned long data[32 / sizeof(unsigned long)];
145 };
146 #endif
147
148 struct sk_buff_head {
149 /* These two members must be first. */
150 struct sk_buff *next;
151 struct sk_buff *prev;
152
153 __u32 qlen;
154 spinlock_t lock;
155 };
156
157 struct sk_buff;
158
159 /* To allow 64K frame to be packed as single skb without frag_list we
160 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
161 * buffers which do not start on a page boundary.
162 *
163 * Since GRO uses frags we allocate at least 16 regardless of page
164 * size.
165 */
166 #if (65536/PAGE_SIZE + 1) < 16
167 #define MAX_SKB_FRAGS 16UL
168 #else
169 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
170 #endif
171
172 typedef struct skb_frag_struct skb_frag_t;
173
174 struct skb_frag_struct {
175 struct {
176 struct page *p;
177 } page;
178 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
179 __u32 page_offset;
180 __u32 size;
181 #else
182 __u16 page_offset;
183 __u16 size;
184 #endif
185 };
186
187 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
188 {
189 return frag->size;
190 }
191
192 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
193 {
194 frag->size = size;
195 }
196
197 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
198 {
199 frag->size += delta;
200 }
201
202 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
203 {
204 frag->size -= delta;
205 }
206
207 #define HAVE_HW_TIME_STAMP
208
209 /**
210 * struct skb_shared_hwtstamps - hardware time stamps
211 * @hwtstamp: hardware time stamp transformed into duration
212 * since arbitrary point in time
213 *
214 * Software time stamps generated by ktime_get_real() are stored in
215 * skb->tstamp.
216 *
217 * hwtstamps can only be compared against other hwtstamps from
218 * the same device.
219 *
220 * This structure is attached to packets as part of the
221 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
222 */
223 struct skb_shared_hwtstamps {
224 ktime_t hwtstamp;
225 };
226
227 /* Definitions for tx_flags in struct skb_shared_info */
228 enum {
229 /* generate hardware time stamp */
230 SKBTX_HW_TSTAMP = 1 << 0,
231
232 /* generate software time stamp */
233 SKBTX_SW_TSTAMP = 1 << 1,
234
235 /* device driver is going to provide hardware time stamp */
236 SKBTX_IN_PROGRESS = 1 << 2,
237
238 /* device driver supports TX zero-copy buffers */
239 SKBTX_DEV_ZEROCOPY = 1 << 3,
240
241 /* generate wifi status information (where possible) */
242 SKBTX_WIFI_STATUS = 1 << 4,
243
244 /* This indicates at least one fragment might be overwritten
245 * (as in vmsplice(), sendfile() ...)
246 * If we need to compute a TX checksum, we'll need to copy
247 * all frags to avoid possible bad checksum
248 */
249 SKBTX_SHARED_FRAG = 1 << 5,
250 };
251
252 /*
253 * The callback notifies userspace to release buffers when skb DMA is done in
254 * lower device, the skb last reference should be 0 when calling this.
255 * The zerocopy_success argument is true if zero copy transmit occurred,
256 * false on data copy or out of memory error caused by data copy attempt.
257 * The ctx field is used to track device context.
258 * The desc field is used to track userspace buffer index.
259 */
260 struct ubuf_info {
261 void (*callback)(struct ubuf_info *, bool zerocopy_success);
262 void *ctx;
263 unsigned long desc;
264 };
265
266 /* This data is invariant across clones and lives at
267 * the end of the header data, ie. at skb->end.
268 */
269 struct skb_shared_info {
270 unsigned char nr_frags;
271 __u8 tx_flags;
272 unsigned short gso_size;
273 /* Warning: this field is not always filled in (UFO)! */
274 unsigned short gso_segs;
275 unsigned short gso_type;
276 struct sk_buff *frag_list;
277 struct skb_shared_hwtstamps hwtstamps;
278 __be32 ip6_frag_id;
279
280 /*
281 * Warning : all fields before dataref are cleared in __alloc_skb()
282 */
283 atomic_t dataref;
284
285 /* Intermediate layers must ensure that destructor_arg
286 * remains valid until skb destructor */
287 void * destructor_arg;
288
289 /* must be last field, see pskb_expand_head() */
290 skb_frag_t frags[MAX_SKB_FRAGS];
291 };
292
293 /* We divide dataref into two halves. The higher 16 bits hold references
294 * to the payload part of skb->data. The lower 16 bits hold references to
295 * the entire skb->data. A clone of a headerless skb holds the length of
296 * the header in skb->hdr_len.
297 *
298 * All users must obey the rule that the skb->data reference count must be
299 * greater than or equal to the payload reference count.
300 *
301 * Holding a reference to the payload part means that the user does not
302 * care about modifications to the header part of skb->data.
303 */
304 #define SKB_DATAREF_SHIFT 16
305 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
306
307
308 enum {
309 SKB_FCLONE_UNAVAILABLE,
310 SKB_FCLONE_ORIG,
311 SKB_FCLONE_CLONE,
312 };
313
314 enum {
315 SKB_GSO_TCPV4 = 1 << 0,
316 SKB_GSO_UDP = 1 << 1,
317
318 /* This indicates the skb is from an untrusted source. */
319 SKB_GSO_DODGY = 1 << 2,
320
321 /* This indicates the tcp segment has CWR set. */
322 SKB_GSO_TCP_ECN = 1 << 3,
323
324 SKB_GSO_TCPV6 = 1 << 4,
325
326 SKB_GSO_FCOE = 1 << 5,
327
328 SKB_GSO_GRE = 1 << 6,
329
330 SKB_GSO_GRE_CSUM = 1 << 7,
331
332 SKB_GSO_IPIP = 1 << 8,
333
334 SKB_GSO_SIT = 1 << 9,
335
336 SKB_GSO_UDP_TUNNEL = 1 << 10,
337
338 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
339
340 SKB_GSO_MPLS = 1 << 12,
341
342 };
343
344 #if BITS_PER_LONG > 32
345 #define NET_SKBUFF_DATA_USES_OFFSET 1
346 #endif
347
348 #ifdef NET_SKBUFF_DATA_USES_OFFSET
349 typedef unsigned int sk_buff_data_t;
350 #else
351 typedef unsigned char *sk_buff_data_t;
352 #endif
353
354 /**
355 * struct skb_mstamp - multi resolution time stamps
356 * @stamp_us: timestamp in us resolution
357 * @stamp_jiffies: timestamp in jiffies
358 */
359 struct skb_mstamp {
360 union {
361 u64 v64;
362 struct {
363 u32 stamp_us;
364 u32 stamp_jiffies;
365 };
366 };
367 };
368
369 /**
370 * skb_mstamp_get - get current timestamp
371 * @cl: place to store timestamps
372 */
373 static inline void skb_mstamp_get(struct skb_mstamp *cl)
374 {
375 u64 val = local_clock();
376
377 do_div(val, NSEC_PER_USEC);
378 cl->stamp_us = (u32)val;
379 cl->stamp_jiffies = (u32)jiffies;
380 }
381
382 /**
383 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
384 * @t1: pointer to newest sample
385 * @t0: pointer to oldest sample
386 */
387 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
388 const struct skb_mstamp *t0)
389 {
390 s32 delta_us = t1->stamp_us - t0->stamp_us;
391 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
392
393 /* If delta_us is negative, this might be because interval is too big,
394 * or local_clock() drift is too big : fallback using jiffies.
395 */
396 if (delta_us <= 0 ||
397 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
398
399 delta_us = jiffies_to_usecs(delta_jiffies);
400
401 return delta_us;
402 }
403
404
405 /**
406 * struct sk_buff - socket buffer
407 * @next: Next buffer in list
408 * @prev: Previous buffer in list
409 * @tstamp: Time we arrived/left
410 * @sk: Socket we are owned by
411 * @dev: Device we arrived on/are leaving by
412 * @cb: Control buffer. Free for use by every layer. Put private vars here
413 * @_skb_refdst: destination entry (with norefcount bit)
414 * @sp: the security path, used for xfrm
415 * @len: Length of actual data
416 * @data_len: Data length
417 * @mac_len: Length of link layer header
418 * @hdr_len: writable header length of cloned skb
419 * @csum: Checksum (must include start/offset pair)
420 * @csum_start: Offset from skb->head where checksumming should start
421 * @csum_offset: Offset from csum_start where checksum should be stored
422 * @priority: Packet queueing priority
423 * @ignore_df: allow local fragmentation
424 * @cloned: Head may be cloned (check refcnt to be sure)
425 * @ip_summed: Driver fed us an IP checksum
426 * @nohdr: Payload reference only, must not modify header
427 * @nfctinfo: Relationship of this skb to the connection
428 * @pkt_type: Packet class
429 * @fclone: skbuff clone status
430 * @ipvs_property: skbuff is owned by ipvs
431 * @peeked: this packet has been seen already, so stats have been
432 * done for it, don't do them again
433 * @nf_trace: netfilter packet trace flag
434 * @protocol: Packet protocol from driver
435 * @destructor: Destruct function
436 * @nfct: Associated connection, if any
437 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
438 * @skb_iif: ifindex of device we arrived on
439 * @tc_index: Traffic control index
440 * @tc_verd: traffic control verdict
441 * @hash: the packet hash
442 * @queue_mapping: Queue mapping for multiqueue devices
443 * @ndisc_nodetype: router type (from link layer)
444 * @ooo_okay: allow the mapping of a socket to a queue to be changed
445 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
446 * ports.
447 * @sw_hash: indicates hash was computed in software stack
448 * @wifi_acked_valid: wifi_acked was set
449 * @wifi_acked: whether frame was acked on wifi or not
450 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
451 * @dma_cookie: a cookie to one of several possible DMA operations
452 * done by skb DMA functions
453 * @napi_id: id of the NAPI struct this skb came from
454 * @secmark: security marking
455 * @mark: Generic packet mark
456 * @dropcount: total number of sk_receive_queue overflows
457 * @vlan_proto: vlan encapsulation protocol
458 * @vlan_tci: vlan tag control information
459 * @inner_protocol: Protocol (encapsulation)
460 * @inner_transport_header: Inner transport layer header (encapsulation)
461 * @inner_network_header: Network layer header (encapsulation)
462 * @inner_mac_header: Link layer header (encapsulation)
463 * @transport_header: Transport layer header
464 * @network_header: Network layer header
465 * @mac_header: Link layer header
466 * @tail: Tail pointer
467 * @end: End pointer
468 * @head: Head of buffer
469 * @data: Data head pointer
470 * @truesize: Buffer size
471 * @users: User count - see {datagram,tcp}.c
472 */
473
474 struct sk_buff {
475 /* These two members must be first. */
476 struct sk_buff *next;
477 struct sk_buff *prev;
478
479 union {
480 ktime_t tstamp;
481 struct skb_mstamp skb_mstamp;
482 };
483
484 struct sock *sk;
485 struct net_device *dev;
486
487 /*
488 * This is the control buffer. It is free to use for every
489 * layer. Please put your private variables there. If you
490 * want to keep them across layers you have to do a skb_clone()
491 * first. This is owned by whoever has the skb queued ATM.
492 */
493 char cb[48] __aligned(8);
494
495 unsigned long _skb_refdst;
496 #ifdef CONFIG_XFRM
497 struct sec_path *sp;
498 #endif
499 unsigned int len,
500 data_len;
501 __u16 mac_len,
502 hdr_len;
503 union {
504 __wsum csum;
505 struct {
506 __u16 csum_start;
507 __u16 csum_offset;
508 };
509 };
510 __u32 priority;
511 kmemcheck_bitfield_begin(flags1);
512 __u8 ignore_df:1,
513 cloned:1,
514 ip_summed:2,
515 nohdr:1,
516 nfctinfo:3;
517 __u8 pkt_type:3,
518 fclone:2,
519 ipvs_property:1,
520 peeked:1,
521 nf_trace:1;
522 kmemcheck_bitfield_end(flags1);
523 __be16 protocol;
524
525 void (*destructor)(struct sk_buff *skb);
526 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
527 struct nf_conntrack *nfct;
528 #endif
529 #ifdef CONFIG_BRIDGE_NETFILTER
530 struct nf_bridge_info *nf_bridge;
531 #endif
532
533 int skb_iif;
534
535 __u32 hash;
536
537 __be16 vlan_proto;
538 __u16 vlan_tci;
539
540 #ifdef CONFIG_NET_SCHED
541 __u16 tc_index; /* traffic control index */
542 #ifdef CONFIG_NET_CLS_ACT
543 __u16 tc_verd; /* traffic control verdict */
544 #endif
545 #endif
546
547 __u16 queue_mapping;
548 kmemcheck_bitfield_begin(flags2);
549 #ifdef CONFIG_IPV6_NDISC_NODETYPE
550 __u8 ndisc_nodetype:2;
551 #endif
552 __u8 pfmemalloc:1;
553 __u8 ooo_okay:1;
554 __u8 l4_hash:1;
555 __u8 sw_hash:1;
556 __u8 wifi_acked_valid:1;
557 __u8 wifi_acked:1;
558 __u8 no_fcs:1;
559 __u8 head_frag:1;
560 /* Encapsulation protocol and NIC drivers should use
561 * this flag to indicate to each other if the skb contains
562 * encapsulated packet or not and maybe use the inner packet
563 * headers if needed
564 */
565 __u8 encapsulation:1;
566 __u8 encap_hdr_csum:1;
567 __u8 csum_valid:1;
568 __u8 csum_complete_sw:1;
569 /* 2/4 bit hole (depending on ndisc_nodetype presence) */
570 kmemcheck_bitfield_end(flags2);
571
572 #if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
573 union {
574 unsigned int napi_id;
575 dma_cookie_t dma_cookie;
576 };
577 #endif
578 #ifdef CONFIG_NETWORK_SECMARK
579 __u32 secmark;
580 #endif
581 union {
582 __u32 mark;
583 __u32 dropcount;
584 __u32 reserved_tailroom;
585 };
586
587 __be16 inner_protocol;
588 __u16 inner_transport_header;
589 __u16 inner_network_header;
590 __u16 inner_mac_header;
591 __u16 transport_header;
592 __u16 network_header;
593 __u16 mac_header;
594 /* These elements must be at the end, see alloc_skb() for details. */
595 sk_buff_data_t tail;
596 sk_buff_data_t end;
597 unsigned char *head,
598 *data;
599 unsigned int truesize;
600 atomic_t users;
601 };
602
603 #ifdef __KERNEL__
604 /*
605 * Handling routines are only of interest to the kernel
606 */
607 #include <linux/slab.h>
608
609
610 #define SKB_ALLOC_FCLONE 0x01
611 #define SKB_ALLOC_RX 0x02
612
613 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
614 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
615 {
616 return unlikely(skb->pfmemalloc);
617 }
618
619 /*
620 * skb might have a dst pointer attached, refcounted or not.
621 * _skb_refdst low order bit is set if refcount was _not_ taken
622 */
623 #define SKB_DST_NOREF 1UL
624 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
625
626 /**
627 * skb_dst - returns skb dst_entry
628 * @skb: buffer
629 *
630 * Returns skb dst_entry, regardless of reference taken or not.
631 */
632 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
633 {
634 /* If refdst was not refcounted, check we still are in a
635 * rcu_read_lock section
636 */
637 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
638 !rcu_read_lock_held() &&
639 !rcu_read_lock_bh_held());
640 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
641 }
642
643 /**
644 * skb_dst_set - sets skb dst
645 * @skb: buffer
646 * @dst: dst entry
647 *
648 * Sets skb dst, assuming a reference was taken on dst and should
649 * be released by skb_dst_drop()
650 */
651 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
652 {
653 skb->_skb_refdst = (unsigned long)dst;
654 }
655
656 void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
657 bool force);
658
659 /**
660 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
661 * @skb: buffer
662 * @dst: dst entry
663 *
664 * Sets skb dst, assuming a reference was not taken on dst.
665 * If dst entry is cached, we do not take reference and dst_release
666 * will be avoided by refdst_drop. If dst entry is not cached, we take
667 * reference, so that last dst_release can destroy the dst immediately.
668 */
669 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
670 {
671 __skb_dst_set_noref(skb, dst, false);
672 }
673
674 /**
675 * skb_dst_set_noref_force - sets skb dst, without taking reference
676 * @skb: buffer
677 * @dst: dst entry
678 *
679 * Sets skb dst, assuming a reference was not taken on dst.
680 * No reference is taken and no dst_release will be called. While for
681 * cached dsts deferred reclaim is a basic feature, for entries that are
682 * not cached it is caller's job to guarantee that last dst_release for
683 * provided dst happens when nobody uses it, eg. after a RCU grace period.
684 */
685 static inline void skb_dst_set_noref_force(struct sk_buff *skb,
686 struct dst_entry *dst)
687 {
688 __skb_dst_set_noref(skb, dst, true);
689 }
690
691 /**
692 * skb_dst_is_noref - Test if skb dst isn't refcounted
693 * @skb: buffer
694 */
695 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
696 {
697 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
698 }
699
700 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
701 {
702 return (struct rtable *)skb_dst(skb);
703 }
704
705 void kfree_skb(struct sk_buff *skb);
706 void kfree_skb_list(struct sk_buff *segs);
707 void skb_tx_error(struct sk_buff *skb);
708 void consume_skb(struct sk_buff *skb);
709 void __kfree_skb(struct sk_buff *skb);
710 extern struct kmem_cache *skbuff_head_cache;
711
712 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
713 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
714 bool *fragstolen, int *delta_truesize);
715
716 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
717 int node);
718 struct sk_buff *build_skb(void *data, unsigned int frag_size);
719 static inline struct sk_buff *alloc_skb(unsigned int size,
720 gfp_t priority)
721 {
722 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
723 }
724
725 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
726 gfp_t priority)
727 {
728 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
729 }
730
731 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
732 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
733 {
734 return __alloc_skb_head(priority, -1);
735 }
736
737 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
738 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
739 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
740 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
741 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
742 gfp_t gfp_mask, bool fclone);
743 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
744 gfp_t gfp_mask)
745 {
746 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
747 }
748
749 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
750 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
751 unsigned int headroom);
752 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
753 int newtailroom, gfp_t priority);
754 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
755 int offset, int len);
756 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
757 int len);
758 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
759 int skb_pad(struct sk_buff *skb, int pad);
760 #define dev_kfree_skb(a) consume_skb(a)
761
762 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
763 int getfrag(void *from, char *to, int offset,
764 int len, int odd, struct sk_buff *skb),
765 void *from, int length);
766
767 struct skb_seq_state {
768 __u32 lower_offset;
769 __u32 upper_offset;
770 __u32 frag_idx;
771 __u32 stepped_offset;
772 struct sk_buff *root_skb;
773 struct sk_buff *cur_skb;
774 __u8 *frag_data;
775 };
776
777 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
778 unsigned int to, struct skb_seq_state *st);
779 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
780 struct skb_seq_state *st);
781 void skb_abort_seq_read(struct skb_seq_state *st);
782
783 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
784 unsigned int to, struct ts_config *config,
785 struct ts_state *state);
786
787 /*
788 * Packet hash types specify the type of hash in skb_set_hash.
789 *
790 * Hash types refer to the protocol layer addresses which are used to
791 * construct a packet's hash. The hashes are used to differentiate or identify
792 * flows of the protocol layer for the hash type. Hash types are either
793 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
794 *
795 * Properties of hashes:
796 *
797 * 1) Two packets in different flows have different hash values
798 * 2) Two packets in the same flow should have the same hash value
799 *
800 * A hash at a higher layer is considered to be more specific. A driver should
801 * set the most specific hash possible.
802 *
803 * A driver cannot indicate a more specific hash than the layer at which a hash
804 * was computed. For instance an L3 hash cannot be set as an L4 hash.
805 *
806 * A driver may indicate a hash level which is less specific than the
807 * actual layer the hash was computed on. For instance, a hash computed
808 * at L4 may be considered an L3 hash. This should only be done if the
809 * driver can't unambiguously determine that the HW computed the hash at
810 * the higher layer. Note that the "should" in the second property above
811 * permits this.
812 */
813 enum pkt_hash_types {
814 PKT_HASH_TYPE_NONE, /* Undefined type */
815 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
816 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
817 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
818 };
819
820 static inline void
821 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
822 {
823 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
824 skb->sw_hash = 0;
825 skb->hash = hash;
826 }
827
828 void __skb_get_hash(struct sk_buff *skb);
829 static inline __u32 skb_get_hash(struct sk_buff *skb)
830 {
831 if (!skb->l4_hash && !skb->sw_hash)
832 __skb_get_hash(skb);
833
834 return skb->hash;
835 }
836
837 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
838 {
839 return skb->hash;
840 }
841
842 static inline void skb_clear_hash(struct sk_buff *skb)
843 {
844 skb->hash = 0;
845 skb->sw_hash = 0;
846 skb->l4_hash = 0;
847 }
848
849 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
850 {
851 if (!skb->l4_hash)
852 skb_clear_hash(skb);
853 }
854
855 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
856 {
857 to->hash = from->hash;
858 to->sw_hash = from->sw_hash;
859 to->l4_hash = from->l4_hash;
860 };
861
862 #ifdef NET_SKBUFF_DATA_USES_OFFSET
863 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
864 {
865 return skb->head + skb->end;
866 }
867
868 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
869 {
870 return skb->end;
871 }
872 #else
873 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
874 {
875 return skb->end;
876 }
877
878 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
879 {
880 return skb->end - skb->head;
881 }
882 #endif
883
884 /* Internal */
885 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
886
887 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
888 {
889 return &skb_shinfo(skb)->hwtstamps;
890 }
891
892 /**
893 * skb_queue_empty - check if a queue is empty
894 * @list: queue head
895 *
896 * Returns true if the queue is empty, false otherwise.
897 */
898 static inline int skb_queue_empty(const struct sk_buff_head *list)
899 {
900 return list->next == (const struct sk_buff *) list;
901 }
902
903 /**
904 * skb_queue_is_last - check if skb is the last entry in the queue
905 * @list: queue head
906 * @skb: buffer
907 *
908 * Returns true if @skb is the last buffer on the list.
909 */
910 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
911 const struct sk_buff *skb)
912 {
913 return skb->next == (const struct sk_buff *) list;
914 }
915
916 /**
917 * skb_queue_is_first - check if skb is the first entry in the queue
918 * @list: queue head
919 * @skb: buffer
920 *
921 * Returns true if @skb is the first buffer on the list.
922 */
923 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
924 const struct sk_buff *skb)
925 {
926 return skb->prev == (const struct sk_buff *) list;
927 }
928
929 /**
930 * skb_queue_next - return the next packet in the queue
931 * @list: queue head
932 * @skb: current buffer
933 *
934 * Return the next packet in @list after @skb. It is only valid to
935 * call this if skb_queue_is_last() evaluates to false.
936 */
937 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
938 const struct sk_buff *skb)
939 {
940 /* This BUG_ON may seem severe, but if we just return then we
941 * are going to dereference garbage.
942 */
943 BUG_ON(skb_queue_is_last(list, skb));
944 return skb->next;
945 }
946
947 /**
948 * skb_queue_prev - return the prev packet in the queue
949 * @list: queue head
950 * @skb: current buffer
951 *
952 * Return the prev packet in @list before @skb. It is only valid to
953 * call this if skb_queue_is_first() evaluates to false.
954 */
955 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
956 const struct sk_buff *skb)
957 {
958 /* This BUG_ON may seem severe, but if we just return then we
959 * are going to dereference garbage.
960 */
961 BUG_ON(skb_queue_is_first(list, skb));
962 return skb->prev;
963 }
964
965 /**
966 * skb_get - reference buffer
967 * @skb: buffer to reference
968 *
969 * Makes another reference to a socket buffer and returns a pointer
970 * to the buffer.
971 */
972 static inline struct sk_buff *skb_get(struct sk_buff *skb)
973 {
974 atomic_inc(&skb->users);
975 return skb;
976 }
977
978 /*
979 * If users == 1, we are the only owner and are can avoid redundant
980 * atomic change.
981 */
982
983 /**
984 * skb_cloned - is the buffer a clone
985 * @skb: buffer to check
986 *
987 * Returns true if the buffer was generated with skb_clone() and is
988 * one of multiple shared copies of the buffer. Cloned buffers are
989 * shared data so must not be written to under normal circumstances.
990 */
991 static inline int skb_cloned(const struct sk_buff *skb)
992 {
993 return skb->cloned &&
994 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
995 }
996
997 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
998 {
999 might_sleep_if(pri & __GFP_WAIT);
1000
1001 if (skb_cloned(skb))
1002 return pskb_expand_head(skb, 0, 0, pri);
1003
1004 return 0;
1005 }
1006
1007 /**
1008 * skb_header_cloned - is the header a clone
1009 * @skb: buffer to check
1010 *
1011 * Returns true if modifying the header part of the buffer requires
1012 * the data to be copied.
1013 */
1014 static inline int skb_header_cloned(const struct sk_buff *skb)
1015 {
1016 int dataref;
1017
1018 if (!skb->cloned)
1019 return 0;
1020
1021 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1022 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1023 return dataref != 1;
1024 }
1025
1026 /**
1027 * skb_header_release - release reference to header
1028 * @skb: buffer to operate on
1029 *
1030 * Drop a reference to the header part of the buffer. This is done
1031 * by acquiring a payload reference. You must not read from the header
1032 * part of skb->data after this.
1033 */
1034 static inline void skb_header_release(struct sk_buff *skb)
1035 {
1036 BUG_ON(skb->nohdr);
1037 skb->nohdr = 1;
1038 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1039 }
1040
1041 /**
1042 * skb_shared - is the buffer shared
1043 * @skb: buffer to check
1044 *
1045 * Returns true if more than one person has a reference to this
1046 * buffer.
1047 */
1048 static inline int skb_shared(const struct sk_buff *skb)
1049 {
1050 return atomic_read(&skb->users) != 1;
1051 }
1052
1053 /**
1054 * skb_share_check - check if buffer is shared and if so clone it
1055 * @skb: buffer to check
1056 * @pri: priority for memory allocation
1057 *
1058 * If the buffer is shared the buffer is cloned and the old copy
1059 * drops a reference. A new clone with a single reference is returned.
1060 * If the buffer is not shared the original buffer is returned. When
1061 * being called from interrupt status or with spinlocks held pri must
1062 * be GFP_ATOMIC.
1063 *
1064 * NULL is returned on a memory allocation failure.
1065 */
1066 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1067 {
1068 might_sleep_if(pri & __GFP_WAIT);
1069 if (skb_shared(skb)) {
1070 struct sk_buff *nskb = skb_clone(skb, pri);
1071
1072 if (likely(nskb))
1073 consume_skb(skb);
1074 else
1075 kfree_skb(skb);
1076 skb = nskb;
1077 }
1078 return skb;
1079 }
1080
1081 /*
1082 * Copy shared buffers into a new sk_buff. We effectively do COW on
1083 * packets to handle cases where we have a local reader and forward
1084 * and a couple of other messy ones. The normal one is tcpdumping
1085 * a packet thats being forwarded.
1086 */
1087
1088 /**
1089 * skb_unshare - make a copy of a shared buffer
1090 * @skb: buffer to check
1091 * @pri: priority for memory allocation
1092 *
1093 * If the socket buffer is a clone then this function creates a new
1094 * copy of the data, drops a reference count on the old copy and returns
1095 * the new copy with the reference count at 1. If the buffer is not a clone
1096 * the original buffer is returned. When called with a spinlock held or
1097 * from interrupt state @pri must be %GFP_ATOMIC
1098 *
1099 * %NULL is returned on a memory allocation failure.
1100 */
1101 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1102 gfp_t pri)
1103 {
1104 might_sleep_if(pri & __GFP_WAIT);
1105 if (skb_cloned(skb)) {
1106 struct sk_buff *nskb = skb_copy(skb, pri);
1107 kfree_skb(skb); /* Free our shared copy */
1108 skb = nskb;
1109 }
1110 return skb;
1111 }
1112
1113 /**
1114 * skb_peek - peek at the head of an &sk_buff_head
1115 * @list_: list to peek at
1116 *
1117 * Peek an &sk_buff. Unlike most other operations you _MUST_
1118 * be careful with this one. A peek leaves the buffer on the
1119 * list and someone else may run off with it. You must hold
1120 * the appropriate locks or have a private queue to do this.
1121 *
1122 * Returns %NULL for an empty list or a pointer to the head element.
1123 * The reference count is not incremented and the reference is therefore
1124 * volatile. Use with caution.
1125 */
1126 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1127 {
1128 struct sk_buff *skb = list_->next;
1129
1130 if (skb == (struct sk_buff *)list_)
1131 skb = NULL;
1132 return skb;
1133 }
1134
1135 /**
1136 * skb_peek_next - peek skb following the given one from a queue
1137 * @skb: skb to start from
1138 * @list_: list to peek at
1139 *
1140 * Returns %NULL when the end of the list is met or a pointer to the
1141 * next element. The reference count is not incremented and the
1142 * reference is therefore volatile. Use with caution.
1143 */
1144 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1145 const struct sk_buff_head *list_)
1146 {
1147 struct sk_buff *next = skb->next;
1148
1149 if (next == (struct sk_buff *)list_)
1150 next = NULL;
1151 return next;
1152 }
1153
1154 /**
1155 * skb_peek_tail - peek at the tail of an &sk_buff_head
1156 * @list_: list to peek at
1157 *
1158 * Peek an &sk_buff. Unlike most other operations you _MUST_
1159 * be careful with this one. A peek leaves the buffer on the
1160 * list and someone else may run off with it. You must hold
1161 * the appropriate locks or have a private queue to do this.
1162 *
1163 * Returns %NULL for an empty list or a pointer to the tail element.
1164 * The reference count is not incremented and the reference is therefore
1165 * volatile. Use with caution.
1166 */
1167 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1168 {
1169 struct sk_buff *skb = list_->prev;
1170
1171 if (skb == (struct sk_buff *)list_)
1172 skb = NULL;
1173 return skb;
1174
1175 }
1176
1177 /**
1178 * skb_queue_len - get queue length
1179 * @list_: list to measure
1180 *
1181 * Return the length of an &sk_buff queue.
1182 */
1183 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1184 {
1185 return list_->qlen;
1186 }
1187
1188 /**
1189 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1190 * @list: queue to initialize
1191 *
1192 * This initializes only the list and queue length aspects of
1193 * an sk_buff_head object. This allows to initialize the list
1194 * aspects of an sk_buff_head without reinitializing things like
1195 * the spinlock. It can also be used for on-stack sk_buff_head
1196 * objects where the spinlock is known to not be used.
1197 */
1198 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1199 {
1200 list->prev = list->next = (struct sk_buff *)list;
1201 list->qlen = 0;
1202 }
1203
1204 /*
1205 * This function creates a split out lock class for each invocation;
1206 * this is needed for now since a whole lot of users of the skb-queue
1207 * infrastructure in drivers have different locking usage (in hardirq)
1208 * than the networking core (in softirq only). In the long run either the
1209 * network layer or drivers should need annotation to consolidate the
1210 * main types of usage into 3 classes.
1211 */
1212 static inline void skb_queue_head_init(struct sk_buff_head *list)
1213 {
1214 spin_lock_init(&list->lock);
1215 __skb_queue_head_init(list);
1216 }
1217
1218 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1219 struct lock_class_key *class)
1220 {
1221 skb_queue_head_init(list);
1222 lockdep_set_class(&list->lock, class);
1223 }
1224
1225 /*
1226 * Insert an sk_buff on a list.
1227 *
1228 * The "__skb_xxxx()" functions are the non-atomic ones that
1229 * can only be called with interrupts disabled.
1230 */
1231 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1232 struct sk_buff_head *list);
1233 static inline void __skb_insert(struct sk_buff *newsk,
1234 struct sk_buff *prev, struct sk_buff *next,
1235 struct sk_buff_head *list)
1236 {
1237 newsk->next = next;
1238 newsk->prev = prev;
1239 next->prev = prev->next = newsk;
1240 list->qlen++;
1241 }
1242
1243 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1244 struct sk_buff *prev,
1245 struct sk_buff *next)
1246 {
1247 struct sk_buff *first = list->next;
1248 struct sk_buff *last = list->prev;
1249
1250 first->prev = prev;
1251 prev->next = first;
1252
1253 last->next = next;
1254 next->prev = last;
1255 }
1256
1257 /**
1258 * skb_queue_splice - join two skb lists, this is designed for stacks
1259 * @list: the new list to add
1260 * @head: the place to add it in the first list
1261 */
1262 static inline void skb_queue_splice(const struct sk_buff_head *list,
1263 struct sk_buff_head *head)
1264 {
1265 if (!skb_queue_empty(list)) {
1266 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1267 head->qlen += list->qlen;
1268 }
1269 }
1270
1271 /**
1272 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1273 * @list: the new list to add
1274 * @head: the place to add it in the first list
1275 *
1276 * The list at @list is reinitialised
1277 */
1278 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1279 struct sk_buff_head *head)
1280 {
1281 if (!skb_queue_empty(list)) {
1282 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1283 head->qlen += list->qlen;
1284 __skb_queue_head_init(list);
1285 }
1286 }
1287
1288 /**
1289 * skb_queue_splice_tail - join two skb lists, each list being a queue
1290 * @list: the new list to add
1291 * @head: the place to add it in the first list
1292 */
1293 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1294 struct sk_buff_head *head)
1295 {
1296 if (!skb_queue_empty(list)) {
1297 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1298 head->qlen += list->qlen;
1299 }
1300 }
1301
1302 /**
1303 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1304 * @list: the new list to add
1305 * @head: the place to add it in the first list
1306 *
1307 * Each of the lists is a queue.
1308 * The list at @list is reinitialised
1309 */
1310 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1311 struct sk_buff_head *head)
1312 {
1313 if (!skb_queue_empty(list)) {
1314 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1315 head->qlen += list->qlen;
1316 __skb_queue_head_init(list);
1317 }
1318 }
1319
1320 /**
1321 * __skb_queue_after - queue a buffer at the list head
1322 * @list: list to use
1323 * @prev: place after this buffer
1324 * @newsk: buffer to queue
1325 *
1326 * Queue a buffer int the middle of a list. This function takes no locks
1327 * and you must therefore hold required locks before calling it.
1328 *
1329 * A buffer cannot be placed on two lists at the same time.
1330 */
1331 static inline void __skb_queue_after(struct sk_buff_head *list,
1332 struct sk_buff *prev,
1333 struct sk_buff *newsk)
1334 {
1335 __skb_insert(newsk, prev, prev->next, list);
1336 }
1337
1338 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1339 struct sk_buff_head *list);
1340
1341 static inline void __skb_queue_before(struct sk_buff_head *list,
1342 struct sk_buff *next,
1343 struct sk_buff *newsk)
1344 {
1345 __skb_insert(newsk, next->prev, next, list);
1346 }
1347
1348 /**
1349 * __skb_queue_head - queue a buffer at the list head
1350 * @list: list to use
1351 * @newsk: buffer to queue
1352 *
1353 * Queue a buffer at the start of a list. This function takes no locks
1354 * and you must therefore hold required locks before calling it.
1355 *
1356 * A buffer cannot be placed on two lists at the same time.
1357 */
1358 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1359 static inline void __skb_queue_head(struct sk_buff_head *list,
1360 struct sk_buff *newsk)
1361 {
1362 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1363 }
1364
1365 /**
1366 * __skb_queue_tail - queue a buffer at the list tail
1367 * @list: list to use
1368 * @newsk: buffer to queue
1369 *
1370 * Queue a buffer at the end of a list. This function takes no locks
1371 * and you must therefore hold required locks before calling it.
1372 *
1373 * A buffer cannot be placed on two lists at the same time.
1374 */
1375 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1376 static inline void __skb_queue_tail(struct sk_buff_head *list,
1377 struct sk_buff *newsk)
1378 {
1379 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1380 }
1381
1382 /*
1383 * remove sk_buff from list. _Must_ be called atomically, and with
1384 * the list known..
1385 */
1386 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1387 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1388 {
1389 struct sk_buff *next, *prev;
1390
1391 list->qlen--;
1392 next = skb->next;
1393 prev = skb->prev;
1394 skb->next = skb->prev = NULL;
1395 next->prev = prev;
1396 prev->next = next;
1397 }
1398
1399 /**
1400 * __skb_dequeue - remove from the head of the queue
1401 * @list: list to dequeue from
1402 *
1403 * Remove the head of the list. This function does not take any locks
1404 * so must be used with appropriate locks held only. The head item is
1405 * returned or %NULL if the list is empty.
1406 */
1407 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1408 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1409 {
1410 struct sk_buff *skb = skb_peek(list);
1411 if (skb)
1412 __skb_unlink(skb, list);
1413 return skb;
1414 }
1415
1416 /**
1417 * __skb_dequeue_tail - remove from the tail of the queue
1418 * @list: list to dequeue from
1419 *
1420 * Remove the tail of the list. This function does not take any locks
1421 * so must be used with appropriate locks held only. The tail item is
1422 * returned or %NULL if the list is empty.
1423 */
1424 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1425 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1426 {
1427 struct sk_buff *skb = skb_peek_tail(list);
1428 if (skb)
1429 __skb_unlink(skb, list);
1430 return skb;
1431 }
1432
1433
1434 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1435 {
1436 return skb->data_len;
1437 }
1438
1439 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1440 {
1441 return skb->len - skb->data_len;
1442 }
1443
1444 static inline int skb_pagelen(const struct sk_buff *skb)
1445 {
1446 int i, len = 0;
1447
1448 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1449 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1450 return len + skb_headlen(skb);
1451 }
1452
1453 /**
1454 * __skb_fill_page_desc - initialise a paged fragment in an skb
1455 * @skb: buffer containing fragment to be initialised
1456 * @i: paged fragment index to initialise
1457 * @page: the page to use for this fragment
1458 * @off: the offset to the data with @page
1459 * @size: the length of the data
1460 *
1461 * Initialises the @i'th fragment of @skb to point to &size bytes at
1462 * offset @off within @page.
1463 *
1464 * Does not take any additional reference on the fragment.
1465 */
1466 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1467 struct page *page, int off, int size)
1468 {
1469 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1470
1471 /*
1472 * Propagate page->pfmemalloc to the skb if we can. The problem is
1473 * that not all callers have unique ownership of the page. If
1474 * pfmemalloc is set, we check the mapping as a mapping implies
1475 * page->index is set (index and pfmemalloc share space).
1476 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1477 * do not lose pfmemalloc information as the pages would not be
1478 * allocated using __GFP_MEMALLOC.
1479 */
1480 frag->page.p = page;
1481 frag->page_offset = off;
1482 skb_frag_size_set(frag, size);
1483
1484 page = compound_head(page);
1485 if (page->pfmemalloc && !page->mapping)
1486 skb->pfmemalloc = true;
1487 }
1488
1489 /**
1490 * skb_fill_page_desc - initialise a paged fragment in an skb
1491 * @skb: buffer containing fragment to be initialised
1492 * @i: paged fragment index to initialise
1493 * @page: the page to use for this fragment
1494 * @off: the offset to the data with @page
1495 * @size: the length of the data
1496 *
1497 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1498 * @skb to point to @size bytes at offset @off within @page. In
1499 * addition updates @skb such that @i is the last fragment.
1500 *
1501 * Does not take any additional reference on the fragment.
1502 */
1503 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1504 struct page *page, int off, int size)
1505 {
1506 __skb_fill_page_desc(skb, i, page, off, size);
1507 skb_shinfo(skb)->nr_frags = i + 1;
1508 }
1509
1510 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1511 int size, unsigned int truesize);
1512
1513 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1514 unsigned int truesize);
1515
1516 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1517 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1518 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1519
1520 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1521 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1522 {
1523 return skb->head + skb->tail;
1524 }
1525
1526 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1527 {
1528 skb->tail = skb->data - skb->head;
1529 }
1530
1531 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1532 {
1533 skb_reset_tail_pointer(skb);
1534 skb->tail += offset;
1535 }
1536
1537 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1538 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1539 {
1540 return skb->tail;
1541 }
1542
1543 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1544 {
1545 skb->tail = skb->data;
1546 }
1547
1548 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1549 {
1550 skb->tail = skb->data + offset;
1551 }
1552
1553 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1554
1555 /*
1556 * Add data to an sk_buff
1557 */
1558 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1559 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1560 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1561 {
1562 unsigned char *tmp = skb_tail_pointer(skb);
1563 SKB_LINEAR_ASSERT(skb);
1564 skb->tail += len;
1565 skb->len += len;
1566 return tmp;
1567 }
1568
1569 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1570 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1571 {
1572 skb->data -= len;
1573 skb->len += len;
1574 return skb->data;
1575 }
1576
1577 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1578 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1579 {
1580 skb->len -= len;
1581 BUG_ON(skb->len < skb->data_len);
1582 return skb->data += len;
1583 }
1584
1585 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1586 {
1587 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1588 }
1589
1590 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1591
1592 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1593 {
1594 if (len > skb_headlen(skb) &&
1595 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1596 return NULL;
1597 skb->len -= len;
1598 return skb->data += len;
1599 }
1600
1601 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1602 {
1603 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1604 }
1605
1606 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1607 {
1608 if (likely(len <= skb_headlen(skb)))
1609 return 1;
1610 if (unlikely(len > skb->len))
1611 return 0;
1612 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1613 }
1614
1615 /**
1616 * skb_headroom - bytes at buffer head
1617 * @skb: buffer to check
1618 *
1619 * Return the number of bytes of free space at the head of an &sk_buff.
1620 */
1621 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1622 {
1623 return skb->data - skb->head;
1624 }
1625
1626 /**
1627 * skb_tailroom - bytes at buffer end
1628 * @skb: buffer to check
1629 *
1630 * Return the number of bytes of free space at the tail of an sk_buff
1631 */
1632 static inline int skb_tailroom(const struct sk_buff *skb)
1633 {
1634 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1635 }
1636
1637 /**
1638 * skb_availroom - bytes at buffer end
1639 * @skb: buffer to check
1640 *
1641 * Return the number of bytes of free space at the tail of an sk_buff
1642 * allocated by sk_stream_alloc()
1643 */
1644 static inline int skb_availroom(const struct sk_buff *skb)
1645 {
1646 if (skb_is_nonlinear(skb))
1647 return 0;
1648
1649 return skb->end - skb->tail - skb->reserved_tailroom;
1650 }
1651
1652 /**
1653 * skb_reserve - adjust headroom
1654 * @skb: buffer to alter
1655 * @len: bytes to move
1656 *
1657 * Increase the headroom of an empty &sk_buff by reducing the tail
1658 * room. This is only allowed for an empty buffer.
1659 */
1660 static inline void skb_reserve(struct sk_buff *skb, int len)
1661 {
1662 skb->data += len;
1663 skb->tail += len;
1664 }
1665
1666 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1667 {
1668 skb->inner_mac_header = skb->mac_header;
1669 skb->inner_network_header = skb->network_header;
1670 skb->inner_transport_header = skb->transport_header;
1671 }
1672
1673 static inline void skb_reset_mac_len(struct sk_buff *skb)
1674 {
1675 skb->mac_len = skb->network_header - skb->mac_header;
1676 }
1677
1678 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1679 *skb)
1680 {
1681 return skb->head + skb->inner_transport_header;
1682 }
1683
1684 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1685 {
1686 skb->inner_transport_header = skb->data - skb->head;
1687 }
1688
1689 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1690 const int offset)
1691 {
1692 skb_reset_inner_transport_header(skb);
1693 skb->inner_transport_header += offset;
1694 }
1695
1696 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1697 {
1698 return skb->head + skb->inner_network_header;
1699 }
1700
1701 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1702 {
1703 skb->inner_network_header = skb->data - skb->head;
1704 }
1705
1706 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1707 const int offset)
1708 {
1709 skb_reset_inner_network_header(skb);
1710 skb->inner_network_header += offset;
1711 }
1712
1713 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1714 {
1715 return skb->head + skb->inner_mac_header;
1716 }
1717
1718 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1719 {
1720 skb->inner_mac_header = skb->data - skb->head;
1721 }
1722
1723 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1724 const int offset)
1725 {
1726 skb_reset_inner_mac_header(skb);
1727 skb->inner_mac_header += offset;
1728 }
1729 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1730 {
1731 return skb->transport_header != (typeof(skb->transport_header))~0U;
1732 }
1733
1734 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1735 {
1736 return skb->head + skb->transport_header;
1737 }
1738
1739 static inline void skb_reset_transport_header(struct sk_buff *skb)
1740 {
1741 skb->transport_header = skb->data - skb->head;
1742 }
1743
1744 static inline void skb_set_transport_header(struct sk_buff *skb,
1745 const int offset)
1746 {
1747 skb_reset_transport_header(skb);
1748 skb->transport_header += offset;
1749 }
1750
1751 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1752 {
1753 return skb->head + skb->network_header;
1754 }
1755
1756 static inline void skb_reset_network_header(struct sk_buff *skb)
1757 {
1758 skb->network_header = skb->data - skb->head;
1759 }
1760
1761 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1762 {
1763 skb_reset_network_header(skb);
1764 skb->network_header += offset;
1765 }
1766
1767 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1768 {
1769 return skb->head + skb->mac_header;
1770 }
1771
1772 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1773 {
1774 return skb->mac_header != (typeof(skb->mac_header))~0U;
1775 }
1776
1777 static inline void skb_reset_mac_header(struct sk_buff *skb)
1778 {
1779 skb->mac_header = skb->data - skb->head;
1780 }
1781
1782 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1783 {
1784 skb_reset_mac_header(skb);
1785 skb->mac_header += offset;
1786 }
1787
1788 static inline void skb_pop_mac_header(struct sk_buff *skb)
1789 {
1790 skb->mac_header = skb->network_header;
1791 }
1792
1793 static inline void skb_probe_transport_header(struct sk_buff *skb,
1794 const int offset_hint)
1795 {
1796 struct flow_keys keys;
1797
1798 if (skb_transport_header_was_set(skb))
1799 return;
1800 else if (skb_flow_dissect(skb, &keys))
1801 skb_set_transport_header(skb, keys.thoff);
1802 else
1803 skb_set_transport_header(skb, offset_hint);
1804 }
1805
1806 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1807 {
1808 if (skb_mac_header_was_set(skb)) {
1809 const unsigned char *old_mac = skb_mac_header(skb);
1810
1811 skb_set_mac_header(skb, -skb->mac_len);
1812 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1813 }
1814 }
1815
1816 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1817 {
1818 return skb->csum_start - skb_headroom(skb);
1819 }
1820
1821 static inline int skb_transport_offset(const struct sk_buff *skb)
1822 {
1823 return skb_transport_header(skb) - skb->data;
1824 }
1825
1826 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1827 {
1828 return skb->transport_header - skb->network_header;
1829 }
1830
1831 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1832 {
1833 return skb->inner_transport_header - skb->inner_network_header;
1834 }
1835
1836 static inline int skb_network_offset(const struct sk_buff *skb)
1837 {
1838 return skb_network_header(skb) - skb->data;
1839 }
1840
1841 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1842 {
1843 return skb_inner_network_header(skb) - skb->data;
1844 }
1845
1846 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1847 {
1848 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1849 }
1850
1851 static inline void skb_pop_rcv_encapsulation(struct sk_buff *skb)
1852 {
1853 /* Only continue with checksum unnecessary if device indicated
1854 * it is valid across encapsulation (skb->encapsulation was set).
1855 */
1856 if (skb->ip_summed == CHECKSUM_UNNECESSARY && !skb->encapsulation)
1857 skb->ip_summed = CHECKSUM_NONE;
1858
1859 skb->encapsulation = 0;
1860 skb->csum_valid = 0;
1861 }
1862
1863 /*
1864 * CPUs often take a performance hit when accessing unaligned memory
1865 * locations. The actual performance hit varies, it can be small if the
1866 * hardware handles it or large if we have to take an exception and fix it
1867 * in software.
1868 *
1869 * Since an ethernet header is 14 bytes network drivers often end up with
1870 * the IP header at an unaligned offset. The IP header can be aligned by
1871 * shifting the start of the packet by 2 bytes. Drivers should do this
1872 * with:
1873 *
1874 * skb_reserve(skb, NET_IP_ALIGN);
1875 *
1876 * The downside to this alignment of the IP header is that the DMA is now
1877 * unaligned. On some architectures the cost of an unaligned DMA is high
1878 * and this cost outweighs the gains made by aligning the IP header.
1879 *
1880 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1881 * to be overridden.
1882 */
1883 #ifndef NET_IP_ALIGN
1884 #define NET_IP_ALIGN 2
1885 #endif
1886
1887 /*
1888 * The networking layer reserves some headroom in skb data (via
1889 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1890 * the header has to grow. In the default case, if the header has to grow
1891 * 32 bytes or less we avoid the reallocation.
1892 *
1893 * Unfortunately this headroom changes the DMA alignment of the resulting
1894 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1895 * on some architectures. An architecture can override this value,
1896 * perhaps setting it to a cacheline in size (since that will maintain
1897 * cacheline alignment of the DMA). It must be a power of 2.
1898 *
1899 * Various parts of the networking layer expect at least 32 bytes of
1900 * headroom, you should not reduce this.
1901 *
1902 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1903 * to reduce average number of cache lines per packet.
1904 * get_rps_cpus() for example only access one 64 bytes aligned block :
1905 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1906 */
1907 #ifndef NET_SKB_PAD
1908 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1909 #endif
1910
1911 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1912
1913 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1914 {
1915 if (unlikely(skb_is_nonlinear(skb))) {
1916 WARN_ON(1);
1917 return;
1918 }
1919 skb->len = len;
1920 skb_set_tail_pointer(skb, len);
1921 }
1922
1923 void skb_trim(struct sk_buff *skb, unsigned int len);
1924
1925 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1926 {
1927 if (skb->data_len)
1928 return ___pskb_trim(skb, len);
1929 __skb_trim(skb, len);
1930 return 0;
1931 }
1932
1933 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1934 {
1935 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1936 }
1937
1938 /**
1939 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1940 * @skb: buffer to alter
1941 * @len: new length
1942 *
1943 * This is identical to pskb_trim except that the caller knows that
1944 * the skb is not cloned so we should never get an error due to out-
1945 * of-memory.
1946 */
1947 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1948 {
1949 int err = pskb_trim(skb, len);
1950 BUG_ON(err);
1951 }
1952
1953 /**
1954 * skb_orphan - orphan a buffer
1955 * @skb: buffer to orphan
1956 *
1957 * If a buffer currently has an owner then we call the owner's
1958 * destructor function and make the @skb unowned. The buffer continues
1959 * to exist but is no longer charged to its former owner.
1960 */
1961 static inline void skb_orphan(struct sk_buff *skb)
1962 {
1963 if (skb->destructor) {
1964 skb->destructor(skb);
1965 skb->destructor = NULL;
1966 skb->sk = NULL;
1967 } else {
1968 BUG_ON(skb->sk);
1969 }
1970 }
1971
1972 /**
1973 * skb_orphan_frags - orphan the frags contained in a buffer
1974 * @skb: buffer to orphan frags from
1975 * @gfp_mask: allocation mask for replacement pages
1976 *
1977 * For each frag in the SKB which needs a destructor (i.e. has an
1978 * owner) create a copy of that frag and release the original
1979 * page by calling the destructor.
1980 */
1981 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1982 {
1983 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1984 return 0;
1985 return skb_copy_ubufs(skb, gfp_mask);
1986 }
1987
1988 /**
1989 * __skb_queue_purge - empty a list
1990 * @list: list to empty
1991 *
1992 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1993 * the list and one reference dropped. This function does not take the
1994 * list lock and the caller must hold the relevant locks to use it.
1995 */
1996 void skb_queue_purge(struct sk_buff_head *list);
1997 static inline void __skb_queue_purge(struct sk_buff_head *list)
1998 {
1999 struct sk_buff *skb;
2000 while ((skb = __skb_dequeue(list)) != NULL)
2001 kfree_skb(skb);
2002 }
2003
2004 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2005 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2006 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2007
2008 void *netdev_alloc_frag(unsigned int fragsz);
2009
2010 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2011 gfp_t gfp_mask);
2012
2013 /**
2014 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2015 * @dev: network device to receive on
2016 * @length: length to allocate
2017 *
2018 * Allocate a new &sk_buff and assign it a usage count of one. The
2019 * buffer has unspecified headroom built in. Users should allocate
2020 * the headroom they think they need without accounting for the
2021 * built in space. The built in space is used for optimisations.
2022 *
2023 * %NULL is returned if there is no free memory. Although this function
2024 * allocates memory it can be called from an interrupt.
2025 */
2026 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2027 unsigned int length)
2028 {
2029 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2030 }
2031
2032 /* legacy helper around __netdev_alloc_skb() */
2033 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2034 gfp_t gfp_mask)
2035 {
2036 return __netdev_alloc_skb(NULL, length, gfp_mask);
2037 }
2038
2039 /* legacy helper around netdev_alloc_skb() */
2040 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2041 {
2042 return netdev_alloc_skb(NULL, length);
2043 }
2044
2045
2046 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2047 unsigned int length, gfp_t gfp)
2048 {
2049 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2050
2051 if (NET_IP_ALIGN && skb)
2052 skb_reserve(skb, NET_IP_ALIGN);
2053 return skb;
2054 }
2055
2056 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2057 unsigned int length)
2058 {
2059 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2060 }
2061
2062 /**
2063 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
2064 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2065 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2066 * @order: size of the allocation
2067 *
2068 * Allocate a new page.
2069 *
2070 * %NULL is returned if there is no free memory.
2071 */
2072 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2073 struct sk_buff *skb,
2074 unsigned int order)
2075 {
2076 struct page *page;
2077
2078 gfp_mask |= __GFP_COLD;
2079
2080 if (!(gfp_mask & __GFP_NOMEMALLOC))
2081 gfp_mask |= __GFP_MEMALLOC;
2082
2083 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2084 if (skb && page && page->pfmemalloc)
2085 skb->pfmemalloc = true;
2086
2087 return page;
2088 }
2089
2090 /**
2091 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2092 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2093 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2094 *
2095 * Allocate a new page.
2096 *
2097 * %NULL is returned if there is no free memory.
2098 */
2099 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2100 struct sk_buff *skb)
2101 {
2102 return __skb_alloc_pages(gfp_mask, skb, 0);
2103 }
2104
2105 /**
2106 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2107 * @page: The page that was allocated from skb_alloc_page
2108 * @skb: The skb that may need pfmemalloc set
2109 */
2110 static inline void skb_propagate_pfmemalloc(struct page *page,
2111 struct sk_buff *skb)
2112 {
2113 if (page && page->pfmemalloc)
2114 skb->pfmemalloc = true;
2115 }
2116
2117 /**
2118 * skb_frag_page - retrieve the page referred to by a paged fragment
2119 * @frag: the paged fragment
2120 *
2121 * Returns the &struct page associated with @frag.
2122 */
2123 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2124 {
2125 return frag->page.p;
2126 }
2127
2128 /**
2129 * __skb_frag_ref - take an addition reference on a paged fragment.
2130 * @frag: the paged fragment
2131 *
2132 * Takes an additional reference on the paged fragment @frag.
2133 */
2134 static inline void __skb_frag_ref(skb_frag_t *frag)
2135 {
2136 get_page(skb_frag_page(frag));
2137 }
2138
2139 /**
2140 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2141 * @skb: the buffer
2142 * @f: the fragment offset.
2143 *
2144 * Takes an additional reference on the @f'th paged fragment of @skb.
2145 */
2146 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2147 {
2148 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2149 }
2150
2151 /**
2152 * __skb_frag_unref - release a reference on a paged fragment.
2153 * @frag: the paged fragment
2154 *
2155 * Releases a reference on the paged fragment @frag.
2156 */
2157 static inline void __skb_frag_unref(skb_frag_t *frag)
2158 {
2159 put_page(skb_frag_page(frag));
2160 }
2161
2162 /**
2163 * skb_frag_unref - release a reference on a paged fragment of an skb.
2164 * @skb: the buffer
2165 * @f: the fragment offset
2166 *
2167 * Releases a reference on the @f'th paged fragment of @skb.
2168 */
2169 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2170 {
2171 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2172 }
2173
2174 /**
2175 * skb_frag_address - gets the address of the data contained in a paged fragment
2176 * @frag: the paged fragment buffer
2177 *
2178 * Returns the address of the data within @frag. The page must already
2179 * be mapped.
2180 */
2181 static inline void *skb_frag_address(const skb_frag_t *frag)
2182 {
2183 return page_address(skb_frag_page(frag)) + frag->page_offset;
2184 }
2185
2186 /**
2187 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2188 * @frag: the paged fragment buffer
2189 *
2190 * Returns the address of the data within @frag. Checks that the page
2191 * is mapped and returns %NULL otherwise.
2192 */
2193 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2194 {
2195 void *ptr = page_address(skb_frag_page(frag));
2196 if (unlikely(!ptr))
2197 return NULL;
2198
2199 return ptr + frag->page_offset;
2200 }
2201
2202 /**
2203 * __skb_frag_set_page - sets the page contained in a paged fragment
2204 * @frag: the paged fragment
2205 * @page: the page to set
2206 *
2207 * Sets the fragment @frag to contain @page.
2208 */
2209 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2210 {
2211 frag->page.p = page;
2212 }
2213
2214 /**
2215 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2216 * @skb: the buffer
2217 * @f: the fragment offset
2218 * @page: the page to set
2219 *
2220 * Sets the @f'th fragment of @skb to contain @page.
2221 */
2222 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2223 struct page *page)
2224 {
2225 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2226 }
2227
2228 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2229
2230 /**
2231 * skb_frag_dma_map - maps a paged fragment via the DMA API
2232 * @dev: the device to map the fragment to
2233 * @frag: the paged fragment to map
2234 * @offset: the offset within the fragment (starting at the
2235 * fragment's own offset)
2236 * @size: the number of bytes to map
2237 * @dir: the direction of the mapping (%PCI_DMA_*)
2238 *
2239 * Maps the page associated with @frag to @device.
2240 */
2241 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2242 const skb_frag_t *frag,
2243 size_t offset, size_t size,
2244 enum dma_data_direction dir)
2245 {
2246 return dma_map_page(dev, skb_frag_page(frag),
2247 frag->page_offset + offset, size, dir);
2248 }
2249
2250 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2251 gfp_t gfp_mask)
2252 {
2253 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2254 }
2255
2256
2257 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2258 gfp_t gfp_mask)
2259 {
2260 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2261 }
2262
2263
2264 /**
2265 * skb_clone_writable - is the header of a clone writable
2266 * @skb: buffer to check
2267 * @len: length up to which to write
2268 *
2269 * Returns true if modifying the header part of the cloned buffer
2270 * does not requires the data to be copied.
2271 */
2272 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2273 {
2274 return !skb_header_cloned(skb) &&
2275 skb_headroom(skb) + len <= skb->hdr_len;
2276 }
2277
2278 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2279 int cloned)
2280 {
2281 int delta = 0;
2282
2283 if (headroom > skb_headroom(skb))
2284 delta = headroom - skb_headroom(skb);
2285
2286 if (delta || cloned)
2287 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2288 GFP_ATOMIC);
2289 return 0;
2290 }
2291
2292 /**
2293 * skb_cow - copy header of skb when it is required
2294 * @skb: buffer to cow
2295 * @headroom: needed headroom
2296 *
2297 * If the skb passed lacks sufficient headroom or its data part
2298 * is shared, data is reallocated. If reallocation fails, an error
2299 * is returned and original skb is not changed.
2300 *
2301 * The result is skb with writable area skb->head...skb->tail
2302 * and at least @headroom of space at head.
2303 */
2304 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2305 {
2306 return __skb_cow(skb, headroom, skb_cloned(skb));
2307 }
2308
2309 /**
2310 * skb_cow_head - skb_cow but only making the head writable
2311 * @skb: buffer to cow
2312 * @headroom: needed headroom
2313 *
2314 * This function is identical to skb_cow except that we replace the
2315 * skb_cloned check by skb_header_cloned. It should be used when
2316 * you only need to push on some header and do not need to modify
2317 * the data.
2318 */
2319 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2320 {
2321 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2322 }
2323
2324 /**
2325 * skb_padto - pad an skbuff up to a minimal size
2326 * @skb: buffer to pad
2327 * @len: minimal length
2328 *
2329 * Pads up a buffer to ensure the trailing bytes exist and are
2330 * blanked. If the buffer already contains sufficient data it
2331 * is untouched. Otherwise it is extended. Returns zero on
2332 * success. The skb is freed on error.
2333 */
2334
2335 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2336 {
2337 unsigned int size = skb->len;
2338 if (likely(size >= len))
2339 return 0;
2340 return skb_pad(skb, len - size);
2341 }
2342
2343 static inline int skb_add_data(struct sk_buff *skb,
2344 char __user *from, int copy)
2345 {
2346 const int off = skb->len;
2347
2348 if (skb->ip_summed == CHECKSUM_NONE) {
2349 int err = 0;
2350 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2351 copy, 0, &err);
2352 if (!err) {
2353 skb->csum = csum_block_add(skb->csum, csum, off);
2354 return 0;
2355 }
2356 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2357 return 0;
2358
2359 __skb_trim(skb, off);
2360 return -EFAULT;
2361 }
2362
2363 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2364 const struct page *page, int off)
2365 {
2366 if (i) {
2367 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2368
2369 return page == skb_frag_page(frag) &&
2370 off == frag->page_offset + skb_frag_size(frag);
2371 }
2372 return false;
2373 }
2374
2375 static inline int __skb_linearize(struct sk_buff *skb)
2376 {
2377 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2378 }
2379
2380 /**
2381 * skb_linearize - convert paged skb to linear one
2382 * @skb: buffer to linarize
2383 *
2384 * If there is no free memory -ENOMEM is returned, otherwise zero
2385 * is returned and the old skb data released.
2386 */
2387 static inline int skb_linearize(struct sk_buff *skb)
2388 {
2389 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2390 }
2391
2392 /**
2393 * skb_has_shared_frag - can any frag be overwritten
2394 * @skb: buffer to test
2395 *
2396 * Return true if the skb has at least one frag that might be modified
2397 * by an external entity (as in vmsplice()/sendfile())
2398 */
2399 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2400 {
2401 return skb_is_nonlinear(skb) &&
2402 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2403 }
2404
2405 /**
2406 * skb_linearize_cow - make sure skb is linear and writable
2407 * @skb: buffer to process
2408 *
2409 * If there is no free memory -ENOMEM is returned, otherwise zero
2410 * is returned and the old skb data released.
2411 */
2412 static inline int skb_linearize_cow(struct sk_buff *skb)
2413 {
2414 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2415 __skb_linearize(skb) : 0;
2416 }
2417
2418 /**
2419 * skb_postpull_rcsum - update checksum for received skb after pull
2420 * @skb: buffer to update
2421 * @start: start of data before pull
2422 * @len: length of data pulled
2423 *
2424 * After doing a pull on a received packet, you need to call this to
2425 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2426 * CHECKSUM_NONE so that it can be recomputed from scratch.
2427 */
2428
2429 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2430 const void *start, unsigned int len)
2431 {
2432 if (skb->ip_summed == CHECKSUM_COMPLETE)
2433 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2434 }
2435
2436 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2437
2438 /**
2439 * pskb_trim_rcsum - trim received skb and update checksum
2440 * @skb: buffer to trim
2441 * @len: new length
2442 *
2443 * This is exactly the same as pskb_trim except that it ensures the
2444 * checksum of received packets are still valid after the operation.
2445 */
2446
2447 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2448 {
2449 if (likely(len >= skb->len))
2450 return 0;
2451 if (skb->ip_summed == CHECKSUM_COMPLETE)
2452 skb->ip_summed = CHECKSUM_NONE;
2453 return __pskb_trim(skb, len);
2454 }
2455
2456 #define skb_queue_walk(queue, skb) \
2457 for (skb = (queue)->next; \
2458 skb != (struct sk_buff *)(queue); \
2459 skb = skb->next)
2460
2461 #define skb_queue_walk_safe(queue, skb, tmp) \
2462 for (skb = (queue)->next, tmp = skb->next; \
2463 skb != (struct sk_buff *)(queue); \
2464 skb = tmp, tmp = skb->next)
2465
2466 #define skb_queue_walk_from(queue, skb) \
2467 for (; skb != (struct sk_buff *)(queue); \
2468 skb = skb->next)
2469
2470 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2471 for (tmp = skb->next; \
2472 skb != (struct sk_buff *)(queue); \
2473 skb = tmp, tmp = skb->next)
2474
2475 #define skb_queue_reverse_walk(queue, skb) \
2476 for (skb = (queue)->prev; \
2477 skb != (struct sk_buff *)(queue); \
2478 skb = skb->prev)
2479
2480 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2481 for (skb = (queue)->prev, tmp = skb->prev; \
2482 skb != (struct sk_buff *)(queue); \
2483 skb = tmp, tmp = skb->prev)
2484
2485 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2486 for (tmp = skb->prev; \
2487 skb != (struct sk_buff *)(queue); \
2488 skb = tmp, tmp = skb->prev)
2489
2490 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2491 {
2492 return skb_shinfo(skb)->frag_list != NULL;
2493 }
2494
2495 static inline void skb_frag_list_init(struct sk_buff *skb)
2496 {
2497 skb_shinfo(skb)->frag_list = NULL;
2498 }
2499
2500 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2501 {
2502 frag->next = skb_shinfo(skb)->frag_list;
2503 skb_shinfo(skb)->frag_list = frag;
2504 }
2505
2506 #define skb_walk_frags(skb, iter) \
2507 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2508
2509 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2510 int *peeked, int *off, int *err);
2511 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2512 int *err);
2513 unsigned int datagram_poll(struct file *file, struct socket *sock,
2514 struct poll_table_struct *wait);
2515 int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2516 struct iovec *to, int size);
2517 int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2518 struct iovec *iov);
2519 int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2520 const struct iovec *from, int from_offset,
2521 int len);
2522 int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2523 int offset, size_t count);
2524 int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2525 const struct iovec *to, int to_offset,
2526 int size);
2527 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2528 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2529 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2530 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2531 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2532 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2533 int len, __wsum csum);
2534 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2535 struct pipe_inode_info *pipe, unsigned int len,
2536 unsigned int flags);
2537 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2538 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2539 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2540 int len, int hlen);
2541 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2542 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2543 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2544 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2545 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2546
2547 struct skb_checksum_ops {
2548 __wsum (*update)(const void *mem, int len, __wsum wsum);
2549 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2550 };
2551
2552 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2553 __wsum csum, const struct skb_checksum_ops *ops);
2554 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2555 __wsum csum);
2556
2557 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2558 int len, void *buffer)
2559 {
2560 int hlen = skb_headlen(skb);
2561
2562 if (hlen - offset >= len)
2563 return skb->data + offset;
2564
2565 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2566 return NULL;
2567
2568 return buffer;
2569 }
2570
2571 /**
2572 * skb_needs_linearize - check if we need to linearize a given skb
2573 * depending on the given device features.
2574 * @skb: socket buffer to check
2575 * @features: net device features
2576 *
2577 * Returns true if either:
2578 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2579 * 2. skb is fragmented and the device does not support SG.
2580 */
2581 static inline bool skb_needs_linearize(struct sk_buff *skb,
2582 netdev_features_t features)
2583 {
2584 return skb_is_nonlinear(skb) &&
2585 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2586 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2587 }
2588
2589 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2590 void *to,
2591 const unsigned int len)
2592 {
2593 memcpy(to, skb->data, len);
2594 }
2595
2596 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2597 const int offset, void *to,
2598 const unsigned int len)
2599 {
2600 memcpy(to, skb->data + offset, len);
2601 }
2602
2603 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2604 const void *from,
2605 const unsigned int len)
2606 {
2607 memcpy(skb->data, from, len);
2608 }
2609
2610 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2611 const int offset,
2612 const void *from,
2613 const unsigned int len)
2614 {
2615 memcpy(skb->data + offset, from, len);
2616 }
2617
2618 void skb_init(void);
2619
2620 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2621 {
2622 return skb->tstamp;
2623 }
2624
2625 /**
2626 * skb_get_timestamp - get timestamp from a skb
2627 * @skb: skb to get stamp from
2628 * @stamp: pointer to struct timeval to store stamp in
2629 *
2630 * Timestamps are stored in the skb as offsets to a base timestamp.
2631 * This function converts the offset back to a struct timeval and stores
2632 * it in stamp.
2633 */
2634 static inline void skb_get_timestamp(const struct sk_buff *skb,
2635 struct timeval *stamp)
2636 {
2637 *stamp = ktime_to_timeval(skb->tstamp);
2638 }
2639
2640 static inline void skb_get_timestampns(const struct sk_buff *skb,
2641 struct timespec *stamp)
2642 {
2643 *stamp = ktime_to_timespec(skb->tstamp);
2644 }
2645
2646 static inline void __net_timestamp(struct sk_buff *skb)
2647 {
2648 skb->tstamp = ktime_get_real();
2649 }
2650
2651 static inline ktime_t net_timedelta(ktime_t t)
2652 {
2653 return ktime_sub(ktime_get_real(), t);
2654 }
2655
2656 static inline ktime_t net_invalid_timestamp(void)
2657 {
2658 return ktime_set(0, 0);
2659 }
2660
2661 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2662
2663 void skb_clone_tx_timestamp(struct sk_buff *skb);
2664 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2665
2666 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2667
2668 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2669 {
2670 }
2671
2672 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2673 {
2674 return false;
2675 }
2676
2677 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2678
2679 /**
2680 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2681 *
2682 * PHY drivers may accept clones of transmitted packets for
2683 * timestamping via their phy_driver.txtstamp method. These drivers
2684 * must call this function to return the skb back to the stack, with
2685 * or without a timestamp.
2686 *
2687 * @skb: clone of the the original outgoing packet
2688 * @hwtstamps: hardware time stamps, may be NULL if not available
2689 *
2690 */
2691 void skb_complete_tx_timestamp(struct sk_buff *skb,
2692 struct skb_shared_hwtstamps *hwtstamps);
2693
2694 /**
2695 * skb_tstamp_tx - queue clone of skb with send time stamps
2696 * @orig_skb: the original outgoing packet
2697 * @hwtstamps: hardware time stamps, may be NULL if not available
2698 *
2699 * If the skb has a socket associated, then this function clones the
2700 * skb (thus sharing the actual data and optional structures), stores
2701 * the optional hardware time stamping information (if non NULL) or
2702 * generates a software time stamp (otherwise), then queues the clone
2703 * to the error queue of the socket. Errors are silently ignored.
2704 */
2705 void skb_tstamp_tx(struct sk_buff *orig_skb,
2706 struct skb_shared_hwtstamps *hwtstamps);
2707
2708 static inline void sw_tx_timestamp(struct sk_buff *skb)
2709 {
2710 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2711 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2712 skb_tstamp_tx(skb, NULL);
2713 }
2714
2715 /**
2716 * skb_tx_timestamp() - Driver hook for transmit timestamping
2717 *
2718 * Ethernet MAC Drivers should call this function in their hard_xmit()
2719 * function immediately before giving the sk_buff to the MAC hardware.
2720 *
2721 * Specifically, one should make absolutely sure that this function is
2722 * called before TX completion of this packet can trigger. Otherwise
2723 * the packet could potentially already be freed.
2724 *
2725 * @skb: A socket buffer.
2726 */
2727 static inline void skb_tx_timestamp(struct sk_buff *skb)
2728 {
2729 skb_clone_tx_timestamp(skb);
2730 sw_tx_timestamp(skb);
2731 }
2732
2733 /**
2734 * skb_complete_wifi_ack - deliver skb with wifi status
2735 *
2736 * @skb: the original outgoing packet
2737 * @acked: ack status
2738 *
2739 */
2740 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2741
2742 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2743 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2744
2745 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2746 {
2747 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
2748 }
2749
2750 /**
2751 * skb_checksum_complete - Calculate checksum of an entire packet
2752 * @skb: packet to process
2753 *
2754 * This function calculates the checksum over the entire packet plus
2755 * the value of skb->csum. The latter can be used to supply the
2756 * checksum of a pseudo header as used by TCP/UDP. It returns the
2757 * checksum.
2758 *
2759 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2760 * this function can be used to verify that checksum on received
2761 * packets. In that case the function should return zero if the
2762 * checksum is correct. In particular, this function will return zero
2763 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2764 * hardware has already verified the correctness of the checksum.
2765 */
2766 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2767 {
2768 return skb_csum_unnecessary(skb) ?
2769 0 : __skb_checksum_complete(skb);
2770 }
2771
2772 /* Check if we need to perform checksum complete validation.
2773 *
2774 * Returns true if checksum complete is needed, false otherwise
2775 * (either checksum is unnecessary or zero checksum is allowed).
2776 */
2777 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2778 bool zero_okay,
2779 __sum16 check)
2780 {
2781 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2782 skb->csum_valid = 1;
2783 return false;
2784 }
2785
2786 return true;
2787 }
2788
2789 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2790 * in checksum_init.
2791 */
2792 #define CHECKSUM_BREAK 76
2793
2794 /* Validate (init) checksum based on checksum complete.
2795 *
2796 * Return values:
2797 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2798 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2799 * checksum is stored in skb->csum for use in __skb_checksum_complete
2800 * non-zero: value of invalid checksum
2801 *
2802 */
2803 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2804 bool complete,
2805 __wsum psum)
2806 {
2807 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2808 if (!csum_fold(csum_add(psum, skb->csum))) {
2809 skb->csum_valid = 1;
2810 return 0;
2811 }
2812 }
2813
2814 skb->csum = psum;
2815
2816 if (complete || skb->len <= CHECKSUM_BREAK) {
2817 __sum16 csum;
2818
2819 csum = __skb_checksum_complete(skb);
2820 skb->csum_valid = !csum;
2821 return csum;
2822 }
2823
2824 return 0;
2825 }
2826
2827 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2828 {
2829 return 0;
2830 }
2831
2832 /* Perform checksum validate (init). Note that this is a macro since we only
2833 * want to calculate the pseudo header which is an input function if necessary.
2834 * First we try to validate without any computation (checksum unnecessary) and
2835 * then calculate based on checksum complete calling the function to compute
2836 * pseudo header.
2837 *
2838 * Return values:
2839 * 0: checksum is validated or try to in skb_checksum_complete
2840 * non-zero: value of invalid checksum
2841 */
2842 #define __skb_checksum_validate(skb, proto, complete, \
2843 zero_okay, check, compute_pseudo) \
2844 ({ \
2845 __sum16 __ret = 0; \
2846 skb->csum_valid = 0; \
2847 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2848 __ret = __skb_checksum_validate_complete(skb, \
2849 complete, compute_pseudo(skb, proto)); \
2850 __ret; \
2851 })
2852
2853 #define skb_checksum_init(skb, proto, compute_pseudo) \
2854 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2855
2856 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2857 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2858
2859 #define skb_checksum_validate(skb, proto, compute_pseudo) \
2860 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2861
2862 #define skb_checksum_validate_zero_check(skb, proto, check, \
2863 compute_pseudo) \
2864 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2865
2866 #define skb_checksum_simple_validate(skb) \
2867 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2868
2869 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2870 void nf_conntrack_destroy(struct nf_conntrack *nfct);
2871 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2872 {
2873 if (nfct && atomic_dec_and_test(&nfct->use))
2874 nf_conntrack_destroy(nfct);
2875 }
2876 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2877 {
2878 if (nfct)
2879 atomic_inc(&nfct->use);
2880 }
2881 #endif
2882 #ifdef CONFIG_BRIDGE_NETFILTER
2883 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2884 {
2885 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2886 kfree(nf_bridge);
2887 }
2888 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2889 {
2890 if (nf_bridge)
2891 atomic_inc(&nf_bridge->use);
2892 }
2893 #endif /* CONFIG_BRIDGE_NETFILTER */
2894 static inline void nf_reset(struct sk_buff *skb)
2895 {
2896 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2897 nf_conntrack_put(skb->nfct);
2898 skb->nfct = NULL;
2899 #endif
2900 #ifdef CONFIG_BRIDGE_NETFILTER
2901 nf_bridge_put(skb->nf_bridge);
2902 skb->nf_bridge = NULL;
2903 #endif
2904 }
2905
2906 static inline void nf_reset_trace(struct sk_buff *skb)
2907 {
2908 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2909 skb->nf_trace = 0;
2910 #endif
2911 }
2912
2913 /* Note: This doesn't put any conntrack and bridge info in dst. */
2914 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2915 {
2916 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2917 dst->nfct = src->nfct;
2918 nf_conntrack_get(src->nfct);
2919 dst->nfctinfo = src->nfctinfo;
2920 #endif
2921 #ifdef CONFIG_BRIDGE_NETFILTER
2922 dst->nf_bridge = src->nf_bridge;
2923 nf_bridge_get(src->nf_bridge);
2924 #endif
2925 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2926 dst->nf_trace = src->nf_trace;
2927 #endif
2928 }
2929
2930 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2931 {
2932 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2933 nf_conntrack_put(dst->nfct);
2934 #endif
2935 #ifdef CONFIG_BRIDGE_NETFILTER
2936 nf_bridge_put(dst->nf_bridge);
2937 #endif
2938 __nf_copy(dst, src);
2939 }
2940
2941 #ifdef CONFIG_NETWORK_SECMARK
2942 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2943 {
2944 to->secmark = from->secmark;
2945 }
2946
2947 static inline void skb_init_secmark(struct sk_buff *skb)
2948 {
2949 skb->secmark = 0;
2950 }
2951 #else
2952 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2953 { }
2954
2955 static inline void skb_init_secmark(struct sk_buff *skb)
2956 { }
2957 #endif
2958
2959 static inline bool skb_irq_freeable(const struct sk_buff *skb)
2960 {
2961 return !skb->destructor &&
2962 #if IS_ENABLED(CONFIG_XFRM)
2963 !skb->sp &&
2964 #endif
2965 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
2966 !skb->nfct &&
2967 #endif
2968 !skb->_skb_refdst &&
2969 !skb_has_frag_list(skb);
2970 }
2971
2972 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2973 {
2974 skb->queue_mapping = queue_mapping;
2975 }
2976
2977 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
2978 {
2979 return skb->queue_mapping;
2980 }
2981
2982 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2983 {
2984 to->queue_mapping = from->queue_mapping;
2985 }
2986
2987 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2988 {
2989 skb->queue_mapping = rx_queue + 1;
2990 }
2991
2992 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
2993 {
2994 return skb->queue_mapping - 1;
2995 }
2996
2997 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
2998 {
2999 return skb->queue_mapping != 0;
3000 }
3001
3002 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3003 unsigned int num_tx_queues);
3004
3005 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3006 {
3007 #ifdef CONFIG_XFRM
3008 return skb->sp;
3009 #else
3010 return NULL;
3011 #endif
3012 }
3013
3014 /* Keeps track of mac header offset relative to skb->head.
3015 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3016 * For non-tunnel skb it points to skb_mac_header() and for
3017 * tunnel skb it points to outer mac header.
3018 * Keeps track of level of encapsulation of network headers.
3019 */
3020 struct skb_gso_cb {
3021 int mac_offset;
3022 int encap_level;
3023 __u16 csum_start;
3024 };
3025 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3026
3027 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3028 {
3029 return (skb_mac_header(inner_skb) - inner_skb->head) -
3030 SKB_GSO_CB(inner_skb)->mac_offset;
3031 }
3032
3033 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3034 {
3035 int new_headroom, headroom;
3036 int ret;
3037
3038 headroom = skb_headroom(skb);
3039 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3040 if (ret)
3041 return ret;
3042
3043 new_headroom = skb_headroom(skb);
3044 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3045 return 0;
3046 }
3047
3048 /* Compute the checksum for a gso segment. First compute the checksum value
3049 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3050 * then add in skb->csum (checksum from csum_start to end of packet).
3051 * skb->csum and csum_start are then updated to reflect the checksum of the
3052 * resultant packet starting from the transport header-- the resultant checksum
3053 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3054 * header.
3055 */
3056 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3057 {
3058 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3059 skb_transport_offset(skb);
3060 __u16 csum;
3061
3062 csum = csum_fold(csum_partial(skb_transport_header(skb),
3063 plen, skb->csum));
3064 skb->csum = res;
3065 SKB_GSO_CB(skb)->csum_start -= plen;
3066
3067 return csum;
3068 }
3069
3070 static inline bool skb_is_gso(const struct sk_buff *skb)
3071 {
3072 return skb_shinfo(skb)->gso_size;
3073 }
3074
3075 /* Note: Should be called only if skb_is_gso(skb) is true */
3076 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
3077 {
3078 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3079 }
3080
3081 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
3082
3083 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3084 {
3085 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3086 * wanted then gso_type will be set. */
3087 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3088
3089 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3090 unlikely(shinfo->gso_type == 0)) {
3091 __skb_warn_lro_forwarding(skb);
3092 return true;
3093 }
3094 return false;
3095 }
3096
3097 static inline void skb_forward_csum(struct sk_buff *skb)
3098 {
3099 /* Unfortunately we don't support this one. Any brave souls? */
3100 if (skb->ip_summed == CHECKSUM_COMPLETE)
3101 skb->ip_summed = CHECKSUM_NONE;
3102 }
3103
3104 /**
3105 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3106 * @skb: skb to check
3107 *
3108 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3109 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3110 * use this helper, to document places where we make this assertion.
3111 */
3112 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
3113 {
3114 #ifdef DEBUG
3115 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3116 #endif
3117 }
3118
3119 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
3120
3121 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3122
3123 u32 __skb_get_poff(const struct sk_buff *skb);
3124
3125 /**
3126 * skb_head_is_locked - Determine if the skb->head is locked down
3127 * @skb: skb to check
3128 *
3129 * The head on skbs build around a head frag can be removed if they are
3130 * not cloned. This function returns true if the skb head is locked down
3131 * due to either being allocated via kmalloc, or by being a clone with
3132 * multiple references to the head.
3133 */
3134 static inline bool skb_head_is_locked(const struct sk_buff *skb)
3135 {
3136 return !skb->head_frag || skb_cloned(skb);
3137 }
3138
3139 /**
3140 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3141 *
3142 * @skb: GSO skb
3143 *
3144 * skb_gso_network_seglen is used to determine the real size of the
3145 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3146 *
3147 * The MAC/L2 header is not accounted for.
3148 */
3149 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3150 {
3151 unsigned int hdr_len = skb_transport_header(skb) -
3152 skb_network_header(skb);
3153 return hdr_len + skb_gso_transport_seglen(skb);
3154 }
3155 #endif /* __KERNEL__ */
3156 #endif /* _LINUX_SKBUFF_H */
This page took 0.097954 seconds and 6 git commands to generate.