net: Update sk_buff flag bit availability comment.
[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 when queueing packet to NIC */
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 /* generate software time stamp when entering packet scheduling */
252 SKBTX_SCHED_TSTAMP = 1 << 6,
253
254 /* generate software timestamp on peer data acknowledgment */
255 SKBTX_ACK_TSTAMP = 1 << 7,
256 };
257
258 #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
259 SKBTX_SCHED_TSTAMP | \
260 SKBTX_ACK_TSTAMP)
261 #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
262
263 /*
264 * The callback notifies userspace to release buffers when skb DMA is done in
265 * lower device, the skb last reference should be 0 when calling this.
266 * The zerocopy_success argument is true if zero copy transmit occurred,
267 * false on data copy or out of memory error caused by data copy attempt.
268 * The ctx field is used to track device context.
269 * The desc field is used to track userspace buffer index.
270 */
271 struct ubuf_info {
272 void (*callback)(struct ubuf_info *, bool zerocopy_success);
273 void *ctx;
274 unsigned long desc;
275 };
276
277 /* This data is invariant across clones and lives at
278 * the end of the header data, ie. at skb->end.
279 */
280 struct skb_shared_info {
281 unsigned char nr_frags;
282 __u8 tx_flags;
283 unsigned short gso_size;
284 /* Warning: this field is not always filled in (UFO)! */
285 unsigned short gso_segs;
286 unsigned short gso_type;
287 struct sk_buff *frag_list;
288 struct skb_shared_hwtstamps hwtstamps;
289 u32 tskey;
290 __be32 ip6_frag_id;
291
292 /*
293 * Warning : all fields before dataref are cleared in __alloc_skb()
294 */
295 atomic_t dataref;
296
297 /* Intermediate layers must ensure that destructor_arg
298 * remains valid until skb destructor */
299 void * destructor_arg;
300
301 /* must be last field, see pskb_expand_head() */
302 skb_frag_t frags[MAX_SKB_FRAGS];
303 };
304
305 /* We divide dataref into two halves. The higher 16 bits hold references
306 * to the payload part of skb->data. The lower 16 bits hold references to
307 * the entire skb->data. A clone of a headerless skb holds the length of
308 * the header in skb->hdr_len.
309 *
310 * All users must obey the rule that the skb->data reference count must be
311 * greater than or equal to the payload reference count.
312 *
313 * Holding a reference to the payload part means that the user does not
314 * care about modifications to the header part of skb->data.
315 */
316 #define SKB_DATAREF_SHIFT 16
317 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
318
319
320 enum {
321 SKB_FCLONE_UNAVAILABLE,
322 SKB_FCLONE_ORIG,
323 SKB_FCLONE_CLONE,
324 };
325
326 enum {
327 SKB_GSO_TCPV4 = 1 << 0,
328 SKB_GSO_UDP = 1 << 1,
329
330 /* This indicates the skb is from an untrusted source. */
331 SKB_GSO_DODGY = 1 << 2,
332
333 /* This indicates the tcp segment has CWR set. */
334 SKB_GSO_TCP_ECN = 1 << 3,
335
336 SKB_GSO_TCPV6 = 1 << 4,
337
338 SKB_GSO_FCOE = 1 << 5,
339
340 SKB_GSO_GRE = 1 << 6,
341
342 SKB_GSO_GRE_CSUM = 1 << 7,
343
344 SKB_GSO_IPIP = 1 << 8,
345
346 SKB_GSO_SIT = 1 << 9,
347
348 SKB_GSO_UDP_TUNNEL = 1 << 10,
349
350 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
351
352 SKB_GSO_MPLS = 1 << 12,
353
354 };
355
356 #if BITS_PER_LONG > 32
357 #define NET_SKBUFF_DATA_USES_OFFSET 1
358 #endif
359
360 #ifdef NET_SKBUFF_DATA_USES_OFFSET
361 typedef unsigned int sk_buff_data_t;
362 #else
363 typedef unsigned char *sk_buff_data_t;
364 #endif
365
366 /**
367 * struct skb_mstamp - multi resolution time stamps
368 * @stamp_us: timestamp in us resolution
369 * @stamp_jiffies: timestamp in jiffies
370 */
371 struct skb_mstamp {
372 union {
373 u64 v64;
374 struct {
375 u32 stamp_us;
376 u32 stamp_jiffies;
377 };
378 };
379 };
380
381 /**
382 * skb_mstamp_get - get current timestamp
383 * @cl: place to store timestamps
384 */
385 static inline void skb_mstamp_get(struct skb_mstamp *cl)
386 {
387 u64 val = local_clock();
388
389 do_div(val, NSEC_PER_USEC);
390 cl->stamp_us = (u32)val;
391 cl->stamp_jiffies = (u32)jiffies;
392 }
393
394 /**
395 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
396 * @t1: pointer to newest sample
397 * @t0: pointer to oldest sample
398 */
399 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
400 const struct skb_mstamp *t0)
401 {
402 s32 delta_us = t1->stamp_us - t0->stamp_us;
403 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
404
405 /* If delta_us is negative, this might be because interval is too big,
406 * or local_clock() drift is too big : fallback using jiffies.
407 */
408 if (delta_us <= 0 ||
409 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
410
411 delta_us = jiffies_to_usecs(delta_jiffies);
412
413 return delta_us;
414 }
415
416
417 /**
418 * struct sk_buff - socket buffer
419 * @next: Next buffer in list
420 * @prev: Previous buffer in list
421 * @tstamp: Time we arrived/left
422 * @sk: Socket we are owned by
423 * @dev: Device we arrived on/are leaving by
424 * @cb: Control buffer. Free for use by every layer. Put private vars here
425 * @_skb_refdst: destination entry (with norefcount bit)
426 * @sp: the security path, used for xfrm
427 * @len: Length of actual data
428 * @data_len: Data length
429 * @mac_len: Length of link layer header
430 * @hdr_len: writable header length of cloned skb
431 * @csum: Checksum (must include start/offset pair)
432 * @csum_start: Offset from skb->head where checksumming should start
433 * @csum_offset: Offset from csum_start where checksum should be stored
434 * @priority: Packet queueing priority
435 * @ignore_df: allow local fragmentation
436 * @cloned: Head may be cloned (check refcnt to be sure)
437 * @ip_summed: Driver fed us an IP checksum
438 * @nohdr: Payload reference only, must not modify header
439 * @nfctinfo: Relationship of this skb to the connection
440 * @pkt_type: Packet class
441 * @fclone: skbuff clone status
442 * @ipvs_property: skbuff is owned by ipvs
443 * @peeked: this packet has been seen already, so stats have been
444 * done for it, don't do them again
445 * @nf_trace: netfilter packet trace flag
446 * @protocol: Packet protocol from driver
447 * @destructor: Destruct function
448 * @nfct: Associated connection, if any
449 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
450 * @skb_iif: ifindex of device we arrived on
451 * @tc_index: Traffic control index
452 * @tc_verd: traffic control verdict
453 * @hash: the packet hash
454 * @queue_mapping: Queue mapping for multiqueue devices
455 * @xmit_more: More SKBs are pending for this queue
456 * @ndisc_nodetype: router type (from link layer)
457 * @ooo_okay: allow the mapping of a socket to a queue to be changed
458 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
459 * ports.
460 * @sw_hash: indicates hash was computed in software stack
461 * @wifi_acked_valid: wifi_acked was set
462 * @wifi_acked: whether frame was acked on wifi or not
463 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
464 * @dma_cookie: a cookie to one of several possible DMA operations
465 * done by skb DMA functions
466 * @napi_id: id of the NAPI struct this skb came from
467 * @secmark: security marking
468 * @mark: Generic packet mark
469 * @dropcount: total number of sk_receive_queue overflows
470 * @vlan_proto: vlan encapsulation protocol
471 * @vlan_tci: vlan tag control information
472 * @inner_protocol: Protocol (encapsulation)
473 * @inner_transport_header: Inner transport layer header (encapsulation)
474 * @inner_network_header: Network layer header (encapsulation)
475 * @inner_mac_header: Link layer header (encapsulation)
476 * @transport_header: Transport layer header
477 * @network_header: Network layer header
478 * @mac_header: Link layer header
479 * @tail: Tail pointer
480 * @end: End pointer
481 * @head: Head of buffer
482 * @data: Data head pointer
483 * @truesize: Buffer size
484 * @users: User count - see {datagram,tcp}.c
485 */
486
487 struct sk_buff {
488 /* These two members must be first. */
489 struct sk_buff *next;
490 struct sk_buff *prev;
491
492 union {
493 ktime_t tstamp;
494 struct skb_mstamp skb_mstamp;
495 };
496
497 struct sock *sk;
498 struct net_device *dev;
499
500 /*
501 * This is the control buffer. It is free to use for every
502 * layer. Please put your private variables there. If you
503 * want to keep them across layers you have to do a skb_clone()
504 * first. This is owned by whoever has the skb queued ATM.
505 */
506 char cb[48] __aligned(8);
507
508 unsigned long _skb_refdst;
509 #ifdef CONFIG_XFRM
510 struct sec_path *sp;
511 #endif
512 unsigned int len,
513 data_len;
514 __u16 mac_len,
515 hdr_len;
516 union {
517 __wsum csum;
518 struct {
519 __u16 csum_start;
520 __u16 csum_offset;
521 };
522 };
523 __u32 priority;
524 kmemcheck_bitfield_begin(flags1);
525 __u8 ignore_df:1,
526 cloned:1,
527 ip_summed:2,
528 nohdr:1,
529 nfctinfo:3;
530 __u8 pkt_type:3,
531 fclone:2,
532 ipvs_property:1,
533 peeked:1,
534 nf_trace:1;
535 kmemcheck_bitfield_end(flags1);
536 __be16 protocol;
537
538 void (*destructor)(struct sk_buff *skb);
539 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
540 struct nf_conntrack *nfct;
541 #endif
542 #ifdef CONFIG_BRIDGE_NETFILTER
543 struct nf_bridge_info *nf_bridge;
544 #endif
545
546 int skb_iif;
547
548 __u32 hash;
549
550 __be16 vlan_proto;
551 __u16 vlan_tci;
552
553 #ifdef CONFIG_NET_SCHED
554 __u16 tc_index; /* traffic control index */
555 #ifdef CONFIG_NET_CLS_ACT
556 __u16 tc_verd; /* traffic control verdict */
557 #endif
558 #endif
559
560 __u16 queue_mapping;
561 kmemcheck_bitfield_begin(flags2);
562 __u8 xmit_more:1;
563 #ifdef CONFIG_IPV6_NDISC_NODETYPE
564 __u8 ndisc_nodetype:2;
565 #endif
566 __u8 pfmemalloc:1;
567 __u8 ooo_okay:1;
568 __u8 l4_hash:1;
569 __u8 sw_hash:1;
570 __u8 wifi_acked_valid:1;
571 __u8 wifi_acked:1;
572 __u8 no_fcs:1;
573 __u8 head_frag:1;
574 /* Encapsulation protocol and NIC drivers should use
575 * this flag to indicate to each other if the skb contains
576 * encapsulated packet or not and maybe use the inner packet
577 * headers if needed
578 */
579 __u8 encapsulation:1;
580 __u8 encap_hdr_csum:1;
581 __u8 csum_valid:1;
582 __u8 csum_complete_sw:1;
583 /* 1/3 bit hole (depending on ndisc_nodetype presence) */
584 kmemcheck_bitfield_end(flags2);
585
586 #if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
587 union {
588 unsigned int napi_id;
589 dma_cookie_t dma_cookie;
590 };
591 #endif
592 #ifdef CONFIG_NETWORK_SECMARK
593 __u32 secmark;
594 #endif
595 union {
596 __u32 mark;
597 __u32 dropcount;
598 __u32 reserved_tailroom;
599 };
600
601 __be16 inner_protocol;
602 __u16 inner_transport_header;
603 __u16 inner_network_header;
604 __u16 inner_mac_header;
605 __u16 transport_header;
606 __u16 network_header;
607 __u16 mac_header;
608 /* These elements must be at the end, see alloc_skb() for details. */
609 sk_buff_data_t tail;
610 sk_buff_data_t end;
611 unsigned char *head,
612 *data;
613 unsigned int truesize;
614 atomic_t users;
615 };
616
617 #ifdef __KERNEL__
618 /*
619 * Handling routines are only of interest to the kernel
620 */
621 #include <linux/slab.h>
622
623
624 #define SKB_ALLOC_FCLONE 0x01
625 #define SKB_ALLOC_RX 0x02
626
627 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
628 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
629 {
630 return unlikely(skb->pfmemalloc);
631 }
632
633 /*
634 * skb might have a dst pointer attached, refcounted or not.
635 * _skb_refdst low order bit is set if refcount was _not_ taken
636 */
637 #define SKB_DST_NOREF 1UL
638 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
639
640 /**
641 * skb_dst - returns skb dst_entry
642 * @skb: buffer
643 *
644 * Returns skb dst_entry, regardless of reference taken or not.
645 */
646 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
647 {
648 /* If refdst was not refcounted, check we still are in a
649 * rcu_read_lock section
650 */
651 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
652 !rcu_read_lock_held() &&
653 !rcu_read_lock_bh_held());
654 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
655 }
656
657 /**
658 * skb_dst_set - sets skb dst
659 * @skb: buffer
660 * @dst: dst entry
661 *
662 * Sets skb dst, assuming a reference was taken on dst and should
663 * be released by skb_dst_drop()
664 */
665 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
666 {
667 skb->_skb_refdst = (unsigned long)dst;
668 }
669
670 void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
671 bool force);
672
673 /**
674 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
675 * @skb: buffer
676 * @dst: dst entry
677 *
678 * Sets skb dst, assuming a reference was not taken on dst.
679 * If dst entry is cached, we do not take reference and dst_release
680 * will be avoided by refdst_drop. If dst entry is not cached, we take
681 * reference, so that last dst_release can destroy the dst immediately.
682 */
683 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
684 {
685 __skb_dst_set_noref(skb, dst, false);
686 }
687
688 /**
689 * skb_dst_set_noref_force - sets skb dst, without taking reference
690 * @skb: buffer
691 * @dst: dst entry
692 *
693 * Sets skb dst, assuming a reference was not taken on dst.
694 * No reference is taken and no dst_release will be called. While for
695 * cached dsts deferred reclaim is a basic feature, for entries that are
696 * not cached it is caller's job to guarantee that last dst_release for
697 * provided dst happens when nobody uses it, eg. after a RCU grace period.
698 */
699 static inline void skb_dst_set_noref_force(struct sk_buff *skb,
700 struct dst_entry *dst)
701 {
702 __skb_dst_set_noref(skb, dst, true);
703 }
704
705 /**
706 * skb_dst_is_noref - Test if skb dst isn't refcounted
707 * @skb: buffer
708 */
709 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
710 {
711 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
712 }
713
714 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
715 {
716 return (struct rtable *)skb_dst(skb);
717 }
718
719 void kfree_skb(struct sk_buff *skb);
720 void kfree_skb_list(struct sk_buff *segs);
721 void skb_tx_error(struct sk_buff *skb);
722 void consume_skb(struct sk_buff *skb);
723 void __kfree_skb(struct sk_buff *skb);
724 extern struct kmem_cache *skbuff_head_cache;
725
726 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
727 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
728 bool *fragstolen, int *delta_truesize);
729
730 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
731 int node);
732 struct sk_buff *build_skb(void *data, unsigned int frag_size);
733 static inline struct sk_buff *alloc_skb(unsigned int size,
734 gfp_t priority)
735 {
736 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
737 }
738
739 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
740 gfp_t priority)
741 {
742 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
743 }
744
745 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
746 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
747 {
748 return __alloc_skb_head(priority, -1);
749 }
750
751 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
752 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
753 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
754 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
755 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
756 gfp_t gfp_mask, bool fclone);
757 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
758 gfp_t gfp_mask)
759 {
760 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
761 }
762
763 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
764 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
765 unsigned int headroom);
766 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
767 int newtailroom, gfp_t priority);
768 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
769 int offset, int len);
770 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
771 int len);
772 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
773 int skb_pad(struct sk_buff *skb, int pad);
774 #define dev_kfree_skb(a) consume_skb(a)
775
776 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
777 int getfrag(void *from, char *to, int offset,
778 int len, int odd, struct sk_buff *skb),
779 void *from, int length);
780
781 struct skb_seq_state {
782 __u32 lower_offset;
783 __u32 upper_offset;
784 __u32 frag_idx;
785 __u32 stepped_offset;
786 struct sk_buff *root_skb;
787 struct sk_buff *cur_skb;
788 __u8 *frag_data;
789 };
790
791 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
792 unsigned int to, struct skb_seq_state *st);
793 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
794 struct skb_seq_state *st);
795 void skb_abort_seq_read(struct skb_seq_state *st);
796
797 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
798 unsigned int to, struct ts_config *config,
799 struct ts_state *state);
800
801 /*
802 * Packet hash types specify the type of hash in skb_set_hash.
803 *
804 * Hash types refer to the protocol layer addresses which are used to
805 * construct a packet's hash. The hashes are used to differentiate or identify
806 * flows of the protocol layer for the hash type. Hash types are either
807 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
808 *
809 * Properties of hashes:
810 *
811 * 1) Two packets in different flows have different hash values
812 * 2) Two packets in the same flow should have the same hash value
813 *
814 * A hash at a higher layer is considered to be more specific. A driver should
815 * set the most specific hash possible.
816 *
817 * A driver cannot indicate a more specific hash than the layer at which a hash
818 * was computed. For instance an L3 hash cannot be set as an L4 hash.
819 *
820 * A driver may indicate a hash level which is less specific than the
821 * actual layer the hash was computed on. For instance, a hash computed
822 * at L4 may be considered an L3 hash. This should only be done if the
823 * driver can't unambiguously determine that the HW computed the hash at
824 * the higher layer. Note that the "should" in the second property above
825 * permits this.
826 */
827 enum pkt_hash_types {
828 PKT_HASH_TYPE_NONE, /* Undefined type */
829 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
830 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
831 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
832 };
833
834 static inline void
835 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
836 {
837 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
838 skb->sw_hash = 0;
839 skb->hash = hash;
840 }
841
842 void __skb_get_hash(struct sk_buff *skb);
843 static inline __u32 skb_get_hash(struct sk_buff *skb)
844 {
845 if (!skb->l4_hash && !skb->sw_hash)
846 __skb_get_hash(skb);
847
848 return skb->hash;
849 }
850
851 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
852 {
853 return skb->hash;
854 }
855
856 static inline void skb_clear_hash(struct sk_buff *skb)
857 {
858 skb->hash = 0;
859 skb->sw_hash = 0;
860 skb->l4_hash = 0;
861 }
862
863 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
864 {
865 if (!skb->l4_hash)
866 skb_clear_hash(skb);
867 }
868
869 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
870 {
871 to->hash = from->hash;
872 to->sw_hash = from->sw_hash;
873 to->l4_hash = from->l4_hash;
874 };
875
876 #ifdef NET_SKBUFF_DATA_USES_OFFSET
877 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
878 {
879 return skb->head + skb->end;
880 }
881
882 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
883 {
884 return skb->end;
885 }
886 #else
887 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
888 {
889 return skb->end;
890 }
891
892 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
893 {
894 return skb->end - skb->head;
895 }
896 #endif
897
898 /* Internal */
899 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
900
901 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
902 {
903 return &skb_shinfo(skb)->hwtstamps;
904 }
905
906 /**
907 * skb_queue_empty - check if a queue is empty
908 * @list: queue head
909 *
910 * Returns true if the queue is empty, false otherwise.
911 */
912 static inline int skb_queue_empty(const struct sk_buff_head *list)
913 {
914 return list->next == (const struct sk_buff *) list;
915 }
916
917 /**
918 * skb_queue_is_last - check if skb is the last entry in the queue
919 * @list: queue head
920 * @skb: buffer
921 *
922 * Returns true if @skb is the last buffer on the list.
923 */
924 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
925 const struct sk_buff *skb)
926 {
927 return skb->next == (const struct sk_buff *) list;
928 }
929
930 /**
931 * skb_queue_is_first - check if skb is the first entry in the queue
932 * @list: queue head
933 * @skb: buffer
934 *
935 * Returns true if @skb is the first buffer on the list.
936 */
937 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
938 const struct sk_buff *skb)
939 {
940 return skb->prev == (const struct sk_buff *) list;
941 }
942
943 /**
944 * skb_queue_next - return the next packet in the queue
945 * @list: queue head
946 * @skb: current buffer
947 *
948 * Return the next packet in @list after @skb. It is only valid to
949 * call this if skb_queue_is_last() evaluates to false.
950 */
951 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
952 const struct sk_buff *skb)
953 {
954 /* This BUG_ON may seem severe, but if we just return then we
955 * are going to dereference garbage.
956 */
957 BUG_ON(skb_queue_is_last(list, skb));
958 return skb->next;
959 }
960
961 /**
962 * skb_queue_prev - return the prev packet in the queue
963 * @list: queue head
964 * @skb: current buffer
965 *
966 * Return the prev packet in @list before @skb. It is only valid to
967 * call this if skb_queue_is_first() evaluates to false.
968 */
969 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
970 const struct sk_buff *skb)
971 {
972 /* This BUG_ON may seem severe, but if we just return then we
973 * are going to dereference garbage.
974 */
975 BUG_ON(skb_queue_is_first(list, skb));
976 return skb->prev;
977 }
978
979 /**
980 * skb_get - reference buffer
981 * @skb: buffer to reference
982 *
983 * Makes another reference to a socket buffer and returns a pointer
984 * to the buffer.
985 */
986 static inline struct sk_buff *skb_get(struct sk_buff *skb)
987 {
988 atomic_inc(&skb->users);
989 return skb;
990 }
991
992 /*
993 * If users == 1, we are the only owner and are can avoid redundant
994 * atomic change.
995 */
996
997 /**
998 * skb_cloned - is the buffer a clone
999 * @skb: buffer to check
1000 *
1001 * Returns true if the buffer was generated with skb_clone() and is
1002 * one of multiple shared copies of the buffer. Cloned buffers are
1003 * shared data so must not be written to under normal circumstances.
1004 */
1005 static inline int skb_cloned(const struct sk_buff *skb)
1006 {
1007 return skb->cloned &&
1008 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1009 }
1010
1011 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1012 {
1013 might_sleep_if(pri & __GFP_WAIT);
1014
1015 if (skb_cloned(skb))
1016 return pskb_expand_head(skb, 0, 0, pri);
1017
1018 return 0;
1019 }
1020
1021 /**
1022 * skb_header_cloned - is the header a clone
1023 * @skb: buffer to check
1024 *
1025 * Returns true if modifying the header part of the buffer requires
1026 * the data to be copied.
1027 */
1028 static inline int skb_header_cloned(const struct sk_buff *skb)
1029 {
1030 int dataref;
1031
1032 if (!skb->cloned)
1033 return 0;
1034
1035 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1036 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1037 return dataref != 1;
1038 }
1039
1040 /**
1041 * skb_header_release - release reference to header
1042 * @skb: buffer to operate on
1043 *
1044 * Drop a reference to the header part of the buffer. This is done
1045 * by acquiring a payload reference. You must not read from the header
1046 * part of skb->data after this.
1047 */
1048 static inline void skb_header_release(struct sk_buff *skb)
1049 {
1050 BUG_ON(skb->nohdr);
1051 skb->nohdr = 1;
1052 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1053 }
1054
1055 /**
1056 * skb_shared - is the buffer shared
1057 * @skb: buffer to check
1058 *
1059 * Returns true if more than one person has a reference to this
1060 * buffer.
1061 */
1062 static inline int skb_shared(const struct sk_buff *skb)
1063 {
1064 return atomic_read(&skb->users) != 1;
1065 }
1066
1067 /**
1068 * skb_share_check - check if buffer is shared and if so clone it
1069 * @skb: buffer to check
1070 * @pri: priority for memory allocation
1071 *
1072 * If the buffer is shared the buffer is cloned and the old copy
1073 * drops a reference. A new clone with a single reference is returned.
1074 * If the buffer is not shared the original buffer is returned. When
1075 * being called from interrupt status or with spinlocks held pri must
1076 * be GFP_ATOMIC.
1077 *
1078 * NULL is returned on a memory allocation failure.
1079 */
1080 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1081 {
1082 might_sleep_if(pri & __GFP_WAIT);
1083 if (skb_shared(skb)) {
1084 struct sk_buff *nskb = skb_clone(skb, pri);
1085
1086 if (likely(nskb))
1087 consume_skb(skb);
1088 else
1089 kfree_skb(skb);
1090 skb = nskb;
1091 }
1092 return skb;
1093 }
1094
1095 /*
1096 * Copy shared buffers into a new sk_buff. We effectively do COW on
1097 * packets to handle cases where we have a local reader and forward
1098 * and a couple of other messy ones. The normal one is tcpdumping
1099 * a packet thats being forwarded.
1100 */
1101
1102 /**
1103 * skb_unshare - make a copy of a shared buffer
1104 * @skb: buffer to check
1105 * @pri: priority for memory allocation
1106 *
1107 * If the socket buffer is a clone then this function creates a new
1108 * copy of the data, drops a reference count on the old copy and returns
1109 * the new copy with the reference count at 1. If the buffer is not a clone
1110 * the original buffer is returned. When called with a spinlock held or
1111 * from interrupt state @pri must be %GFP_ATOMIC
1112 *
1113 * %NULL is returned on a memory allocation failure.
1114 */
1115 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1116 gfp_t pri)
1117 {
1118 might_sleep_if(pri & __GFP_WAIT);
1119 if (skb_cloned(skb)) {
1120 struct sk_buff *nskb = skb_copy(skb, pri);
1121 kfree_skb(skb); /* Free our shared copy */
1122 skb = nskb;
1123 }
1124 return skb;
1125 }
1126
1127 /**
1128 * skb_peek - peek at the head of an &sk_buff_head
1129 * @list_: list to peek at
1130 *
1131 * Peek an &sk_buff. Unlike most other operations you _MUST_
1132 * be careful with this one. A peek leaves the buffer on the
1133 * list and someone else may run off with it. You must hold
1134 * the appropriate locks or have a private queue to do this.
1135 *
1136 * Returns %NULL for an empty list or a pointer to the head element.
1137 * The reference count is not incremented and the reference is therefore
1138 * volatile. Use with caution.
1139 */
1140 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1141 {
1142 struct sk_buff *skb = list_->next;
1143
1144 if (skb == (struct sk_buff *)list_)
1145 skb = NULL;
1146 return skb;
1147 }
1148
1149 /**
1150 * skb_peek_next - peek skb following the given one from a queue
1151 * @skb: skb to start from
1152 * @list_: list to peek at
1153 *
1154 * Returns %NULL when the end of the list is met or a pointer to the
1155 * next element. The reference count is not incremented and the
1156 * reference is therefore volatile. Use with caution.
1157 */
1158 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1159 const struct sk_buff_head *list_)
1160 {
1161 struct sk_buff *next = skb->next;
1162
1163 if (next == (struct sk_buff *)list_)
1164 next = NULL;
1165 return next;
1166 }
1167
1168 /**
1169 * skb_peek_tail - peek at the tail of an &sk_buff_head
1170 * @list_: list to peek at
1171 *
1172 * Peek an &sk_buff. Unlike most other operations you _MUST_
1173 * be careful with this one. A peek leaves the buffer on the
1174 * list and someone else may run off with it. You must hold
1175 * the appropriate locks or have a private queue to do this.
1176 *
1177 * Returns %NULL for an empty list or a pointer to the tail element.
1178 * The reference count is not incremented and the reference is therefore
1179 * volatile. Use with caution.
1180 */
1181 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1182 {
1183 struct sk_buff *skb = list_->prev;
1184
1185 if (skb == (struct sk_buff *)list_)
1186 skb = NULL;
1187 return skb;
1188
1189 }
1190
1191 /**
1192 * skb_queue_len - get queue length
1193 * @list_: list to measure
1194 *
1195 * Return the length of an &sk_buff queue.
1196 */
1197 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1198 {
1199 return list_->qlen;
1200 }
1201
1202 /**
1203 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1204 * @list: queue to initialize
1205 *
1206 * This initializes only the list and queue length aspects of
1207 * an sk_buff_head object. This allows to initialize the list
1208 * aspects of an sk_buff_head without reinitializing things like
1209 * the spinlock. It can also be used for on-stack sk_buff_head
1210 * objects where the spinlock is known to not be used.
1211 */
1212 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1213 {
1214 list->prev = list->next = (struct sk_buff *)list;
1215 list->qlen = 0;
1216 }
1217
1218 /*
1219 * This function creates a split out lock class for each invocation;
1220 * this is needed for now since a whole lot of users of the skb-queue
1221 * infrastructure in drivers have different locking usage (in hardirq)
1222 * than the networking core (in softirq only). In the long run either the
1223 * network layer or drivers should need annotation to consolidate the
1224 * main types of usage into 3 classes.
1225 */
1226 static inline void skb_queue_head_init(struct sk_buff_head *list)
1227 {
1228 spin_lock_init(&list->lock);
1229 __skb_queue_head_init(list);
1230 }
1231
1232 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1233 struct lock_class_key *class)
1234 {
1235 skb_queue_head_init(list);
1236 lockdep_set_class(&list->lock, class);
1237 }
1238
1239 /*
1240 * Insert an sk_buff on a list.
1241 *
1242 * The "__skb_xxxx()" functions are the non-atomic ones that
1243 * can only be called with interrupts disabled.
1244 */
1245 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1246 struct sk_buff_head *list);
1247 static inline void __skb_insert(struct sk_buff *newsk,
1248 struct sk_buff *prev, struct sk_buff *next,
1249 struct sk_buff_head *list)
1250 {
1251 newsk->next = next;
1252 newsk->prev = prev;
1253 next->prev = prev->next = newsk;
1254 list->qlen++;
1255 }
1256
1257 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1258 struct sk_buff *prev,
1259 struct sk_buff *next)
1260 {
1261 struct sk_buff *first = list->next;
1262 struct sk_buff *last = list->prev;
1263
1264 first->prev = prev;
1265 prev->next = first;
1266
1267 last->next = next;
1268 next->prev = last;
1269 }
1270
1271 /**
1272 * skb_queue_splice - join two skb lists, this is designed for stacks
1273 * @list: the new list to add
1274 * @head: the place to add it in the first list
1275 */
1276 static inline void skb_queue_splice(const struct sk_buff_head *list,
1277 struct sk_buff_head *head)
1278 {
1279 if (!skb_queue_empty(list)) {
1280 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1281 head->qlen += list->qlen;
1282 }
1283 }
1284
1285 /**
1286 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1287 * @list: the new list to add
1288 * @head: the place to add it in the first list
1289 *
1290 * The list at @list is reinitialised
1291 */
1292 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1293 struct sk_buff_head *head)
1294 {
1295 if (!skb_queue_empty(list)) {
1296 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1297 head->qlen += list->qlen;
1298 __skb_queue_head_init(list);
1299 }
1300 }
1301
1302 /**
1303 * skb_queue_splice_tail - join two skb lists, each list being a queue
1304 * @list: the new list to add
1305 * @head: the place to add it in the first list
1306 */
1307 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1308 struct sk_buff_head *head)
1309 {
1310 if (!skb_queue_empty(list)) {
1311 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1312 head->qlen += list->qlen;
1313 }
1314 }
1315
1316 /**
1317 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1318 * @list: the new list to add
1319 * @head: the place to add it in the first list
1320 *
1321 * Each of the lists is a queue.
1322 * The list at @list is reinitialised
1323 */
1324 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1325 struct sk_buff_head *head)
1326 {
1327 if (!skb_queue_empty(list)) {
1328 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1329 head->qlen += list->qlen;
1330 __skb_queue_head_init(list);
1331 }
1332 }
1333
1334 /**
1335 * __skb_queue_after - queue a buffer at the list head
1336 * @list: list to use
1337 * @prev: place after this buffer
1338 * @newsk: buffer to queue
1339 *
1340 * Queue a buffer int the middle of a list. This function takes no locks
1341 * and you must therefore hold required locks before calling it.
1342 *
1343 * A buffer cannot be placed on two lists at the same time.
1344 */
1345 static inline void __skb_queue_after(struct sk_buff_head *list,
1346 struct sk_buff *prev,
1347 struct sk_buff *newsk)
1348 {
1349 __skb_insert(newsk, prev, prev->next, list);
1350 }
1351
1352 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1353 struct sk_buff_head *list);
1354
1355 static inline void __skb_queue_before(struct sk_buff_head *list,
1356 struct sk_buff *next,
1357 struct sk_buff *newsk)
1358 {
1359 __skb_insert(newsk, next->prev, next, list);
1360 }
1361
1362 /**
1363 * __skb_queue_head - queue a buffer at the list head
1364 * @list: list to use
1365 * @newsk: buffer to queue
1366 *
1367 * Queue a buffer at the start of a list. This function takes no locks
1368 * and you must therefore hold required locks before calling it.
1369 *
1370 * A buffer cannot be placed on two lists at the same time.
1371 */
1372 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1373 static inline void __skb_queue_head(struct sk_buff_head *list,
1374 struct sk_buff *newsk)
1375 {
1376 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1377 }
1378
1379 /**
1380 * __skb_queue_tail - queue a buffer at the list tail
1381 * @list: list to use
1382 * @newsk: buffer to queue
1383 *
1384 * Queue a buffer at the end of a list. This function takes no locks
1385 * and you must therefore hold required locks before calling it.
1386 *
1387 * A buffer cannot be placed on two lists at the same time.
1388 */
1389 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1390 static inline void __skb_queue_tail(struct sk_buff_head *list,
1391 struct sk_buff *newsk)
1392 {
1393 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1394 }
1395
1396 /*
1397 * remove sk_buff from list. _Must_ be called atomically, and with
1398 * the list known..
1399 */
1400 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1401 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1402 {
1403 struct sk_buff *next, *prev;
1404
1405 list->qlen--;
1406 next = skb->next;
1407 prev = skb->prev;
1408 skb->next = skb->prev = NULL;
1409 next->prev = prev;
1410 prev->next = next;
1411 }
1412
1413 /**
1414 * __skb_dequeue - remove from the head of the queue
1415 * @list: list to dequeue from
1416 *
1417 * Remove the head of the list. This function does not take any locks
1418 * so must be used with appropriate locks held only. The head item is
1419 * returned or %NULL if the list is empty.
1420 */
1421 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1422 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1423 {
1424 struct sk_buff *skb = skb_peek(list);
1425 if (skb)
1426 __skb_unlink(skb, list);
1427 return skb;
1428 }
1429
1430 /**
1431 * __skb_dequeue_tail - remove from the tail of the queue
1432 * @list: list to dequeue from
1433 *
1434 * Remove the tail of the list. This function does not take any locks
1435 * so must be used with appropriate locks held only. The tail item is
1436 * returned or %NULL if the list is empty.
1437 */
1438 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1439 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1440 {
1441 struct sk_buff *skb = skb_peek_tail(list);
1442 if (skb)
1443 __skb_unlink(skb, list);
1444 return skb;
1445 }
1446
1447
1448 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1449 {
1450 return skb->data_len;
1451 }
1452
1453 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1454 {
1455 return skb->len - skb->data_len;
1456 }
1457
1458 static inline int skb_pagelen(const struct sk_buff *skb)
1459 {
1460 int i, len = 0;
1461
1462 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1463 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1464 return len + skb_headlen(skb);
1465 }
1466
1467 /**
1468 * __skb_fill_page_desc - initialise a paged fragment in an skb
1469 * @skb: buffer containing fragment to be initialised
1470 * @i: paged fragment index to initialise
1471 * @page: the page to use for this fragment
1472 * @off: the offset to the data with @page
1473 * @size: the length of the data
1474 *
1475 * Initialises the @i'th fragment of @skb to point to &size bytes at
1476 * offset @off within @page.
1477 *
1478 * Does not take any additional reference on the fragment.
1479 */
1480 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1481 struct page *page, int off, int size)
1482 {
1483 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1484
1485 /*
1486 * Propagate page->pfmemalloc to the skb if we can. The problem is
1487 * that not all callers have unique ownership of the page. If
1488 * pfmemalloc is set, we check the mapping as a mapping implies
1489 * page->index is set (index and pfmemalloc share space).
1490 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1491 * do not lose pfmemalloc information as the pages would not be
1492 * allocated using __GFP_MEMALLOC.
1493 */
1494 frag->page.p = page;
1495 frag->page_offset = off;
1496 skb_frag_size_set(frag, size);
1497
1498 page = compound_head(page);
1499 if (page->pfmemalloc && !page->mapping)
1500 skb->pfmemalloc = true;
1501 }
1502
1503 /**
1504 * skb_fill_page_desc - initialise a paged fragment in an skb
1505 * @skb: buffer containing fragment to be initialised
1506 * @i: paged fragment index to initialise
1507 * @page: the page to use for this fragment
1508 * @off: the offset to the data with @page
1509 * @size: the length of the data
1510 *
1511 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1512 * @skb to point to @size bytes at offset @off within @page. In
1513 * addition updates @skb such that @i is the last fragment.
1514 *
1515 * Does not take any additional reference on the fragment.
1516 */
1517 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1518 struct page *page, int off, int size)
1519 {
1520 __skb_fill_page_desc(skb, i, page, off, size);
1521 skb_shinfo(skb)->nr_frags = i + 1;
1522 }
1523
1524 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1525 int size, unsigned int truesize);
1526
1527 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1528 unsigned int truesize);
1529
1530 #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
1531 #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1532 #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1533
1534 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1535 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1536 {
1537 return skb->head + skb->tail;
1538 }
1539
1540 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1541 {
1542 skb->tail = skb->data - skb->head;
1543 }
1544
1545 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1546 {
1547 skb_reset_tail_pointer(skb);
1548 skb->tail += offset;
1549 }
1550
1551 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1552 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1553 {
1554 return skb->tail;
1555 }
1556
1557 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1558 {
1559 skb->tail = skb->data;
1560 }
1561
1562 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1563 {
1564 skb->tail = skb->data + offset;
1565 }
1566
1567 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1568
1569 /*
1570 * Add data to an sk_buff
1571 */
1572 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1573 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1574 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1575 {
1576 unsigned char *tmp = skb_tail_pointer(skb);
1577 SKB_LINEAR_ASSERT(skb);
1578 skb->tail += len;
1579 skb->len += len;
1580 return tmp;
1581 }
1582
1583 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1584 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1585 {
1586 skb->data -= len;
1587 skb->len += len;
1588 return skb->data;
1589 }
1590
1591 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1592 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1593 {
1594 skb->len -= len;
1595 BUG_ON(skb->len < skb->data_len);
1596 return skb->data += len;
1597 }
1598
1599 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1600 {
1601 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1602 }
1603
1604 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1605
1606 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1607 {
1608 if (len > skb_headlen(skb) &&
1609 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1610 return NULL;
1611 skb->len -= len;
1612 return skb->data += len;
1613 }
1614
1615 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1616 {
1617 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1618 }
1619
1620 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1621 {
1622 if (likely(len <= skb_headlen(skb)))
1623 return 1;
1624 if (unlikely(len > skb->len))
1625 return 0;
1626 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1627 }
1628
1629 /**
1630 * skb_headroom - bytes at buffer head
1631 * @skb: buffer to check
1632 *
1633 * Return the number of bytes of free space at the head of an &sk_buff.
1634 */
1635 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1636 {
1637 return skb->data - skb->head;
1638 }
1639
1640 /**
1641 * skb_tailroom - bytes at buffer end
1642 * @skb: buffer to check
1643 *
1644 * Return the number of bytes of free space at the tail of an sk_buff
1645 */
1646 static inline int skb_tailroom(const struct sk_buff *skb)
1647 {
1648 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1649 }
1650
1651 /**
1652 * skb_availroom - bytes at buffer end
1653 * @skb: buffer to check
1654 *
1655 * Return the number of bytes of free space at the tail of an sk_buff
1656 * allocated by sk_stream_alloc()
1657 */
1658 static inline int skb_availroom(const struct sk_buff *skb)
1659 {
1660 if (skb_is_nonlinear(skb))
1661 return 0;
1662
1663 return skb->end - skb->tail - skb->reserved_tailroom;
1664 }
1665
1666 /**
1667 * skb_reserve - adjust headroom
1668 * @skb: buffer to alter
1669 * @len: bytes to move
1670 *
1671 * Increase the headroom of an empty &sk_buff by reducing the tail
1672 * room. This is only allowed for an empty buffer.
1673 */
1674 static inline void skb_reserve(struct sk_buff *skb, int len)
1675 {
1676 skb->data += len;
1677 skb->tail += len;
1678 }
1679
1680 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1681 {
1682 skb->inner_mac_header = skb->mac_header;
1683 skb->inner_network_header = skb->network_header;
1684 skb->inner_transport_header = skb->transport_header;
1685 }
1686
1687 static inline void skb_reset_mac_len(struct sk_buff *skb)
1688 {
1689 skb->mac_len = skb->network_header - skb->mac_header;
1690 }
1691
1692 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1693 *skb)
1694 {
1695 return skb->head + skb->inner_transport_header;
1696 }
1697
1698 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1699 {
1700 skb->inner_transport_header = skb->data - skb->head;
1701 }
1702
1703 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1704 const int offset)
1705 {
1706 skb_reset_inner_transport_header(skb);
1707 skb->inner_transport_header += offset;
1708 }
1709
1710 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1711 {
1712 return skb->head + skb->inner_network_header;
1713 }
1714
1715 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1716 {
1717 skb->inner_network_header = skb->data - skb->head;
1718 }
1719
1720 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1721 const int offset)
1722 {
1723 skb_reset_inner_network_header(skb);
1724 skb->inner_network_header += offset;
1725 }
1726
1727 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1728 {
1729 return skb->head + skb->inner_mac_header;
1730 }
1731
1732 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1733 {
1734 skb->inner_mac_header = skb->data - skb->head;
1735 }
1736
1737 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1738 const int offset)
1739 {
1740 skb_reset_inner_mac_header(skb);
1741 skb->inner_mac_header += offset;
1742 }
1743 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1744 {
1745 return skb->transport_header != (typeof(skb->transport_header))~0U;
1746 }
1747
1748 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1749 {
1750 return skb->head + skb->transport_header;
1751 }
1752
1753 static inline void skb_reset_transport_header(struct sk_buff *skb)
1754 {
1755 skb->transport_header = skb->data - skb->head;
1756 }
1757
1758 static inline void skb_set_transport_header(struct sk_buff *skb,
1759 const int offset)
1760 {
1761 skb_reset_transport_header(skb);
1762 skb->transport_header += offset;
1763 }
1764
1765 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1766 {
1767 return skb->head + skb->network_header;
1768 }
1769
1770 static inline void skb_reset_network_header(struct sk_buff *skb)
1771 {
1772 skb->network_header = skb->data - skb->head;
1773 }
1774
1775 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1776 {
1777 skb_reset_network_header(skb);
1778 skb->network_header += offset;
1779 }
1780
1781 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1782 {
1783 return skb->head + skb->mac_header;
1784 }
1785
1786 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1787 {
1788 return skb->mac_header != (typeof(skb->mac_header))~0U;
1789 }
1790
1791 static inline void skb_reset_mac_header(struct sk_buff *skb)
1792 {
1793 skb->mac_header = skb->data - skb->head;
1794 }
1795
1796 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1797 {
1798 skb_reset_mac_header(skb);
1799 skb->mac_header += offset;
1800 }
1801
1802 static inline void skb_pop_mac_header(struct sk_buff *skb)
1803 {
1804 skb->mac_header = skb->network_header;
1805 }
1806
1807 static inline void skb_probe_transport_header(struct sk_buff *skb,
1808 const int offset_hint)
1809 {
1810 struct flow_keys keys;
1811
1812 if (skb_transport_header_was_set(skb))
1813 return;
1814 else if (skb_flow_dissect(skb, &keys))
1815 skb_set_transport_header(skb, keys.thoff);
1816 else
1817 skb_set_transport_header(skb, offset_hint);
1818 }
1819
1820 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1821 {
1822 if (skb_mac_header_was_set(skb)) {
1823 const unsigned char *old_mac = skb_mac_header(skb);
1824
1825 skb_set_mac_header(skb, -skb->mac_len);
1826 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1827 }
1828 }
1829
1830 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1831 {
1832 return skb->csum_start - skb_headroom(skb);
1833 }
1834
1835 static inline int skb_transport_offset(const struct sk_buff *skb)
1836 {
1837 return skb_transport_header(skb) - skb->data;
1838 }
1839
1840 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1841 {
1842 return skb->transport_header - skb->network_header;
1843 }
1844
1845 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1846 {
1847 return skb->inner_transport_header - skb->inner_network_header;
1848 }
1849
1850 static inline int skb_network_offset(const struct sk_buff *skb)
1851 {
1852 return skb_network_header(skb) - skb->data;
1853 }
1854
1855 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1856 {
1857 return skb_inner_network_header(skb) - skb->data;
1858 }
1859
1860 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1861 {
1862 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1863 }
1864
1865 static inline void skb_pop_rcv_encapsulation(struct sk_buff *skb)
1866 {
1867 /* Only continue with checksum unnecessary if device indicated
1868 * it is valid across encapsulation (skb->encapsulation was set).
1869 */
1870 if (skb->ip_summed == CHECKSUM_UNNECESSARY && !skb->encapsulation)
1871 skb->ip_summed = CHECKSUM_NONE;
1872
1873 skb->encapsulation = 0;
1874 skb->csum_valid = 0;
1875 }
1876
1877 /*
1878 * CPUs often take a performance hit when accessing unaligned memory
1879 * locations. The actual performance hit varies, it can be small if the
1880 * hardware handles it or large if we have to take an exception and fix it
1881 * in software.
1882 *
1883 * Since an ethernet header is 14 bytes network drivers often end up with
1884 * the IP header at an unaligned offset. The IP header can be aligned by
1885 * shifting the start of the packet by 2 bytes. Drivers should do this
1886 * with:
1887 *
1888 * skb_reserve(skb, NET_IP_ALIGN);
1889 *
1890 * The downside to this alignment of the IP header is that the DMA is now
1891 * unaligned. On some architectures the cost of an unaligned DMA is high
1892 * and this cost outweighs the gains made by aligning the IP header.
1893 *
1894 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1895 * to be overridden.
1896 */
1897 #ifndef NET_IP_ALIGN
1898 #define NET_IP_ALIGN 2
1899 #endif
1900
1901 /*
1902 * The networking layer reserves some headroom in skb data (via
1903 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1904 * the header has to grow. In the default case, if the header has to grow
1905 * 32 bytes or less we avoid the reallocation.
1906 *
1907 * Unfortunately this headroom changes the DMA alignment of the resulting
1908 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1909 * on some architectures. An architecture can override this value,
1910 * perhaps setting it to a cacheline in size (since that will maintain
1911 * cacheline alignment of the DMA). It must be a power of 2.
1912 *
1913 * Various parts of the networking layer expect at least 32 bytes of
1914 * headroom, you should not reduce this.
1915 *
1916 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1917 * to reduce average number of cache lines per packet.
1918 * get_rps_cpus() for example only access one 64 bytes aligned block :
1919 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
1920 */
1921 #ifndef NET_SKB_PAD
1922 #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
1923 #endif
1924
1925 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1926
1927 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1928 {
1929 if (unlikely(skb_is_nonlinear(skb))) {
1930 WARN_ON(1);
1931 return;
1932 }
1933 skb->len = len;
1934 skb_set_tail_pointer(skb, len);
1935 }
1936
1937 void skb_trim(struct sk_buff *skb, unsigned int len);
1938
1939 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1940 {
1941 if (skb->data_len)
1942 return ___pskb_trim(skb, len);
1943 __skb_trim(skb, len);
1944 return 0;
1945 }
1946
1947 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1948 {
1949 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1950 }
1951
1952 /**
1953 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1954 * @skb: buffer to alter
1955 * @len: new length
1956 *
1957 * This is identical to pskb_trim except that the caller knows that
1958 * the skb is not cloned so we should never get an error due to out-
1959 * of-memory.
1960 */
1961 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1962 {
1963 int err = pskb_trim(skb, len);
1964 BUG_ON(err);
1965 }
1966
1967 /**
1968 * skb_orphan - orphan a buffer
1969 * @skb: buffer to orphan
1970 *
1971 * If a buffer currently has an owner then we call the owner's
1972 * destructor function and make the @skb unowned. The buffer continues
1973 * to exist but is no longer charged to its former owner.
1974 */
1975 static inline void skb_orphan(struct sk_buff *skb)
1976 {
1977 if (skb->destructor) {
1978 skb->destructor(skb);
1979 skb->destructor = NULL;
1980 skb->sk = NULL;
1981 } else {
1982 BUG_ON(skb->sk);
1983 }
1984 }
1985
1986 /**
1987 * skb_orphan_frags - orphan the frags contained in a buffer
1988 * @skb: buffer to orphan frags from
1989 * @gfp_mask: allocation mask for replacement pages
1990 *
1991 * For each frag in the SKB which needs a destructor (i.e. has an
1992 * owner) create a copy of that frag and release the original
1993 * page by calling the destructor.
1994 */
1995 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1996 {
1997 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1998 return 0;
1999 return skb_copy_ubufs(skb, gfp_mask);
2000 }
2001
2002 /**
2003 * __skb_queue_purge - empty a list
2004 * @list: list to empty
2005 *
2006 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2007 * the list and one reference dropped. This function does not take the
2008 * list lock and the caller must hold the relevant locks to use it.
2009 */
2010 void skb_queue_purge(struct sk_buff_head *list);
2011 static inline void __skb_queue_purge(struct sk_buff_head *list)
2012 {
2013 struct sk_buff *skb;
2014 while ((skb = __skb_dequeue(list)) != NULL)
2015 kfree_skb(skb);
2016 }
2017
2018 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2019 #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2020 #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2021
2022 void *netdev_alloc_frag(unsigned int fragsz);
2023
2024 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2025 gfp_t gfp_mask);
2026
2027 /**
2028 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2029 * @dev: network device to receive on
2030 * @length: length to allocate
2031 *
2032 * Allocate a new &sk_buff and assign it a usage count of one. The
2033 * buffer has unspecified headroom built in. Users should allocate
2034 * the headroom they think they need without accounting for the
2035 * built in space. The built in space is used for optimisations.
2036 *
2037 * %NULL is returned if there is no free memory. Although this function
2038 * allocates memory it can be called from an interrupt.
2039 */
2040 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2041 unsigned int length)
2042 {
2043 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2044 }
2045
2046 /* legacy helper around __netdev_alloc_skb() */
2047 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2048 gfp_t gfp_mask)
2049 {
2050 return __netdev_alloc_skb(NULL, length, gfp_mask);
2051 }
2052
2053 /* legacy helper around netdev_alloc_skb() */
2054 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2055 {
2056 return netdev_alloc_skb(NULL, length);
2057 }
2058
2059
2060 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2061 unsigned int length, gfp_t gfp)
2062 {
2063 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2064
2065 if (NET_IP_ALIGN && skb)
2066 skb_reserve(skb, NET_IP_ALIGN);
2067 return skb;
2068 }
2069
2070 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2071 unsigned int length)
2072 {
2073 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2074 }
2075
2076 /**
2077 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
2078 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2079 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2080 * @order: size of the allocation
2081 *
2082 * Allocate a new page.
2083 *
2084 * %NULL is returned if there is no free memory.
2085 */
2086 static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2087 struct sk_buff *skb,
2088 unsigned int order)
2089 {
2090 struct page *page;
2091
2092 gfp_mask |= __GFP_COLD;
2093
2094 if (!(gfp_mask & __GFP_NOMEMALLOC))
2095 gfp_mask |= __GFP_MEMALLOC;
2096
2097 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2098 if (skb && page && page->pfmemalloc)
2099 skb->pfmemalloc = true;
2100
2101 return page;
2102 }
2103
2104 /**
2105 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2106 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2107 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2108 *
2109 * Allocate a new page.
2110 *
2111 * %NULL is returned if there is no free memory.
2112 */
2113 static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2114 struct sk_buff *skb)
2115 {
2116 return __skb_alloc_pages(gfp_mask, skb, 0);
2117 }
2118
2119 /**
2120 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2121 * @page: The page that was allocated from skb_alloc_page
2122 * @skb: The skb that may need pfmemalloc set
2123 */
2124 static inline void skb_propagate_pfmemalloc(struct page *page,
2125 struct sk_buff *skb)
2126 {
2127 if (page && page->pfmemalloc)
2128 skb->pfmemalloc = true;
2129 }
2130
2131 /**
2132 * skb_frag_page - retrieve the page referred to by a paged fragment
2133 * @frag: the paged fragment
2134 *
2135 * Returns the &struct page associated with @frag.
2136 */
2137 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2138 {
2139 return frag->page.p;
2140 }
2141
2142 /**
2143 * __skb_frag_ref - take an addition reference on a paged fragment.
2144 * @frag: the paged fragment
2145 *
2146 * Takes an additional reference on the paged fragment @frag.
2147 */
2148 static inline void __skb_frag_ref(skb_frag_t *frag)
2149 {
2150 get_page(skb_frag_page(frag));
2151 }
2152
2153 /**
2154 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2155 * @skb: the buffer
2156 * @f: the fragment offset.
2157 *
2158 * Takes an additional reference on the @f'th paged fragment of @skb.
2159 */
2160 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2161 {
2162 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2163 }
2164
2165 /**
2166 * __skb_frag_unref - release a reference on a paged fragment.
2167 * @frag: the paged fragment
2168 *
2169 * Releases a reference on the paged fragment @frag.
2170 */
2171 static inline void __skb_frag_unref(skb_frag_t *frag)
2172 {
2173 put_page(skb_frag_page(frag));
2174 }
2175
2176 /**
2177 * skb_frag_unref - release a reference on a paged fragment of an skb.
2178 * @skb: the buffer
2179 * @f: the fragment offset
2180 *
2181 * Releases a reference on the @f'th paged fragment of @skb.
2182 */
2183 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2184 {
2185 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2186 }
2187
2188 /**
2189 * skb_frag_address - gets the address of the data contained in a paged fragment
2190 * @frag: the paged fragment buffer
2191 *
2192 * Returns the address of the data within @frag. The page must already
2193 * be mapped.
2194 */
2195 static inline void *skb_frag_address(const skb_frag_t *frag)
2196 {
2197 return page_address(skb_frag_page(frag)) + frag->page_offset;
2198 }
2199
2200 /**
2201 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2202 * @frag: the paged fragment buffer
2203 *
2204 * Returns the address of the data within @frag. Checks that the page
2205 * is mapped and returns %NULL otherwise.
2206 */
2207 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2208 {
2209 void *ptr = page_address(skb_frag_page(frag));
2210 if (unlikely(!ptr))
2211 return NULL;
2212
2213 return ptr + frag->page_offset;
2214 }
2215
2216 /**
2217 * __skb_frag_set_page - sets the page contained in a paged fragment
2218 * @frag: the paged fragment
2219 * @page: the page to set
2220 *
2221 * Sets the fragment @frag to contain @page.
2222 */
2223 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2224 {
2225 frag->page.p = page;
2226 }
2227
2228 /**
2229 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2230 * @skb: the buffer
2231 * @f: the fragment offset
2232 * @page: the page to set
2233 *
2234 * Sets the @f'th fragment of @skb to contain @page.
2235 */
2236 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2237 struct page *page)
2238 {
2239 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2240 }
2241
2242 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2243
2244 /**
2245 * skb_frag_dma_map - maps a paged fragment via the DMA API
2246 * @dev: the device to map the fragment to
2247 * @frag: the paged fragment to map
2248 * @offset: the offset within the fragment (starting at the
2249 * fragment's own offset)
2250 * @size: the number of bytes to map
2251 * @dir: the direction of the mapping (%PCI_DMA_*)
2252 *
2253 * Maps the page associated with @frag to @device.
2254 */
2255 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2256 const skb_frag_t *frag,
2257 size_t offset, size_t size,
2258 enum dma_data_direction dir)
2259 {
2260 return dma_map_page(dev, skb_frag_page(frag),
2261 frag->page_offset + offset, size, dir);
2262 }
2263
2264 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2265 gfp_t gfp_mask)
2266 {
2267 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2268 }
2269
2270
2271 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2272 gfp_t gfp_mask)
2273 {
2274 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2275 }
2276
2277
2278 /**
2279 * skb_clone_writable - is the header of a clone writable
2280 * @skb: buffer to check
2281 * @len: length up to which to write
2282 *
2283 * Returns true if modifying the header part of the cloned buffer
2284 * does not requires the data to be copied.
2285 */
2286 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2287 {
2288 return !skb_header_cloned(skb) &&
2289 skb_headroom(skb) + len <= skb->hdr_len;
2290 }
2291
2292 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2293 int cloned)
2294 {
2295 int delta = 0;
2296
2297 if (headroom > skb_headroom(skb))
2298 delta = headroom - skb_headroom(skb);
2299
2300 if (delta || cloned)
2301 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2302 GFP_ATOMIC);
2303 return 0;
2304 }
2305
2306 /**
2307 * skb_cow - copy header of skb when it is required
2308 * @skb: buffer to cow
2309 * @headroom: needed headroom
2310 *
2311 * If the skb passed lacks sufficient headroom or its data part
2312 * is shared, data is reallocated. If reallocation fails, an error
2313 * is returned and original skb is not changed.
2314 *
2315 * The result is skb with writable area skb->head...skb->tail
2316 * and at least @headroom of space at head.
2317 */
2318 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2319 {
2320 return __skb_cow(skb, headroom, skb_cloned(skb));
2321 }
2322
2323 /**
2324 * skb_cow_head - skb_cow but only making the head writable
2325 * @skb: buffer to cow
2326 * @headroom: needed headroom
2327 *
2328 * This function is identical to skb_cow except that we replace the
2329 * skb_cloned check by skb_header_cloned. It should be used when
2330 * you only need to push on some header and do not need to modify
2331 * the data.
2332 */
2333 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2334 {
2335 return __skb_cow(skb, headroom, skb_header_cloned(skb));
2336 }
2337
2338 /**
2339 * skb_padto - pad an skbuff up to a minimal size
2340 * @skb: buffer to pad
2341 * @len: minimal length
2342 *
2343 * Pads up a buffer to ensure the trailing bytes exist and are
2344 * blanked. If the buffer already contains sufficient data it
2345 * is untouched. Otherwise it is extended. Returns zero on
2346 * success. The skb is freed on error.
2347 */
2348
2349 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2350 {
2351 unsigned int size = skb->len;
2352 if (likely(size >= len))
2353 return 0;
2354 return skb_pad(skb, len - size);
2355 }
2356
2357 static inline int skb_add_data(struct sk_buff *skb,
2358 char __user *from, int copy)
2359 {
2360 const int off = skb->len;
2361
2362 if (skb->ip_summed == CHECKSUM_NONE) {
2363 int err = 0;
2364 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
2365 copy, 0, &err);
2366 if (!err) {
2367 skb->csum = csum_block_add(skb->csum, csum, off);
2368 return 0;
2369 }
2370 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2371 return 0;
2372
2373 __skb_trim(skb, off);
2374 return -EFAULT;
2375 }
2376
2377 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2378 const struct page *page, int off)
2379 {
2380 if (i) {
2381 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2382
2383 return page == skb_frag_page(frag) &&
2384 off == frag->page_offset + skb_frag_size(frag);
2385 }
2386 return false;
2387 }
2388
2389 static inline int __skb_linearize(struct sk_buff *skb)
2390 {
2391 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2392 }
2393
2394 /**
2395 * skb_linearize - convert paged skb to linear one
2396 * @skb: buffer to linarize
2397 *
2398 * If there is no free memory -ENOMEM is returned, otherwise zero
2399 * is returned and the old skb data released.
2400 */
2401 static inline int skb_linearize(struct sk_buff *skb)
2402 {
2403 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2404 }
2405
2406 /**
2407 * skb_has_shared_frag - can any frag be overwritten
2408 * @skb: buffer to test
2409 *
2410 * Return true if the skb has at least one frag that might be modified
2411 * by an external entity (as in vmsplice()/sendfile())
2412 */
2413 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2414 {
2415 return skb_is_nonlinear(skb) &&
2416 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2417 }
2418
2419 /**
2420 * skb_linearize_cow - make sure skb is linear and writable
2421 * @skb: buffer to process
2422 *
2423 * If there is no free memory -ENOMEM is returned, otherwise zero
2424 * is returned and the old skb data released.
2425 */
2426 static inline int skb_linearize_cow(struct sk_buff *skb)
2427 {
2428 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2429 __skb_linearize(skb) : 0;
2430 }
2431
2432 /**
2433 * skb_postpull_rcsum - update checksum for received skb after pull
2434 * @skb: buffer to update
2435 * @start: start of data before pull
2436 * @len: length of data pulled
2437 *
2438 * After doing a pull on a received packet, you need to call this to
2439 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2440 * CHECKSUM_NONE so that it can be recomputed from scratch.
2441 */
2442
2443 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2444 const void *start, unsigned int len)
2445 {
2446 if (skb->ip_summed == CHECKSUM_COMPLETE)
2447 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2448 }
2449
2450 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2451
2452 /**
2453 * pskb_trim_rcsum - trim received skb and update checksum
2454 * @skb: buffer to trim
2455 * @len: new length
2456 *
2457 * This is exactly the same as pskb_trim except that it ensures the
2458 * checksum of received packets are still valid after the operation.
2459 */
2460
2461 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2462 {
2463 if (likely(len >= skb->len))
2464 return 0;
2465 if (skb->ip_summed == CHECKSUM_COMPLETE)
2466 skb->ip_summed = CHECKSUM_NONE;
2467 return __pskb_trim(skb, len);
2468 }
2469
2470 #define skb_queue_walk(queue, skb) \
2471 for (skb = (queue)->next; \
2472 skb != (struct sk_buff *)(queue); \
2473 skb = skb->next)
2474
2475 #define skb_queue_walk_safe(queue, skb, tmp) \
2476 for (skb = (queue)->next, tmp = skb->next; \
2477 skb != (struct sk_buff *)(queue); \
2478 skb = tmp, tmp = skb->next)
2479
2480 #define skb_queue_walk_from(queue, skb) \
2481 for (; skb != (struct sk_buff *)(queue); \
2482 skb = skb->next)
2483
2484 #define skb_queue_walk_from_safe(queue, skb, tmp) \
2485 for (tmp = skb->next; \
2486 skb != (struct sk_buff *)(queue); \
2487 skb = tmp, tmp = skb->next)
2488
2489 #define skb_queue_reverse_walk(queue, skb) \
2490 for (skb = (queue)->prev; \
2491 skb != (struct sk_buff *)(queue); \
2492 skb = skb->prev)
2493
2494 #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2495 for (skb = (queue)->prev, tmp = skb->prev; \
2496 skb != (struct sk_buff *)(queue); \
2497 skb = tmp, tmp = skb->prev)
2498
2499 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2500 for (tmp = skb->prev; \
2501 skb != (struct sk_buff *)(queue); \
2502 skb = tmp, tmp = skb->prev)
2503
2504 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2505 {
2506 return skb_shinfo(skb)->frag_list != NULL;
2507 }
2508
2509 static inline void skb_frag_list_init(struct sk_buff *skb)
2510 {
2511 skb_shinfo(skb)->frag_list = NULL;
2512 }
2513
2514 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2515 {
2516 frag->next = skb_shinfo(skb)->frag_list;
2517 skb_shinfo(skb)->frag_list = frag;
2518 }
2519
2520 #define skb_walk_frags(skb, iter) \
2521 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2522
2523 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2524 int *peeked, int *off, int *err);
2525 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2526 int *err);
2527 unsigned int datagram_poll(struct file *file, struct socket *sock,
2528 struct poll_table_struct *wait);
2529 int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2530 struct iovec *to, int size);
2531 int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2532 struct iovec *iov);
2533 int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2534 const struct iovec *from, int from_offset,
2535 int len);
2536 int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2537 int offset, size_t count);
2538 int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2539 const struct iovec *to, int to_offset,
2540 int size);
2541 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2542 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2543 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2544 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2545 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2546 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2547 int len, __wsum csum);
2548 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2549 struct pipe_inode_info *pipe, unsigned int len,
2550 unsigned int flags);
2551 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2552 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2553 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2554 int len, int hlen);
2555 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2556 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2557 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2558 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2559 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2560 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
2561
2562 struct skb_checksum_ops {
2563 __wsum (*update)(const void *mem, int len, __wsum wsum);
2564 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2565 };
2566
2567 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2568 __wsum csum, const struct skb_checksum_ops *ops);
2569 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2570 __wsum csum);
2571
2572 static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2573 int len, void *data, int hlen, void *buffer)
2574 {
2575 if (hlen - offset >= len)
2576 return data + offset;
2577
2578 if (!skb ||
2579 skb_copy_bits(skb, offset, buffer, len) < 0)
2580 return NULL;
2581
2582 return buffer;
2583 }
2584
2585 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2586 int len, void *buffer)
2587 {
2588 return __skb_header_pointer(skb, offset, len, skb->data,
2589 skb_headlen(skb), buffer);
2590 }
2591
2592 /**
2593 * skb_needs_linearize - check if we need to linearize a given skb
2594 * depending on the given device features.
2595 * @skb: socket buffer to check
2596 * @features: net device features
2597 *
2598 * Returns true if either:
2599 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2600 * 2. skb is fragmented and the device does not support SG.
2601 */
2602 static inline bool skb_needs_linearize(struct sk_buff *skb,
2603 netdev_features_t features)
2604 {
2605 return skb_is_nonlinear(skb) &&
2606 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2607 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2608 }
2609
2610 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2611 void *to,
2612 const unsigned int len)
2613 {
2614 memcpy(to, skb->data, len);
2615 }
2616
2617 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2618 const int offset, void *to,
2619 const unsigned int len)
2620 {
2621 memcpy(to, skb->data + offset, len);
2622 }
2623
2624 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2625 const void *from,
2626 const unsigned int len)
2627 {
2628 memcpy(skb->data, from, len);
2629 }
2630
2631 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2632 const int offset,
2633 const void *from,
2634 const unsigned int len)
2635 {
2636 memcpy(skb->data + offset, from, len);
2637 }
2638
2639 void skb_init(void);
2640
2641 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2642 {
2643 return skb->tstamp;
2644 }
2645
2646 /**
2647 * skb_get_timestamp - get timestamp from a skb
2648 * @skb: skb to get stamp from
2649 * @stamp: pointer to struct timeval to store stamp in
2650 *
2651 * Timestamps are stored in the skb as offsets to a base timestamp.
2652 * This function converts the offset back to a struct timeval and stores
2653 * it in stamp.
2654 */
2655 static inline void skb_get_timestamp(const struct sk_buff *skb,
2656 struct timeval *stamp)
2657 {
2658 *stamp = ktime_to_timeval(skb->tstamp);
2659 }
2660
2661 static inline void skb_get_timestampns(const struct sk_buff *skb,
2662 struct timespec *stamp)
2663 {
2664 *stamp = ktime_to_timespec(skb->tstamp);
2665 }
2666
2667 static inline void __net_timestamp(struct sk_buff *skb)
2668 {
2669 skb->tstamp = ktime_get_real();
2670 }
2671
2672 static inline ktime_t net_timedelta(ktime_t t)
2673 {
2674 return ktime_sub(ktime_get_real(), t);
2675 }
2676
2677 static inline ktime_t net_invalid_timestamp(void)
2678 {
2679 return ktime_set(0, 0);
2680 }
2681
2682 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2683
2684 void skb_clone_tx_timestamp(struct sk_buff *skb);
2685 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2686
2687 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2688
2689 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2690 {
2691 }
2692
2693 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2694 {
2695 return false;
2696 }
2697
2698 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2699
2700 /**
2701 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2702 *
2703 * PHY drivers may accept clones of transmitted packets for
2704 * timestamping via their phy_driver.txtstamp method. These drivers
2705 * must call this function to return the skb back to the stack, with
2706 * or without a timestamp.
2707 *
2708 * @skb: clone of the the original outgoing packet
2709 * @hwtstamps: hardware time stamps, may be NULL if not available
2710 *
2711 */
2712 void skb_complete_tx_timestamp(struct sk_buff *skb,
2713 struct skb_shared_hwtstamps *hwtstamps);
2714
2715 void __skb_tstamp_tx(struct sk_buff *orig_skb,
2716 struct skb_shared_hwtstamps *hwtstamps,
2717 struct sock *sk, int tstype);
2718
2719 /**
2720 * skb_tstamp_tx - queue clone of skb with send time stamps
2721 * @orig_skb: the original outgoing packet
2722 * @hwtstamps: hardware time stamps, may be NULL if not available
2723 *
2724 * If the skb has a socket associated, then this function clones the
2725 * skb (thus sharing the actual data and optional structures), stores
2726 * the optional hardware time stamping information (if non NULL) or
2727 * generates a software time stamp (otherwise), then queues the clone
2728 * to the error queue of the socket. Errors are silently ignored.
2729 */
2730 void skb_tstamp_tx(struct sk_buff *orig_skb,
2731 struct skb_shared_hwtstamps *hwtstamps);
2732
2733 static inline void sw_tx_timestamp(struct sk_buff *skb)
2734 {
2735 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2736 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2737 skb_tstamp_tx(skb, NULL);
2738 }
2739
2740 /**
2741 * skb_tx_timestamp() - Driver hook for transmit timestamping
2742 *
2743 * Ethernet MAC Drivers should call this function in their hard_xmit()
2744 * function immediately before giving the sk_buff to the MAC hardware.
2745 *
2746 * Specifically, one should make absolutely sure that this function is
2747 * called before TX completion of this packet can trigger. Otherwise
2748 * the packet could potentially already be freed.
2749 *
2750 * @skb: A socket buffer.
2751 */
2752 static inline void skb_tx_timestamp(struct sk_buff *skb)
2753 {
2754 skb_clone_tx_timestamp(skb);
2755 sw_tx_timestamp(skb);
2756 }
2757
2758 /**
2759 * skb_complete_wifi_ack - deliver skb with wifi status
2760 *
2761 * @skb: the original outgoing packet
2762 * @acked: ack status
2763 *
2764 */
2765 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2766
2767 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2768 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2769
2770 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2771 {
2772 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
2773 }
2774
2775 /**
2776 * skb_checksum_complete - Calculate checksum of an entire packet
2777 * @skb: packet to process
2778 *
2779 * This function calculates the checksum over the entire packet plus
2780 * the value of skb->csum. The latter can be used to supply the
2781 * checksum of a pseudo header as used by TCP/UDP. It returns the
2782 * checksum.
2783 *
2784 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2785 * this function can be used to verify that checksum on received
2786 * packets. In that case the function should return zero if the
2787 * checksum is correct. In particular, this function will return zero
2788 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2789 * hardware has already verified the correctness of the checksum.
2790 */
2791 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2792 {
2793 return skb_csum_unnecessary(skb) ?
2794 0 : __skb_checksum_complete(skb);
2795 }
2796
2797 /* Check if we need to perform checksum complete validation.
2798 *
2799 * Returns true if checksum complete is needed, false otherwise
2800 * (either checksum is unnecessary or zero checksum is allowed).
2801 */
2802 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2803 bool zero_okay,
2804 __sum16 check)
2805 {
2806 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2807 skb->csum_valid = 1;
2808 return false;
2809 }
2810
2811 return true;
2812 }
2813
2814 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2815 * in checksum_init.
2816 */
2817 #define CHECKSUM_BREAK 76
2818
2819 /* Validate (init) checksum based on checksum complete.
2820 *
2821 * Return values:
2822 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2823 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2824 * checksum is stored in skb->csum for use in __skb_checksum_complete
2825 * non-zero: value of invalid checksum
2826 *
2827 */
2828 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2829 bool complete,
2830 __wsum psum)
2831 {
2832 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2833 if (!csum_fold(csum_add(psum, skb->csum))) {
2834 skb->csum_valid = 1;
2835 return 0;
2836 }
2837 }
2838
2839 skb->csum = psum;
2840
2841 if (complete || skb->len <= CHECKSUM_BREAK) {
2842 __sum16 csum;
2843
2844 csum = __skb_checksum_complete(skb);
2845 skb->csum_valid = !csum;
2846 return csum;
2847 }
2848
2849 return 0;
2850 }
2851
2852 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2853 {
2854 return 0;
2855 }
2856
2857 /* Perform checksum validate (init). Note that this is a macro since we only
2858 * want to calculate the pseudo header which is an input function if necessary.
2859 * First we try to validate without any computation (checksum unnecessary) and
2860 * then calculate based on checksum complete calling the function to compute
2861 * pseudo header.
2862 *
2863 * Return values:
2864 * 0: checksum is validated or try to in skb_checksum_complete
2865 * non-zero: value of invalid checksum
2866 */
2867 #define __skb_checksum_validate(skb, proto, complete, \
2868 zero_okay, check, compute_pseudo) \
2869 ({ \
2870 __sum16 __ret = 0; \
2871 skb->csum_valid = 0; \
2872 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2873 __ret = __skb_checksum_validate_complete(skb, \
2874 complete, compute_pseudo(skb, proto)); \
2875 __ret; \
2876 })
2877
2878 #define skb_checksum_init(skb, proto, compute_pseudo) \
2879 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2880
2881 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2882 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2883
2884 #define skb_checksum_validate(skb, proto, compute_pseudo) \
2885 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2886
2887 #define skb_checksum_validate_zero_check(skb, proto, check, \
2888 compute_pseudo) \
2889 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2890
2891 #define skb_checksum_simple_validate(skb) \
2892 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2893
2894 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2895 void nf_conntrack_destroy(struct nf_conntrack *nfct);
2896 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2897 {
2898 if (nfct && atomic_dec_and_test(&nfct->use))
2899 nf_conntrack_destroy(nfct);
2900 }
2901 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2902 {
2903 if (nfct)
2904 atomic_inc(&nfct->use);
2905 }
2906 #endif
2907 #ifdef CONFIG_BRIDGE_NETFILTER
2908 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2909 {
2910 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2911 kfree(nf_bridge);
2912 }
2913 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2914 {
2915 if (nf_bridge)
2916 atomic_inc(&nf_bridge->use);
2917 }
2918 #endif /* CONFIG_BRIDGE_NETFILTER */
2919 static inline void nf_reset(struct sk_buff *skb)
2920 {
2921 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2922 nf_conntrack_put(skb->nfct);
2923 skb->nfct = NULL;
2924 #endif
2925 #ifdef CONFIG_BRIDGE_NETFILTER
2926 nf_bridge_put(skb->nf_bridge);
2927 skb->nf_bridge = NULL;
2928 #endif
2929 }
2930
2931 static inline void nf_reset_trace(struct sk_buff *skb)
2932 {
2933 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2934 skb->nf_trace = 0;
2935 #endif
2936 }
2937
2938 /* Note: This doesn't put any conntrack and bridge info in dst. */
2939 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2940 {
2941 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2942 dst->nfct = src->nfct;
2943 nf_conntrack_get(src->nfct);
2944 dst->nfctinfo = src->nfctinfo;
2945 #endif
2946 #ifdef CONFIG_BRIDGE_NETFILTER
2947 dst->nf_bridge = src->nf_bridge;
2948 nf_bridge_get(src->nf_bridge);
2949 #endif
2950 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2951 dst->nf_trace = src->nf_trace;
2952 #endif
2953 }
2954
2955 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2956 {
2957 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2958 nf_conntrack_put(dst->nfct);
2959 #endif
2960 #ifdef CONFIG_BRIDGE_NETFILTER
2961 nf_bridge_put(dst->nf_bridge);
2962 #endif
2963 __nf_copy(dst, src);
2964 }
2965
2966 #ifdef CONFIG_NETWORK_SECMARK
2967 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2968 {
2969 to->secmark = from->secmark;
2970 }
2971
2972 static inline void skb_init_secmark(struct sk_buff *skb)
2973 {
2974 skb->secmark = 0;
2975 }
2976 #else
2977 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2978 { }
2979
2980 static inline void skb_init_secmark(struct sk_buff *skb)
2981 { }
2982 #endif
2983
2984 static inline bool skb_irq_freeable(const struct sk_buff *skb)
2985 {
2986 return !skb->destructor &&
2987 #if IS_ENABLED(CONFIG_XFRM)
2988 !skb->sp &&
2989 #endif
2990 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
2991 !skb->nfct &&
2992 #endif
2993 !skb->_skb_refdst &&
2994 !skb_has_frag_list(skb);
2995 }
2996
2997 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2998 {
2999 skb->queue_mapping = queue_mapping;
3000 }
3001
3002 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3003 {
3004 return skb->queue_mapping;
3005 }
3006
3007 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3008 {
3009 to->queue_mapping = from->queue_mapping;
3010 }
3011
3012 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3013 {
3014 skb->queue_mapping = rx_queue + 1;
3015 }
3016
3017 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3018 {
3019 return skb->queue_mapping - 1;
3020 }
3021
3022 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3023 {
3024 return skb->queue_mapping != 0;
3025 }
3026
3027 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3028 unsigned int num_tx_queues);
3029
3030 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3031 {
3032 #ifdef CONFIG_XFRM
3033 return skb->sp;
3034 #else
3035 return NULL;
3036 #endif
3037 }
3038
3039 /* Keeps track of mac header offset relative to skb->head.
3040 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3041 * For non-tunnel skb it points to skb_mac_header() and for
3042 * tunnel skb it points to outer mac header.
3043 * Keeps track of level of encapsulation of network headers.
3044 */
3045 struct skb_gso_cb {
3046 int mac_offset;
3047 int encap_level;
3048 __u16 csum_start;
3049 };
3050 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3051
3052 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3053 {
3054 return (skb_mac_header(inner_skb) - inner_skb->head) -
3055 SKB_GSO_CB(inner_skb)->mac_offset;
3056 }
3057
3058 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3059 {
3060 int new_headroom, headroom;
3061 int ret;
3062
3063 headroom = skb_headroom(skb);
3064 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3065 if (ret)
3066 return ret;
3067
3068 new_headroom = skb_headroom(skb);
3069 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3070 return 0;
3071 }
3072
3073 /* Compute the checksum for a gso segment. First compute the checksum value
3074 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3075 * then add in skb->csum (checksum from csum_start to end of packet).
3076 * skb->csum and csum_start are then updated to reflect the checksum of the
3077 * resultant packet starting from the transport header-- the resultant checksum
3078 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3079 * header.
3080 */
3081 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3082 {
3083 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3084 skb_transport_offset(skb);
3085 __u16 csum;
3086
3087 csum = csum_fold(csum_partial(skb_transport_header(skb),
3088 plen, skb->csum));
3089 skb->csum = res;
3090 SKB_GSO_CB(skb)->csum_start -= plen;
3091
3092 return csum;
3093 }
3094
3095 static inline bool skb_is_gso(const struct sk_buff *skb)
3096 {
3097 return skb_shinfo(skb)->gso_size;
3098 }
3099
3100 /* Note: Should be called only if skb_is_gso(skb) is true */
3101 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
3102 {
3103 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3104 }
3105
3106 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
3107
3108 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3109 {
3110 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3111 * wanted then gso_type will be set. */
3112 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3113
3114 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3115 unlikely(shinfo->gso_type == 0)) {
3116 __skb_warn_lro_forwarding(skb);
3117 return true;
3118 }
3119 return false;
3120 }
3121
3122 static inline void skb_forward_csum(struct sk_buff *skb)
3123 {
3124 /* Unfortunately we don't support this one. Any brave souls? */
3125 if (skb->ip_summed == CHECKSUM_COMPLETE)
3126 skb->ip_summed = CHECKSUM_NONE;
3127 }
3128
3129 /**
3130 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3131 * @skb: skb to check
3132 *
3133 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3134 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3135 * use this helper, to document places where we make this assertion.
3136 */
3137 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
3138 {
3139 #ifdef DEBUG
3140 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3141 #endif
3142 }
3143
3144 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
3145
3146 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3147
3148 u32 __skb_get_poff(const struct sk_buff *skb);
3149
3150 /**
3151 * skb_head_is_locked - Determine if the skb->head is locked down
3152 * @skb: skb to check
3153 *
3154 * The head on skbs build around a head frag can be removed if they are
3155 * not cloned. This function returns true if the skb head is locked down
3156 * due to either being allocated via kmalloc, or by being a clone with
3157 * multiple references to the head.
3158 */
3159 static inline bool skb_head_is_locked(const struct sk_buff *skb)
3160 {
3161 return !skb->head_frag || skb_cloned(skb);
3162 }
3163
3164 /**
3165 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3166 *
3167 * @skb: GSO skb
3168 *
3169 * skb_gso_network_seglen is used to determine the real size of the
3170 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3171 *
3172 * The MAC/L2 header is not accounted for.
3173 */
3174 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3175 {
3176 unsigned int hdr_len = skb_transport_header(skb) -
3177 skb_network_header(skb);
3178 return hdr_len + skb_gso_transport_seglen(skb);
3179 }
3180 #endif /* __KERNEL__ */
3181 #endif /* _LINUX_SKBUFF_H */
This page took 0.109704 seconds and 5 git commands to generate.