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