• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  *	Definitions for the 'struct sk_buff' memory handlers.
4  *
5  *	Authors:
6  *		Alan Cox, <gw4pts@gw4pts.ampr.org>
7  *		Florian La Roche, <rzsfl@rz.uni-sb.de>
8  */
9 
10 #ifndef _LINUX_SKBUFF_H
11 #define _LINUX_SKBUFF_H
12 
13 #include <linux/kernel.h>
14 #include <linux/compiler.h>
15 #include <linux/time.h>
16 #include <linux/bug.h>
17 #include <linux/bvec.h>
18 #include <linux/cache.h>
19 #include <linux/rbtree.h>
20 #include <linux/socket.h>
21 #include <linux/refcount.h>
22 
23 #include <linux/atomic.h>
24 #include <asm/types.h>
25 #include <linux/spinlock.h>
26 #include <linux/net.h>
27 #include <linux/textsearch.h>
28 #include <net/checksum.h>
29 #include <linux/rcupdate.h>
30 #include <linux/hrtimer.h>
31 #include <linux/dma-mapping.h>
32 #include <linux/netdev_features.h>
33 #include <linux/sched.h>
34 #include <linux/sched/clock.h>
35 #include <net/flow_dissector.h>
36 #include <linux/splice.h>
37 #include <linux/in6.h>
38 #include <linux/if_packet.h>
39 #include <net/flow.h>
40 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
41 #include <linux/netfilter/nf_conntrack_common.h>
42 #endif
43 
44 /* The interface for checksum offload between the stack and networking drivers
45  * is as follows...
46  *
47  * A. IP checksum related features
48  *
49  * Drivers advertise checksum offload capabilities in the features of a device.
50  * From the stack's point of view these are capabilities offered by the driver.
51  * A driver typically only advertises features that it is capable of offloading
52  * to its device.
53  *
54  * The checksum related features are:
55  *
56  *	NETIF_F_HW_CSUM	- The driver (or its device) is able to compute one
57  *			  IP (one's complement) checksum for any combination
58  *			  of protocols or protocol layering. The checksum is
59  *			  computed and set in a packet per the CHECKSUM_PARTIAL
60  *			  interface (see below).
61  *
62  *	NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63  *			  TCP or UDP packets over IPv4. These are specifically
64  *			  unencapsulated packets of the form IPv4|TCP or
65  *			  IPv4|UDP where the Protocol field in the IPv4 header
66  *			  is TCP or UDP. The IPv4 header may contain IP options.
67  *			  This feature cannot be set in features for a device
68  *			  with NETIF_F_HW_CSUM also set. This feature is being
69  *			  DEPRECATED (see below).
70  *
71  *	NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72  *			  TCP or UDP packets over IPv6. These are specifically
73  *			  unencapsulated packets of the form IPv6|TCP or
74  *			  IPv6|UDP where the Next Header field in the IPv6
75  *			  header is either TCP or UDP. IPv6 extension headers
76  *			  are not supported with this feature. This feature
77  *			  cannot be set in features for a device with
78  *			  NETIF_F_HW_CSUM also set. This feature is being
79  *			  DEPRECATED (see below).
80  *
81  *	NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82  *			 This flag is only used to disable the RX checksum
83  *			 feature for a device. The stack will accept receive
84  *			 checksum indication in packets received on a device
85  *			 regardless of whether NETIF_F_RXCSUM is set.
86  *
87  * B. Checksumming of received packets by device. Indication of checksum
88  *    verification is set in skb->ip_summed. Possible values are:
89  *
90  * CHECKSUM_NONE:
91  *
92  *   Device did not checksum this packet e.g. due to lack of capabilities.
93  *   The packet contains full (though not verified) checksum in packet but
94  *   not in skb->csum. Thus, skb->csum is undefined in this case.
95  *
96  * CHECKSUM_UNNECESSARY:
97  *
98  *   The hardware you're dealing with doesn't calculate the full checksum
99  *   (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
100  *   for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101  *   if their checksums are okay. skb->csum is still undefined in this case
102  *   though. A driver or device must never modify the checksum field in the
103  *   packet even if checksum is verified.
104  *
105  *   CHECKSUM_UNNECESSARY is applicable to following protocols:
106  *     TCP: IPv6 and IPv4.
107  *     UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108  *       zero UDP checksum for either IPv4 or IPv6, the networking stack
109  *       may perform further validation in this case.
110  *     GRE: only if the checksum is present in the header.
111  *     SCTP: indicates the CRC in SCTP header has been validated.
112  *     FCOE: indicates the CRC in FC frame has been validated.
113  *
114  *   skb->csum_level indicates the number of consecutive checksums found in
115  *   the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116  *   For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117  *   and a device is able to verify the checksums for UDP (possibly zero),
118  *   GRE (checksum flag is set) and TCP, skb->csum_level would be set to
119  *   two. If the device were only able to verify the UDP checksum and not
120  *   GRE, either because it doesn't support GRE checksum or because GRE
121  *   checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122  *   not considered in this case).
123  *
124  * CHECKSUM_COMPLETE:
125  *
126  *   This is the most generic way. The device supplied checksum of the _whole_
127  *   packet as seen by netif_rx() and fills in skb->csum. This means the
128  *   hardware doesn't need to parse L3/L4 headers to implement this.
129  *
130  *   Notes:
131  *   - Even if device supports only some protocols, but is able to produce
132  *     skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133  *   - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
134  *
135  * CHECKSUM_PARTIAL:
136  *
137  *   A checksum is set up to be offloaded to a device as described in the
138  *   output description for CHECKSUM_PARTIAL. This may occur on a packet
139  *   received directly from another Linux OS, e.g., a virtualized Linux kernel
140  *   on the same host, or it may be set in the input path in GRO or remote
141  *   checksum offload. For the purposes of checksum verification, the checksum
142  *   referred to by skb->csum_start + skb->csum_offset and any preceding
143  *   checksums in the packet are considered verified. Any checksums in the
144  *   packet that are after the checksum being offloaded are not considered to
145  *   be verified.
146  *
147  * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148  *    in the skb->ip_summed for a packet. Values are:
149  *
150  * CHECKSUM_PARTIAL:
151  *
152  *   The driver is required to checksum the packet as seen by hard_start_xmit()
153  *   from skb->csum_start up to the end, and to record/write the checksum at
154  *   offset skb->csum_start + skb->csum_offset. A driver may verify that the
155  *   csum_start and csum_offset values are valid values given the length and
156  *   offset of the packet, but it should not attempt to validate that the
157  *   checksum refers to a legitimate transport layer checksum -- it is the
158  *   purview of the stack to validate that csum_start and csum_offset are set
159  *   correctly.
160  *
161  *   When the stack requests checksum offload for a packet, the driver MUST
162  *   ensure that the checksum is set correctly. A driver can either offload the
163  *   checksum calculation to the device, or call skb_checksum_help (in the case
164  *   that the device does not support offload for a particular checksum).
165  *
166  *   NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167  *   NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
168  *   checksum offload capability.
169  *   skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170  *   on network device checksumming capabilities: if a packet does not match
171  *   them, skb_checksum_help or skb_crc32c_help (depending on the value of
172  *   csum_not_inet, see item D.) is called to resolve the checksum.
173  *
174  * CHECKSUM_NONE:
175  *
176  *   The skb was already checksummed by the protocol, or a checksum is not
177  *   required.
178  *
179  * CHECKSUM_UNNECESSARY:
180  *
181  *   This has the same meaning as CHECKSUM_NONE for checksum offload on
182  *   output.
183  *
184  * CHECKSUM_COMPLETE:
185  *   Not used in checksum output. If a driver observes a packet with this value
186  *   set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
187  *
188  * D. Non-IP checksum (CRC) offloads
189  *
190  *   NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191  *     offloading the SCTP CRC in a packet. To perform this offload the stack
192  *     will set csum_start and csum_offset accordingly, set ip_summed to
193  *     CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194  *     the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195  *     A driver that supports both IP checksum offload and SCTP CRC32c offload
196  *     must verify which offload is configured for a packet by testing the
197  *     value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198  *     CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
199  *
200  *   NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201  *     offloading the FCOE CRC in a packet. To perform this offload the stack
202  *     will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203  *     accordingly. Note that there is no indication in the skbuff that the
204  *     CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
205  *     both IP checksum offload and FCOE CRC offload must verify which offload
206  *     is configured for a packet, presumably by inspecting packet headers.
207  *
208  * E. Checksumming on output with GSO.
209  *
210  * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211  * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212  * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213  * part of the GSO operation is implied. If a checksum is being offloaded
214  * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
215  * csum_offset are set to refer to the outermost checksum being offloaded
216  * (two offloaded checksums are possible with UDP encapsulation).
217  */
218 
219 /* Don't change this without changing skb_csum_unnecessary! */
220 #define CHECKSUM_NONE		0
221 #define CHECKSUM_UNNECESSARY	1
222 #define CHECKSUM_COMPLETE	2
223 #define CHECKSUM_PARTIAL	3
224 
225 /* Maximum value in skb->csum_level */
226 #define SKB_MAX_CSUM_LEVEL	3
227 
228 #define SKB_DATA_ALIGN(X)	ALIGN(X, SMP_CACHE_BYTES)
229 #define SKB_WITH_OVERHEAD(X)	\
230 	((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
231 #define SKB_MAX_ORDER(X, ORDER) \
232 	SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
233 #define SKB_MAX_HEAD(X)		(SKB_MAX_ORDER((X), 0))
234 #define SKB_MAX_ALLOC		(SKB_MAX_ORDER(0, 2))
235 
236 /* return minimum truesize of one skb containing X bytes of data */
237 #define SKB_TRUESIZE(X) ((X) +						\
238 			 SKB_DATA_ALIGN(sizeof(struct sk_buff)) +	\
239 			 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240 
241 struct ahash_request;
242 struct net_device;
243 struct scatterlist;
244 struct pipe_inode_info;
245 struct iov_iter;
246 struct napi_struct;
247 struct bpf_prog;
248 union bpf_attr;
249 struct skb_ext;
250 
251 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
252 struct nf_bridge_info {
253 	enum {
254 		BRNF_PROTO_UNCHANGED,
255 		BRNF_PROTO_8021Q,
256 		BRNF_PROTO_PPPOE
257 	} orig_proto:8;
258 	u8			pkt_otherhost:1;
259 	u8			in_prerouting:1;
260 	u8			bridged_dnat:1;
261 	__u16			frag_max_size;
262 	struct net_device	*physindev;
263 
264 	/* always valid & non-NULL from FORWARD on, for physdev match */
265 	struct net_device	*physoutdev;
266 	union {
267 		/* prerouting: detect dnat in orig/reply direction */
268 		__be32          ipv4_daddr;
269 		struct in6_addr ipv6_daddr;
270 
271 		/* after prerouting + nat detected: store original source
272 		 * mac since neigh resolution overwrites it, only used while
273 		 * skb is out in neigh layer.
274 		 */
275 		char neigh_header[8];
276 	};
277 };
278 #endif
279 
280 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
281 /* Chain in tc_skb_ext will be used to share the tc chain with
282  * ovs recirc_id. It will be set to the current chain by tc
283  * and read by ovs to recirc_id.
284  */
285 struct tc_skb_ext {
286 	__u32 chain;
287 	__u16 mru;
288 };
289 #endif
290 
291 struct sk_buff_head {
292 	/* These two members must be first. */
293 	struct sk_buff	*next;
294 	struct sk_buff	*prev;
295 
296 	__u32		qlen;
297 	spinlock_t	lock;
298 };
299 
300 struct sk_buff;
301 
302 /* To allow 64K frame to be packed as single skb without frag_list we
303  * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
304  * buffers which do not start on a page boundary.
305  *
306  * Since GRO uses frags we allocate at least 16 regardless of page
307  * size.
308  */
309 #if (65536/PAGE_SIZE + 1) < 16
310 #define MAX_SKB_FRAGS 16UL
311 #else
312 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
313 #endif
314 extern int sysctl_max_skb_frags;
315 
316 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
317  * segment using its current segmentation instead.
318  */
319 #define GSO_BY_FRAGS	0xFFFF
320 
321 typedef struct bio_vec skb_frag_t;
322 
323 /**
324  * skb_frag_size() - Returns the size of a skb fragment
325  * @frag: skb fragment
326  */
skb_frag_size(const skb_frag_t * frag)327 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
328 {
329 	return frag->bv_len;
330 }
331 
332 /**
333  * skb_frag_size_set() - Sets the size of a skb fragment
334  * @frag: skb fragment
335  * @size: size of fragment
336  */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)337 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
338 {
339 	frag->bv_len = size;
340 }
341 
342 /**
343  * skb_frag_size_add() - Increments the size of a skb fragment by @delta
344  * @frag: skb fragment
345  * @delta: value to add
346  */
skb_frag_size_add(skb_frag_t * frag,int delta)347 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
348 {
349 	frag->bv_len += delta;
350 }
351 
352 /**
353  * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
354  * @frag: skb fragment
355  * @delta: value to subtract
356  */
skb_frag_size_sub(skb_frag_t * frag,int delta)357 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
358 {
359 	frag->bv_len -= delta;
360 }
361 
362 /**
363  * skb_frag_must_loop - Test if %p is a high memory page
364  * @p: fragment's page
365  */
skb_frag_must_loop(struct page * p)366 static inline bool skb_frag_must_loop(struct page *p)
367 {
368 #if defined(CONFIG_HIGHMEM)
369 	if (PageHighMem(p))
370 		return true;
371 #endif
372 	return false;
373 }
374 
375 /**
376  *	skb_frag_foreach_page - loop over pages in a fragment
377  *
378  *	@f:		skb frag to operate on
379  *	@f_off:		offset from start of f->bv_page
380  *	@f_len:		length from f_off to loop over
381  *	@p:		(temp var) current page
382  *	@p_off:		(temp var) offset from start of current page,
383  *	                           non-zero only on first page.
384  *	@p_len:		(temp var) length in current page,
385  *				   < PAGE_SIZE only on first and last page.
386  *	@copied:	(temp var) length so far, excluding current p_len.
387  *
388  *	A fragment can hold a compound page, in which case per-page
389  *	operations, notably kmap_atomic, must be called for each
390  *	regular page.
391  */
392 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied)	\
393 	for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT),		\
394 	     p_off = (f_off) & (PAGE_SIZE - 1),				\
395 	     p_len = skb_frag_must_loop(p) ?				\
396 	     min_t(u32, f_len, PAGE_SIZE - p_off) : f_len,		\
397 	     copied = 0;						\
398 	     copied < f_len;						\
399 	     copied += p_len, p++, p_off = 0,				\
400 	     p_len = min_t(u32, f_len - copied, PAGE_SIZE))		\
401 
402 #define HAVE_HW_TIME_STAMP
403 
404 /**
405  * struct skb_shared_hwtstamps - hardware time stamps
406  * @hwtstamp:	hardware time stamp transformed into duration
407  *		since arbitrary point in time
408  *
409  * Software time stamps generated by ktime_get_real() are stored in
410  * skb->tstamp.
411  *
412  * hwtstamps can only be compared against other hwtstamps from
413  * the same device.
414  *
415  * This structure is attached to packets as part of the
416  * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
417  */
418 struct skb_shared_hwtstamps {
419 	ktime_t	hwtstamp;
420 };
421 
422 /* Definitions for tx_flags in struct skb_shared_info */
423 enum {
424 	/* generate hardware time stamp */
425 	SKBTX_HW_TSTAMP = 1 << 0,
426 
427 	/* generate software time stamp when queueing packet to NIC */
428 	SKBTX_SW_TSTAMP = 1 << 1,
429 
430 	/* device driver is going to provide hardware time stamp */
431 	SKBTX_IN_PROGRESS = 1 << 2,
432 
433 	/* device driver supports TX zero-copy buffers */
434 	SKBTX_DEV_ZEROCOPY = 1 << 3,
435 
436 	/* generate wifi status information (where possible) */
437 	SKBTX_WIFI_STATUS = 1 << 4,
438 
439 	/* This indicates at least one fragment might be overwritten
440 	 * (as in vmsplice(), sendfile() ...)
441 	 * If we need to compute a TX checksum, we'll need to copy
442 	 * all frags to avoid possible bad checksum
443 	 */
444 	SKBTX_SHARED_FRAG = 1 << 5,
445 
446 	/* generate software time stamp when entering packet scheduling */
447 	SKBTX_SCHED_TSTAMP = 1 << 6,
448 };
449 
450 #define SKBTX_ZEROCOPY_FRAG	(SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
451 #define SKBTX_ANY_SW_TSTAMP	(SKBTX_SW_TSTAMP    | \
452 				 SKBTX_SCHED_TSTAMP)
453 #define SKBTX_ANY_TSTAMP	(SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
454 
455 /*
456  * The callback notifies userspace to release buffers when skb DMA is done in
457  * lower device, the skb last reference should be 0 when calling this.
458  * The zerocopy_success argument is true if zero copy transmit occurred,
459  * false on data copy or out of memory error caused by data copy attempt.
460  * The ctx field is used to track device context.
461  * The desc field is used to track userspace buffer index.
462  */
463 struct ubuf_info {
464 	void (*callback)(struct ubuf_info *, bool zerocopy_success);
465 	union {
466 		struct {
467 			unsigned long desc;
468 			void *ctx;
469 		};
470 		struct {
471 			u32 id;
472 			u16 len;
473 			u16 zerocopy:1;
474 			u32 bytelen;
475 		};
476 	};
477 	refcount_t refcnt;
478 
479 	struct mmpin {
480 		struct user_struct *user;
481 		unsigned int num_pg;
482 	} mmp;
483 };
484 
485 #define skb_uarg(SKB)	((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
486 
487 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
488 void mm_unaccount_pinned_pages(struct mmpin *mmp);
489 
490 struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
491 struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
492 					struct ubuf_info *uarg);
493 
sock_zerocopy_get(struct ubuf_info * uarg)494 static inline void sock_zerocopy_get(struct ubuf_info *uarg)
495 {
496 	refcount_inc(&uarg->refcnt);
497 }
498 
499 void sock_zerocopy_put(struct ubuf_info *uarg);
500 void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
501 
502 void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
503 
504 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
505 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
506 			     struct msghdr *msg, int len,
507 			     struct ubuf_info *uarg);
508 
509 /* This data is invariant across clones and lives at
510  * the end of the header data, ie. at skb->end.
511  */
512 struct skb_shared_info {
513 	__u8		__unused;
514 	__u8		meta_len;
515 	__u8		nr_frags;
516 	__u8		tx_flags;
517 	unsigned short	gso_size;
518 	/* Warning: this field is not always filled in (UFO)! */
519 	unsigned short	gso_segs;
520 	struct sk_buff	*frag_list;
521 	struct skb_shared_hwtstamps hwtstamps;
522 	unsigned int	gso_type;
523 	u32		tskey;
524 
525 	/*
526 	 * Warning : all fields before dataref are cleared in __alloc_skb()
527 	 */
528 	atomic_t	dataref;
529 
530 	/* Intermediate layers must ensure that destructor_arg
531 	 * remains valid until skb destructor */
532 	void *		destructor_arg;
533 
534 	/* must be last field, see pskb_expand_head() */
535 	skb_frag_t	frags[MAX_SKB_FRAGS];
536 };
537 
538 /* We divide dataref into two halves.  The higher 16 bits hold references
539  * to the payload part of skb->data.  The lower 16 bits hold references to
540  * the entire skb->data.  A clone of a headerless skb holds the length of
541  * the header in skb->hdr_len.
542  *
543  * All users must obey the rule that the skb->data reference count must be
544  * greater than or equal to the payload reference count.
545  *
546  * Holding a reference to the payload part means that the user does not
547  * care about modifications to the header part of skb->data.
548  */
549 #define SKB_DATAREF_SHIFT 16
550 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
551 
552 
553 enum {
554 	SKB_FCLONE_UNAVAILABLE,	/* skb has no fclone (from head_cache) */
555 	SKB_FCLONE_ORIG,	/* orig skb (from fclone_cache) */
556 	SKB_FCLONE_CLONE,	/* companion fclone skb (from fclone_cache) */
557 };
558 
559 enum {
560 	SKB_GSO_TCPV4 = 1 << 0,
561 
562 	/* This indicates the skb is from an untrusted source. */
563 	SKB_GSO_DODGY = 1 << 1,
564 
565 	/* This indicates the tcp segment has CWR set. */
566 	SKB_GSO_TCP_ECN = 1 << 2,
567 
568 	SKB_GSO_TCP_FIXEDID = 1 << 3,
569 
570 	SKB_GSO_TCPV6 = 1 << 4,
571 
572 	SKB_GSO_FCOE = 1 << 5,
573 
574 	SKB_GSO_GRE = 1 << 6,
575 
576 	SKB_GSO_GRE_CSUM = 1 << 7,
577 
578 	SKB_GSO_IPXIP4 = 1 << 8,
579 
580 	SKB_GSO_IPXIP6 = 1 << 9,
581 
582 	SKB_GSO_UDP_TUNNEL = 1 << 10,
583 
584 	SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
585 
586 	SKB_GSO_PARTIAL = 1 << 12,
587 
588 	SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
589 
590 	SKB_GSO_SCTP = 1 << 14,
591 
592 	SKB_GSO_ESP = 1 << 15,
593 
594 	SKB_GSO_UDP = 1 << 16,
595 
596 	SKB_GSO_UDP_L4 = 1 << 17,
597 
598 	SKB_GSO_FRAGLIST = 1 << 18,
599 };
600 
601 #if BITS_PER_LONG > 32
602 #define NET_SKBUFF_DATA_USES_OFFSET 1
603 #endif
604 
605 #ifdef NET_SKBUFF_DATA_USES_OFFSET
606 typedef unsigned int sk_buff_data_t;
607 #else
608 typedef unsigned char *sk_buff_data_t;
609 #endif
610 
611 /**
612  *	struct sk_buff - socket buffer
613  *	@next: Next buffer in list
614  *	@prev: Previous buffer in list
615  *	@tstamp: Time we arrived/left
616  *	@skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
617  *		for retransmit timer
618  *	@rbnode: RB tree node, alternative to next/prev for netem/tcp
619  *	@list: queue head
620  *	@sk: Socket we are owned by
621  *	@ip_defrag_offset: (aka @sk) alternate use of @sk, used in
622  *		fragmentation management
623  *	@dev: Device we arrived on/are leaving by
624  *	@dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
625  *	@cb: Control buffer. Free for use by every layer. Put private vars here
626  *	@_skb_refdst: destination entry (with norefcount bit)
627  *	@sp: the security path, used for xfrm
628  *	@len: Length of actual data
629  *	@data_len: Data length
630  *	@mac_len: Length of link layer header
631  *	@hdr_len: writable header length of cloned skb
632  *	@csum: Checksum (must include start/offset pair)
633  *	@csum_start: Offset from skb->head where checksumming should start
634  *	@csum_offset: Offset from csum_start where checksum should be stored
635  *	@priority: Packet queueing priority
636  *	@ignore_df: allow local fragmentation
637  *	@cloned: Head may be cloned (check refcnt to be sure)
638  *	@ip_summed: Driver fed us an IP checksum
639  *	@nohdr: Payload reference only, must not modify header
640  *	@pkt_type: Packet class
641  *	@fclone: skbuff clone status
642  *	@ipvs_property: skbuff is owned by ipvs
643  *	@inner_protocol_type: whether the inner protocol is
644  *		ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
645  *	@remcsum_offload: remote checksum offload is enabled
646  *	@offload_fwd_mark: Packet was L2-forwarded in hardware
647  *	@offload_l3_fwd_mark: Packet was L3-forwarded in hardware
648  *	@tc_skip_classify: do not classify packet. set by IFB device
649  *	@tc_at_ingress: used within tc_classify to distinguish in/egress
650  *	@redirected: packet was redirected by packet classifier
651  *	@from_ingress: packet was redirected from the ingress path
652  *	@peeked: this packet has been seen already, so stats have been
653  *		done for it, don't do them again
654  *	@nf_trace: netfilter packet trace flag
655  *	@protocol: Packet protocol from driver
656  *	@destructor: Destruct function
657  *	@tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
658  *	@_nfct: Associated connection, if any (with nfctinfo bits)
659  *	@nf_bridge: Saved data about a bridged frame - see br_netfilter.c
660  *	@skb_iif: ifindex of device we arrived on
661  *	@tc_index: Traffic control index
662  *	@hash: the packet hash
663  *	@queue_mapping: Queue mapping for multiqueue devices
664  *	@head_frag: skb was allocated from page fragments,
665  *		not allocated by kmalloc() or vmalloc().
666  *	@pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
667  *	@active_extensions: active extensions (skb_ext_id types)
668  *	@ndisc_nodetype: router type (from link layer)
669  *	@ooo_okay: allow the mapping of a socket to a queue to be changed
670  *	@l4_hash: indicate hash is a canonical 4-tuple hash over transport
671  *		ports.
672  *	@sw_hash: indicates hash was computed in software stack
673  *	@wifi_acked_valid: wifi_acked was set
674  *	@wifi_acked: whether frame was acked on wifi or not
675  *	@no_fcs:  Request NIC to treat last 4 bytes as Ethernet FCS
676  *	@encapsulation: indicates the inner headers in the skbuff are valid
677  *	@encap_hdr_csum: software checksum is needed
678  *	@csum_valid: checksum is already valid
679  *	@csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
680  *	@csum_complete_sw: checksum was completed by software
681  *	@csum_level: indicates the number of consecutive checksums found in
682  *		the packet minus one that have been verified as
683  *		CHECKSUM_UNNECESSARY (max 3)
684  *	@scm_io_uring: SKB holds io_uring registered files
685  *	@dst_pending_confirm: need to confirm neighbour
686  *	@decrypted: Decrypted SKB
687  *	@napi_id: id of the NAPI struct this skb came from
688  *	@sender_cpu: (aka @napi_id) source CPU in XPS
689  *	@secmark: security marking
690  *	@mark: Generic packet mark
691  *	@reserved_tailroom: (aka @mark) number of bytes of free space available
692  *		at the tail of an sk_buff
693  *	@vlan_present: VLAN tag is present
694  *	@vlan_proto: vlan encapsulation protocol
695  *	@vlan_tci: vlan tag control information
696  *	@inner_protocol: Protocol (encapsulation)
697  *	@inner_ipproto: (aka @inner_protocol) stores ipproto when
698  *		skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
699  *	@inner_transport_header: Inner transport layer header (encapsulation)
700  *	@inner_network_header: Network layer header (encapsulation)
701  *	@inner_mac_header: Link layer header (encapsulation)
702  *	@transport_header: Transport layer header
703  *	@network_header: Network layer header
704  *	@mac_header: Link layer header
705  *	@tail: Tail pointer
706  *	@end: End pointer
707  *	@head: Head of buffer
708  *	@data: Data head pointer
709  *	@truesize: Buffer size
710  *	@users: User count - see {datagram,tcp}.c
711  *	@extensions: allocated extensions, valid if active_extensions is nonzero
712  */
713 
714 struct sk_buff {
715 	union {
716 		struct {
717 			/* These two members must be first. */
718 			struct sk_buff		*next;
719 			struct sk_buff		*prev;
720 
721 			union {
722 				struct net_device	*dev;
723 				/* Some protocols might use this space to store information,
724 				 * while device pointer would be NULL.
725 				 * UDP receive path is one user.
726 				 */
727 				unsigned long		dev_scratch;
728 			};
729 		};
730 		struct rb_node		rbnode; /* used in netem, ip4 defrag, and tcp stack */
731 		struct list_head	list;
732 	};
733 
734 	union {
735 		struct sock		*sk;
736 		int			ip_defrag_offset;
737 	};
738 
739 	union {
740 		ktime_t		tstamp;
741 		u64		skb_mstamp_ns; /* earliest departure time */
742 	};
743 	/*
744 	 * This is the control buffer. It is free to use for every
745 	 * layer. Please put your private variables there. If you
746 	 * want to keep them across layers you have to do a skb_clone()
747 	 * first. This is owned by whoever has the skb queued ATM.
748 	 */
749 	char			cb[48] __aligned(8);
750 
751 	union {
752 		struct {
753 			unsigned long	_skb_refdst;
754 			void		(*destructor)(struct sk_buff *skb);
755 		};
756 		struct list_head	tcp_tsorted_anchor;
757 	};
758 
759 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
760 	unsigned long		 _nfct;
761 #endif
762 	unsigned int		len,
763 				data_len;
764 	__u16			mac_len,
765 				hdr_len;
766 
767 	/* Following fields are _not_ copied in __copy_skb_header()
768 	 * Note that queue_mapping is here mostly to fill a hole.
769 	 */
770 	__u16			queue_mapping;
771 
772 /* if you move cloned around you also must adapt those constants */
773 #ifdef __BIG_ENDIAN_BITFIELD
774 #define CLONED_MASK	(1 << 7)
775 #else
776 #define CLONED_MASK	1
777 #endif
778 #define CLONED_OFFSET()		offsetof(struct sk_buff, __cloned_offset)
779 
780 	/* private: */
781 	__u8			__cloned_offset[0];
782 	/* public: */
783 	__u8			cloned:1,
784 				nohdr:1,
785 				fclone:2,
786 				peeked:1,
787 				head_frag:1,
788 				pfmemalloc:1;
789 #ifdef CONFIG_SKB_EXTENSIONS
790 	__u8			active_extensions;
791 #endif
792 	/* fields enclosed in headers_start/headers_end are copied
793 	 * using a single memcpy() in __copy_skb_header()
794 	 */
795 	/* private: */
796 	__u32			headers_start[0];
797 	/* public: */
798 
799 /* if you move pkt_type around you also must adapt those constants */
800 #ifdef __BIG_ENDIAN_BITFIELD
801 #define PKT_TYPE_MAX	(7 << 5)
802 #else
803 #define PKT_TYPE_MAX	7
804 #endif
805 #define PKT_TYPE_OFFSET()	offsetof(struct sk_buff, __pkt_type_offset)
806 
807 	/* private: */
808 	__u8			__pkt_type_offset[0];
809 	/* public: */
810 	__u8			pkt_type:3;
811 	__u8			ignore_df:1;
812 	__u8			nf_trace:1;
813 	__u8			ip_summed:2;
814 	__u8			ooo_okay:1;
815 
816 	__u8			l4_hash:1;
817 	__u8			sw_hash:1;
818 	__u8			wifi_acked_valid:1;
819 	__u8			wifi_acked:1;
820 	__u8			no_fcs:1;
821 	/* Indicates the inner headers are valid in the skbuff. */
822 	__u8			encapsulation:1;
823 	__u8			encap_hdr_csum:1;
824 	__u8			csum_valid:1;
825 
826 #ifdef __BIG_ENDIAN_BITFIELD
827 #define PKT_VLAN_PRESENT_BIT	7
828 #else
829 #define PKT_VLAN_PRESENT_BIT	0
830 #endif
831 #define PKT_VLAN_PRESENT_OFFSET()	offsetof(struct sk_buff, __pkt_vlan_present_offset)
832 	/* private: */
833 	__u8			__pkt_vlan_present_offset[0];
834 	/* public: */
835 	__u8			vlan_present:1;
836 	__u8			csum_complete_sw:1;
837 	__u8			csum_level:2;
838 	__u8			csum_not_inet:1;
839 	__u8			dst_pending_confirm:1;
840 #ifdef CONFIG_IPV6_NDISC_NODETYPE
841 	__u8			ndisc_nodetype:2;
842 #endif
843 
844 	__u8			ipvs_property:1;
845 	__u8			inner_protocol_type:1;
846 	__u8			remcsum_offload:1;
847 #ifdef CONFIG_NET_SWITCHDEV
848 	__u8			offload_fwd_mark:1;
849 	__u8			offload_l3_fwd_mark:1;
850 #endif
851 #ifdef CONFIG_NET_CLS_ACT
852 	__u8			tc_skip_classify:1;
853 	__u8			tc_at_ingress:1;
854 #endif
855 #ifdef CONFIG_NET_REDIRECT
856 	__u8			redirected:1;
857 	__u8			from_ingress:1;
858 #endif
859 #ifdef CONFIG_TLS_DEVICE
860 	__u8			decrypted:1;
861 #endif
862 	__u8			scm_io_uring:1;
863 
864 #ifdef CONFIG_NET_SCHED
865 	__u16			tc_index;	/* traffic control index */
866 #endif
867 
868 	union {
869 		__wsum		csum;
870 		struct {
871 			__u16	csum_start;
872 			__u16	csum_offset;
873 		};
874 	};
875 	__u32			priority;
876 	int			skb_iif;
877 	__u32			hash;
878 	__be16			vlan_proto;
879 	__u16			vlan_tci;
880 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
881 	union {
882 		unsigned int	napi_id;
883 		unsigned int	sender_cpu;
884 	};
885 #endif
886 #ifdef CONFIG_NETWORK_SECMARK
887 	__u32		secmark;
888 #endif
889 
890 	union {
891 		__u32		mark;
892 		__u32		reserved_tailroom;
893 	};
894 
895 	union {
896 		__be16		inner_protocol;
897 		__u8		inner_ipproto;
898 	};
899 
900 	__u16			inner_transport_header;
901 	__u16			inner_network_header;
902 	__u16			inner_mac_header;
903 
904 	__be16			protocol;
905 	__u16			transport_header;
906 	__u16			network_header;
907 	__u16			mac_header;
908 
909 	/* private: */
910 	__u32			headers_end[0];
911 	/* public: */
912 
913 	/* These elements must be at the end, see alloc_skb() for details.  */
914 	sk_buff_data_t		tail;
915 	sk_buff_data_t		end;
916 	unsigned char		*head,
917 				*data;
918 	unsigned int		truesize;
919 	refcount_t		users;
920 
921 #ifdef CONFIG_SKB_EXTENSIONS
922 	/* only useable after checking ->active_extensions != 0 */
923 	struct skb_ext		*extensions;
924 #endif
925 };
926 
927 #ifdef __KERNEL__
928 /*
929  *	Handling routines are only of interest to the kernel
930  */
931 
932 #define SKB_ALLOC_FCLONE	0x01
933 #define SKB_ALLOC_RX		0x02
934 #define SKB_ALLOC_NAPI		0x04
935 
936 /**
937  * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
938  * @skb: buffer
939  */
skb_pfmemalloc(const struct sk_buff * skb)940 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
941 {
942 	return unlikely(skb->pfmemalloc);
943 }
944 
945 /*
946  * skb might have a dst pointer attached, refcounted or not.
947  * _skb_refdst low order bit is set if refcount was _not_ taken
948  */
949 #define SKB_DST_NOREF	1UL
950 #define SKB_DST_PTRMASK	~(SKB_DST_NOREF)
951 
952 /**
953  * skb_dst - returns skb dst_entry
954  * @skb: buffer
955  *
956  * Returns skb dst_entry, regardless of reference taken or not.
957  */
skb_dst(const struct sk_buff * skb)958 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
959 {
960 	/* If refdst was not refcounted, check we still are in a
961 	 * rcu_read_lock section
962 	 */
963 	WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
964 		!rcu_read_lock_held() &&
965 		!rcu_read_lock_bh_held());
966 	return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
967 }
968 
969 /**
970  * skb_dst_set - sets skb dst
971  * @skb: buffer
972  * @dst: dst entry
973  *
974  * Sets skb dst, assuming a reference was taken on dst and should
975  * be released by skb_dst_drop()
976  */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)977 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
978 {
979 	skb->_skb_refdst = (unsigned long)dst;
980 }
981 
982 /**
983  * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
984  * @skb: buffer
985  * @dst: dst entry
986  *
987  * Sets skb dst, assuming a reference was not taken on dst.
988  * If dst entry is cached, we do not take reference and dst_release
989  * will be avoided by refdst_drop. If dst entry is not cached, we take
990  * reference, so that last dst_release can destroy the dst immediately.
991  */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)992 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
993 {
994 	WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
995 	skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
996 }
997 
998 /**
999  * skb_dst_is_noref - Test if skb dst isn't refcounted
1000  * @skb: buffer
1001  */
skb_dst_is_noref(const struct sk_buff * skb)1002 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1003 {
1004 	return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1005 }
1006 
1007 /**
1008  * skb_rtable - Returns the skb &rtable
1009  * @skb: buffer
1010  */
skb_rtable(const struct sk_buff * skb)1011 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1012 {
1013 	return (struct rtable *)skb_dst(skb);
1014 }
1015 
1016 /* For mangling skb->pkt_type from user space side from applications
1017  * such as nft, tc, etc, we only allow a conservative subset of
1018  * possible pkt_types to be set.
1019 */
skb_pkt_type_ok(u32 ptype)1020 static inline bool skb_pkt_type_ok(u32 ptype)
1021 {
1022 	return ptype <= PACKET_OTHERHOST;
1023 }
1024 
1025 /**
1026  * skb_napi_id - Returns the skb's NAPI id
1027  * @skb: buffer
1028  */
skb_napi_id(const struct sk_buff * skb)1029 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1030 {
1031 #ifdef CONFIG_NET_RX_BUSY_POLL
1032 	return skb->napi_id;
1033 #else
1034 	return 0;
1035 #endif
1036 }
1037 
1038 /**
1039  * skb_unref - decrement the skb's reference count
1040  * @skb: buffer
1041  *
1042  * Returns true if we can free the skb.
1043  */
skb_unref(struct sk_buff * skb)1044 static inline bool skb_unref(struct sk_buff *skb)
1045 {
1046 	if (unlikely(!skb))
1047 		return false;
1048 	if (likely(refcount_read(&skb->users) == 1))
1049 		smp_rmb();
1050 	else if (likely(!refcount_dec_and_test(&skb->users)))
1051 		return false;
1052 
1053 	return true;
1054 }
1055 
1056 void skb_release_head_state(struct sk_buff *skb);
1057 void kfree_skb(struct sk_buff *skb);
1058 void kfree_skb_list(struct sk_buff *segs);
1059 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1060 void skb_tx_error(struct sk_buff *skb);
1061 
1062 #ifdef CONFIG_TRACEPOINTS
1063 void consume_skb(struct sk_buff *skb);
1064 #else
consume_skb(struct sk_buff * skb)1065 static inline void consume_skb(struct sk_buff *skb)
1066 {
1067 	return kfree_skb(skb);
1068 }
1069 #endif
1070 
1071 void __consume_stateless_skb(struct sk_buff *skb);
1072 void  __kfree_skb(struct sk_buff *skb);
1073 extern struct kmem_cache *skbuff_head_cache;
1074 
1075 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1076 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1077 		      bool *fragstolen, int *delta_truesize);
1078 
1079 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1080 			    int node);
1081 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1082 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1083 struct sk_buff *build_skb_around(struct sk_buff *skb,
1084 				 void *data, unsigned int frag_size);
1085 
1086 /**
1087  * alloc_skb - allocate a network buffer
1088  * @size: size to allocate
1089  * @priority: allocation mask
1090  *
1091  * This function is a convenient wrapper around __alloc_skb().
1092  */
alloc_skb(unsigned int size,gfp_t priority)1093 static inline struct sk_buff *alloc_skb(unsigned int size,
1094 					gfp_t priority)
1095 {
1096 	return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1097 }
1098 
1099 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1100 				     unsigned long data_len,
1101 				     int max_page_order,
1102 				     int *errcode,
1103 				     gfp_t gfp_mask);
1104 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1105 
1106 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1107 struct sk_buff_fclones {
1108 	struct sk_buff	skb1;
1109 
1110 	struct sk_buff	skb2;
1111 
1112 	refcount_t	fclone_ref;
1113 };
1114 
1115 /**
1116  *	skb_fclone_busy - check if fclone is busy
1117  *	@sk: socket
1118  *	@skb: buffer
1119  *
1120  * Returns true if skb is a fast clone, and its clone is not freed.
1121  * Some drivers call skb_orphan() in their ndo_start_xmit(),
1122  * so we also check that this didnt happen.
1123  */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1124 static inline bool skb_fclone_busy(const struct sock *sk,
1125 				   const struct sk_buff *skb)
1126 {
1127 	const struct sk_buff_fclones *fclones;
1128 
1129 	fclones = container_of(skb, struct sk_buff_fclones, skb1);
1130 
1131 	return skb->fclone == SKB_FCLONE_ORIG &&
1132 	       refcount_read(&fclones->fclone_ref) > 1 &&
1133 	       fclones->skb2.sk == sk;
1134 }
1135 
1136 /**
1137  * alloc_skb_fclone - allocate a network buffer from fclone cache
1138  * @size: size to allocate
1139  * @priority: allocation mask
1140  *
1141  * This function is a convenient wrapper around __alloc_skb().
1142  */
alloc_skb_fclone(unsigned int size,gfp_t priority)1143 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1144 					       gfp_t priority)
1145 {
1146 	return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1147 }
1148 
1149 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1150 void skb_headers_offset_update(struct sk_buff *skb, int off);
1151 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1152 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1153 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1154 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1155 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1156 				   gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1157 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1158 					  gfp_t gfp_mask)
1159 {
1160 	return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1161 }
1162 
1163 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1164 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1165 				     unsigned int headroom);
1166 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1167 				int newtailroom, gfp_t priority);
1168 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1169 				     int offset, int len);
1170 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1171 			      int offset, int len);
1172 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1173 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1174 
1175 /**
1176  *	skb_pad			-	zero pad the tail of an skb
1177  *	@skb: buffer to pad
1178  *	@pad: space to pad
1179  *
1180  *	Ensure that a buffer is followed by a padding area that is zero
1181  *	filled. Used by network drivers which may DMA or transfer data
1182  *	beyond the buffer end onto the wire.
1183  *
1184  *	May return error in out of memory cases. The skb is freed on error.
1185  */
skb_pad(struct sk_buff * skb,int pad)1186 static inline int skb_pad(struct sk_buff *skb, int pad)
1187 {
1188 	return __skb_pad(skb, pad, true);
1189 }
1190 #define dev_kfree_skb(a)	consume_skb(a)
1191 
1192 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1193 			 int offset, size_t size);
1194 
1195 struct skb_seq_state {
1196 	__u32		lower_offset;
1197 	__u32		upper_offset;
1198 	__u32		frag_idx;
1199 	__u32		stepped_offset;
1200 	struct sk_buff	*root_skb;
1201 	struct sk_buff	*cur_skb;
1202 	__u8		*frag_data;
1203 };
1204 
1205 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1206 			  unsigned int to, struct skb_seq_state *st);
1207 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1208 			  struct skb_seq_state *st);
1209 void skb_abort_seq_read(struct skb_seq_state *st);
1210 
1211 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1212 			   unsigned int to, struct ts_config *config);
1213 
1214 /*
1215  * Packet hash types specify the type of hash in skb_set_hash.
1216  *
1217  * Hash types refer to the protocol layer addresses which are used to
1218  * construct a packet's hash. The hashes are used to differentiate or identify
1219  * flows of the protocol layer for the hash type. Hash types are either
1220  * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1221  *
1222  * Properties of hashes:
1223  *
1224  * 1) Two packets in different flows have different hash values
1225  * 2) Two packets in the same flow should have the same hash value
1226  *
1227  * A hash at a higher layer is considered to be more specific. A driver should
1228  * set the most specific hash possible.
1229  *
1230  * A driver cannot indicate a more specific hash than the layer at which a hash
1231  * was computed. For instance an L3 hash cannot be set as an L4 hash.
1232  *
1233  * A driver may indicate a hash level which is less specific than the
1234  * actual layer the hash was computed on. For instance, a hash computed
1235  * at L4 may be considered an L3 hash. This should only be done if the
1236  * driver can't unambiguously determine that the HW computed the hash at
1237  * the higher layer. Note that the "should" in the second property above
1238  * permits this.
1239  */
1240 enum pkt_hash_types {
1241 	PKT_HASH_TYPE_NONE,	/* Undefined type */
1242 	PKT_HASH_TYPE_L2,	/* Input: src_MAC, dest_MAC */
1243 	PKT_HASH_TYPE_L3,	/* Input: src_IP, dst_IP */
1244 	PKT_HASH_TYPE_L4,	/* Input: src_IP, dst_IP, src_port, dst_port */
1245 };
1246 
skb_clear_hash(struct sk_buff * skb)1247 static inline void skb_clear_hash(struct sk_buff *skb)
1248 {
1249 	skb->hash = 0;
1250 	skb->sw_hash = 0;
1251 	skb->l4_hash = 0;
1252 }
1253 
skb_clear_hash_if_not_l4(struct sk_buff * skb)1254 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1255 {
1256 	if (!skb->l4_hash)
1257 		skb_clear_hash(skb);
1258 }
1259 
1260 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1261 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1262 {
1263 	skb->l4_hash = is_l4;
1264 	skb->sw_hash = is_sw;
1265 	skb->hash = hash;
1266 }
1267 
1268 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1269 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1270 {
1271 	/* Used by drivers to set hash from HW */
1272 	__skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1273 }
1274 
1275 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1276 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1277 {
1278 	__skb_set_hash(skb, hash, true, is_l4);
1279 }
1280 
1281 void __skb_get_hash(struct sk_buff *skb);
1282 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1283 u32 skb_get_poff(const struct sk_buff *skb);
1284 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1285 		   const struct flow_keys_basic *keys, int hlen);
1286 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1287 			    void *data, int hlen_proto);
1288 
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto)1289 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1290 					int thoff, u8 ip_proto)
1291 {
1292 	return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1293 }
1294 
1295 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1296 			     const struct flow_dissector_key *key,
1297 			     unsigned int key_count);
1298 
1299 struct bpf_flow_dissector;
1300 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1301 		      __be16 proto, int nhoff, int hlen, unsigned int flags);
1302 
1303 bool __skb_flow_dissect(const struct net *net,
1304 			const struct sk_buff *skb,
1305 			struct flow_dissector *flow_dissector,
1306 			void *target_container,
1307 			void *data, __be16 proto, int nhoff, int hlen,
1308 			unsigned int flags);
1309 
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1310 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1311 				    struct flow_dissector *flow_dissector,
1312 				    void *target_container, unsigned int flags)
1313 {
1314 	return __skb_flow_dissect(NULL, skb, flow_dissector,
1315 				  target_container, NULL, 0, 0, 0, flags);
1316 }
1317 
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1318 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1319 					      struct flow_keys *flow,
1320 					      unsigned int flags)
1321 {
1322 	memset(flow, 0, sizeof(*flow));
1323 	return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1324 				  flow, NULL, 0, 0, 0, flags);
1325 }
1326 
1327 static inline bool
skb_flow_dissect_flow_keys_basic(const struct net * net,const struct sk_buff * skb,struct flow_keys_basic * flow,void * data,__be16 proto,int nhoff,int hlen,unsigned int flags)1328 skb_flow_dissect_flow_keys_basic(const struct net *net,
1329 				 const struct sk_buff *skb,
1330 				 struct flow_keys_basic *flow, void *data,
1331 				 __be16 proto, int nhoff, int hlen,
1332 				 unsigned int flags)
1333 {
1334 	memset(flow, 0, sizeof(*flow));
1335 	return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1336 				  data, proto, nhoff, hlen, flags);
1337 }
1338 
1339 void skb_flow_dissect_meta(const struct sk_buff *skb,
1340 			   struct flow_dissector *flow_dissector,
1341 			   void *target_container);
1342 
1343 /* Gets a skb connection tracking info, ctinfo map should be a
1344  * map of mapsize to translate enum ip_conntrack_info states
1345  * to user states.
1346  */
1347 void
1348 skb_flow_dissect_ct(const struct sk_buff *skb,
1349 		    struct flow_dissector *flow_dissector,
1350 		    void *target_container,
1351 		    u16 *ctinfo_map,
1352 		    size_t mapsize);
1353 void
1354 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1355 			     struct flow_dissector *flow_dissector,
1356 			     void *target_container);
1357 
1358 void skb_flow_dissect_hash(const struct sk_buff *skb,
1359 			   struct flow_dissector *flow_dissector,
1360 			   void *target_container);
1361 
skb_get_hash(struct sk_buff * skb)1362 static inline __u32 skb_get_hash(struct sk_buff *skb)
1363 {
1364 	if (!skb->l4_hash && !skb->sw_hash)
1365 		__skb_get_hash(skb);
1366 
1367 	return skb->hash;
1368 }
1369 
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1370 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1371 {
1372 	if (!skb->l4_hash && !skb->sw_hash) {
1373 		struct flow_keys keys;
1374 		__u32 hash = __get_hash_from_flowi6(fl6, &keys);
1375 
1376 		__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1377 	}
1378 
1379 	return skb->hash;
1380 }
1381 
1382 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1383 			   const siphash_key_t *perturb);
1384 
skb_get_hash_raw(const struct sk_buff * skb)1385 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1386 {
1387 	return skb->hash;
1388 }
1389 
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1390 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1391 {
1392 	to->hash = from->hash;
1393 	to->sw_hash = from->sw_hash;
1394 	to->l4_hash = from->l4_hash;
1395 };
1396 
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1397 static inline void skb_copy_decrypted(struct sk_buff *to,
1398 				      const struct sk_buff *from)
1399 {
1400 #ifdef CONFIG_TLS_DEVICE
1401 	to->decrypted = from->decrypted;
1402 #endif
1403 }
1404 
1405 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1406 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1407 {
1408 	return skb->head + skb->end;
1409 }
1410 
skb_end_offset(const struct sk_buff * skb)1411 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1412 {
1413 	return skb->end;
1414 }
1415 #else
skb_end_pointer(const struct sk_buff * skb)1416 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1417 {
1418 	return skb->end;
1419 }
1420 
skb_end_offset(const struct sk_buff * skb)1421 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1422 {
1423 	return skb->end - skb->head;
1424 }
1425 #endif
1426 
1427 /* Internal */
1428 #define skb_shinfo(SKB)	((struct skb_shared_info *)(skb_end_pointer(SKB)))
1429 
skb_hwtstamps(struct sk_buff * skb)1430 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1431 {
1432 	return &skb_shinfo(skb)->hwtstamps;
1433 }
1434 
skb_zcopy(struct sk_buff * skb)1435 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1436 {
1437 	bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1438 
1439 	return is_zcopy ? skb_uarg(skb) : NULL;
1440 }
1441 
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1442 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1443 				 bool *have_ref)
1444 {
1445 	if (skb && uarg && !skb_zcopy(skb)) {
1446 		if (unlikely(have_ref && *have_ref))
1447 			*have_ref = false;
1448 		else
1449 			sock_zerocopy_get(uarg);
1450 		skb_shinfo(skb)->destructor_arg = uarg;
1451 		skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1452 	}
1453 }
1454 
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1455 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1456 {
1457 	skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1458 	skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1459 }
1460 
skb_zcopy_is_nouarg(struct sk_buff * skb)1461 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1462 {
1463 	return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1464 }
1465 
skb_zcopy_get_nouarg(struct sk_buff * skb)1466 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1467 {
1468 	return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1469 }
1470 
1471 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy)1472 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1473 {
1474 	struct ubuf_info *uarg = skb_zcopy(skb);
1475 
1476 	if (uarg) {
1477 		if (skb_zcopy_is_nouarg(skb)) {
1478 			/* no notification callback */
1479 		} else if (uarg->callback == sock_zerocopy_callback) {
1480 			uarg->zerocopy = uarg->zerocopy && zerocopy;
1481 			sock_zerocopy_put(uarg);
1482 		} else {
1483 			uarg->callback(uarg, zerocopy);
1484 		}
1485 
1486 		skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1487 	}
1488 }
1489 
1490 /* Abort a zerocopy operation and revert zckey on error in send syscall */
skb_zcopy_abort(struct sk_buff * skb)1491 static inline void skb_zcopy_abort(struct sk_buff *skb)
1492 {
1493 	struct ubuf_info *uarg = skb_zcopy(skb);
1494 
1495 	if (uarg) {
1496 		sock_zerocopy_put_abort(uarg, false);
1497 		skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1498 	}
1499 }
1500 
skb_mark_not_on_list(struct sk_buff * skb)1501 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1502 {
1503 	skb->next = NULL;
1504 }
1505 
1506 /* Iterate through singly-linked GSO fragments of an skb. */
1507 #define skb_list_walk_safe(first, skb, next_skb)                               \
1508 	for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb);  \
1509 	     (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1510 
skb_list_del_init(struct sk_buff * skb)1511 static inline void skb_list_del_init(struct sk_buff *skb)
1512 {
1513 	__list_del_entry(&skb->list);
1514 	skb_mark_not_on_list(skb);
1515 }
1516 
1517 /**
1518  *	skb_queue_empty - check if a queue is empty
1519  *	@list: queue head
1520  *
1521  *	Returns true if the queue is empty, false otherwise.
1522  */
skb_queue_empty(const struct sk_buff_head * list)1523 static inline int skb_queue_empty(const struct sk_buff_head *list)
1524 {
1525 	return list->next == (const struct sk_buff *) list;
1526 }
1527 
1528 /**
1529  *	skb_queue_empty_lockless - check if a queue is empty
1530  *	@list: queue head
1531  *
1532  *	Returns true if the queue is empty, false otherwise.
1533  *	This variant can be used in lockless contexts.
1534  */
skb_queue_empty_lockless(const struct sk_buff_head * list)1535 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1536 {
1537 	return READ_ONCE(list->next) == (const struct sk_buff *) list;
1538 }
1539 
1540 
1541 /**
1542  *	skb_queue_is_last - check if skb is the last entry in the queue
1543  *	@list: queue head
1544  *	@skb: buffer
1545  *
1546  *	Returns true if @skb is the last buffer on the list.
1547  */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1548 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1549 				     const struct sk_buff *skb)
1550 {
1551 	return skb->next == (const struct sk_buff *) list;
1552 }
1553 
1554 /**
1555  *	skb_queue_is_first - check if skb is the first entry in the queue
1556  *	@list: queue head
1557  *	@skb: buffer
1558  *
1559  *	Returns true if @skb is the first buffer on the list.
1560  */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1561 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1562 				      const struct sk_buff *skb)
1563 {
1564 	return skb->prev == (const struct sk_buff *) list;
1565 }
1566 
1567 /**
1568  *	skb_queue_next - return the next packet in the queue
1569  *	@list: queue head
1570  *	@skb: current buffer
1571  *
1572  *	Return the next packet in @list after @skb.  It is only valid to
1573  *	call this if skb_queue_is_last() evaluates to false.
1574  */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1575 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1576 					     const struct sk_buff *skb)
1577 {
1578 	/* This BUG_ON may seem severe, but if we just return then we
1579 	 * are going to dereference garbage.
1580 	 */
1581 	BUG_ON(skb_queue_is_last(list, skb));
1582 	return skb->next;
1583 }
1584 
1585 /**
1586  *	skb_queue_prev - return the prev packet in the queue
1587  *	@list: queue head
1588  *	@skb: current buffer
1589  *
1590  *	Return the prev packet in @list before @skb.  It is only valid to
1591  *	call this if skb_queue_is_first() evaluates to false.
1592  */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1593 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1594 					     const struct sk_buff *skb)
1595 {
1596 	/* This BUG_ON may seem severe, but if we just return then we
1597 	 * are going to dereference garbage.
1598 	 */
1599 	BUG_ON(skb_queue_is_first(list, skb));
1600 	return skb->prev;
1601 }
1602 
1603 /**
1604  *	skb_get - reference buffer
1605  *	@skb: buffer to reference
1606  *
1607  *	Makes another reference to a socket buffer and returns a pointer
1608  *	to the buffer.
1609  */
skb_get(struct sk_buff * skb)1610 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1611 {
1612 	refcount_inc(&skb->users);
1613 	return skb;
1614 }
1615 
1616 /*
1617  * If users == 1, we are the only owner and can avoid redundant atomic changes.
1618  */
1619 
1620 /**
1621  *	skb_cloned - is the buffer a clone
1622  *	@skb: buffer to check
1623  *
1624  *	Returns true if the buffer was generated with skb_clone() and is
1625  *	one of multiple shared copies of the buffer. Cloned buffers are
1626  *	shared data so must not be written to under normal circumstances.
1627  */
skb_cloned(const struct sk_buff * skb)1628 static inline int skb_cloned(const struct sk_buff *skb)
1629 {
1630 	return skb->cloned &&
1631 	       (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1632 }
1633 
skb_unclone(struct sk_buff * skb,gfp_t pri)1634 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1635 {
1636 	might_sleep_if(gfpflags_allow_blocking(pri));
1637 
1638 	if (skb_cloned(skb))
1639 		return pskb_expand_head(skb, 0, 0, pri);
1640 
1641 	return 0;
1642 }
1643 
1644 /**
1645  *	skb_header_cloned - is the header a clone
1646  *	@skb: buffer to check
1647  *
1648  *	Returns true if modifying the header part of the buffer requires
1649  *	the data to be copied.
1650  */
skb_header_cloned(const struct sk_buff * skb)1651 static inline int skb_header_cloned(const struct sk_buff *skb)
1652 {
1653 	int dataref;
1654 
1655 	if (!skb->cloned)
1656 		return 0;
1657 
1658 	dataref = atomic_read(&skb_shinfo(skb)->dataref);
1659 	dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1660 	return dataref != 1;
1661 }
1662 
skb_header_unclone(struct sk_buff * skb,gfp_t pri)1663 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1664 {
1665 	might_sleep_if(gfpflags_allow_blocking(pri));
1666 
1667 	if (skb_header_cloned(skb))
1668 		return pskb_expand_head(skb, 0, 0, pri);
1669 
1670 	return 0;
1671 }
1672 
1673 /**
1674  *	__skb_header_release - release reference to header
1675  *	@skb: buffer to operate on
1676  */
__skb_header_release(struct sk_buff * skb)1677 static inline void __skb_header_release(struct sk_buff *skb)
1678 {
1679 	skb->nohdr = 1;
1680 	atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1681 }
1682 
1683 
1684 /**
1685  *	skb_shared - is the buffer shared
1686  *	@skb: buffer to check
1687  *
1688  *	Returns true if more than one person has a reference to this
1689  *	buffer.
1690  */
skb_shared(const struct sk_buff * skb)1691 static inline int skb_shared(const struct sk_buff *skb)
1692 {
1693 	return refcount_read(&skb->users) != 1;
1694 }
1695 
1696 /**
1697  *	skb_share_check - check if buffer is shared and if so clone it
1698  *	@skb: buffer to check
1699  *	@pri: priority for memory allocation
1700  *
1701  *	If the buffer is shared the buffer is cloned and the old copy
1702  *	drops a reference. A new clone with a single reference is returned.
1703  *	If the buffer is not shared the original buffer is returned. When
1704  *	being called from interrupt status or with spinlocks held pri must
1705  *	be GFP_ATOMIC.
1706  *
1707  *	NULL is returned on a memory allocation failure.
1708  */
skb_share_check(struct sk_buff * skb,gfp_t pri)1709 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1710 {
1711 	might_sleep_if(gfpflags_allow_blocking(pri));
1712 	if (skb_shared(skb)) {
1713 		struct sk_buff *nskb = skb_clone(skb, pri);
1714 
1715 		if (likely(nskb))
1716 			consume_skb(skb);
1717 		else
1718 			kfree_skb(skb);
1719 		skb = nskb;
1720 	}
1721 	return skb;
1722 }
1723 
1724 /*
1725  *	Copy shared buffers into a new sk_buff. We effectively do COW on
1726  *	packets to handle cases where we have a local reader and forward
1727  *	and a couple of other messy ones. The normal one is tcpdumping
1728  *	a packet thats being forwarded.
1729  */
1730 
1731 /**
1732  *	skb_unshare - make a copy of a shared buffer
1733  *	@skb: buffer to check
1734  *	@pri: priority for memory allocation
1735  *
1736  *	If the socket buffer is a clone then this function creates a new
1737  *	copy of the data, drops a reference count on the old copy and returns
1738  *	the new copy with the reference count at 1. If the buffer is not a clone
1739  *	the original buffer is returned. When called with a spinlock held or
1740  *	from interrupt state @pri must be %GFP_ATOMIC
1741  *
1742  *	%NULL is returned on a memory allocation failure.
1743  */
skb_unshare(struct sk_buff * skb,gfp_t pri)1744 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1745 					  gfp_t pri)
1746 {
1747 	might_sleep_if(gfpflags_allow_blocking(pri));
1748 	if (skb_cloned(skb)) {
1749 		struct sk_buff *nskb = skb_copy(skb, pri);
1750 
1751 		/* Free our shared copy */
1752 		if (likely(nskb))
1753 			consume_skb(skb);
1754 		else
1755 			kfree_skb(skb);
1756 		skb = nskb;
1757 	}
1758 	return skb;
1759 }
1760 
1761 /**
1762  *	skb_peek - peek at the head of an &sk_buff_head
1763  *	@list_: list to peek at
1764  *
1765  *	Peek an &sk_buff. Unlike most other operations you _MUST_
1766  *	be careful with this one. A peek leaves the buffer on the
1767  *	list and someone else may run off with it. You must hold
1768  *	the appropriate locks or have a private queue to do this.
1769  *
1770  *	Returns %NULL for an empty list or a pointer to the head element.
1771  *	The reference count is not incremented and the reference is therefore
1772  *	volatile. Use with caution.
1773  */
skb_peek(const struct sk_buff_head * list_)1774 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1775 {
1776 	struct sk_buff *skb = list_->next;
1777 
1778 	if (skb == (struct sk_buff *)list_)
1779 		skb = NULL;
1780 	return skb;
1781 }
1782 
1783 /**
1784  *	__skb_peek - peek at the head of a non-empty &sk_buff_head
1785  *	@list_: list to peek at
1786  *
1787  *	Like skb_peek(), but the caller knows that the list is not empty.
1788  */
__skb_peek(const struct sk_buff_head * list_)1789 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1790 {
1791 	return list_->next;
1792 }
1793 
1794 /**
1795  *	skb_peek_next - peek skb following the given one from a queue
1796  *	@skb: skb to start from
1797  *	@list_: list to peek at
1798  *
1799  *	Returns %NULL when the end of the list is met or a pointer to the
1800  *	next element. The reference count is not incremented and the
1801  *	reference is therefore volatile. Use with caution.
1802  */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)1803 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1804 		const struct sk_buff_head *list_)
1805 {
1806 	struct sk_buff *next = skb->next;
1807 
1808 	if (next == (struct sk_buff *)list_)
1809 		next = NULL;
1810 	return next;
1811 }
1812 
1813 /**
1814  *	skb_peek_tail - peek at the tail of an &sk_buff_head
1815  *	@list_: list to peek at
1816  *
1817  *	Peek an &sk_buff. Unlike most other operations you _MUST_
1818  *	be careful with this one. A peek leaves the buffer on the
1819  *	list and someone else may run off with it. You must hold
1820  *	the appropriate locks or have a private queue to do this.
1821  *
1822  *	Returns %NULL for an empty list or a pointer to the tail element.
1823  *	The reference count is not incremented and the reference is therefore
1824  *	volatile. Use with caution.
1825  */
skb_peek_tail(const struct sk_buff_head * list_)1826 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1827 {
1828 	struct sk_buff *skb = READ_ONCE(list_->prev);
1829 
1830 	if (skb == (struct sk_buff *)list_)
1831 		skb = NULL;
1832 	return skb;
1833 
1834 }
1835 
1836 /**
1837  *	skb_queue_len	- get queue length
1838  *	@list_: list to measure
1839  *
1840  *	Return the length of an &sk_buff queue.
1841  */
skb_queue_len(const struct sk_buff_head * list_)1842 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1843 {
1844 	return list_->qlen;
1845 }
1846 
1847 /**
1848  *	skb_queue_len_lockless	- get queue length
1849  *	@list_: list to measure
1850  *
1851  *	Return the length of an &sk_buff queue.
1852  *	This variant can be used in lockless contexts.
1853  */
skb_queue_len_lockless(const struct sk_buff_head * list_)1854 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1855 {
1856 	return READ_ONCE(list_->qlen);
1857 }
1858 
1859 /**
1860  *	__skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1861  *	@list: queue to initialize
1862  *
1863  *	This initializes only the list and queue length aspects of
1864  *	an sk_buff_head object.  This allows to initialize the list
1865  *	aspects of an sk_buff_head without reinitializing things like
1866  *	the spinlock.  It can also be used for on-stack sk_buff_head
1867  *	objects where the spinlock is known to not be used.
1868  */
__skb_queue_head_init(struct sk_buff_head * list)1869 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1870 {
1871 	list->prev = list->next = (struct sk_buff *)list;
1872 	list->qlen = 0;
1873 }
1874 
1875 /*
1876  * This function creates a split out lock class for each invocation;
1877  * this is needed for now since a whole lot of users of the skb-queue
1878  * infrastructure in drivers have different locking usage (in hardirq)
1879  * than the networking core (in softirq only). In the long run either the
1880  * network layer or drivers should need annotation to consolidate the
1881  * main types of usage into 3 classes.
1882  */
skb_queue_head_init(struct sk_buff_head * list)1883 static inline void skb_queue_head_init(struct sk_buff_head *list)
1884 {
1885 	spin_lock_init(&list->lock);
1886 	__skb_queue_head_init(list);
1887 }
1888 
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)1889 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1890 		struct lock_class_key *class)
1891 {
1892 	skb_queue_head_init(list);
1893 	lockdep_set_class(&list->lock, class);
1894 }
1895 
1896 /*
1897  *	Insert an sk_buff on a list.
1898  *
1899  *	The "__skb_xxxx()" functions are the non-atomic ones that
1900  *	can only be called with interrupts disabled.
1901  */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)1902 static inline void __skb_insert(struct sk_buff *newsk,
1903 				struct sk_buff *prev, struct sk_buff *next,
1904 				struct sk_buff_head *list)
1905 {
1906 	/* See skb_queue_empty_lockless() and skb_peek_tail()
1907 	 * for the opposite READ_ONCE()
1908 	 */
1909 	WRITE_ONCE(newsk->next, next);
1910 	WRITE_ONCE(newsk->prev, prev);
1911 	WRITE_ONCE(next->prev, newsk);
1912 	WRITE_ONCE(prev->next, newsk);
1913 	WRITE_ONCE(list->qlen, list->qlen + 1);
1914 }
1915 
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)1916 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1917 				      struct sk_buff *prev,
1918 				      struct sk_buff *next)
1919 {
1920 	struct sk_buff *first = list->next;
1921 	struct sk_buff *last = list->prev;
1922 
1923 	WRITE_ONCE(first->prev, prev);
1924 	WRITE_ONCE(prev->next, first);
1925 
1926 	WRITE_ONCE(last->next, next);
1927 	WRITE_ONCE(next->prev, last);
1928 }
1929 
1930 /**
1931  *	skb_queue_splice - join two skb lists, this is designed for stacks
1932  *	@list: the new list to add
1933  *	@head: the place to add it in the first list
1934  */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)1935 static inline void skb_queue_splice(const struct sk_buff_head *list,
1936 				    struct sk_buff_head *head)
1937 {
1938 	if (!skb_queue_empty(list)) {
1939 		__skb_queue_splice(list, (struct sk_buff *) head, head->next);
1940 		head->qlen += list->qlen;
1941 	}
1942 }
1943 
1944 /**
1945  *	skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1946  *	@list: the new list to add
1947  *	@head: the place to add it in the first list
1948  *
1949  *	The list at @list is reinitialised
1950  */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)1951 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1952 					 struct sk_buff_head *head)
1953 {
1954 	if (!skb_queue_empty(list)) {
1955 		__skb_queue_splice(list, (struct sk_buff *) head, head->next);
1956 		head->qlen += list->qlen;
1957 		__skb_queue_head_init(list);
1958 	}
1959 }
1960 
1961 /**
1962  *	skb_queue_splice_tail - join two skb lists, each list being a queue
1963  *	@list: the new list to add
1964  *	@head: the place to add it in the first list
1965  */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)1966 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1967 					 struct sk_buff_head *head)
1968 {
1969 	if (!skb_queue_empty(list)) {
1970 		__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1971 		head->qlen += list->qlen;
1972 	}
1973 }
1974 
1975 /**
1976  *	skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1977  *	@list: the new list to add
1978  *	@head: the place to add it in the first list
1979  *
1980  *	Each of the lists is a queue.
1981  *	The list at @list is reinitialised
1982  */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)1983 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1984 					      struct sk_buff_head *head)
1985 {
1986 	if (!skb_queue_empty(list)) {
1987 		__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1988 		head->qlen += list->qlen;
1989 		__skb_queue_head_init(list);
1990 	}
1991 }
1992 
1993 /**
1994  *	__skb_queue_after - queue a buffer at the list head
1995  *	@list: list to use
1996  *	@prev: place after this buffer
1997  *	@newsk: buffer to queue
1998  *
1999  *	Queue a buffer int the middle of a list. This function takes no locks
2000  *	and you must therefore hold required locks before calling it.
2001  *
2002  *	A buffer cannot be placed on two lists at the same time.
2003  */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)2004 static inline void __skb_queue_after(struct sk_buff_head *list,
2005 				     struct sk_buff *prev,
2006 				     struct sk_buff *newsk)
2007 {
2008 	__skb_insert(newsk, prev, prev->next, list);
2009 }
2010 
2011 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2012 		struct sk_buff_head *list);
2013 
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)2014 static inline void __skb_queue_before(struct sk_buff_head *list,
2015 				      struct sk_buff *next,
2016 				      struct sk_buff *newsk)
2017 {
2018 	__skb_insert(newsk, next->prev, next, list);
2019 }
2020 
2021 /**
2022  *	__skb_queue_head - queue a buffer at the list head
2023  *	@list: list to use
2024  *	@newsk: buffer to queue
2025  *
2026  *	Queue a buffer at the start of a list. This function takes no locks
2027  *	and you must therefore hold required locks before calling it.
2028  *
2029  *	A buffer cannot be placed on two lists at the same time.
2030  */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)2031 static inline void __skb_queue_head(struct sk_buff_head *list,
2032 				    struct sk_buff *newsk)
2033 {
2034 	__skb_queue_after(list, (struct sk_buff *)list, newsk);
2035 }
2036 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2037 
2038 /**
2039  *	__skb_queue_tail - queue a buffer at the list tail
2040  *	@list: list to use
2041  *	@newsk: buffer to queue
2042  *
2043  *	Queue a buffer at the end of a list. This function takes no locks
2044  *	and you must therefore hold required locks before calling it.
2045  *
2046  *	A buffer cannot be placed on two lists at the same time.
2047  */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2048 static inline void __skb_queue_tail(struct sk_buff_head *list,
2049 				   struct sk_buff *newsk)
2050 {
2051 	__skb_queue_before(list, (struct sk_buff *)list, newsk);
2052 }
2053 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2054 
2055 /*
2056  * remove sk_buff from list. _Must_ be called atomically, and with
2057  * the list known..
2058  */
2059 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2060 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2061 {
2062 	struct sk_buff *next, *prev;
2063 
2064 	WRITE_ONCE(list->qlen, list->qlen - 1);
2065 	next	   = skb->next;
2066 	prev	   = skb->prev;
2067 	skb->next  = skb->prev = NULL;
2068 	WRITE_ONCE(next->prev, prev);
2069 	WRITE_ONCE(prev->next, next);
2070 }
2071 
2072 /**
2073  *	__skb_dequeue - remove from the head of the queue
2074  *	@list: list to dequeue from
2075  *
2076  *	Remove the head of the list. This function does not take any locks
2077  *	so must be used with appropriate locks held only. The head item is
2078  *	returned or %NULL if the list is empty.
2079  */
__skb_dequeue(struct sk_buff_head * list)2080 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2081 {
2082 	struct sk_buff *skb = skb_peek(list);
2083 	if (skb)
2084 		__skb_unlink(skb, list);
2085 	return skb;
2086 }
2087 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2088 
2089 /**
2090  *	__skb_dequeue_tail - remove from the tail of the queue
2091  *	@list: list to dequeue from
2092  *
2093  *	Remove the tail of the list. This function does not take any locks
2094  *	so must be used with appropriate locks held only. The tail item is
2095  *	returned or %NULL if the list is empty.
2096  */
__skb_dequeue_tail(struct sk_buff_head * list)2097 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2098 {
2099 	struct sk_buff *skb = skb_peek_tail(list);
2100 	if (skb)
2101 		__skb_unlink(skb, list);
2102 	return skb;
2103 }
2104 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2105 
2106 
skb_is_nonlinear(const struct sk_buff * skb)2107 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2108 {
2109 	return skb->data_len;
2110 }
2111 
skb_headlen(const struct sk_buff * skb)2112 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2113 {
2114 	return skb->len - skb->data_len;
2115 }
2116 
__skb_pagelen(const struct sk_buff * skb)2117 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2118 {
2119 	unsigned int i, len = 0;
2120 
2121 	for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2122 		len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2123 	return len;
2124 }
2125 
skb_pagelen(const struct sk_buff * skb)2126 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2127 {
2128 	return skb_headlen(skb) + __skb_pagelen(skb);
2129 }
2130 
2131 /**
2132  * __skb_fill_page_desc - initialise a paged fragment in an skb
2133  * @skb: buffer containing fragment to be initialised
2134  * @i: paged fragment index to initialise
2135  * @page: the page to use for this fragment
2136  * @off: the offset to the data with @page
2137  * @size: the length of the data
2138  *
2139  * Initialises the @i'th fragment of @skb to point to &size bytes at
2140  * offset @off within @page.
2141  *
2142  * Does not take any additional reference on the fragment.
2143  */
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2144 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2145 					struct page *page, int off, int size)
2146 {
2147 	skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2148 
2149 	/*
2150 	 * Propagate page pfmemalloc to the skb if we can. The problem is
2151 	 * that not all callers have unique ownership of the page but rely
2152 	 * on page_is_pfmemalloc doing the right thing(tm).
2153 	 */
2154 	frag->bv_page		  = page;
2155 	frag->bv_offset		  = off;
2156 	skb_frag_size_set(frag, size);
2157 
2158 	page = compound_head(page);
2159 	if (page_is_pfmemalloc(page))
2160 		skb->pfmemalloc	= true;
2161 }
2162 
2163 /**
2164  * skb_fill_page_desc - initialise a paged fragment in an skb
2165  * @skb: buffer containing fragment to be initialised
2166  * @i: paged fragment index to initialise
2167  * @page: the page to use for this fragment
2168  * @off: the offset to the data with @page
2169  * @size: the length of the data
2170  *
2171  * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2172  * @skb to point to @size bytes at offset @off within @page. In
2173  * addition updates @skb such that @i is the last fragment.
2174  *
2175  * Does not take any additional reference on the fragment.
2176  */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2177 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2178 				      struct page *page, int off, int size)
2179 {
2180 	__skb_fill_page_desc(skb, i, page, off, size);
2181 	skb_shinfo(skb)->nr_frags = i + 1;
2182 }
2183 
2184 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2185 		     int size, unsigned int truesize);
2186 
2187 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2188 			  unsigned int truesize);
2189 
2190 #define SKB_LINEAR_ASSERT(skb)  BUG_ON(skb_is_nonlinear(skb))
2191 
2192 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2193 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2194 {
2195 	return skb->head + skb->tail;
2196 }
2197 
skb_reset_tail_pointer(struct sk_buff * skb)2198 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2199 {
2200 	skb->tail = skb->data - skb->head;
2201 }
2202 
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2203 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2204 {
2205 	skb_reset_tail_pointer(skb);
2206 	skb->tail += offset;
2207 }
2208 
2209 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2210 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2211 {
2212 	return skb->tail;
2213 }
2214 
skb_reset_tail_pointer(struct sk_buff * skb)2215 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2216 {
2217 	skb->tail = skb->data;
2218 }
2219 
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2220 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2221 {
2222 	skb->tail = skb->data + offset;
2223 }
2224 
2225 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2226 
2227 /*
2228  *	Add data to an sk_buff
2229  */
2230 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2231 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2232 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2233 {
2234 	void *tmp = skb_tail_pointer(skb);
2235 	SKB_LINEAR_ASSERT(skb);
2236 	skb->tail += len;
2237 	skb->len  += len;
2238 	return tmp;
2239 }
2240 
__skb_put_zero(struct sk_buff * skb,unsigned int len)2241 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2242 {
2243 	void *tmp = __skb_put(skb, len);
2244 
2245 	memset(tmp, 0, len);
2246 	return tmp;
2247 }
2248 
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2249 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2250 				   unsigned int len)
2251 {
2252 	void *tmp = __skb_put(skb, len);
2253 
2254 	memcpy(tmp, data, len);
2255 	return tmp;
2256 }
2257 
__skb_put_u8(struct sk_buff * skb,u8 val)2258 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2259 {
2260 	*(u8 *)__skb_put(skb, 1) = val;
2261 }
2262 
skb_put_zero(struct sk_buff * skb,unsigned int len)2263 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2264 {
2265 	void *tmp = skb_put(skb, len);
2266 
2267 	memset(tmp, 0, len);
2268 
2269 	return tmp;
2270 }
2271 
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2272 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2273 				 unsigned int len)
2274 {
2275 	void *tmp = skb_put(skb, len);
2276 
2277 	memcpy(tmp, data, len);
2278 
2279 	return tmp;
2280 }
2281 
skb_put_u8(struct sk_buff * skb,u8 val)2282 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2283 {
2284 	*(u8 *)skb_put(skb, 1) = val;
2285 }
2286 
2287 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2288 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2289 {
2290 	skb->data -= len;
2291 	skb->len  += len;
2292 	return skb->data;
2293 }
2294 
2295 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2296 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2297 {
2298 	skb->len -= len;
2299 	BUG_ON(skb->len < skb->data_len);
2300 	return skb->data += len;
2301 }
2302 
skb_pull_inline(struct sk_buff * skb,unsigned int len)2303 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2304 {
2305 	return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2306 }
2307 
2308 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2309 
__pskb_pull(struct sk_buff * skb,unsigned int len)2310 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2311 {
2312 	if (len > skb_headlen(skb) &&
2313 	    !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2314 		return NULL;
2315 	skb->len -= len;
2316 	return skb->data += len;
2317 }
2318 
pskb_pull(struct sk_buff * skb,unsigned int len)2319 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2320 {
2321 	return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2322 }
2323 
pskb_may_pull(struct sk_buff * skb,unsigned int len)2324 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2325 {
2326 	if (likely(len <= skb_headlen(skb)))
2327 		return true;
2328 	if (unlikely(len > skb->len))
2329 		return false;
2330 	return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2331 }
2332 
2333 void skb_condense(struct sk_buff *skb);
2334 
2335 /**
2336  *	skb_headroom - bytes at buffer head
2337  *	@skb: buffer to check
2338  *
2339  *	Return the number of bytes of free space at the head of an &sk_buff.
2340  */
skb_headroom(const struct sk_buff * skb)2341 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2342 {
2343 	return skb->data - skb->head;
2344 }
2345 
2346 /**
2347  *	skb_tailroom - bytes at buffer end
2348  *	@skb: buffer to check
2349  *
2350  *	Return the number of bytes of free space at the tail of an sk_buff
2351  */
skb_tailroom(const struct sk_buff * skb)2352 static inline int skb_tailroom(const struct sk_buff *skb)
2353 {
2354 	return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2355 }
2356 
2357 /**
2358  *	skb_availroom - bytes at buffer end
2359  *	@skb: buffer to check
2360  *
2361  *	Return the number of bytes of free space at the tail of an sk_buff
2362  *	allocated by sk_stream_alloc()
2363  */
skb_availroom(const struct sk_buff * skb)2364 static inline int skb_availroom(const struct sk_buff *skb)
2365 {
2366 	if (skb_is_nonlinear(skb))
2367 		return 0;
2368 
2369 	return skb->end - skb->tail - skb->reserved_tailroom;
2370 }
2371 
2372 /**
2373  *	skb_reserve - adjust headroom
2374  *	@skb: buffer to alter
2375  *	@len: bytes to move
2376  *
2377  *	Increase the headroom of an empty &sk_buff by reducing the tail
2378  *	room. This is only allowed for an empty buffer.
2379  */
skb_reserve(struct sk_buff * skb,int len)2380 static inline void skb_reserve(struct sk_buff *skb, int len)
2381 {
2382 	skb->data += len;
2383 	skb->tail += len;
2384 }
2385 
2386 /**
2387  *	skb_tailroom_reserve - adjust reserved_tailroom
2388  *	@skb: buffer to alter
2389  *	@mtu: maximum amount of headlen permitted
2390  *	@needed_tailroom: minimum amount of reserved_tailroom
2391  *
2392  *	Set reserved_tailroom so that headlen can be as large as possible but
2393  *	not larger than mtu and tailroom cannot be smaller than
2394  *	needed_tailroom.
2395  *	The required headroom should already have been reserved before using
2396  *	this function.
2397  */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2398 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2399 					unsigned int needed_tailroom)
2400 {
2401 	SKB_LINEAR_ASSERT(skb);
2402 	if (mtu < skb_tailroom(skb) - needed_tailroom)
2403 		/* use at most mtu */
2404 		skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2405 	else
2406 		/* use up to all available space */
2407 		skb->reserved_tailroom = needed_tailroom;
2408 }
2409 
2410 #define ENCAP_TYPE_ETHER	0
2411 #define ENCAP_TYPE_IPPROTO	1
2412 
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2413 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2414 					  __be16 protocol)
2415 {
2416 	skb->inner_protocol = protocol;
2417 	skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2418 }
2419 
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2420 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2421 					 __u8 ipproto)
2422 {
2423 	skb->inner_ipproto = ipproto;
2424 	skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2425 }
2426 
skb_reset_inner_headers(struct sk_buff * skb)2427 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2428 {
2429 	skb->inner_mac_header = skb->mac_header;
2430 	skb->inner_network_header = skb->network_header;
2431 	skb->inner_transport_header = skb->transport_header;
2432 }
2433 
skb_reset_mac_len(struct sk_buff * skb)2434 static inline void skb_reset_mac_len(struct sk_buff *skb)
2435 {
2436 	skb->mac_len = skb->network_header - skb->mac_header;
2437 }
2438 
skb_inner_transport_header(const struct sk_buff * skb)2439 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2440 							*skb)
2441 {
2442 	return skb->head + skb->inner_transport_header;
2443 }
2444 
skb_inner_transport_offset(const struct sk_buff * skb)2445 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2446 {
2447 	return skb_inner_transport_header(skb) - skb->data;
2448 }
2449 
skb_reset_inner_transport_header(struct sk_buff * skb)2450 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2451 {
2452 	skb->inner_transport_header = skb->data - skb->head;
2453 }
2454 
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2455 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2456 						   const int offset)
2457 {
2458 	skb_reset_inner_transport_header(skb);
2459 	skb->inner_transport_header += offset;
2460 }
2461 
skb_inner_network_header(const struct sk_buff * skb)2462 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2463 {
2464 	return skb->head + skb->inner_network_header;
2465 }
2466 
skb_reset_inner_network_header(struct sk_buff * skb)2467 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2468 {
2469 	skb->inner_network_header = skb->data - skb->head;
2470 }
2471 
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2472 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2473 						const int offset)
2474 {
2475 	skb_reset_inner_network_header(skb);
2476 	skb->inner_network_header += offset;
2477 }
2478 
skb_inner_mac_header(const struct sk_buff * skb)2479 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2480 {
2481 	return skb->head + skb->inner_mac_header;
2482 }
2483 
skb_reset_inner_mac_header(struct sk_buff * skb)2484 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2485 {
2486 	skb->inner_mac_header = skb->data - skb->head;
2487 }
2488 
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)2489 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2490 					    const int offset)
2491 {
2492 	skb_reset_inner_mac_header(skb);
2493 	skb->inner_mac_header += offset;
2494 }
skb_transport_header_was_set(const struct sk_buff * skb)2495 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2496 {
2497 	return skb->transport_header != (typeof(skb->transport_header))~0U;
2498 }
2499 
skb_transport_header(const struct sk_buff * skb)2500 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2501 {
2502 	return skb->head + skb->transport_header;
2503 }
2504 
skb_reset_transport_header(struct sk_buff * skb)2505 static inline void skb_reset_transport_header(struct sk_buff *skb)
2506 {
2507 	skb->transport_header = skb->data - skb->head;
2508 }
2509 
skb_set_transport_header(struct sk_buff * skb,const int offset)2510 static inline void skb_set_transport_header(struct sk_buff *skb,
2511 					    const int offset)
2512 {
2513 	skb_reset_transport_header(skb);
2514 	skb->transport_header += offset;
2515 }
2516 
skb_network_header(const struct sk_buff * skb)2517 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2518 {
2519 	return skb->head + skb->network_header;
2520 }
2521 
skb_reset_network_header(struct sk_buff * skb)2522 static inline void skb_reset_network_header(struct sk_buff *skb)
2523 {
2524 	skb->network_header = skb->data - skb->head;
2525 }
2526 
skb_set_network_header(struct sk_buff * skb,const int offset)2527 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2528 {
2529 	skb_reset_network_header(skb);
2530 	skb->network_header += offset;
2531 }
2532 
skb_mac_header(const struct sk_buff * skb)2533 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2534 {
2535 	return skb->head + skb->mac_header;
2536 }
2537 
skb_mac_offset(const struct sk_buff * skb)2538 static inline int skb_mac_offset(const struct sk_buff *skb)
2539 {
2540 	return skb_mac_header(skb) - skb->data;
2541 }
2542 
skb_mac_header_len(const struct sk_buff * skb)2543 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2544 {
2545 	return skb->network_header - skb->mac_header;
2546 }
2547 
skb_mac_header_was_set(const struct sk_buff * skb)2548 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2549 {
2550 	return skb->mac_header != (typeof(skb->mac_header))~0U;
2551 }
2552 
skb_unset_mac_header(struct sk_buff * skb)2553 static inline void skb_unset_mac_header(struct sk_buff *skb)
2554 {
2555 	skb->mac_header = (typeof(skb->mac_header))~0U;
2556 }
2557 
skb_reset_mac_header(struct sk_buff * skb)2558 static inline void skb_reset_mac_header(struct sk_buff *skb)
2559 {
2560 	skb->mac_header = skb->data - skb->head;
2561 }
2562 
skb_set_mac_header(struct sk_buff * skb,const int offset)2563 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2564 {
2565 	skb_reset_mac_header(skb);
2566 	skb->mac_header += offset;
2567 }
2568 
skb_pop_mac_header(struct sk_buff * skb)2569 static inline void skb_pop_mac_header(struct sk_buff *skb)
2570 {
2571 	skb->mac_header = skb->network_header;
2572 }
2573 
skb_probe_transport_header(struct sk_buff * skb)2574 static inline void skb_probe_transport_header(struct sk_buff *skb)
2575 {
2576 	struct flow_keys_basic keys;
2577 
2578 	if (skb_transport_header_was_set(skb))
2579 		return;
2580 
2581 	if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2582 					     NULL, 0, 0, 0, 0))
2583 		skb_set_transport_header(skb, keys.control.thoff);
2584 }
2585 
skb_mac_header_rebuild(struct sk_buff * skb)2586 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2587 {
2588 	if (skb_mac_header_was_set(skb)) {
2589 		const unsigned char *old_mac = skb_mac_header(skb);
2590 
2591 		skb_set_mac_header(skb, -skb->mac_len);
2592 		memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2593 	}
2594 }
2595 
skb_checksum_start_offset(const struct sk_buff * skb)2596 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2597 {
2598 	return skb->csum_start - skb_headroom(skb);
2599 }
2600 
skb_checksum_start(const struct sk_buff * skb)2601 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2602 {
2603 	return skb->head + skb->csum_start;
2604 }
2605 
skb_transport_offset(const struct sk_buff * skb)2606 static inline int skb_transport_offset(const struct sk_buff *skb)
2607 {
2608 	return skb_transport_header(skb) - skb->data;
2609 }
2610 
skb_network_header_len(const struct sk_buff * skb)2611 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2612 {
2613 	return skb->transport_header - skb->network_header;
2614 }
2615 
skb_inner_network_header_len(const struct sk_buff * skb)2616 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2617 {
2618 	return skb->inner_transport_header - skb->inner_network_header;
2619 }
2620 
skb_network_offset(const struct sk_buff * skb)2621 static inline int skb_network_offset(const struct sk_buff *skb)
2622 {
2623 	return skb_network_header(skb) - skb->data;
2624 }
2625 
skb_inner_network_offset(const struct sk_buff * skb)2626 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2627 {
2628 	return skb_inner_network_header(skb) - skb->data;
2629 }
2630 
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)2631 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2632 {
2633 	return pskb_may_pull(skb, skb_network_offset(skb) + len);
2634 }
2635 
2636 /*
2637  * CPUs often take a performance hit when accessing unaligned memory
2638  * locations. The actual performance hit varies, it can be small if the
2639  * hardware handles it or large if we have to take an exception and fix it
2640  * in software.
2641  *
2642  * Since an ethernet header is 14 bytes network drivers often end up with
2643  * the IP header at an unaligned offset. The IP header can be aligned by
2644  * shifting the start of the packet by 2 bytes. Drivers should do this
2645  * with:
2646  *
2647  * skb_reserve(skb, NET_IP_ALIGN);
2648  *
2649  * The downside to this alignment of the IP header is that the DMA is now
2650  * unaligned. On some architectures the cost of an unaligned DMA is high
2651  * and this cost outweighs the gains made by aligning the IP header.
2652  *
2653  * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2654  * to be overridden.
2655  */
2656 #ifndef NET_IP_ALIGN
2657 #define NET_IP_ALIGN	2
2658 #endif
2659 
2660 /*
2661  * The networking layer reserves some headroom in skb data (via
2662  * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2663  * the header has to grow. In the default case, if the header has to grow
2664  * 32 bytes or less we avoid the reallocation.
2665  *
2666  * Unfortunately this headroom changes the DMA alignment of the resulting
2667  * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2668  * on some architectures. An architecture can override this value,
2669  * perhaps setting it to a cacheline in size (since that will maintain
2670  * cacheline alignment of the DMA). It must be a power of 2.
2671  *
2672  * Various parts of the networking layer expect at least 32 bytes of
2673  * headroom, you should not reduce this.
2674  *
2675  * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2676  * to reduce average number of cache lines per packet.
2677  * get_rps_cpu() for example only access one 64 bytes aligned block :
2678  * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2679  */
2680 #ifndef NET_SKB_PAD
2681 #define NET_SKB_PAD	max(32, L1_CACHE_BYTES)
2682 #endif
2683 
2684 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2685 
__skb_set_length(struct sk_buff * skb,unsigned int len)2686 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2687 {
2688 	if (WARN_ON(skb_is_nonlinear(skb)))
2689 		return;
2690 	skb->len = len;
2691 	skb_set_tail_pointer(skb, len);
2692 }
2693 
__skb_trim(struct sk_buff * skb,unsigned int len)2694 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2695 {
2696 	__skb_set_length(skb, len);
2697 }
2698 
2699 void skb_trim(struct sk_buff *skb, unsigned int len);
2700 
__pskb_trim(struct sk_buff * skb,unsigned int len)2701 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2702 {
2703 	if (skb->data_len)
2704 		return ___pskb_trim(skb, len);
2705 	__skb_trim(skb, len);
2706 	return 0;
2707 }
2708 
pskb_trim(struct sk_buff * skb,unsigned int len)2709 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2710 {
2711 	return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2712 }
2713 
2714 /**
2715  *	pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2716  *	@skb: buffer to alter
2717  *	@len: new length
2718  *
2719  *	This is identical to pskb_trim except that the caller knows that
2720  *	the skb is not cloned so we should never get an error due to out-
2721  *	of-memory.
2722  */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)2723 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2724 {
2725 	int err = pskb_trim(skb, len);
2726 	BUG_ON(err);
2727 }
2728 
__skb_grow(struct sk_buff * skb,unsigned int len)2729 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2730 {
2731 	unsigned int diff = len - skb->len;
2732 
2733 	if (skb_tailroom(skb) < diff) {
2734 		int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2735 					   GFP_ATOMIC);
2736 		if (ret)
2737 			return ret;
2738 	}
2739 	__skb_set_length(skb, len);
2740 	return 0;
2741 }
2742 
2743 /**
2744  *	skb_orphan - orphan a buffer
2745  *	@skb: buffer to orphan
2746  *
2747  *	If a buffer currently has an owner then we call the owner's
2748  *	destructor function and make the @skb unowned. The buffer continues
2749  *	to exist but is no longer charged to its former owner.
2750  */
skb_orphan(struct sk_buff * skb)2751 static inline void skb_orphan(struct sk_buff *skb)
2752 {
2753 	if (skb->destructor) {
2754 		skb->destructor(skb);
2755 		skb->destructor = NULL;
2756 		skb->sk		= NULL;
2757 	} else {
2758 		BUG_ON(skb->sk);
2759 	}
2760 }
2761 
2762 /**
2763  *	skb_orphan_frags - orphan the frags contained in a buffer
2764  *	@skb: buffer to orphan frags from
2765  *	@gfp_mask: allocation mask for replacement pages
2766  *
2767  *	For each frag in the SKB which needs a destructor (i.e. has an
2768  *	owner) create a copy of that frag and release the original
2769  *	page by calling the destructor.
2770  */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)2771 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2772 {
2773 	if (likely(!skb_zcopy(skb)))
2774 		return 0;
2775 	if (!skb_zcopy_is_nouarg(skb) &&
2776 	    skb_uarg(skb)->callback == sock_zerocopy_callback)
2777 		return 0;
2778 	return skb_copy_ubufs(skb, gfp_mask);
2779 }
2780 
2781 /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
skb_orphan_frags_rx(struct sk_buff * skb,gfp_t gfp_mask)2782 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2783 {
2784 	if (likely(!skb_zcopy(skb)))
2785 		return 0;
2786 	return skb_copy_ubufs(skb, gfp_mask);
2787 }
2788 
2789 /**
2790  *	__skb_queue_purge - empty a list
2791  *	@list: list to empty
2792  *
2793  *	Delete all buffers on an &sk_buff list. Each buffer is removed from
2794  *	the list and one reference dropped. This function does not take the
2795  *	list lock and the caller must hold the relevant locks to use it.
2796  */
__skb_queue_purge(struct sk_buff_head * list)2797 static inline void __skb_queue_purge(struct sk_buff_head *list)
2798 {
2799 	struct sk_buff *skb;
2800 	while ((skb = __skb_dequeue(list)) != NULL)
2801 		kfree_skb(skb);
2802 }
2803 void skb_queue_purge(struct sk_buff_head *list);
2804 
2805 unsigned int skb_rbtree_purge(struct rb_root *root);
2806 
2807 void *netdev_alloc_frag(unsigned int fragsz);
2808 
2809 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2810 				   gfp_t gfp_mask);
2811 
2812 /**
2813  *	netdev_alloc_skb - allocate an skbuff for rx on a specific device
2814  *	@dev: network device to receive on
2815  *	@length: length to allocate
2816  *
2817  *	Allocate a new &sk_buff and assign it a usage count of one. The
2818  *	buffer has unspecified headroom built in. Users should allocate
2819  *	the headroom they think they need without accounting for the
2820  *	built in space. The built in space is used for optimisations.
2821  *
2822  *	%NULL is returned if there is no free memory. Although this function
2823  *	allocates memory it can be called from an interrupt.
2824  */
netdev_alloc_skb(struct net_device * dev,unsigned int length)2825 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2826 					       unsigned int length)
2827 {
2828 	return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2829 }
2830 
2831 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)2832 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2833 					      gfp_t gfp_mask)
2834 {
2835 	return __netdev_alloc_skb(NULL, length, gfp_mask);
2836 }
2837 
2838 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)2839 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2840 {
2841 	return netdev_alloc_skb(NULL, length);
2842 }
2843 
2844 
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)2845 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2846 		unsigned int length, gfp_t gfp)
2847 {
2848 	struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2849 
2850 	if (NET_IP_ALIGN && skb)
2851 		skb_reserve(skb, NET_IP_ALIGN);
2852 	return skb;
2853 }
2854 
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)2855 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2856 		unsigned int length)
2857 {
2858 	return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2859 }
2860 
skb_free_frag(void * addr)2861 static inline void skb_free_frag(void *addr)
2862 {
2863 	page_frag_free(addr);
2864 }
2865 
2866 void *napi_alloc_frag(unsigned int fragsz);
2867 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2868 				 unsigned int length, gfp_t gfp_mask);
napi_alloc_skb(struct napi_struct * napi,unsigned int length)2869 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2870 					     unsigned int length)
2871 {
2872 	return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2873 }
2874 void napi_consume_skb(struct sk_buff *skb, int budget);
2875 
2876 void __kfree_skb_flush(void);
2877 void __kfree_skb_defer(struct sk_buff *skb);
2878 
2879 /**
2880  * __dev_alloc_pages - allocate page for network Rx
2881  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2882  * @order: size of the allocation
2883  *
2884  * Allocate a new page.
2885  *
2886  * %NULL is returned if there is no free memory.
2887 */
__dev_alloc_pages(gfp_t gfp_mask,unsigned int order)2888 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2889 					     unsigned int order)
2890 {
2891 	/* This piece of code contains several assumptions.
2892 	 * 1.  This is for device Rx, therefor a cold page is preferred.
2893 	 * 2.  The expectation is the user wants a compound page.
2894 	 * 3.  If requesting a order 0 page it will not be compound
2895 	 *     due to the check to see if order has a value in prep_new_page
2896 	 * 4.  __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2897 	 *     code in gfp_to_alloc_flags that should be enforcing this.
2898 	 */
2899 	gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2900 
2901 	return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2902 }
2903 
dev_alloc_pages(unsigned int order)2904 static inline struct page *dev_alloc_pages(unsigned int order)
2905 {
2906 	return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2907 }
2908 
2909 /**
2910  * __dev_alloc_page - allocate a page for network Rx
2911  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2912  *
2913  * Allocate a new page.
2914  *
2915  * %NULL is returned if there is no free memory.
2916  */
__dev_alloc_page(gfp_t gfp_mask)2917 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2918 {
2919 	return __dev_alloc_pages(gfp_mask, 0);
2920 }
2921 
dev_alloc_page(void)2922 static inline struct page *dev_alloc_page(void)
2923 {
2924 	return dev_alloc_pages(0);
2925 }
2926 
2927 /**
2928  *	skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2929  *	@page: The page that was allocated from skb_alloc_page
2930  *	@skb: The skb that may need pfmemalloc set
2931  */
skb_propagate_pfmemalloc(struct page * page,struct sk_buff * skb)2932 static inline void skb_propagate_pfmemalloc(struct page *page,
2933 					     struct sk_buff *skb)
2934 {
2935 	if (page_is_pfmemalloc(page))
2936 		skb->pfmemalloc = true;
2937 }
2938 
2939 /**
2940  * skb_frag_off() - Returns the offset of a skb fragment
2941  * @frag: the paged fragment
2942  */
skb_frag_off(const skb_frag_t * frag)2943 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2944 {
2945 	return frag->bv_offset;
2946 }
2947 
2948 /**
2949  * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2950  * @frag: skb fragment
2951  * @delta: value to add
2952  */
skb_frag_off_add(skb_frag_t * frag,int delta)2953 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2954 {
2955 	frag->bv_offset += delta;
2956 }
2957 
2958 /**
2959  * skb_frag_off_set() - Sets the offset of a skb fragment
2960  * @frag: skb fragment
2961  * @offset: offset of fragment
2962  */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)2963 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2964 {
2965 	frag->bv_offset = offset;
2966 }
2967 
2968 /**
2969  * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2970  * @fragto: skb fragment where offset is set
2971  * @fragfrom: skb fragment offset is copied from
2972  */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)2973 static inline void skb_frag_off_copy(skb_frag_t *fragto,
2974 				     const skb_frag_t *fragfrom)
2975 {
2976 	fragto->bv_offset = fragfrom->bv_offset;
2977 }
2978 
2979 /**
2980  * skb_frag_page - retrieve the page referred to by a paged fragment
2981  * @frag: the paged fragment
2982  *
2983  * Returns the &struct page associated with @frag.
2984  */
skb_frag_page(const skb_frag_t * frag)2985 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2986 {
2987 	return frag->bv_page;
2988 }
2989 
2990 /**
2991  * __skb_frag_ref - take an addition reference on a paged fragment.
2992  * @frag: the paged fragment
2993  *
2994  * Takes an additional reference on the paged fragment @frag.
2995  */
__skb_frag_ref(skb_frag_t * frag)2996 static inline void __skb_frag_ref(skb_frag_t *frag)
2997 {
2998 	get_page(skb_frag_page(frag));
2999 }
3000 
3001 /**
3002  * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3003  * @skb: the buffer
3004  * @f: the fragment offset.
3005  *
3006  * Takes an additional reference on the @f'th paged fragment of @skb.
3007  */
skb_frag_ref(struct sk_buff * skb,int f)3008 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3009 {
3010 	__skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3011 }
3012 
3013 /**
3014  * __skb_frag_unref - release a reference on a paged fragment.
3015  * @frag: the paged fragment
3016  *
3017  * Releases a reference on the paged fragment @frag.
3018  */
__skb_frag_unref(skb_frag_t * frag)3019 static inline void __skb_frag_unref(skb_frag_t *frag)
3020 {
3021 	put_page(skb_frag_page(frag));
3022 }
3023 
3024 /**
3025  * skb_frag_unref - release a reference on a paged fragment of an skb.
3026  * @skb: the buffer
3027  * @f: the fragment offset
3028  *
3029  * Releases a reference on the @f'th paged fragment of @skb.
3030  */
skb_frag_unref(struct sk_buff * skb,int f)3031 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3032 {
3033 	__skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3034 }
3035 
3036 /**
3037  * skb_frag_address - gets the address of the data contained in a paged fragment
3038  * @frag: the paged fragment buffer
3039  *
3040  * Returns the address of the data within @frag. The page must already
3041  * be mapped.
3042  */
skb_frag_address(const skb_frag_t * frag)3043 static inline void *skb_frag_address(const skb_frag_t *frag)
3044 {
3045 	return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3046 }
3047 
3048 /**
3049  * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3050  * @frag: the paged fragment buffer
3051  *
3052  * Returns the address of the data within @frag. Checks that the page
3053  * is mapped and returns %NULL otherwise.
3054  */
skb_frag_address_safe(const skb_frag_t * frag)3055 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3056 {
3057 	void *ptr = page_address(skb_frag_page(frag));
3058 	if (unlikely(!ptr))
3059 		return NULL;
3060 
3061 	return ptr + skb_frag_off(frag);
3062 }
3063 
3064 /**
3065  * skb_frag_page_copy() - sets the page in a fragment from another fragment
3066  * @fragto: skb fragment where page is set
3067  * @fragfrom: skb fragment page is copied from
3068  */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3069 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3070 				      const skb_frag_t *fragfrom)
3071 {
3072 	fragto->bv_page = fragfrom->bv_page;
3073 }
3074 
3075 /**
3076  * __skb_frag_set_page - sets the page contained in a paged fragment
3077  * @frag: the paged fragment
3078  * @page: the page to set
3079  *
3080  * Sets the fragment @frag to contain @page.
3081  */
__skb_frag_set_page(skb_frag_t * frag,struct page * page)3082 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3083 {
3084 	frag->bv_page = page;
3085 }
3086 
3087 /**
3088  * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3089  * @skb: the buffer
3090  * @f: the fragment offset
3091  * @page: the page to set
3092  *
3093  * Sets the @f'th fragment of @skb to contain @page.
3094  */
skb_frag_set_page(struct sk_buff * skb,int f,struct page * page)3095 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3096 				     struct page *page)
3097 {
3098 	__skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3099 }
3100 
3101 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3102 
3103 /**
3104  * skb_frag_dma_map - maps a paged fragment via the DMA API
3105  * @dev: the device to map the fragment to
3106  * @frag: the paged fragment to map
3107  * @offset: the offset within the fragment (starting at the
3108  *          fragment's own offset)
3109  * @size: the number of bytes to map
3110  * @dir: the direction of the mapping (``PCI_DMA_*``)
3111  *
3112  * Maps the page associated with @frag to @device.
3113  */
skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3114 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3115 					  const skb_frag_t *frag,
3116 					  size_t offset, size_t size,
3117 					  enum dma_data_direction dir)
3118 {
3119 	return dma_map_page(dev, skb_frag_page(frag),
3120 			    skb_frag_off(frag) + offset, size, dir);
3121 }
3122 
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3123 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3124 					gfp_t gfp_mask)
3125 {
3126 	return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3127 }
3128 
3129 
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3130 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3131 						  gfp_t gfp_mask)
3132 {
3133 	return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3134 }
3135 
3136 
3137 /**
3138  *	skb_clone_writable - is the header of a clone writable
3139  *	@skb: buffer to check
3140  *	@len: length up to which to write
3141  *
3142  *	Returns true if modifying the header part of the cloned buffer
3143  *	does not requires the data to be copied.
3144  */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3145 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3146 {
3147 	return !skb_header_cloned(skb) &&
3148 	       skb_headroom(skb) + len <= skb->hdr_len;
3149 }
3150 
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3151 static inline int skb_try_make_writable(struct sk_buff *skb,
3152 					unsigned int write_len)
3153 {
3154 	return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3155 	       pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3156 }
3157 
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3158 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3159 			    int cloned)
3160 {
3161 	int delta = 0;
3162 
3163 	if (headroom > skb_headroom(skb))
3164 		delta = headroom - skb_headroom(skb);
3165 
3166 	if (delta || cloned)
3167 		return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3168 					GFP_ATOMIC);
3169 	return 0;
3170 }
3171 
3172 /**
3173  *	skb_cow - copy header of skb when it is required
3174  *	@skb: buffer to cow
3175  *	@headroom: needed headroom
3176  *
3177  *	If the skb passed lacks sufficient headroom or its data part
3178  *	is shared, data is reallocated. If reallocation fails, an error
3179  *	is returned and original skb is not changed.
3180  *
3181  *	The result is skb with writable area skb->head...skb->tail
3182  *	and at least @headroom of space at head.
3183  */
skb_cow(struct sk_buff * skb,unsigned int headroom)3184 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3185 {
3186 	return __skb_cow(skb, headroom, skb_cloned(skb));
3187 }
3188 
3189 /**
3190  *	skb_cow_head - skb_cow but only making the head writable
3191  *	@skb: buffer to cow
3192  *	@headroom: needed headroom
3193  *
3194  *	This function is identical to skb_cow except that we replace the
3195  *	skb_cloned check by skb_header_cloned.  It should be used when
3196  *	you only need to push on some header and do not need to modify
3197  *	the data.
3198  */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3199 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3200 {
3201 	return __skb_cow(skb, headroom, skb_header_cloned(skb));
3202 }
3203 
3204 /**
3205  *	skb_padto	- pad an skbuff up to a minimal size
3206  *	@skb: buffer to pad
3207  *	@len: minimal length
3208  *
3209  *	Pads up a buffer to ensure the trailing bytes exist and are
3210  *	blanked. If the buffer already contains sufficient data it
3211  *	is untouched. Otherwise it is extended. Returns zero on
3212  *	success. The skb is freed on error.
3213  */
skb_padto(struct sk_buff * skb,unsigned int len)3214 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3215 {
3216 	unsigned int size = skb->len;
3217 	if (likely(size >= len))
3218 		return 0;
3219 	return skb_pad(skb, len - size);
3220 }
3221 
3222 /**
3223  *	__skb_put_padto - increase size and pad an skbuff up to a minimal size
3224  *	@skb: buffer to pad
3225  *	@len: minimal length
3226  *	@free_on_error: free buffer on error
3227  *
3228  *	Pads up a buffer to ensure the trailing bytes exist and are
3229  *	blanked. If the buffer already contains sufficient data it
3230  *	is untouched. Otherwise it is extended. Returns zero on
3231  *	success. The skb is freed on error if @free_on_error is true.
3232  */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3233 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3234 					       unsigned int len,
3235 					       bool free_on_error)
3236 {
3237 	unsigned int size = skb->len;
3238 
3239 	if (unlikely(size < len)) {
3240 		len -= size;
3241 		if (__skb_pad(skb, len, free_on_error))
3242 			return -ENOMEM;
3243 		__skb_put(skb, len);
3244 	}
3245 	return 0;
3246 }
3247 
3248 /**
3249  *	skb_put_padto - increase size and pad an skbuff up to a minimal size
3250  *	@skb: buffer to pad
3251  *	@len: minimal length
3252  *
3253  *	Pads up a buffer to ensure the trailing bytes exist and are
3254  *	blanked. If the buffer already contains sufficient data it
3255  *	is untouched. Otherwise it is extended. Returns zero on
3256  *	success. The skb is freed on error.
3257  */
skb_put_padto(struct sk_buff * skb,unsigned int len)3258 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3259 {
3260 	return __skb_put_padto(skb, len, true);
3261 }
3262 
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3263 static inline int skb_add_data(struct sk_buff *skb,
3264 			       struct iov_iter *from, int copy)
3265 {
3266 	const int off = skb->len;
3267 
3268 	if (skb->ip_summed == CHECKSUM_NONE) {
3269 		__wsum csum = 0;
3270 		if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3271 					         &csum, from)) {
3272 			skb->csum = csum_block_add(skb->csum, csum, off);
3273 			return 0;
3274 		}
3275 	} else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3276 		return 0;
3277 
3278 	__skb_trim(skb, off);
3279 	return -EFAULT;
3280 }
3281 
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3282 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3283 				    const struct page *page, int off)
3284 {
3285 	if (skb_zcopy(skb))
3286 		return false;
3287 	if (i) {
3288 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3289 
3290 		return page == skb_frag_page(frag) &&
3291 		       off == skb_frag_off(frag) + skb_frag_size(frag);
3292 	}
3293 	return false;
3294 }
3295 
__skb_linearize(struct sk_buff * skb)3296 static inline int __skb_linearize(struct sk_buff *skb)
3297 {
3298 	return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3299 }
3300 
3301 /**
3302  *	skb_linearize - convert paged skb to linear one
3303  *	@skb: buffer to linarize
3304  *
3305  *	If there is no free memory -ENOMEM is returned, otherwise zero
3306  *	is returned and the old skb data released.
3307  */
skb_linearize(struct sk_buff * skb)3308 static inline int skb_linearize(struct sk_buff *skb)
3309 {
3310 	return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3311 }
3312 
3313 /**
3314  * skb_has_shared_frag - can any frag be overwritten
3315  * @skb: buffer to test
3316  *
3317  * Return true if the skb has at least one frag that might be modified
3318  * by an external entity (as in vmsplice()/sendfile())
3319  */
skb_has_shared_frag(const struct sk_buff * skb)3320 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3321 {
3322 	return skb_is_nonlinear(skb) &&
3323 	       skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3324 }
3325 
3326 /**
3327  *	skb_linearize_cow - make sure skb is linear and writable
3328  *	@skb: buffer to process
3329  *
3330  *	If there is no free memory -ENOMEM is returned, otherwise zero
3331  *	is returned and the old skb data released.
3332  */
skb_linearize_cow(struct sk_buff * skb)3333 static inline int skb_linearize_cow(struct sk_buff *skb)
3334 {
3335 	return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3336 	       __skb_linearize(skb) : 0;
3337 }
3338 
3339 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3340 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3341 		     unsigned int off)
3342 {
3343 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3344 		skb->csum = csum_block_sub(skb->csum,
3345 					   csum_partial(start, len, 0), off);
3346 	else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3347 		 skb_checksum_start_offset(skb) < 0)
3348 		skb->ip_summed = CHECKSUM_NONE;
3349 }
3350 
3351 /**
3352  *	skb_postpull_rcsum - update checksum for received skb after pull
3353  *	@skb: buffer to update
3354  *	@start: start of data before pull
3355  *	@len: length of data pulled
3356  *
3357  *	After doing a pull on a received packet, you need to call this to
3358  *	update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3359  *	CHECKSUM_NONE so that it can be recomputed from scratch.
3360  */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3361 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3362 				      const void *start, unsigned int len)
3363 {
3364 	__skb_postpull_rcsum(skb, start, len, 0);
3365 }
3366 
3367 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3368 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3369 		     unsigned int off)
3370 {
3371 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3372 		skb->csum = csum_block_add(skb->csum,
3373 					   csum_partial(start, len, 0), off);
3374 }
3375 
3376 /**
3377  *	skb_postpush_rcsum - update checksum for received skb after push
3378  *	@skb: buffer to update
3379  *	@start: start of data after push
3380  *	@len: length of data pushed
3381  *
3382  *	After doing a push on a received packet, you need to call this to
3383  *	update the CHECKSUM_COMPLETE checksum.
3384  */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3385 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3386 				      const void *start, unsigned int len)
3387 {
3388 	__skb_postpush_rcsum(skb, start, len, 0);
3389 }
3390 
3391 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3392 
3393 /**
3394  *	skb_push_rcsum - push skb and update receive checksum
3395  *	@skb: buffer to update
3396  *	@len: length of data pulled
3397  *
3398  *	This function performs an skb_push on the packet and updates
3399  *	the CHECKSUM_COMPLETE checksum.  It should be used on
3400  *	receive path processing instead of skb_push unless you know
3401  *	that the checksum difference is zero (e.g., a valid IP header)
3402  *	or you are setting ip_summed to CHECKSUM_NONE.
3403  */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3404 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3405 {
3406 	skb_push(skb, len);
3407 	skb_postpush_rcsum(skb, skb->data, len);
3408 	return skb->data;
3409 }
3410 
3411 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3412 /**
3413  *	pskb_trim_rcsum - trim received skb and update checksum
3414  *	@skb: buffer to trim
3415  *	@len: new length
3416  *
3417  *	This is exactly the same as pskb_trim except that it ensures the
3418  *	checksum of received packets are still valid after the operation.
3419  *	It can change skb pointers.
3420  */
3421 
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)3422 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3423 {
3424 	if (likely(len >= skb->len))
3425 		return 0;
3426 	return pskb_trim_rcsum_slow(skb, len);
3427 }
3428 
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)3429 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3430 {
3431 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3432 		skb->ip_summed = CHECKSUM_NONE;
3433 	__skb_trim(skb, len);
3434 	return 0;
3435 }
3436 
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)3437 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3438 {
3439 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3440 		skb->ip_summed = CHECKSUM_NONE;
3441 	return __skb_grow(skb, len);
3442 }
3443 
3444 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3445 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3446 #define skb_rb_last(root)  rb_to_skb(rb_last(root))
3447 #define skb_rb_next(skb)   rb_to_skb(rb_next(&(skb)->rbnode))
3448 #define skb_rb_prev(skb)   rb_to_skb(rb_prev(&(skb)->rbnode))
3449 
3450 #define skb_queue_walk(queue, skb) \
3451 		for (skb = (queue)->next;					\
3452 		     skb != (struct sk_buff *)(queue);				\
3453 		     skb = skb->next)
3454 
3455 #define skb_queue_walk_safe(queue, skb, tmp)					\
3456 		for (skb = (queue)->next, tmp = skb->next;			\
3457 		     skb != (struct sk_buff *)(queue);				\
3458 		     skb = tmp, tmp = skb->next)
3459 
3460 #define skb_queue_walk_from(queue, skb)						\
3461 		for (; skb != (struct sk_buff *)(queue);			\
3462 		     skb = skb->next)
3463 
3464 #define skb_rbtree_walk(skb, root)						\
3465 		for (skb = skb_rb_first(root); skb != NULL;			\
3466 		     skb = skb_rb_next(skb))
3467 
3468 #define skb_rbtree_walk_from(skb)						\
3469 		for (; skb != NULL;						\
3470 		     skb = skb_rb_next(skb))
3471 
3472 #define skb_rbtree_walk_from_safe(skb, tmp)					\
3473 		for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL);	\
3474 		     skb = tmp)
3475 
3476 #define skb_queue_walk_from_safe(queue, skb, tmp)				\
3477 		for (tmp = skb->next;						\
3478 		     skb != (struct sk_buff *)(queue);				\
3479 		     skb = tmp, tmp = skb->next)
3480 
3481 #define skb_queue_reverse_walk(queue, skb) \
3482 		for (skb = (queue)->prev;					\
3483 		     skb != (struct sk_buff *)(queue);				\
3484 		     skb = skb->prev)
3485 
3486 #define skb_queue_reverse_walk_safe(queue, skb, tmp)				\
3487 		for (skb = (queue)->prev, tmp = skb->prev;			\
3488 		     skb != (struct sk_buff *)(queue);				\
3489 		     skb = tmp, tmp = skb->prev)
3490 
3491 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp)			\
3492 		for (tmp = skb->prev;						\
3493 		     skb != (struct sk_buff *)(queue);				\
3494 		     skb = tmp, tmp = skb->prev)
3495 
skb_has_frag_list(const struct sk_buff * skb)3496 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3497 {
3498 	return skb_shinfo(skb)->frag_list != NULL;
3499 }
3500 
skb_frag_list_init(struct sk_buff * skb)3501 static inline void skb_frag_list_init(struct sk_buff *skb)
3502 {
3503 	skb_shinfo(skb)->frag_list = NULL;
3504 }
3505 
3506 #define skb_walk_frags(skb, iter)	\
3507 	for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3508 
3509 
3510 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3511 				int *err, long *timeo_p,
3512 				const struct sk_buff *skb);
3513 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3514 					  struct sk_buff_head *queue,
3515 					  unsigned int flags,
3516 					  int *off, int *err,
3517 					  struct sk_buff **last);
3518 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3519 					struct sk_buff_head *queue,
3520 					unsigned int flags, int *off, int *err,
3521 					struct sk_buff **last);
3522 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3523 				    struct sk_buff_head *sk_queue,
3524 				    unsigned int flags, int *off, int *err);
3525 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3526 				  int *err);
3527 __poll_t datagram_poll(struct file *file, struct socket *sock,
3528 			   struct poll_table_struct *wait);
3529 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3530 			   struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)3531 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3532 					struct msghdr *msg, int size)
3533 {
3534 	return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3535 }
3536 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3537 				   struct msghdr *msg);
3538 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3539 			   struct iov_iter *to, int len,
3540 			   struct ahash_request *hash);
3541 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3542 				 struct iov_iter *from, int len);
3543 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3544 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3545 void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
skb_free_datagram_locked(struct sock * sk,struct sk_buff * skb)3546 static inline void skb_free_datagram_locked(struct sock *sk,
3547 					    struct sk_buff *skb)
3548 {
3549 	__skb_free_datagram_locked(sk, skb, 0);
3550 }
3551 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3552 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3553 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3554 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3555 			      int len);
3556 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3557 		    struct pipe_inode_info *pipe, unsigned int len,
3558 		    unsigned int flags);
3559 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3560 			 int len);
3561 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3562 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3563 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3564 		 int len, int hlen);
3565 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3566 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3567 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3568 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3569 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3570 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3571 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3572 				 unsigned int offset);
3573 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3574 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3575 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3576 int skb_vlan_pop(struct sk_buff *skb);
3577 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3578 int skb_eth_pop(struct sk_buff *skb);
3579 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3580 		 const unsigned char *src);
3581 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3582 		  int mac_len, bool ethernet);
3583 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3584 		 bool ethernet);
3585 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3586 int skb_mpls_dec_ttl(struct sk_buff *skb);
3587 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3588 			     gfp_t gfp);
3589 
memcpy_from_msg(void * data,struct msghdr * msg,int len)3590 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3591 {
3592 	return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3593 }
3594 
memcpy_to_msg(struct msghdr * msg,void * data,int len)3595 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3596 {
3597 	return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3598 }
3599 
3600 struct skb_checksum_ops {
3601 	__wsum (*update)(const void *mem, int len, __wsum wsum);
3602 	__wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3603 };
3604 
3605 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3606 
3607 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3608 		      __wsum csum, const struct skb_checksum_ops *ops);
3609 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3610 		    __wsum csum);
3611 
3612 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * data,int hlen,void * buffer)3613 __skb_header_pointer(const struct sk_buff *skb, int offset,
3614 		     int len, void *data, int hlen, void *buffer)
3615 {
3616 	if (hlen - offset >= len)
3617 		return data + offset;
3618 
3619 	if (!skb ||
3620 	    skb_copy_bits(skb, offset, buffer, len) < 0)
3621 		return NULL;
3622 
3623 	return buffer;
3624 }
3625 
3626 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)3627 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3628 {
3629 	return __skb_header_pointer(skb, offset, len, skb->data,
3630 				    skb_headlen(skb), buffer);
3631 }
3632 
3633 /**
3634  *	skb_needs_linearize - check if we need to linearize a given skb
3635  *			      depending on the given device features.
3636  *	@skb: socket buffer to check
3637  *	@features: net device features
3638  *
3639  *	Returns true if either:
3640  *	1. skb has frag_list and the device doesn't support FRAGLIST, or
3641  *	2. skb is fragmented and the device does not support SG.
3642  */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)3643 static inline bool skb_needs_linearize(struct sk_buff *skb,
3644 				       netdev_features_t features)
3645 {
3646 	return skb_is_nonlinear(skb) &&
3647 	       ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3648 		(skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3649 }
3650 
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)3651 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3652 					     void *to,
3653 					     const unsigned int len)
3654 {
3655 	memcpy(to, skb->data, len);
3656 }
3657 
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)3658 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3659 						    const int offset, void *to,
3660 						    const unsigned int len)
3661 {
3662 	memcpy(to, skb->data + offset, len);
3663 }
3664 
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)3665 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3666 					   const void *from,
3667 					   const unsigned int len)
3668 {
3669 	memcpy(skb->data, from, len);
3670 }
3671 
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)3672 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3673 						  const int offset,
3674 						  const void *from,
3675 						  const unsigned int len)
3676 {
3677 	memcpy(skb->data + offset, from, len);
3678 }
3679 
3680 void skb_init(void);
3681 
skb_get_ktime(const struct sk_buff * skb)3682 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3683 {
3684 	return skb->tstamp;
3685 }
3686 
3687 /**
3688  *	skb_get_timestamp - get timestamp from a skb
3689  *	@skb: skb to get stamp from
3690  *	@stamp: pointer to struct __kernel_old_timeval to store stamp in
3691  *
3692  *	Timestamps are stored in the skb as offsets to a base timestamp.
3693  *	This function converts the offset back to a struct timeval and stores
3694  *	it in stamp.
3695  */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)3696 static inline void skb_get_timestamp(const struct sk_buff *skb,
3697 				     struct __kernel_old_timeval *stamp)
3698 {
3699 	*stamp = ns_to_kernel_old_timeval(skb->tstamp);
3700 }
3701 
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)3702 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3703 					 struct __kernel_sock_timeval *stamp)
3704 {
3705 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3706 
3707 	stamp->tv_sec = ts.tv_sec;
3708 	stamp->tv_usec = ts.tv_nsec / 1000;
3709 }
3710 
skb_get_timestampns(const struct sk_buff * skb,struct __kernel_old_timespec * stamp)3711 static inline void skb_get_timestampns(const struct sk_buff *skb,
3712 				       struct __kernel_old_timespec *stamp)
3713 {
3714 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3715 
3716 	stamp->tv_sec = ts.tv_sec;
3717 	stamp->tv_nsec = ts.tv_nsec;
3718 }
3719 
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)3720 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3721 					   struct __kernel_timespec *stamp)
3722 {
3723 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3724 
3725 	stamp->tv_sec = ts.tv_sec;
3726 	stamp->tv_nsec = ts.tv_nsec;
3727 }
3728 
__net_timestamp(struct sk_buff * skb)3729 static inline void __net_timestamp(struct sk_buff *skb)
3730 {
3731 	skb->tstamp = ktime_get_real();
3732 }
3733 
net_timedelta(ktime_t t)3734 static inline ktime_t net_timedelta(ktime_t t)
3735 {
3736 	return ktime_sub(ktime_get_real(), t);
3737 }
3738 
net_invalid_timestamp(void)3739 static inline ktime_t net_invalid_timestamp(void)
3740 {
3741 	return 0;
3742 }
3743 
skb_metadata_len(const struct sk_buff * skb)3744 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3745 {
3746 	return skb_shinfo(skb)->meta_len;
3747 }
3748 
skb_metadata_end(const struct sk_buff * skb)3749 static inline void *skb_metadata_end(const struct sk_buff *skb)
3750 {
3751 	return skb_mac_header(skb);
3752 }
3753 
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)3754 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3755 					  const struct sk_buff *skb_b,
3756 					  u8 meta_len)
3757 {
3758 	const void *a = skb_metadata_end(skb_a);
3759 	const void *b = skb_metadata_end(skb_b);
3760 	/* Using more efficient varaiant than plain call to memcmp(). */
3761 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3762 	u64 diffs = 0;
3763 
3764 	switch (meta_len) {
3765 #define __it(x, op) (x -= sizeof(u##op))
3766 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3767 	case 32: diffs |= __it_diff(a, b, 64);
3768 		fallthrough;
3769 	case 24: diffs |= __it_diff(a, b, 64);
3770 		fallthrough;
3771 	case 16: diffs |= __it_diff(a, b, 64);
3772 		fallthrough;
3773 	case  8: diffs |= __it_diff(a, b, 64);
3774 		break;
3775 	case 28: diffs |= __it_diff(a, b, 64);
3776 		fallthrough;
3777 	case 20: diffs |= __it_diff(a, b, 64);
3778 		fallthrough;
3779 	case 12: diffs |= __it_diff(a, b, 64);
3780 		fallthrough;
3781 	case  4: diffs |= __it_diff(a, b, 32);
3782 		break;
3783 	}
3784 	return diffs;
3785 #else
3786 	return memcmp(a - meta_len, b - meta_len, meta_len);
3787 #endif
3788 }
3789 
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)3790 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3791 					const struct sk_buff *skb_b)
3792 {
3793 	u8 len_a = skb_metadata_len(skb_a);
3794 	u8 len_b = skb_metadata_len(skb_b);
3795 
3796 	if (!(len_a | len_b))
3797 		return false;
3798 
3799 	return len_a != len_b ?
3800 	       true : __skb_metadata_differs(skb_a, skb_b, len_a);
3801 }
3802 
skb_metadata_set(struct sk_buff * skb,u8 meta_len)3803 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3804 {
3805 	skb_shinfo(skb)->meta_len = meta_len;
3806 }
3807 
skb_metadata_clear(struct sk_buff * skb)3808 static inline void skb_metadata_clear(struct sk_buff *skb)
3809 {
3810 	skb_metadata_set(skb, 0);
3811 }
3812 
3813 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3814 
3815 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3816 
3817 void skb_clone_tx_timestamp(struct sk_buff *skb);
3818 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3819 
3820 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3821 
skb_clone_tx_timestamp(struct sk_buff * skb)3822 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3823 {
3824 }
3825 
skb_defer_rx_timestamp(struct sk_buff * skb)3826 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3827 {
3828 	return false;
3829 }
3830 
3831 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3832 
3833 /**
3834  * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3835  *
3836  * PHY drivers may accept clones of transmitted packets for
3837  * timestamping via their phy_driver.txtstamp method. These drivers
3838  * must call this function to return the skb back to the stack with a
3839  * timestamp.
3840  *
3841  * @skb: clone of the original outgoing packet
3842  * @hwtstamps: hardware time stamps
3843  *
3844  */
3845 void skb_complete_tx_timestamp(struct sk_buff *skb,
3846 			       struct skb_shared_hwtstamps *hwtstamps);
3847 
3848 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3849 		     struct skb_shared_hwtstamps *hwtstamps,
3850 		     struct sock *sk, int tstype);
3851 
3852 /**
3853  * skb_tstamp_tx - queue clone of skb with send time stamps
3854  * @orig_skb:	the original outgoing packet
3855  * @hwtstamps:	hardware time stamps, may be NULL if not available
3856  *
3857  * If the skb has a socket associated, then this function clones the
3858  * skb (thus sharing the actual data and optional structures), stores
3859  * the optional hardware time stamping information (if non NULL) or
3860  * generates a software time stamp (otherwise), then queues the clone
3861  * to the error queue of the socket.  Errors are silently ignored.
3862  */
3863 void skb_tstamp_tx(struct sk_buff *orig_skb,
3864 		   struct skb_shared_hwtstamps *hwtstamps);
3865 
3866 /**
3867  * skb_tx_timestamp() - Driver hook for transmit timestamping
3868  *
3869  * Ethernet MAC Drivers should call this function in their hard_xmit()
3870  * function immediately before giving the sk_buff to the MAC hardware.
3871  *
3872  * Specifically, one should make absolutely sure that this function is
3873  * called before TX completion of this packet can trigger.  Otherwise
3874  * the packet could potentially already be freed.
3875  *
3876  * @skb: A socket buffer.
3877  */
skb_tx_timestamp(struct sk_buff * skb)3878 static inline void skb_tx_timestamp(struct sk_buff *skb)
3879 {
3880 	skb_clone_tx_timestamp(skb);
3881 	if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3882 		skb_tstamp_tx(skb, NULL);
3883 }
3884 
3885 /**
3886  * skb_complete_wifi_ack - deliver skb with wifi status
3887  *
3888  * @skb: the original outgoing packet
3889  * @acked: ack status
3890  *
3891  */
3892 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3893 
3894 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3895 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3896 
skb_csum_unnecessary(const struct sk_buff * skb)3897 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3898 {
3899 	return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3900 		skb->csum_valid ||
3901 		(skb->ip_summed == CHECKSUM_PARTIAL &&
3902 		 skb_checksum_start_offset(skb) >= 0));
3903 }
3904 
3905 /**
3906  *	skb_checksum_complete - Calculate checksum of an entire packet
3907  *	@skb: packet to process
3908  *
3909  *	This function calculates the checksum over the entire packet plus
3910  *	the value of skb->csum.  The latter can be used to supply the
3911  *	checksum of a pseudo header as used by TCP/UDP.  It returns the
3912  *	checksum.
3913  *
3914  *	For protocols that contain complete checksums such as ICMP/TCP/UDP,
3915  *	this function can be used to verify that checksum on received
3916  *	packets.  In that case the function should return zero if the
3917  *	checksum is correct.  In particular, this function will return zero
3918  *	if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3919  *	hardware has already verified the correctness of the checksum.
3920  */
skb_checksum_complete(struct sk_buff * skb)3921 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3922 {
3923 	return skb_csum_unnecessary(skb) ?
3924 	       0 : __skb_checksum_complete(skb);
3925 }
3926 
__skb_decr_checksum_unnecessary(struct sk_buff * skb)3927 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3928 {
3929 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3930 		if (skb->csum_level == 0)
3931 			skb->ip_summed = CHECKSUM_NONE;
3932 		else
3933 			skb->csum_level--;
3934 	}
3935 }
3936 
__skb_incr_checksum_unnecessary(struct sk_buff * skb)3937 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3938 {
3939 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3940 		if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3941 			skb->csum_level++;
3942 	} else if (skb->ip_summed == CHECKSUM_NONE) {
3943 		skb->ip_summed = CHECKSUM_UNNECESSARY;
3944 		skb->csum_level = 0;
3945 	}
3946 }
3947 
__skb_reset_checksum_unnecessary(struct sk_buff * skb)3948 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
3949 {
3950 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3951 		skb->ip_summed = CHECKSUM_NONE;
3952 		skb->csum_level = 0;
3953 	}
3954 }
3955 
3956 /* Check if we need to perform checksum complete validation.
3957  *
3958  * Returns true if checksum complete is needed, false otherwise
3959  * (either checksum is unnecessary or zero checksum is allowed).
3960  */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)3961 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3962 						  bool zero_okay,
3963 						  __sum16 check)
3964 {
3965 	if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3966 		skb->csum_valid = 1;
3967 		__skb_decr_checksum_unnecessary(skb);
3968 		return false;
3969 	}
3970 
3971 	return true;
3972 }
3973 
3974 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
3975  * in checksum_init.
3976  */
3977 #define CHECKSUM_BREAK 76
3978 
3979 /* Unset checksum-complete
3980  *
3981  * Unset checksum complete can be done when packet is being modified
3982  * (uncompressed for instance) and checksum-complete value is
3983  * invalidated.
3984  */
skb_checksum_complete_unset(struct sk_buff * skb)3985 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3986 {
3987 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3988 		skb->ip_summed = CHECKSUM_NONE;
3989 }
3990 
3991 /* Validate (init) checksum based on checksum complete.
3992  *
3993  * Return values:
3994  *   0: checksum is validated or try to in skb_checksum_complete. In the latter
3995  *	case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3996  *	checksum is stored in skb->csum for use in __skb_checksum_complete
3997  *   non-zero: value of invalid checksum
3998  *
3999  */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)4000 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4001 						       bool complete,
4002 						       __wsum psum)
4003 {
4004 	if (skb->ip_summed == CHECKSUM_COMPLETE) {
4005 		if (!csum_fold(csum_add(psum, skb->csum))) {
4006 			skb->csum_valid = 1;
4007 			return 0;
4008 		}
4009 	}
4010 
4011 	skb->csum = psum;
4012 
4013 	if (complete || skb->len <= CHECKSUM_BREAK) {
4014 		__sum16 csum;
4015 
4016 		csum = __skb_checksum_complete(skb);
4017 		skb->csum_valid = !csum;
4018 		return csum;
4019 	}
4020 
4021 	return 0;
4022 }
4023 
null_compute_pseudo(struct sk_buff * skb,int proto)4024 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4025 {
4026 	return 0;
4027 }
4028 
4029 /* Perform checksum validate (init). Note that this is a macro since we only
4030  * want to calculate the pseudo header which is an input function if necessary.
4031  * First we try to validate without any computation (checksum unnecessary) and
4032  * then calculate based on checksum complete calling the function to compute
4033  * pseudo header.
4034  *
4035  * Return values:
4036  *   0: checksum is validated or try to in skb_checksum_complete
4037  *   non-zero: value of invalid checksum
4038  */
4039 #define __skb_checksum_validate(skb, proto, complete,			\
4040 				zero_okay, check, compute_pseudo)	\
4041 ({									\
4042 	__sum16 __ret = 0;						\
4043 	skb->csum_valid = 0;						\
4044 	if (__skb_checksum_validate_needed(skb, zero_okay, check))	\
4045 		__ret = __skb_checksum_validate_complete(skb,		\
4046 				complete, compute_pseudo(skb, proto));	\
4047 	__ret;								\
4048 })
4049 
4050 #define skb_checksum_init(skb, proto, compute_pseudo)			\
4051 	__skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4052 
4053 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo)	\
4054 	__skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4055 
4056 #define skb_checksum_validate(skb, proto, compute_pseudo)		\
4057 	__skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4058 
4059 #define skb_checksum_validate_zero_check(skb, proto, check,		\
4060 					 compute_pseudo)		\
4061 	__skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4062 
4063 #define skb_checksum_simple_validate(skb)				\
4064 	__skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4065 
__skb_checksum_convert_check(struct sk_buff * skb)4066 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4067 {
4068 	return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4069 }
4070 
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4071 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4072 {
4073 	skb->csum = ~pseudo;
4074 	skb->ip_summed = CHECKSUM_COMPLETE;
4075 }
4076 
4077 #define skb_checksum_try_convert(skb, proto, compute_pseudo)	\
4078 do {									\
4079 	if (__skb_checksum_convert_check(skb))				\
4080 		__skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4081 } while (0)
4082 
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4083 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4084 					      u16 start, u16 offset)
4085 {
4086 	skb->ip_summed = CHECKSUM_PARTIAL;
4087 	skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4088 	skb->csum_offset = offset - start;
4089 }
4090 
4091 /* Update skbuf and packet to reflect the remote checksum offload operation.
4092  * When called, ptr indicates the starting point for skb->csum when
4093  * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4094  * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4095  */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4096 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4097 				       int start, int offset, bool nopartial)
4098 {
4099 	__wsum delta;
4100 
4101 	if (!nopartial) {
4102 		skb_remcsum_adjust_partial(skb, ptr, start, offset);
4103 		return;
4104 	}
4105 
4106 	 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4107 		__skb_checksum_complete(skb);
4108 		skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4109 	}
4110 
4111 	delta = remcsum_adjust(ptr, skb->csum, start, offset);
4112 
4113 	/* Adjust skb->csum since we changed the packet */
4114 	skb->csum = csum_add(skb->csum, delta);
4115 }
4116 
skb_nfct(const struct sk_buff * skb)4117 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4118 {
4119 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4120 	return (void *)(skb->_nfct & NFCT_PTRMASK);
4121 #else
4122 	return NULL;
4123 #endif
4124 }
4125 
skb_get_nfct(const struct sk_buff * skb)4126 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4127 {
4128 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4129 	return skb->_nfct;
4130 #else
4131 	return 0UL;
4132 #endif
4133 }
4134 
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4135 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4136 {
4137 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4138 	skb->_nfct = nfct;
4139 #endif
4140 }
4141 
4142 #ifdef CONFIG_SKB_EXTENSIONS
4143 enum skb_ext_id {
4144 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4145 	SKB_EXT_BRIDGE_NF,
4146 #endif
4147 #ifdef CONFIG_XFRM
4148 	SKB_EXT_SEC_PATH,
4149 #endif
4150 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4151 	TC_SKB_EXT,
4152 #endif
4153 #if IS_ENABLED(CONFIG_MPTCP)
4154 	SKB_EXT_MPTCP,
4155 #endif
4156 #if IS_ENABLED(CONFIG_KCOV)
4157 	SKB_EXT_KCOV_HANDLE,
4158 #endif
4159 	SKB_EXT_NUM, /* must be last */
4160 };
4161 
4162 /**
4163  *	struct skb_ext - sk_buff extensions
4164  *	@refcnt: 1 on allocation, deallocated on 0
4165  *	@offset: offset to add to @data to obtain extension address
4166  *	@chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4167  *	@data: start of extension data, variable sized
4168  *
4169  *	Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4170  *	to use 'u8' types while allowing up to 2kb worth of extension data.
4171  */
4172 struct skb_ext {
4173 	refcount_t refcnt;
4174 	u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4175 	u8 chunks;		/* same */
4176 	char data[] __aligned(8);
4177 };
4178 
4179 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4180 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4181 		    struct skb_ext *ext);
4182 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4183 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4184 void __skb_ext_put(struct skb_ext *ext);
4185 
skb_ext_put(struct sk_buff * skb)4186 static inline void skb_ext_put(struct sk_buff *skb)
4187 {
4188 	if (skb->active_extensions)
4189 		__skb_ext_put(skb->extensions);
4190 }
4191 
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4192 static inline void __skb_ext_copy(struct sk_buff *dst,
4193 				  const struct sk_buff *src)
4194 {
4195 	dst->active_extensions = src->active_extensions;
4196 
4197 	if (src->active_extensions) {
4198 		struct skb_ext *ext = src->extensions;
4199 
4200 		refcount_inc(&ext->refcnt);
4201 		dst->extensions = ext;
4202 	}
4203 }
4204 
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4205 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4206 {
4207 	skb_ext_put(dst);
4208 	__skb_ext_copy(dst, src);
4209 }
4210 
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4211 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4212 {
4213 	return !!ext->offset[i];
4214 }
4215 
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4216 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4217 {
4218 	return skb->active_extensions & (1 << id);
4219 }
4220 
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4221 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4222 {
4223 	if (skb_ext_exist(skb, id))
4224 		__skb_ext_del(skb, id);
4225 }
4226 
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4227 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4228 {
4229 	if (skb_ext_exist(skb, id)) {
4230 		struct skb_ext *ext = skb->extensions;
4231 
4232 		return (void *)ext + (ext->offset[id] << 3);
4233 	}
4234 
4235 	return NULL;
4236 }
4237 
skb_ext_reset(struct sk_buff * skb)4238 static inline void skb_ext_reset(struct sk_buff *skb)
4239 {
4240 	if (unlikely(skb->active_extensions)) {
4241 		__skb_ext_put(skb->extensions);
4242 		skb->active_extensions = 0;
4243 	}
4244 }
4245 
skb_has_extensions(struct sk_buff * skb)4246 static inline bool skb_has_extensions(struct sk_buff *skb)
4247 {
4248 	return unlikely(skb->active_extensions);
4249 }
4250 #else
skb_ext_put(struct sk_buff * skb)4251 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4252 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4253 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4254 static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * s)4255 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4256 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4257 #endif /* CONFIG_SKB_EXTENSIONS */
4258 
nf_reset_ct(struct sk_buff * skb)4259 static inline void nf_reset_ct(struct sk_buff *skb)
4260 {
4261 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4262 	nf_conntrack_put(skb_nfct(skb));
4263 	skb->_nfct = 0;
4264 #endif
4265 }
4266 
nf_reset_trace(struct sk_buff * skb)4267 static inline void nf_reset_trace(struct sk_buff *skb)
4268 {
4269 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4270 	skb->nf_trace = 0;
4271 #endif
4272 }
4273 
ipvs_reset(struct sk_buff * skb)4274 static inline void ipvs_reset(struct sk_buff *skb)
4275 {
4276 #if IS_ENABLED(CONFIG_IP_VS)
4277 	skb->ipvs_property = 0;
4278 #endif
4279 }
4280 
4281 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4282 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4283 			     bool copy)
4284 {
4285 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4286 	dst->_nfct = src->_nfct;
4287 	nf_conntrack_get(skb_nfct(src));
4288 #endif
4289 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4290 	if (copy)
4291 		dst->nf_trace = src->nf_trace;
4292 #endif
4293 }
4294 
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4295 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4296 {
4297 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4298 	nf_conntrack_put(skb_nfct(dst));
4299 #endif
4300 	__nf_copy(dst, src, true);
4301 }
4302 
4303 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4304 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4305 {
4306 	to->secmark = from->secmark;
4307 }
4308 
skb_init_secmark(struct sk_buff * skb)4309 static inline void skb_init_secmark(struct sk_buff *skb)
4310 {
4311 	skb->secmark = 0;
4312 }
4313 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4314 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4315 { }
4316 
skb_init_secmark(struct sk_buff * skb)4317 static inline void skb_init_secmark(struct sk_buff *skb)
4318 { }
4319 #endif
4320 
secpath_exists(const struct sk_buff * skb)4321 static inline int secpath_exists(const struct sk_buff *skb)
4322 {
4323 #ifdef CONFIG_XFRM
4324 	return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4325 #else
4326 	return 0;
4327 #endif
4328 }
4329 
skb_irq_freeable(const struct sk_buff * skb)4330 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4331 {
4332 	return !skb->destructor &&
4333 		!secpath_exists(skb) &&
4334 		!skb_nfct(skb) &&
4335 		!skb->_skb_refdst &&
4336 		!skb_has_frag_list(skb);
4337 }
4338 
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)4339 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4340 {
4341 	skb->queue_mapping = queue_mapping;
4342 }
4343 
skb_get_queue_mapping(const struct sk_buff * skb)4344 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4345 {
4346 	return skb->queue_mapping;
4347 }
4348 
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)4349 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4350 {
4351 	to->queue_mapping = from->queue_mapping;
4352 }
4353 
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)4354 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4355 {
4356 	skb->queue_mapping = rx_queue + 1;
4357 }
4358 
skb_get_rx_queue(const struct sk_buff * skb)4359 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4360 {
4361 	return skb->queue_mapping - 1;
4362 }
4363 
skb_rx_queue_recorded(const struct sk_buff * skb)4364 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4365 {
4366 	return skb->queue_mapping != 0;
4367 }
4368 
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)4369 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4370 {
4371 	skb->dst_pending_confirm = val;
4372 }
4373 
skb_get_dst_pending_confirm(const struct sk_buff * skb)4374 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4375 {
4376 	return skb->dst_pending_confirm != 0;
4377 }
4378 
skb_sec_path(const struct sk_buff * skb)4379 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4380 {
4381 #ifdef CONFIG_XFRM
4382 	return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4383 #else
4384 	return NULL;
4385 #endif
4386 }
4387 
4388 /* Keeps track of mac header offset relative to skb->head.
4389  * It is useful for TSO of Tunneling protocol. e.g. GRE.
4390  * For non-tunnel skb it points to skb_mac_header() and for
4391  * tunnel skb it points to outer mac header.
4392  * Keeps track of level of encapsulation of network headers.
4393  */
4394 struct skb_gso_cb {
4395 	union {
4396 		int	mac_offset;
4397 		int	data_offset;
4398 	};
4399 	int	encap_level;
4400 	__wsum	csum;
4401 	__u16	csum_start;
4402 };
4403 #define SKB_GSO_CB_OFFSET	32
4404 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4405 
skb_tnl_header_len(const struct sk_buff * inner_skb)4406 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4407 {
4408 	return (skb_mac_header(inner_skb) - inner_skb->head) -
4409 		SKB_GSO_CB(inner_skb)->mac_offset;
4410 }
4411 
gso_pskb_expand_head(struct sk_buff * skb,int extra)4412 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4413 {
4414 	int new_headroom, headroom;
4415 	int ret;
4416 
4417 	headroom = skb_headroom(skb);
4418 	ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4419 	if (ret)
4420 		return ret;
4421 
4422 	new_headroom = skb_headroom(skb);
4423 	SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4424 	return 0;
4425 }
4426 
gso_reset_checksum(struct sk_buff * skb,__wsum res)4427 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4428 {
4429 	/* Do not update partial checksums if remote checksum is enabled. */
4430 	if (skb->remcsum_offload)
4431 		return;
4432 
4433 	SKB_GSO_CB(skb)->csum = res;
4434 	SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4435 }
4436 
4437 /* Compute the checksum for a gso segment. First compute the checksum value
4438  * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4439  * then add in skb->csum (checksum from csum_start to end of packet).
4440  * skb->csum and csum_start are then updated to reflect the checksum of the
4441  * resultant packet starting from the transport header-- the resultant checksum
4442  * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4443  * header.
4444  */
gso_make_checksum(struct sk_buff * skb,__wsum res)4445 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4446 {
4447 	unsigned char *csum_start = skb_transport_header(skb);
4448 	int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4449 	__wsum partial = SKB_GSO_CB(skb)->csum;
4450 
4451 	SKB_GSO_CB(skb)->csum = res;
4452 	SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4453 
4454 	return csum_fold(csum_partial(csum_start, plen, partial));
4455 }
4456 
skb_is_gso(const struct sk_buff * skb)4457 static inline bool skb_is_gso(const struct sk_buff *skb)
4458 {
4459 	return skb_shinfo(skb)->gso_size;
4460 }
4461 
4462 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)4463 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4464 {
4465 	return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4466 }
4467 
4468 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)4469 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4470 {
4471 	return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4472 }
4473 
4474 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)4475 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4476 {
4477 	return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4478 }
4479 
skb_gso_reset(struct sk_buff * skb)4480 static inline void skb_gso_reset(struct sk_buff *skb)
4481 {
4482 	skb_shinfo(skb)->gso_size = 0;
4483 	skb_shinfo(skb)->gso_segs = 0;
4484 	skb_shinfo(skb)->gso_type = 0;
4485 }
4486 
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)4487 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4488 					 u16 increment)
4489 {
4490 	if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4491 		return;
4492 	shinfo->gso_size += increment;
4493 }
4494 
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)4495 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4496 					 u16 decrement)
4497 {
4498 	if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4499 		return;
4500 	shinfo->gso_size -= decrement;
4501 }
4502 
4503 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4504 
skb_warn_if_lro(const struct sk_buff * skb)4505 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4506 {
4507 	/* LRO sets gso_size but not gso_type, whereas if GSO is really
4508 	 * wanted then gso_type will be set. */
4509 	const struct skb_shared_info *shinfo = skb_shinfo(skb);
4510 
4511 	if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4512 	    unlikely(shinfo->gso_type == 0)) {
4513 		__skb_warn_lro_forwarding(skb);
4514 		return true;
4515 	}
4516 	return false;
4517 }
4518 
skb_forward_csum(struct sk_buff * skb)4519 static inline void skb_forward_csum(struct sk_buff *skb)
4520 {
4521 	/* Unfortunately we don't support this one.  Any brave souls? */
4522 	if (skb->ip_summed == CHECKSUM_COMPLETE)
4523 		skb->ip_summed = CHECKSUM_NONE;
4524 }
4525 
4526 /**
4527  * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4528  * @skb: skb to check
4529  *
4530  * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4531  * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4532  * use this helper, to document places where we make this assertion.
4533  */
skb_checksum_none_assert(const struct sk_buff * skb)4534 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4535 {
4536 #ifdef DEBUG
4537 	BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4538 #endif
4539 }
4540 
4541 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4542 
4543 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4544 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4545 				     unsigned int transport_len,
4546 				     __sum16(*skb_chkf)(struct sk_buff *skb));
4547 
4548 /**
4549  * skb_head_is_locked - Determine if the skb->head is locked down
4550  * @skb: skb to check
4551  *
4552  * The head on skbs build around a head frag can be removed if they are
4553  * not cloned.  This function returns true if the skb head is locked down
4554  * due to either being allocated via kmalloc, or by being a clone with
4555  * multiple references to the head.
4556  */
skb_head_is_locked(const struct sk_buff * skb)4557 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4558 {
4559 	return !skb->head_frag || skb_cloned(skb);
4560 }
4561 
4562 /* Local Checksum Offload.
4563  * Compute outer checksum based on the assumption that the
4564  * inner checksum will be offloaded later.
4565  * See Documentation/networking/checksum-offloads.rst for
4566  * explanation of how this works.
4567  * Fill in outer checksum adjustment (e.g. with sum of outer
4568  * pseudo-header) before calling.
4569  * Also ensure that inner checksum is in linear data area.
4570  */
lco_csum(struct sk_buff * skb)4571 static inline __wsum lco_csum(struct sk_buff *skb)
4572 {
4573 	unsigned char *csum_start = skb_checksum_start(skb);
4574 	unsigned char *l4_hdr = skb_transport_header(skb);
4575 	__wsum partial;
4576 
4577 	/* Start with complement of inner checksum adjustment */
4578 	partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4579 						    skb->csum_offset));
4580 
4581 	/* Add in checksum of our headers (incl. outer checksum
4582 	 * adjustment filled in by caller) and return result.
4583 	 */
4584 	return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4585 }
4586 
skb_is_redirected(const struct sk_buff * skb)4587 static inline bool skb_is_redirected(const struct sk_buff *skb)
4588 {
4589 #ifdef CONFIG_NET_REDIRECT
4590 	return skb->redirected;
4591 #else
4592 	return false;
4593 #endif
4594 }
4595 
skb_set_redirected(struct sk_buff * skb,bool from_ingress)4596 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4597 {
4598 #ifdef CONFIG_NET_REDIRECT
4599 	skb->redirected = 1;
4600 	skb->from_ingress = from_ingress;
4601 	if (skb->from_ingress)
4602 		skb->tstamp = 0;
4603 #endif
4604 }
4605 
skb_reset_redirect(struct sk_buff * skb)4606 static inline void skb_reset_redirect(struct sk_buff *skb)
4607 {
4608 #ifdef CONFIG_NET_REDIRECT
4609 	skb->redirected = 0;
4610 #endif
4611 }
4612 
4613 #if IS_ENABLED(CONFIG_KCOV) && IS_ENABLED(CONFIG_SKB_EXTENSIONS)
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)4614 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4615 				       const u64 kcov_handle)
4616 {
4617 	/* Do not allocate skb extensions only to set kcov_handle to zero
4618 	 * (as it is zero by default). However, if the extensions are
4619 	 * already allocated, update kcov_handle anyway since
4620 	 * skb_set_kcov_handle can be called to zero a previously set
4621 	 * value.
4622 	 */
4623 	if (skb_has_extensions(skb) || kcov_handle) {
4624 		u64 *kcov_handle_ptr = skb_ext_add(skb, SKB_EXT_KCOV_HANDLE);
4625 
4626 		if (kcov_handle_ptr)
4627 			*kcov_handle_ptr = kcov_handle;
4628 	}
4629 }
4630 
skb_get_kcov_handle(struct sk_buff * skb)4631 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4632 {
4633 	u64 *kcov_handle = skb_ext_find(skb, SKB_EXT_KCOV_HANDLE);
4634 
4635 	return kcov_handle ? *kcov_handle : 0;
4636 }
4637 #else
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)4638 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4639 				       const u64 kcov_handle) { }
skb_get_kcov_handle(struct sk_buff * skb)4640 static inline u64 skb_get_kcov_handle(struct sk_buff *skb) { return 0; }
4641 #endif /* CONFIG_KCOV && CONFIG_SKB_EXTENSIONS */
4642 
4643 #endif	/* __KERNEL__ */
4644 #endif	/* _LINUX_SKBUFF_H */
4645