• 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 	u8			sabotage_in_done:1;
262 	__u16			frag_max_size;
263 	struct net_device	*physindev;
264 
265 	/* always valid & non-NULL from FORWARD on, for physdev match */
266 	struct net_device	*physoutdev;
267 	union {
268 		/* prerouting: detect dnat in orig/reply direction */
269 		__be32          ipv4_daddr;
270 		struct in6_addr ipv6_daddr;
271 
272 		/* after prerouting + nat detected: store original source
273 		 * mac since neigh resolution overwrites it, only used while
274 		 * skb is out in neigh layer.
275 		 */
276 		char neigh_header[8];
277 	};
278 };
279 #endif
280 
281 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
282 /* Chain in tc_skb_ext will be used to share the tc chain with
283  * ovs recirc_id. It will be set to the current chain by tc
284  * and read by ovs to recirc_id.
285  */
286 struct tc_skb_ext {
287 	__u32 chain;
288 	__u16 mru;
289 };
290 #endif
291 
292 struct sk_buff_head {
293 	/* These two members must be first. */
294 	struct sk_buff	*next;
295 	struct sk_buff	*prev;
296 
297 	__u32		qlen;
298 	spinlock_t	lock;
299 };
300 
301 struct sk_buff;
302 
303 /* To allow 64K frame to be packed as single skb without frag_list we
304  * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
305  * buffers which do not start on a page boundary.
306  *
307  * Since GRO uses frags we allocate at least 16 regardless of page
308  * size.
309  */
310 #if (65536/PAGE_SIZE + 1) < 16
311 #define MAX_SKB_FRAGS 16UL
312 #else
313 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
314 #endif
315 extern int sysctl_max_skb_frags;
316 
317 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
318  * segment using its current segmentation instead.
319  */
320 #define GSO_BY_FRAGS	0xFFFF
321 
322 typedef struct bio_vec skb_frag_t;
323 
324 /**
325  * skb_frag_size() - Returns the size of a skb fragment
326  * @frag: skb fragment
327  */
skb_frag_size(const skb_frag_t * frag)328 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329 {
330 	return frag->bv_len;
331 }
332 
333 /**
334  * skb_frag_size_set() - Sets the size of a skb fragment
335  * @frag: skb fragment
336  * @size: size of fragment
337  */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)338 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
339 {
340 	frag->bv_len = size;
341 }
342 
343 /**
344  * skb_frag_size_add() - Increments the size of a skb fragment by @delta
345  * @frag: skb fragment
346  * @delta: value to add
347  */
skb_frag_size_add(skb_frag_t * frag,int delta)348 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
349 {
350 	frag->bv_len += delta;
351 }
352 
353 /**
354  * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
355  * @frag: skb fragment
356  * @delta: value to subtract
357  */
skb_frag_size_sub(skb_frag_t * frag,int delta)358 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
359 {
360 	frag->bv_len -= delta;
361 }
362 
363 /**
364  * skb_frag_must_loop - Test if %p is a high memory page
365  * @p: fragment's page
366  */
skb_frag_must_loop(struct page * p)367 static inline bool skb_frag_must_loop(struct page *p)
368 {
369 #if defined(CONFIG_HIGHMEM)
370 	if (PageHighMem(p))
371 		return true;
372 #endif
373 	return false;
374 }
375 
376 /**
377  *	skb_frag_foreach_page - loop over pages in a fragment
378  *
379  *	@f:		skb frag to operate on
380  *	@f_off:		offset from start of f->bv_page
381  *	@f_len:		length from f_off to loop over
382  *	@p:		(temp var) current page
383  *	@p_off:		(temp var) offset from start of current page,
384  *	                           non-zero only on first page.
385  *	@p_len:		(temp var) length in current page,
386  *				   < PAGE_SIZE only on first and last page.
387  *	@copied:	(temp var) length so far, excluding current p_len.
388  *
389  *	A fragment can hold a compound page, in which case per-page
390  *	operations, notably kmap_atomic, must be called for each
391  *	regular page.
392  */
393 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied)	\
394 	for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT),		\
395 	     p_off = (f_off) & (PAGE_SIZE - 1),				\
396 	     p_len = skb_frag_must_loop(p) ?				\
397 	     min_t(u32, f_len, PAGE_SIZE - p_off) : f_len,		\
398 	     copied = 0;						\
399 	     copied < f_len;						\
400 	     copied += p_len, p++, p_off = 0,				\
401 	     p_len = min_t(u32, f_len - copied, PAGE_SIZE))		\
402 
403 #define HAVE_HW_TIME_STAMP
404 
405 /**
406  * struct skb_shared_hwtstamps - hardware time stamps
407  * @hwtstamp:	hardware time stamp transformed into duration
408  *		since arbitrary point in time
409  *
410  * Software time stamps generated by ktime_get_real() are stored in
411  * skb->tstamp.
412  *
413  * hwtstamps can only be compared against other hwtstamps from
414  * the same device.
415  *
416  * This structure is attached to packets as part of the
417  * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
418  */
419 struct skb_shared_hwtstamps {
420 	ktime_t	hwtstamp;
421 };
422 
423 /* Definitions for tx_flags in struct skb_shared_info */
424 enum {
425 	/* generate hardware time stamp */
426 	SKBTX_HW_TSTAMP = 1 << 0,
427 
428 	/* generate software time stamp when queueing packet to NIC */
429 	SKBTX_SW_TSTAMP = 1 << 1,
430 
431 	/* device driver is going to provide hardware time stamp */
432 	SKBTX_IN_PROGRESS = 1 << 2,
433 
434 	/* device driver supports TX zero-copy buffers */
435 	SKBTX_DEV_ZEROCOPY = 1 << 3,
436 
437 	/* generate wifi status information (where possible) */
438 	SKBTX_WIFI_STATUS = 1 << 4,
439 
440 	/* This indicates at least one fragment might be overwritten
441 	 * (as in vmsplice(), sendfile() ...)
442 	 * If we need to compute a TX checksum, we'll need to copy
443 	 * all frags to avoid possible bad checksum
444 	 */
445 	SKBTX_SHARED_FRAG = 1 << 5,
446 
447 	/* generate software time stamp when entering packet scheduling */
448 	SKBTX_SCHED_TSTAMP = 1 << 6,
449 };
450 
451 #define SKBTX_ZEROCOPY_FRAG	(SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
452 #define SKBTX_ANY_SW_TSTAMP	(SKBTX_SW_TSTAMP    | \
453 				 SKBTX_SCHED_TSTAMP)
454 #define SKBTX_ANY_TSTAMP	(SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
455 
456 /*
457  * The callback notifies userspace to release buffers when skb DMA is done in
458  * lower device, the skb last reference should be 0 when calling this.
459  * The zerocopy_success argument is true if zero copy transmit occurred,
460  * false on data copy or out of memory error caused by data copy attempt.
461  * The ctx field is used to track device context.
462  * The desc field is used to track userspace buffer index.
463  */
464 struct ubuf_info {
465 	void (*callback)(struct ubuf_info *, bool zerocopy_success);
466 	union {
467 		struct {
468 			unsigned long desc;
469 			void *ctx;
470 		};
471 		struct {
472 			u32 id;
473 			u16 len;
474 			u16 zerocopy:1;
475 			u32 bytelen;
476 		};
477 	};
478 	refcount_t refcnt;
479 
480 	struct mmpin {
481 		struct user_struct *user;
482 		unsigned int num_pg;
483 	} mmp;
484 };
485 
486 #define skb_uarg(SKB)	((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
487 
488 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
489 void mm_unaccount_pinned_pages(struct mmpin *mmp);
490 
491 struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
492 struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
493 					struct ubuf_info *uarg);
494 
sock_zerocopy_get(struct ubuf_info * uarg)495 static inline void sock_zerocopy_get(struct ubuf_info *uarg)
496 {
497 	refcount_inc(&uarg->refcnt);
498 }
499 
500 void sock_zerocopy_put(struct ubuf_info *uarg);
501 void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
502 
503 void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
504 
505 int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
506 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
507 			     struct msghdr *msg, int len,
508 			     struct ubuf_info *uarg);
509 
510 /* This data is invariant across clones and lives at
511  * the end of the header data, ie. at skb->end.
512  */
513 struct skb_shared_info {
514 	__u8		__unused;
515 	__u8		meta_len;
516 	__u8		nr_frags;
517 	__u8		tx_flags;
518 	unsigned short	gso_size;
519 	/* Warning: this field is not always filled in (UFO)! */
520 	unsigned short	gso_segs;
521 	struct sk_buff	*frag_list;
522 	struct skb_shared_hwtstamps hwtstamps;
523 	unsigned int	gso_type;
524 	u32		tskey;
525 
526 	/*
527 	 * Warning : all fields before dataref are cleared in __alloc_skb()
528 	 */
529 	atomic_t	dataref;
530 
531 	/* Intermediate layers must ensure that destructor_arg
532 	 * remains valid until skb destructor */
533 	void *		destructor_arg;
534 
535 	/* must be last field, see pskb_expand_head() */
536 	skb_frag_t	frags[MAX_SKB_FRAGS];
537 };
538 
539 /* We divide dataref into two halves.  The higher 16 bits hold references
540  * to the payload part of skb->data.  The lower 16 bits hold references to
541  * the entire skb->data.  A clone of a headerless skb holds the length of
542  * the header in skb->hdr_len.
543  *
544  * All users must obey the rule that the skb->data reference count must be
545  * greater than or equal to the payload reference count.
546  *
547  * Holding a reference to the payload part means that the user does not
548  * care about modifications to the header part of skb->data.
549  */
550 #define SKB_DATAREF_SHIFT 16
551 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
552 
553 
554 enum {
555 	SKB_FCLONE_UNAVAILABLE,	/* skb has no fclone (from head_cache) */
556 	SKB_FCLONE_ORIG,	/* orig skb (from fclone_cache) */
557 	SKB_FCLONE_CLONE,	/* companion fclone skb (from fclone_cache) */
558 };
559 
560 enum {
561 	SKB_GSO_TCPV4 = 1 << 0,
562 
563 	/* This indicates the skb is from an untrusted source. */
564 	SKB_GSO_DODGY = 1 << 1,
565 
566 	/* This indicates the tcp segment has CWR set. */
567 	SKB_GSO_TCP_ECN = 1 << 2,
568 
569 	SKB_GSO_TCP_FIXEDID = 1 << 3,
570 
571 	SKB_GSO_TCPV6 = 1 << 4,
572 
573 	SKB_GSO_FCOE = 1 << 5,
574 
575 	SKB_GSO_GRE = 1 << 6,
576 
577 	SKB_GSO_GRE_CSUM = 1 << 7,
578 
579 	SKB_GSO_IPXIP4 = 1 << 8,
580 
581 	SKB_GSO_IPXIP6 = 1 << 9,
582 
583 	SKB_GSO_UDP_TUNNEL = 1 << 10,
584 
585 	SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
586 
587 	SKB_GSO_PARTIAL = 1 << 12,
588 
589 	SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
590 
591 	SKB_GSO_SCTP = 1 << 14,
592 
593 	SKB_GSO_ESP = 1 << 15,
594 
595 	SKB_GSO_UDP = 1 << 16,
596 
597 	SKB_GSO_UDP_L4 = 1 << 17,
598 
599 	SKB_GSO_FRAGLIST = 1 << 18,
600 };
601 
602 #if BITS_PER_LONG > 32
603 #define NET_SKBUFF_DATA_USES_OFFSET 1
604 #endif
605 
606 #ifdef NET_SKBUFF_DATA_USES_OFFSET
607 typedef unsigned int sk_buff_data_t;
608 #else
609 typedef unsigned char *sk_buff_data_t;
610 #endif
611 
612 /**
613  *	struct sk_buff - socket buffer
614  *	@next: Next buffer in list
615  *	@prev: Previous buffer in list
616  *	@tstamp: Time we arrived/left
617  *	@skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
618  *		for retransmit timer
619  *	@rbnode: RB tree node, alternative to next/prev for netem/tcp
620  *	@list: queue head
621  *	@sk: Socket we are owned by
622  *	@ip_defrag_offset: (aka @sk) alternate use of @sk, used in
623  *		fragmentation management
624  *	@dev: Device we arrived on/are leaving by
625  *	@dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
626  *	@cb: Control buffer. Free for use by every layer. Put private vars here
627  *	@_skb_refdst: destination entry (with norefcount bit)
628  *	@sp: the security path, used for xfrm
629  *	@len: Length of actual data
630  *	@data_len: Data length
631  *	@mac_len: Length of link layer header
632  *	@hdr_len: writable header length of cloned skb
633  *	@csum: Checksum (must include start/offset pair)
634  *	@csum_start: Offset from skb->head where checksumming should start
635  *	@csum_offset: Offset from csum_start where checksum should be stored
636  *	@priority: Packet queueing priority
637  *	@ignore_df: allow local fragmentation
638  *	@cloned: Head may be cloned (check refcnt to be sure)
639  *	@ip_summed: Driver fed us an IP checksum
640  *	@nohdr: Payload reference only, must not modify header
641  *	@pkt_type: Packet class
642  *	@fclone: skbuff clone status
643  *	@ipvs_property: skbuff is owned by ipvs
644  *	@inner_protocol_type: whether the inner protocol is
645  *		ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
646  *	@remcsum_offload: remote checksum offload is enabled
647  *	@offload_fwd_mark: Packet was L2-forwarded in hardware
648  *	@offload_l3_fwd_mark: Packet was L3-forwarded in hardware
649  *	@tc_skip_classify: do not classify packet. set by IFB device
650  *	@tc_at_ingress: used within tc_classify to distinguish in/egress
651  *	@redirected: packet was redirected by packet classifier
652  *	@from_ingress: packet was redirected from the ingress path
653  *	@peeked: this packet has been seen already, so stats have been
654  *		done for it, don't do them again
655  *	@nf_trace: netfilter packet trace flag
656  *	@protocol: Packet protocol from driver
657  *	@destructor: Destruct function
658  *	@tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
659  *	@_nfct: Associated connection, if any (with nfctinfo bits)
660  *	@nf_bridge: Saved data about a bridged frame - see br_netfilter.c
661  *	@skb_iif: ifindex of device we arrived on
662  *	@tc_index: Traffic control index
663  *	@hash: the packet hash
664  *	@queue_mapping: Queue mapping for multiqueue devices
665  *	@head_frag: skb was allocated from page fragments,
666  *		not allocated by kmalloc() or vmalloc().
667  *	@pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
668  *	@active_extensions: active extensions (skb_ext_id types)
669  *	@ndisc_nodetype: router type (from link layer)
670  *	@ooo_okay: allow the mapping of a socket to a queue to be changed
671  *	@l4_hash: indicate hash is a canonical 4-tuple hash over transport
672  *		ports.
673  *	@sw_hash: indicates hash was computed in software stack
674  *	@wifi_acked_valid: wifi_acked was set
675  *	@wifi_acked: whether frame was acked on wifi or not
676  *	@no_fcs:  Request NIC to treat last 4 bytes as Ethernet FCS
677  *	@encapsulation: indicates the inner headers in the skbuff are valid
678  *	@encap_hdr_csum: software checksum is needed
679  *	@csum_valid: checksum is already valid
680  *	@csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
681  *	@csum_complete_sw: checksum was completed by software
682  *	@csum_level: indicates the number of consecutive checksums found in
683  *		the packet minus one that have been verified as
684  *		CHECKSUM_UNNECESSARY (max 3)
685  *	@scm_io_uring: SKB holds io_uring registered files
686  *	@dst_pending_confirm: need to confirm neighbour
687  *	@decrypted: Decrypted SKB
688  *	@napi_id: id of the NAPI struct this skb came from
689  *	@sender_cpu: (aka @napi_id) source CPU in XPS
690  *	@secmark: security marking
691  *	@mark: Generic packet mark
692  *	@reserved_tailroom: (aka @mark) number of bytes of free space available
693  *		at the tail of an sk_buff
694  *	@vlan_present: VLAN tag is present
695  *	@vlan_proto: vlan encapsulation protocol
696  *	@vlan_tci: vlan tag control information
697  *	@inner_protocol: Protocol (encapsulation)
698  *	@inner_ipproto: (aka @inner_protocol) stores ipproto when
699  *		skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
700  *	@inner_transport_header: Inner transport layer header (encapsulation)
701  *	@inner_network_header: Network layer header (encapsulation)
702  *	@inner_mac_header: Link layer header (encapsulation)
703  *	@transport_header: Transport layer header
704  *	@network_header: Network layer header
705  *	@mac_header: Link layer header
706  *	@kcov_handle: KCOV remote handle for remote coverage collection
707  *	@tail: Tail pointer
708  *	@end: End pointer
709  *	@head: Head of buffer
710  *	@data: Data head pointer
711  *	@truesize: Buffer size
712  *	@users: User count - see {datagram,tcp}.c
713  *	@extensions: allocated extensions, valid if active_extensions is nonzero
714  */
715 
716 struct sk_buff {
717 	union {
718 		struct {
719 			/* These two members must be first. */
720 			struct sk_buff		*next;
721 			struct sk_buff		*prev;
722 
723 			union {
724 				struct net_device	*dev;
725 				/* Some protocols might use this space to store information,
726 				 * while device pointer would be NULL.
727 				 * UDP receive path is one user.
728 				 */
729 				unsigned long		dev_scratch;
730 			};
731 		};
732 		struct rb_node		rbnode; /* used in netem, ip4 defrag, and tcp stack */
733 		struct list_head	list;
734 	};
735 
736 	union {
737 		struct sock		*sk;
738 		int			ip_defrag_offset;
739 	};
740 
741 	union {
742 		ktime_t		tstamp;
743 		u64		skb_mstamp_ns; /* earliest departure time */
744 	};
745 	/*
746 	 * This is the control buffer. It is free to use for every
747 	 * layer. Please put your private variables there. If you
748 	 * want to keep them across layers you have to do a skb_clone()
749 	 * first. This is owned by whoever has the skb queued ATM.
750 	 */
751 	char			cb[48] __aligned(8);
752 
753 	union {
754 		struct {
755 			unsigned long	_skb_refdst;
756 			void		(*destructor)(struct sk_buff *skb);
757 		};
758 		struct list_head	tcp_tsorted_anchor;
759 	};
760 
761 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
762 	unsigned long		 _nfct;
763 #endif
764 	unsigned int		len,
765 				data_len;
766 	__u16			mac_len,
767 				hdr_len;
768 
769 	/* Following fields are _not_ copied in __copy_skb_header()
770 	 * Note that queue_mapping is here mostly to fill a hole.
771 	 */
772 	__u16			queue_mapping;
773 
774 /* if you move cloned around you also must adapt those constants */
775 #ifdef __BIG_ENDIAN_BITFIELD
776 #define CLONED_MASK	(1 << 7)
777 #else
778 #define CLONED_MASK	1
779 #endif
780 #define CLONED_OFFSET()		offsetof(struct sk_buff, __cloned_offset)
781 
782 	/* private: */
783 	__u8			__cloned_offset[0];
784 	/* public: */
785 	__u8			cloned:1,
786 				nohdr:1,
787 				fclone:2,
788 				peeked:1,
789 				head_frag:1,
790 				pfmemalloc:1;
791 #ifdef CONFIG_SKB_EXTENSIONS
792 	__u8			active_extensions;
793 #endif
794 	/* fields enclosed in headers_start/headers_end are copied
795 	 * using a single memcpy() in __copy_skb_header()
796 	 */
797 	/* private: */
798 	__u32			headers_start[0];
799 	/* public: */
800 
801 /* if you move pkt_type around you also must adapt those constants */
802 #ifdef __BIG_ENDIAN_BITFIELD
803 #define PKT_TYPE_MAX	(7 << 5)
804 #else
805 #define PKT_TYPE_MAX	7
806 #endif
807 #define PKT_TYPE_OFFSET()	offsetof(struct sk_buff, __pkt_type_offset)
808 
809 	/* private: */
810 	__u8			__pkt_type_offset[0];
811 	/* public: */
812 	__u8			pkt_type:3;
813 	__u8			ignore_df:1;
814 	__u8			nf_trace:1;
815 	__u8			ip_summed:2;
816 	__u8			ooo_okay:1;
817 
818 	__u8			l4_hash:1;
819 	__u8			sw_hash:1;
820 	__u8			wifi_acked_valid:1;
821 	__u8			wifi_acked:1;
822 	__u8			no_fcs:1;
823 	/* Indicates the inner headers are valid in the skbuff. */
824 	__u8			encapsulation:1;
825 	__u8			encap_hdr_csum:1;
826 	__u8			csum_valid:1;
827 
828 #ifdef __BIG_ENDIAN_BITFIELD
829 #define PKT_VLAN_PRESENT_BIT	7
830 #else
831 #define PKT_VLAN_PRESENT_BIT	0
832 #endif
833 #define PKT_VLAN_PRESENT_OFFSET()	offsetof(struct sk_buff, __pkt_vlan_present_offset)
834 	/* private: */
835 	__u8			__pkt_vlan_present_offset[0];
836 	/* public: */
837 	__u8			vlan_present:1;
838 	__u8			csum_complete_sw:1;
839 	__u8			csum_level:2;
840 	__u8			csum_not_inet:1;
841 	__u8			dst_pending_confirm:1;
842 #ifdef CONFIG_IPV6_NDISC_NODETYPE
843 	__u8			ndisc_nodetype:2;
844 #endif
845 
846 	__u8			ipvs_property:1;
847 	__u8			inner_protocol_type:1;
848 	__u8			remcsum_offload:1;
849 #ifdef CONFIG_NET_SWITCHDEV
850 	__u8			offload_fwd_mark:1;
851 	__u8			offload_l3_fwd_mark:1;
852 #endif
853 #ifdef CONFIG_NET_CLS_ACT
854 	__u8			tc_skip_classify:1;
855 	__u8			tc_at_ingress:1;
856 #endif
857 #ifdef CONFIG_NET_REDIRECT
858 	__u8			redirected:1;
859 	__u8			from_ingress:1;
860 #endif
861 #ifdef CONFIG_TLS_DEVICE
862 	__u8			decrypted:1;
863 #endif
864 	__u8			scm_io_uring:1;
865 
866 #ifdef CONFIG_NET_SCHED
867 	__u16			tc_index;	/* traffic control index */
868 #endif
869 
870 	union {
871 		__wsum		csum;
872 		struct {
873 			__u16	csum_start;
874 			__u16	csum_offset;
875 		};
876 	};
877 	__u32			priority;
878 	int			skb_iif;
879 	__u32			hash;
880 	__be16			vlan_proto;
881 	__u16			vlan_tci;
882 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
883 	union {
884 		unsigned int	napi_id;
885 		unsigned int	sender_cpu;
886 	};
887 #endif
888 #ifdef CONFIG_NETWORK_SECMARK
889 	__u32		secmark;
890 #endif
891 
892 	union {
893 		__u32		mark;
894 		__u32		reserved_tailroom;
895 	};
896 
897 	union {
898 		__be16		inner_protocol;
899 		__u8		inner_ipproto;
900 	};
901 
902 	__u16			inner_transport_header;
903 	__u16			inner_network_header;
904 	__u16			inner_mac_header;
905 
906 	__be16			protocol;
907 	__u16			transport_header;
908 	__u16			network_header;
909 	__u16			mac_header;
910 
911 #ifdef CONFIG_KCOV
912 	u64			kcov_handle;
913 #endif
914 
915 	/* private: */
916 	__u32			headers_end[0];
917 	/* public: */
918 
919 	/* These elements must be at the end, see alloc_skb() for details.  */
920 	sk_buff_data_t		tail;
921 	sk_buff_data_t		end;
922 	unsigned char		*head,
923 				*data;
924 	unsigned int		truesize;
925 	refcount_t		users;
926 
927 #ifdef CONFIG_SKB_EXTENSIONS
928 	/* only useable after checking ->active_extensions != 0 */
929 	struct skb_ext		*extensions;
930 #endif
931 };
932 
933 #ifdef __KERNEL__
934 /*
935  *	Handling routines are only of interest to the kernel
936  */
937 
938 #define SKB_ALLOC_FCLONE	0x01
939 #define SKB_ALLOC_RX		0x02
940 #define SKB_ALLOC_NAPI		0x04
941 
942 /**
943  * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
944  * @skb: buffer
945  */
skb_pfmemalloc(const struct sk_buff * skb)946 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
947 {
948 	return unlikely(skb->pfmemalloc);
949 }
950 
951 /*
952  * skb might have a dst pointer attached, refcounted or not.
953  * _skb_refdst low order bit is set if refcount was _not_ taken
954  */
955 #define SKB_DST_NOREF	1UL
956 #define SKB_DST_PTRMASK	~(SKB_DST_NOREF)
957 
958 /**
959  * skb_dst - returns skb dst_entry
960  * @skb: buffer
961  *
962  * Returns skb dst_entry, regardless of reference taken or not.
963  */
skb_dst(const struct sk_buff * skb)964 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
965 {
966 	/* If refdst was not refcounted, check we still are in a
967 	 * rcu_read_lock section
968 	 */
969 	WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
970 		!rcu_read_lock_held() &&
971 		!rcu_read_lock_bh_held());
972 	return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
973 }
974 
975 /**
976  * skb_dst_set - sets skb dst
977  * @skb: buffer
978  * @dst: dst entry
979  *
980  * Sets skb dst, assuming a reference was taken on dst and should
981  * be released by skb_dst_drop()
982  */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)983 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
984 {
985 	skb->_skb_refdst = (unsigned long)dst;
986 }
987 
988 /**
989  * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
990  * @skb: buffer
991  * @dst: dst entry
992  *
993  * Sets skb dst, assuming a reference was not taken on dst.
994  * If dst entry is cached, we do not take reference and dst_release
995  * will be avoided by refdst_drop. If dst entry is not cached, we take
996  * reference, so that last dst_release can destroy the dst immediately.
997  */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)998 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
999 {
1000 	WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1001 	skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1002 }
1003 
1004 /**
1005  * skb_dst_is_noref - Test if skb dst isn't refcounted
1006  * @skb: buffer
1007  */
skb_dst_is_noref(const struct sk_buff * skb)1008 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1009 {
1010 	return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1011 }
1012 
1013 /**
1014  * skb_rtable - Returns the skb &rtable
1015  * @skb: buffer
1016  */
skb_rtable(const struct sk_buff * skb)1017 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1018 {
1019 	return (struct rtable *)skb_dst(skb);
1020 }
1021 
1022 /* For mangling skb->pkt_type from user space side from applications
1023  * such as nft, tc, etc, we only allow a conservative subset of
1024  * possible pkt_types to be set.
1025 */
skb_pkt_type_ok(u32 ptype)1026 static inline bool skb_pkt_type_ok(u32 ptype)
1027 {
1028 	return ptype <= PACKET_OTHERHOST;
1029 }
1030 
1031 /**
1032  * skb_napi_id - Returns the skb's NAPI id
1033  * @skb: buffer
1034  */
skb_napi_id(const struct sk_buff * skb)1035 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1036 {
1037 #ifdef CONFIG_NET_RX_BUSY_POLL
1038 	return skb->napi_id;
1039 #else
1040 	return 0;
1041 #endif
1042 }
1043 
1044 /**
1045  * skb_unref - decrement the skb's reference count
1046  * @skb: buffer
1047  *
1048  * Returns true if we can free the skb.
1049  */
skb_unref(struct sk_buff * skb)1050 static inline bool skb_unref(struct sk_buff *skb)
1051 {
1052 	if (unlikely(!skb))
1053 		return false;
1054 	if (likely(refcount_read(&skb->users) == 1))
1055 		smp_rmb();
1056 	else if (likely(!refcount_dec_and_test(&skb->users)))
1057 		return false;
1058 
1059 	return true;
1060 }
1061 
1062 void skb_release_head_state(struct sk_buff *skb);
1063 void kfree_skb(struct sk_buff *skb);
1064 void kfree_skb_list(struct sk_buff *segs);
1065 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1066 void skb_tx_error(struct sk_buff *skb);
1067 
1068 #ifdef CONFIG_TRACEPOINTS
1069 void consume_skb(struct sk_buff *skb);
1070 #else
consume_skb(struct sk_buff * skb)1071 static inline void consume_skb(struct sk_buff *skb)
1072 {
1073 	return kfree_skb(skb);
1074 }
1075 #endif
1076 
1077 void __consume_stateless_skb(struct sk_buff *skb);
1078 void  __kfree_skb(struct sk_buff *skb);
1079 extern struct kmem_cache *skbuff_head_cache;
1080 
1081 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1082 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1083 		      bool *fragstolen, int *delta_truesize);
1084 
1085 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1086 			    int node);
1087 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1088 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1089 struct sk_buff *build_skb_around(struct sk_buff *skb,
1090 				 void *data, unsigned int frag_size);
1091 
1092 /**
1093  * alloc_skb - allocate a network buffer
1094  * @size: size to allocate
1095  * @priority: allocation mask
1096  *
1097  * This function is a convenient wrapper around __alloc_skb().
1098  */
alloc_skb(unsigned int size,gfp_t priority)1099 static inline struct sk_buff *alloc_skb(unsigned int size,
1100 					gfp_t priority)
1101 {
1102 	return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1103 }
1104 
1105 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1106 				     unsigned long data_len,
1107 				     int max_page_order,
1108 				     int *errcode,
1109 				     gfp_t gfp_mask);
1110 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1111 
1112 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1113 struct sk_buff_fclones {
1114 	struct sk_buff	skb1;
1115 
1116 	struct sk_buff	skb2;
1117 
1118 	refcount_t	fclone_ref;
1119 };
1120 
1121 /**
1122  *	skb_fclone_busy - check if fclone is busy
1123  *	@sk: socket
1124  *	@skb: buffer
1125  *
1126  * Returns true if skb is a fast clone, and its clone is not freed.
1127  * Some drivers call skb_orphan() in their ndo_start_xmit(),
1128  * so we also check that this didnt happen.
1129  */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1130 static inline bool skb_fclone_busy(const struct sock *sk,
1131 				   const struct sk_buff *skb)
1132 {
1133 	const struct sk_buff_fclones *fclones;
1134 
1135 	fclones = container_of(skb, struct sk_buff_fclones, skb1);
1136 
1137 	return skb->fclone == SKB_FCLONE_ORIG &&
1138 	       refcount_read(&fclones->fclone_ref) > 1 &&
1139 	       fclones->skb2.sk == sk;
1140 }
1141 
1142 /**
1143  * alloc_skb_fclone - allocate a network buffer from fclone cache
1144  * @size: size to allocate
1145  * @priority: allocation mask
1146  *
1147  * This function is a convenient wrapper around __alloc_skb().
1148  */
alloc_skb_fclone(unsigned int size,gfp_t priority)1149 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1150 					       gfp_t priority)
1151 {
1152 	return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1153 }
1154 
1155 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1156 void skb_headers_offset_update(struct sk_buff *skb, int off);
1157 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1158 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1159 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1160 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1161 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1162 				   gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1163 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1164 					  gfp_t gfp_mask)
1165 {
1166 	return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1167 }
1168 
1169 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1170 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1171 				     unsigned int headroom);
1172 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1173 				int newtailroom, gfp_t priority);
1174 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1175 				     int offset, int len);
1176 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1177 			      int offset, int len);
1178 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1179 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1180 
1181 /**
1182  *	skb_pad			-	zero pad the tail of an skb
1183  *	@skb: buffer to pad
1184  *	@pad: space to pad
1185  *
1186  *	Ensure that a buffer is followed by a padding area that is zero
1187  *	filled. Used by network drivers which may DMA or transfer data
1188  *	beyond the buffer end onto the wire.
1189  *
1190  *	May return error in out of memory cases. The skb is freed on error.
1191  */
skb_pad(struct sk_buff * skb,int pad)1192 static inline int skb_pad(struct sk_buff *skb, int pad)
1193 {
1194 	return __skb_pad(skb, pad, true);
1195 }
1196 #define dev_kfree_skb(a)	consume_skb(a)
1197 
1198 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1199 			 int offset, size_t size);
1200 
1201 struct skb_seq_state {
1202 	__u32		lower_offset;
1203 	__u32		upper_offset;
1204 	__u32		frag_idx;
1205 	__u32		stepped_offset;
1206 	struct sk_buff	*root_skb;
1207 	struct sk_buff	*cur_skb;
1208 	__u8		*frag_data;
1209 };
1210 
1211 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1212 			  unsigned int to, struct skb_seq_state *st);
1213 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1214 			  struct skb_seq_state *st);
1215 void skb_abort_seq_read(struct skb_seq_state *st);
1216 
1217 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1218 			   unsigned int to, struct ts_config *config);
1219 
1220 /*
1221  * Packet hash types specify the type of hash in skb_set_hash.
1222  *
1223  * Hash types refer to the protocol layer addresses which are used to
1224  * construct a packet's hash. The hashes are used to differentiate or identify
1225  * flows of the protocol layer for the hash type. Hash types are either
1226  * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1227  *
1228  * Properties of hashes:
1229  *
1230  * 1) Two packets in different flows have different hash values
1231  * 2) Two packets in the same flow should have the same hash value
1232  *
1233  * A hash at a higher layer is considered to be more specific. A driver should
1234  * set the most specific hash possible.
1235  *
1236  * A driver cannot indicate a more specific hash than the layer at which a hash
1237  * was computed. For instance an L3 hash cannot be set as an L4 hash.
1238  *
1239  * A driver may indicate a hash level which is less specific than the
1240  * actual layer the hash was computed on. For instance, a hash computed
1241  * at L4 may be considered an L3 hash. This should only be done if the
1242  * driver can't unambiguously determine that the HW computed the hash at
1243  * the higher layer. Note that the "should" in the second property above
1244  * permits this.
1245  */
1246 enum pkt_hash_types {
1247 	PKT_HASH_TYPE_NONE,	/* Undefined type */
1248 	PKT_HASH_TYPE_L2,	/* Input: src_MAC, dest_MAC */
1249 	PKT_HASH_TYPE_L3,	/* Input: src_IP, dst_IP */
1250 	PKT_HASH_TYPE_L4,	/* Input: src_IP, dst_IP, src_port, dst_port */
1251 };
1252 
skb_clear_hash(struct sk_buff * skb)1253 static inline void skb_clear_hash(struct sk_buff *skb)
1254 {
1255 	skb->hash = 0;
1256 	skb->sw_hash = 0;
1257 	skb->l4_hash = 0;
1258 }
1259 
skb_clear_hash_if_not_l4(struct sk_buff * skb)1260 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1261 {
1262 	if (!skb->l4_hash)
1263 		skb_clear_hash(skb);
1264 }
1265 
1266 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1267 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1268 {
1269 	skb->l4_hash = is_l4;
1270 	skb->sw_hash = is_sw;
1271 	skb->hash = hash;
1272 }
1273 
1274 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1275 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1276 {
1277 	/* Used by drivers to set hash from HW */
1278 	__skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1279 }
1280 
1281 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1282 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1283 {
1284 	__skb_set_hash(skb, hash, true, is_l4);
1285 }
1286 
1287 void __skb_get_hash(struct sk_buff *skb);
1288 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1289 u32 skb_get_poff(const struct sk_buff *skb);
1290 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1291 		   const struct flow_keys_basic *keys, int hlen);
1292 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1293 			    void *data, int hlen_proto);
1294 
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto)1295 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1296 					int thoff, u8 ip_proto)
1297 {
1298 	return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1299 }
1300 
1301 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1302 			     const struct flow_dissector_key *key,
1303 			     unsigned int key_count);
1304 
1305 struct bpf_flow_dissector;
1306 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1307 		      __be16 proto, int nhoff, int hlen, unsigned int flags);
1308 
1309 bool __skb_flow_dissect(const struct net *net,
1310 			const struct sk_buff *skb,
1311 			struct flow_dissector *flow_dissector,
1312 			void *target_container,
1313 			void *data, __be16 proto, int nhoff, int hlen,
1314 			unsigned int flags);
1315 
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1316 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1317 				    struct flow_dissector *flow_dissector,
1318 				    void *target_container, unsigned int flags)
1319 {
1320 	return __skb_flow_dissect(NULL, skb, flow_dissector,
1321 				  target_container, NULL, 0, 0, 0, flags);
1322 }
1323 
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1324 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1325 					      struct flow_keys *flow,
1326 					      unsigned int flags)
1327 {
1328 	memset(flow, 0, sizeof(*flow));
1329 	return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1330 				  flow, NULL, 0, 0, 0, flags);
1331 }
1332 
1333 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)1334 skb_flow_dissect_flow_keys_basic(const struct net *net,
1335 				 const struct sk_buff *skb,
1336 				 struct flow_keys_basic *flow, void *data,
1337 				 __be16 proto, int nhoff, int hlen,
1338 				 unsigned int flags)
1339 {
1340 	memset(flow, 0, sizeof(*flow));
1341 	return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1342 				  data, proto, nhoff, hlen, flags);
1343 }
1344 
1345 void skb_flow_dissect_meta(const struct sk_buff *skb,
1346 			   struct flow_dissector *flow_dissector,
1347 			   void *target_container);
1348 
1349 /* Gets a skb connection tracking info, ctinfo map should be a
1350  * map of mapsize to translate enum ip_conntrack_info states
1351  * to user states.
1352  */
1353 void
1354 skb_flow_dissect_ct(const struct sk_buff *skb,
1355 		    struct flow_dissector *flow_dissector,
1356 		    void *target_container,
1357 		    u16 *ctinfo_map,
1358 		    size_t mapsize);
1359 void
1360 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1361 			     struct flow_dissector *flow_dissector,
1362 			     void *target_container);
1363 
1364 void skb_flow_dissect_hash(const struct sk_buff *skb,
1365 			   struct flow_dissector *flow_dissector,
1366 			   void *target_container);
1367 
skb_get_hash(struct sk_buff * skb)1368 static inline __u32 skb_get_hash(struct sk_buff *skb)
1369 {
1370 	if (!skb->l4_hash && !skb->sw_hash)
1371 		__skb_get_hash(skb);
1372 
1373 	return skb->hash;
1374 }
1375 
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1376 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1377 {
1378 	if (!skb->l4_hash && !skb->sw_hash) {
1379 		struct flow_keys keys;
1380 		__u32 hash = __get_hash_from_flowi6(fl6, &keys);
1381 
1382 		__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1383 	}
1384 
1385 	return skb->hash;
1386 }
1387 
1388 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1389 			   const siphash_key_t *perturb);
1390 
skb_get_hash_raw(const struct sk_buff * skb)1391 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1392 {
1393 	return skb->hash;
1394 }
1395 
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1396 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1397 {
1398 	to->hash = from->hash;
1399 	to->sw_hash = from->sw_hash;
1400 	to->l4_hash = from->l4_hash;
1401 };
1402 
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1403 static inline void skb_copy_decrypted(struct sk_buff *to,
1404 				      const struct sk_buff *from)
1405 {
1406 #ifdef CONFIG_TLS_DEVICE
1407 	to->decrypted = from->decrypted;
1408 #endif
1409 }
1410 
1411 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1412 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1413 {
1414 	return skb->head + skb->end;
1415 }
1416 
skb_end_offset(const struct sk_buff * skb)1417 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1418 {
1419 	return skb->end;
1420 }
1421 #else
skb_end_pointer(const struct sk_buff * skb)1422 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1423 {
1424 	return skb->end;
1425 }
1426 
skb_end_offset(const struct sk_buff * skb)1427 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1428 {
1429 	return skb->end - skb->head;
1430 }
1431 #endif
1432 
1433 /* Internal */
1434 #define skb_shinfo(SKB)	((struct skb_shared_info *)(skb_end_pointer(SKB)))
1435 
skb_hwtstamps(struct sk_buff * skb)1436 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1437 {
1438 	return &skb_shinfo(skb)->hwtstamps;
1439 }
1440 
skb_zcopy(struct sk_buff * skb)1441 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1442 {
1443 	bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1444 
1445 	return is_zcopy ? skb_uarg(skb) : NULL;
1446 }
1447 
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1448 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1449 				 bool *have_ref)
1450 {
1451 	if (skb && uarg && !skb_zcopy(skb)) {
1452 		if (unlikely(have_ref && *have_ref))
1453 			*have_ref = false;
1454 		else
1455 			sock_zerocopy_get(uarg);
1456 		skb_shinfo(skb)->destructor_arg = uarg;
1457 		skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1458 	}
1459 }
1460 
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1461 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1462 {
1463 	skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1464 	skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1465 }
1466 
skb_zcopy_is_nouarg(struct sk_buff * skb)1467 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1468 {
1469 	return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1470 }
1471 
skb_zcopy_get_nouarg(struct sk_buff * skb)1472 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1473 {
1474 	return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1475 }
1476 
1477 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy)1478 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1479 {
1480 	struct ubuf_info *uarg = skb_zcopy(skb);
1481 
1482 	if (uarg) {
1483 		if (skb_zcopy_is_nouarg(skb)) {
1484 			/* no notification callback */
1485 		} else if (uarg->callback == sock_zerocopy_callback) {
1486 			uarg->zerocopy = uarg->zerocopy && zerocopy;
1487 			sock_zerocopy_put(uarg);
1488 		} else {
1489 			uarg->callback(uarg, zerocopy);
1490 		}
1491 
1492 		skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1493 	}
1494 }
1495 
1496 /* Abort a zerocopy operation and revert zckey on error in send syscall */
skb_zcopy_abort(struct sk_buff * skb)1497 static inline void skb_zcopy_abort(struct sk_buff *skb)
1498 {
1499 	struct ubuf_info *uarg = skb_zcopy(skb);
1500 
1501 	if (uarg) {
1502 		sock_zerocopy_put_abort(uarg, false);
1503 		skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1504 	}
1505 }
1506 
skb_mark_not_on_list(struct sk_buff * skb)1507 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1508 {
1509 	skb->next = NULL;
1510 }
1511 
1512 /* Iterate through singly-linked GSO fragments of an skb. */
1513 #define skb_list_walk_safe(first, skb, next_skb)                               \
1514 	for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb);  \
1515 	     (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1516 
skb_list_del_init(struct sk_buff * skb)1517 static inline void skb_list_del_init(struct sk_buff *skb)
1518 {
1519 	__list_del_entry(&skb->list);
1520 	skb_mark_not_on_list(skb);
1521 }
1522 
1523 /**
1524  *	skb_queue_empty - check if a queue is empty
1525  *	@list: queue head
1526  *
1527  *	Returns true if the queue is empty, false otherwise.
1528  */
skb_queue_empty(const struct sk_buff_head * list)1529 static inline int skb_queue_empty(const struct sk_buff_head *list)
1530 {
1531 	return list->next == (const struct sk_buff *) list;
1532 }
1533 
1534 /**
1535  *	skb_queue_empty_lockless - check if a queue is empty
1536  *	@list: queue head
1537  *
1538  *	Returns true if the queue is empty, false otherwise.
1539  *	This variant can be used in lockless contexts.
1540  */
skb_queue_empty_lockless(const struct sk_buff_head * list)1541 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1542 {
1543 	return READ_ONCE(list->next) == (const struct sk_buff *) list;
1544 }
1545 
1546 
1547 /**
1548  *	skb_queue_is_last - check if skb is the last entry in the queue
1549  *	@list: queue head
1550  *	@skb: buffer
1551  *
1552  *	Returns true if @skb is the last buffer on the list.
1553  */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1554 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1555 				     const struct sk_buff *skb)
1556 {
1557 	return skb->next == (const struct sk_buff *) list;
1558 }
1559 
1560 /**
1561  *	skb_queue_is_first - check if skb is the first entry in the queue
1562  *	@list: queue head
1563  *	@skb: buffer
1564  *
1565  *	Returns true if @skb is the first buffer on the list.
1566  */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1567 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1568 				      const struct sk_buff *skb)
1569 {
1570 	return skb->prev == (const struct sk_buff *) list;
1571 }
1572 
1573 /**
1574  *	skb_queue_next - return the next packet in the queue
1575  *	@list: queue head
1576  *	@skb: current buffer
1577  *
1578  *	Return the next packet in @list after @skb.  It is only valid to
1579  *	call this if skb_queue_is_last() evaluates to false.
1580  */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1581 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1582 					     const struct sk_buff *skb)
1583 {
1584 	/* This BUG_ON may seem severe, but if we just return then we
1585 	 * are going to dereference garbage.
1586 	 */
1587 	BUG_ON(skb_queue_is_last(list, skb));
1588 	return skb->next;
1589 }
1590 
1591 /**
1592  *	skb_queue_prev - return the prev packet in the queue
1593  *	@list: queue head
1594  *	@skb: current buffer
1595  *
1596  *	Return the prev packet in @list before @skb.  It is only valid to
1597  *	call this if skb_queue_is_first() evaluates to false.
1598  */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1599 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1600 					     const struct sk_buff *skb)
1601 {
1602 	/* This BUG_ON may seem severe, but if we just return then we
1603 	 * are going to dereference garbage.
1604 	 */
1605 	BUG_ON(skb_queue_is_first(list, skb));
1606 	return skb->prev;
1607 }
1608 
1609 /**
1610  *	skb_get - reference buffer
1611  *	@skb: buffer to reference
1612  *
1613  *	Makes another reference to a socket buffer and returns a pointer
1614  *	to the buffer.
1615  */
skb_get(struct sk_buff * skb)1616 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1617 {
1618 	refcount_inc(&skb->users);
1619 	return skb;
1620 }
1621 
1622 /*
1623  * If users == 1, we are the only owner and can avoid redundant atomic changes.
1624  */
1625 
1626 /**
1627  *	skb_cloned - is the buffer a clone
1628  *	@skb: buffer to check
1629  *
1630  *	Returns true if the buffer was generated with skb_clone() and is
1631  *	one of multiple shared copies of the buffer. Cloned buffers are
1632  *	shared data so must not be written to under normal circumstances.
1633  */
skb_cloned(const struct sk_buff * skb)1634 static inline int skb_cloned(const struct sk_buff *skb)
1635 {
1636 	return skb->cloned &&
1637 	       (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1638 }
1639 
skb_unclone(struct sk_buff * skb,gfp_t pri)1640 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1641 {
1642 	might_sleep_if(gfpflags_allow_blocking(pri));
1643 
1644 	if (skb_cloned(skb))
1645 		return pskb_expand_head(skb, 0, 0, pri);
1646 
1647 	return 0;
1648 }
1649 
1650 /**
1651  *	skb_header_cloned - is the header a clone
1652  *	@skb: buffer to check
1653  *
1654  *	Returns true if modifying the header part of the buffer requires
1655  *	the data to be copied.
1656  */
skb_header_cloned(const struct sk_buff * skb)1657 static inline int skb_header_cloned(const struct sk_buff *skb)
1658 {
1659 	int dataref;
1660 
1661 	if (!skb->cloned)
1662 		return 0;
1663 
1664 	dataref = atomic_read(&skb_shinfo(skb)->dataref);
1665 	dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1666 	return dataref != 1;
1667 }
1668 
skb_header_unclone(struct sk_buff * skb,gfp_t pri)1669 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1670 {
1671 	might_sleep_if(gfpflags_allow_blocking(pri));
1672 
1673 	if (skb_header_cloned(skb))
1674 		return pskb_expand_head(skb, 0, 0, pri);
1675 
1676 	return 0;
1677 }
1678 
1679 /**
1680  *	__skb_header_release - release reference to header
1681  *	@skb: buffer to operate on
1682  */
__skb_header_release(struct sk_buff * skb)1683 static inline void __skb_header_release(struct sk_buff *skb)
1684 {
1685 	skb->nohdr = 1;
1686 	atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1687 }
1688 
1689 
1690 /**
1691  *	skb_shared - is the buffer shared
1692  *	@skb: buffer to check
1693  *
1694  *	Returns true if more than one person has a reference to this
1695  *	buffer.
1696  */
skb_shared(const struct sk_buff * skb)1697 static inline int skb_shared(const struct sk_buff *skb)
1698 {
1699 	return refcount_read(&skb->users) != 1;
1700 }
1701 
1702 /**
1703  *	skb_share_check - check if buffer is shared and if so clone it
1704  *	@skb: buffer to check
1705  *	@pri: priority for memory allocation
1706  *
1707  *	If the buffer is shared the buffer is cloned and the old copy
1708  *	drops a reference. A new clone with a single reference is returned.
1709  *	If the buffer is not shared the original buffer is returned. When
1710  *	being called from interrupt status or with spinlocks held pri must
1711  *	be GFP_ATOMIC.
1712  *
1713  *	NULL is returned on a memory allocation failure.
1714  */
skb_share_check(struct sk_buff * skb,gfp_t pri)1715 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1716 {
1717 	might_sleep_if(gfpflags_allow_blocking(pri));
1718 	if (skb_shared(skb)) {
1719 		struct sk_buff *nskb = skb_clone(skb, pri);
1720 
1721 		if (likely(nskb))
1722 			consume_skb(skb);
1723 		else
1724 			kfree_skb(skb);
1725 		skb = nskb;
1726 	}
1727 	return skb;
1728 }
1729 
1730 /*
1731  *	Copy shared buffers into a new sk_buff. We effectively do COW on
1732  *	packets to handle cases where we have a local reader and forward
1733  *	and a couple of other messy ones. The normal one is tcpdumping
1734  *	a packet thats being forwarded.
1735  */
1736 
1737 /**
1738  *	skb_unshare - make a copy of a shared buffer
1739  *	@skb: buffer to check
1740  *	@pri: priority for memory allocation
1741  *
1742  *	If the socket buffer is a clone then this function creates a new
1743  *	copy of the data, drops a reference count on the old copy and returns
1744  *	the new copy with the reference count at 1. If the buffer is not a clone
1745  *	the original buffer is returned. When called with a spinlock held or
1746  *	from interrupt state @pri must be %GFP_ATOMIC
1747  *
1748  *	%NULL is returned on a memory allocation failure.
1749  */
skb_unshare(struct sk_buff * skb,gfp_t pri)1750 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1751 					  gfp_t pri)
1752 {
1753 	might_sleep_if(gfpflags_allow_blocking(pri));
1754 	if (skb_cloned(skb)) {
1755 		struct sk_buff *nskb = skb_copy(skb, pri);
1756 
1757 		/* Free our shared copy */
1758 		if (likely(nskb))
1759 			consume_skb(skb);
1760 		else
1761 			kfree_skb(skb);
1762 		skb = nskb;
1763 	}
1764 	return skb;
1765 }
1766 
1767 /**
1768  *	skb_peek - peek at the head of an &sk_buff_head
1769  *	@list_: list to peek at
1770  *
1771  *	Peek an &sk_buff. Unlike most other operations you _MUST_
1772  *	be careful with this one. A peek leaves the buffer on the
1773  *	list and someone else may run off with it. You must hold
1774  *	the appropriate locks or have a private queue to do this.
1775  *
1776  *	Returns %NULL for an empty list or a pointer to the head element.
1777  *	The reference count is not incremented and the reference is therefore
1778  *	volatile. Use with caution.
1779  */
skb_peek(const struct sk_buff_head * list_)1780 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1781 {
1782 	struct sk_buff *skb = list_->next;
1783 
1784 	if (skb == (struct sk_buff *)list_)
1785 		skb = NULL;
1786 	return skb;
1787 }
1788 
1789 /**
1790  *	__skb_peek - peek at the head of a non-empty &sk_buff_head
1791  *	@list_: list to peek at
1792  *
1793  *	Like skb_peek(), but the caller knows that the list is not empty.
1794  */
__skb_peek(const struct sk_buff_head * list_)1795 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1796 {
1797 	return list_->next;
1798 }
1799 
1800 /**
1801  *	skb_peek_next - peek skb following the given one from a queue
1802  *	@skb: skb to start from
1803  *	@list_: list to peek at
1804  *
1805  *	Returns %NULL when the end of the list is met or a pointer to the
1806  *	next element. The reference count is not incremented and the
1807  *	reference is therefore volatile. Use with caution.
1808  */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)1809 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1810 		const struct sk_buff_head *list_)
1811 {
1812 	struct sk_buff *next = skb->next;
1813 
1814 	if (next == (struct sk_buff *)list_)
1815 		next = NULL;
1816 	return next;
1817 }
1818 
1819 /**
1820  *	skb_peek_tail - peek at the tail of an &sk_buff_head
1821  *	@list_: list to peek at
1822  *
1823  *	Peek an &sk_buff. Unlike most other operations you _MUST_
1824  *	be careful with this one. A peek leaves the buffer on the
1825  *	list and someone else may run off with it. You must hold
1826  *	the appropriate locks or have a private queue to do this.
1827  *
1828  *	Returns %NULL for an empty list or a pointer to the tail element.
1829  *	The reference count is not incremented and the reference is therefore
1830  *	volatile. Use with caution.
1831  */
skb_peek_tail(const struct sk_buff_head * list_)1832 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1833 {
1834 	struct sk_buff *skb = READ_ONCE(list_->prev);
1835 
1836 	if (skb == (struct sk_buff *)list_)
1837 		skb = NULL;
1838 	return skb;
1839 
1840 }
1841 
1842 /**
1843  *	skb_queue_len	- get queue length
1844  *	@list_: list to measure
1845  *
1846  *	Return the length of an &sk_buff queue.
1847  */
skb_queue_len(const struct sk_buff_head * list_)1848 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1849 {
1850 	return list_->qlen;
1851 }
1852 
1853 /**
1854  *	skb_queue_len_lockless	- get queue length
1855  *	@list_: list to measure
1856  *
1857  *	Return the length of an &sk_buff queue.
1858  *	This variant can be used in lockless contexts.
1859  */
skb_queue_len_lockless(const struct sk_buff_head * list_)1860 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1861 {
1862 	return READ_ONCE(list_->qlen);
1863 }
1864 
1865 /**
1866  *	__skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1867  *	@list: queue to initialize
1868  *
1869  *	This initializes only the list and queue length aspects of
1870  *	an sk_buff_head object.  This allows to initialize the list
1871  *	aspects of an sk_buff_head without reinitializing things like
1872  *	the spinlock.  It can also be used for on-stack sk_buff_head
1873  *	objects where the spinlock is known to not be used.
1874  */
__skb_queue_head_init(struct sk_buff_head * list)1875 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1876 {
1877 	list->prev = list->next = (struct sk_buff *)list;
1878 	list->qlen = 0;
1879 }
1880 
1881 /*
1882  * This function creates a split out lock class for each invocation;
1883  * this is needed for now since a whole lot of users of the skb-queue
1884  * infrastructure in drivers have different locking usage (in hardirq)
1885  * than the networking core (in softirq only). In the long run either the
1886  * network layer or drivers should need annotation to consolidate the
1887  * main types of usage into 3 classes.
1888  */
skb_queue_head_init(struct sk_buff_head * list)1889 static inline void skb_queue_head_init(struct sk_buff_head *list)
1890 {
1891 	spin_lock_init(&list->lock);
1892 	__skb_queue_head_init(list);
1893 }
1894 
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)1895 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1896 		struct lock_class_key *class)
1897 {
1898 	skb_queue_head_init(list);
1899 	lockdep_set_class(&list->lock, class);
1900 }
1901 
1902 /*
1903  *	Insert an sk_buff on a list.
1904  *
1905  *	The "__skb_xxxx()" functions are the non-atomic ones that
1906  *	can only be called with interrupts disabled.
1907  */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)1908 static inline void __skb_insert(struct sk_buff *newsk,
1909 				struct sk_buff *prev, struct sk_buff *next,
1910 				struct sk_buff_head *list)
1911 {
1912 	/* See skb_queue_empty_lockless() and skb_peek_tail()
1913 	 * for the opposite READ_ONCE()
1914 	 */
1915 	WRITE_ONCE(newsk->next, next);
1916 	WRITE_ONCE(newsk->prev, prev);
1917 	WRITE_ONCE(next->prev, newsk);
1918 	WRITE_ONCE(prev->next, newsk);
1919 	WRITE_ONCE(list->qlen, list->qlen + 1);
1920 }
1921 
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)1922 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1923 				      struct sk_buff *prev,
1924 				      struct sk_buff *next)
1925 {
1926 	struct sk_buff *first = list->next;
1927 	struct sk_buff *last = list->prev;
1928 
1929 	WRITE_ONCE(first->prev, prev);
1930 	WRITE_ONCE(prev->next, first);
1931 
1932 	WRITE_ONCE(last->next, next);
1933 	WRITE_ONCE(next->prev, last);
1934 }
1935 
1936 /**
1937  *	skb_queue_splice - join two skb lists, this is designed for stacks
1938  *	@list: the new list to add
1939  *	@head: the place to add it in the first list
1940  */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)1941 static inline void skb_queue_splice(const struct sk_buff_head *list,
1942 				    struct sk_buff_head *head)
1943 {
1944 	if (!skb_queue_empty(list)) {
1945 		__skb_queue_splice(list, (struct sk_buff *) head, head->next);
1946 		head->qlen += list->qlen;
1947 	}
1948 }
1949 
1950 /**
1951  *	skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1952  *	@list: the new list to add
1953  *	@head: the place to add it in the first list
1954  *
1955  *	The list at @list is reinitialised
1956  */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)1957 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1958 					 struct sk_buff_head *head)
1959 {
1960 	if (!skb_queue_empty(list)) {
1961 		__skb_queue_splice(list, (struct sk_buff *) head, head->next);
1962 		head->qlen += list->qlen;
1963 		__skb_queue_head_init(list);
1964 	}
1965 }
1966 
1967 /**
1968  *	skb_queue_splice_tail - join two skb lists, each list being a queue
1969  *	@list: the new list to add
1970  *	@head: the place to add it in the first list
1971  */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)1972 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1973 					 struct sk_buff_head *head)
1974 {
1975 	if (!skb_queue_empty(list)) {
1976 		__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1977 		head->qlen += list->qlen;
1978 	}
1979 }
1980 
1981 /**
1982  *	skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1983  *	@list: the new list to add
1984  *	@head: the place to add it in the first list
1985  *
1986  *	Each of the lists is a queue.
1987  *	The list at @list is reinitialised
1988  */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)1989 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1990 					      struct sk_buff_head *head)
1991 {
1992 	if (!skb_queue_empty(list)) {
1993 		__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1994 		head->qlen += list->qlen;
1995 		__skb_queue_head_init(list);
1996 	}
1997 }
1998 
1999 /**
2000  *	__skb_queue_after - queue a buffer at the list head
2001  *	@list: list to use
2002  *	@prev: place after this buffer
2003  *	@newsk: buffer to queue
2004  *
2005  *	Queue a buffer int the middle of a list. This function takes no locks
2006  *	and you must therefore hold required locks before calling it.
2007  *
2008  *	A buffer cannot be placed on two lists at the same time.
2009  */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)2010 static inline void __skb_queue_after(struct sk_buff_head *list,
2011 				     struct sk_buff *prev,
2012 				     struct sk_buff *newsk)
2013 {
2014 	__skb_insert(newsk, prev, prev->next, list);
2015 }
2016 
2017 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2018 		struct sk_buff_head *list);
2019 
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)2020 static inline void __skb_queue_before(struct sk_buff_head *list,
2021 				      struct sk_buff *next,
2022 				      struct sk_buff *newsk)
2023 {
2024 	__skb_insert(newsk, next->prev, next, list);
2025 }
2026 
2027 /**
2028  *	__skb_queue_head - queue a buffer at the list head
2029  *	@list: list to use
2030  *	@newsk: buffer to queue
2031  *
2032  *	Queue a buffer at the start of a list. This function takes no locks
2033  *	and you must therefore hold required locks before calling it.
2034  *
2035  *	A buffer cannot be placed on two lists at the same time.
2036  */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)2037 static inline void __skb_queue_head(struct sk_buff_head *list,
2038 				    struct sk_buff *newsk)
2039 {
2040 	__skb_queue_after(list, (struct sk_buff *)list, newsk);
2041 }
2042 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2043 
2044 /**
2045  *	__skb_queue_tail - queue a buffer at the list tail
2046  *	@list: list to use
2047  *	@newsk: buffer to queue
2048  *
2049  *	Queue a buffer at the end of a list. This function takes no locks
2050  *	and you must therefore hold required locks before calling it.
2051  *
2052  *	A buffer cannot be placed on two lists at the same time.
2053  */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2054 static inline void __skb_queue_tail(struct sk_buff_head *list,
2055 				   struct sk_buff *newsk)
2056 {
2057 	__skb_queue_before(list, (struct sk_buff *)list, newsk);
2058 }
2059 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2060 
2061 /*
2062  * remove sk_buff from list. _Must_ be called atomically, and with
2063  * the list known..
2064  */
2065 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2066 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2067 {
2068 	struct sk_buff *next, *prev;
2069 
2070 	WRITE_ONCE(list->qlen, list->qlen - 1);
2071 	next	   = skb->next;
2072 	prev	   = skb->prev;
2073 	skb->next  = skb->prev = NULL;
2074 	WRITE_ONCE(next->prev, prev);
2075 	WRITE_ONCE(prev->next, next);
2076 }
2077 
2078 /**
2079  *	__skb_dequeue - remove from the head of the queue
2080  *	@list: list to dequeue from
2081  *
2082  *	Remove the head of the list. This function does not take any locks
2083  *	so must be used with appropriate locks held only. The head item is
2084  *	returned or %NULL if the list is empty.
2085  */
__skb_dequeue(struct sk_buff_head * list)2086 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2087 {
2088 	struct sk_buff *skb = skb_peek(list);
2089 	if (skb)
2090 		__skb_unlink(skb, list);
2091 	return skb;
2092 }
2093 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2094 
2095 /**
2096  *	__skb_dequeue_tail - remove from the tail of the queue
2097  *	@list: list to dequeue from
2098  *
2099  *	Remove the tail of the list. This function does not take any locks
2100  *	so must be used with appropriate locks held only. The tail item is
2101  *	returned or %NULL if the list is empty.
2102  */
__skb_dequeue_tail(struct sk_buff_head * list)2103 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2104 {
2105 	struct sk_buff *skb = skb_peek_tail(list);
2106 	if (skb)
2107 		__skb_unlink(skb, list);
2108 	return skb;
2109 }
2110 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2111 
2112 
skb_is_nonlinear(const struct sk_buff * skb)2113 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2114 {
2115 	return skb->data_len;
2116 }
2117 
skb_headlen(const struct sk_buff * skb)2118 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2119 {
2120 	return skb->len - skb->data_len;
2121 }
2122 
__skb_pagelen(const struct sk_buff * skb)2123 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2124 {
2125 	unsigned int i, len = 0;
2126 
2127 	for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2128 		len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2129 	return len;
2130 }
2131 
skb_pagelen(const struct sk_buff * skb)2132 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2133 {
2134 	return skb_headlen(skb) + __skb_pagelen(skb);
2135 }
2136 
2137 /**
2138  * __skb_fill_page_desc - initialise a paged fragment in an skb
2139  * @skb: buffer containing fragment to be initialised
2140  * @i: paged fragment index to initialise
2141  * @page: the page to use for this fragment
2142  * @off: the offset to the data with @page
2143  * @size: the length of the data
2144  *
2145  * Initialises the @i'th fragment of @skb to point to &size bytes at
2146  * offset @off within @page.
2147  *
2148  * Does not take any additional reference on the fragment.
2149  */
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2150 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2151 					struct page *page, int off, int size)
2152 {
2153 	skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2154 
2155 	/*
2156 	 * Propagate page pfmemalloc to the skb if we can. The problem is
2157 	 * that not all callers have unique ownership of the page but rely
2158 	 * on page_is_pfmemalloc doing the right thing(tm).
2159 	 */
2160 	frag->bv_page		  = page;
2161 	frag->bv_offset		  = off;
2162 	skb_frag_size_set(frag, size);
2163 
2164 	page = compound_head(page);
2165 	if (page_is_pfmemalloc(page))
2166 		skb->pfmemalloc	= true;
2167 }
2168 
2169 /**
2170  * skb_fill_page_desc - initialise a paged fragment in an skb
2171  * @skb: buffer containing fragment to be initialised
2172  * @i: paged fragment index to initialise
2173  * @page: the page to use for this fragment
2174  * @off: the offset to the data with @page
2175  * @size: the length of the data
2176  *
2177  * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2178  * @skb to point to @size bytes at offset @off within @page. In
2179  * addition updates @skb such that @i is the last fragment.
2180  *
2181  * Does not take any additional reference on the fragment.
2182  */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2183 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2184 				      struct page *page, int off, int size)
2185 {
2186 	__skb_fill_page_desc(skb, i, page, off, size);
2187 	skb_shinfo(skb)->nr_frags = i + 1;
2188 }
2189 
2190 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2191 		     int size, unsigned int truesize);
2192 
2193 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2194 			  unsigned int truesize);
2195 
2196 #define SKB_LINEAR_ASSERT(skb)  BUG_ON(skb_is_nonlinear(skb))
2197 
2198 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2199 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2200 {
2201 	return skb->head + skb->tail;
2202 }
2203 
skb_reset_tail_pointer(struct sk_buff * skb)2204 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2205 {
2206 	skb->tail = skb->data - skb->head;
2207 }
2208 
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2209 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2210 {
2211 	skb_reset_tail_pointer(skb);
2212 	skb->tail += offset;
2213 }
2214 
2215 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2216 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2217 {
2218 	return skb->tail;
2219 }
2220 
skb_reset_tail_pointer(struct sk_buff * skb)2221 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2222 {
2223 	skb->tail = skb->data;
2224 }
2225 
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2226 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2227 {
2228 	skb->tail = skb->data + offset;
2229 }
2230 
2231 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2232 
skb_assert_len(struct sk_buff * skb)2233 static inline void skb_assert_len(struct sk_buff *skb)
2234 {
2235 #ifdef CONFIG_DEBUG_NET
2236 	if (WARN_ONCE(!skb->len, "%s\n", __func__))
2237 		DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2238 #endif /* CONFIG_DEBUG_NET */
2239 }
2240 
2241 /*
2242  *	Add data to an sk_buff
2243  */
2244 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2245 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2246 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2247 {
2248 	void *tmp = skb_tail_pointer(skb);
2249 	SKB_LINEAR_ASSERT(skb);
2250 	skb->tail += len;
2251 	skb->len  += len;
2252 	return tmp;
2253 }
2254 
__skb_put_zero(struct sk_buff * skb,unsigned int len)2255 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2256 {
2257 	void *tmp = __skb_put(skb, len);
2258 
2259 	memset(tmp, 0, len);
2260 	return tmp;
2261 }
2262 
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2263 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2264 				   unsigned int len)
2265 {
2266 	void *tmp = __skb_put(skb, len);
2267 
2268 	memcpy(tmp, data, len);
2269 	return tmp;
2270 }
2271 
__skb_put_u8(struct sk_buff * skb,u8 val)2272 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2273 {
2274 	*(u8 *)__skb_put(skb, 1) = val;
2275 }
2276 
skb_put_zero(struct sk_buff * skb,unsigned int len)2277 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2278 {
2279 	void *tmp = skb_put(skb, len);
2280 
2281 	memset(tmp, 0, len);
2282 
2283 	return tmp;
2284 }
2285 
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2286 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2287 				 unsigned int len)
2288 {
2289 	void *tmp = skb_put(skb, len);
2290 
2291 	memcpy(tmp, data, len);
2292 
2293 	return tmp;
2294 }
2295 
skb_put_u8(struct sk_buff * skb,u8 val)2296 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2297 {
2298 	*(u8 *)skb_put(skb, 1) = val;
2299 }
2300 
2301 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2302 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2303 {
2304 	skb->data -= len;
2305 	skb->len  += len;
2306 	return skb->data;
2307 }
2308 
2309 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2310 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2311 {
2312 	skb->len -= len;
2313 	BUG_ON(skb->len < skb->data_len);
2314 	return skb->data += len;
2315 }
2316 
skb_pull_inline(struct sk_buff * skb,unsigned int len)2317 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2318 {
2319 	return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2320 }
2321 
2322 void *skb_pull_data(struct sk_buff *skb, size_t len);
2323 
2324 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2325 
__pskb_pull(struct sk_buff * skb,unsigned int len)2326 static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2327 {
2328 	if (len > skb_headlen(skb) &&
2329 	    !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2330 		return NULL;
2331 	skb->len -= len;
2332 	return skb->data += len;
2333 }
2334 
pskb_pull(struct sk_buff * skb,unsigned int len)2335 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2336 {
2337 	return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2338 }
2339 
pskb_may_pull(struct sk_buff * skb,unsigned int len)2340 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2341 {
2342 	if (likely(len <= skb_headlen(skb)))
2343 		return true;
2344 	if (unlikely(len > skb->len))
2345 		return false;
2346 	return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2347 }
2348 
2349 void skb_condense(struct sk_buff *skb);
2350 
2351 /**
2352  *	skb_headroom - bytes at buffer head
2353  *	@skb: buffer to check
2354  *
2355  *	Return the number of bytes of free space at the head of an &sk_buff.
2356  */
skb_headroom(const struct sk_buff * skb)2357 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2358 {
2359 	return skb->data - skb->head;
2360 }
2361 
2362 /**
2363  *	skb_tailroom - bytes at buffer end
2364  *	@skb: buffer to check
2365  *
2366  *	Return the number of bytes of free space at the tail of an sk_buff
2367  */
skb_tailroom(const struct sk_buff * skb)2368 static inline int skb_tailroom(const struct sk_buff *skb)
2369 {
2370 	return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2371 }
2372 
2373 /**
2374  *	skb_availroom - bytes at buffer end
2375  *	@skb: buffer to check
2376  *
2377  *	Return the number of bytes of free space at the tail of an sk_buff
2378  *	allocated by sk_stream_alloc()
2379  */
skb_availroom(const struct sk_buff * skb)2380 static inline int skb_availroom(const struct sk_buff *skb)
2381 {
2382 	if (skb_is_nonlinear(skb))
2383 		return 0;
2384 
2385 	return skb->end - skb->tail - skb->reserved_tailroom;
2386 }
2387 
2388 /**
2389  *	skb_reserve - adjust headroom
2390  *	@skb: buffer to alter
2391  *	@len: bytes to move
2392  *
2393  *	Increase the headroom of an empty &sk_buff by reducing the tail
2394  *	room. This is only allowed for an empty buffer.
2395  */
skb_reserve(struct sk_buff * skb,int len)2396 static inline void skb_reserve(struct sk_buff *skb, int len)
2397 {
2398 	skb->data += len;
2399 	skb->tail += len;
2400 }
2401 
2402 /**
2403  *	skb_tailroom_reserve - adjust reserved_tailroom
2404  *	@skb: buffer to alter
2405  *	@mtu: maximum amount of headlen permitted
2406  *	@needed_tailroom: minimum amount of reserved_tailroom
2407  *
2408  *	Set reserved_tailroom so that headlen can be as large as possible but
2409  *	not larger than mtu and tailroom cannot be smaller than
2410  *	needed_tailroom.
2411  *	The required headroom should already have been reserved before using
2412  *	this function.
2413  */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2414 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2415 					unsigned int needed_tailroom)
2416 {
2417 	SKB_LINEAR_ASSERT(skb);
2418 	if (mtu < skb_tailroom(skb) - needed_tailroom)
2419 		/* use at most mtu */
2420 		skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2421 	else
2422 		/* use up to all available space */
2423 		skb->reserved_tailroom = needed_tailroom;
2424 }
2425 
2426 #define ENCAP_TYPE_ETHER	0
2427 #define ENCAP_TYPE_IPPROTO	1
2428 
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2429 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2430 					  __be16 protocol)
2431 {
2432 	skb->inner_protocol = protocol;
2433 	skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2434 }
2435 
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2436 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2437 					 __u8 ipproto)
2438 {
2439 	skb->inner_ipproto = ipproto;
2440 	skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2441 }
2442 
skb_reset_inner_headers(struct sk_buff * skb)2443 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2444 {
2445 	skb->inner_mac_header = skb->mac_header;
2446 	skb->inner_network_header = skb->network_header;
2447 	skb->inner_transport_header = skb->transport_header;
2448 }
2449 
skb_reset_mac_len(struct sk_buff * skb)2450 static inline void skb_reset_mac_len(struct sk_buff *skb)
2451 {
2452 	skb->mac_len = skb->network_header - skb->mac_header;
2453 }
2454 
skb_inner_transport_header(const struct sk_buff * skb)2455 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2456 							*skb)
2457 {
2458 	return skb->head + skb->inner_transport_header;
2459 }
2460 
skb_inner_transport_offset(const struct sk_buff * skb)2461 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2462 {
2463 	return skb_inner_transport_header(skb) - skb->data;
2464 }
2465 
skb_reset_inner_transport_header(struct sk_buff * skb)2466 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2467 {
2468 	skb->inner_transport_header = skb->data - skb->head;
2469 }
2470 
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2471 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2472 						   const int offset)
2473 {
2474 	skb_reset_inner_transport_header(skb);
2475 	skb->inner_transport_header += offset;
2476 }
2477 
skb_inner_network_header(const struct sk_buff * skb)2478 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2479 {
2480 	return skb->head + skb->inner_network_header;
2481 }
2482 
skb_reset_inner_network_header(struct sk_buff * skb)2483 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2484 {
2485 	skb->inner_network_header = skb->data - skb->head;
2486 }
2487 
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2488 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2489 						const int offset)
2490 {
2491 	skb_reset_inner_network_header(skb);
2492 	skb->inner_network_header += offset;
2493 }
2494 
skb_inner_mac_header(const struct sk_buff * skb)2495 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2496 {
2497 	return skb->head + skb->inner_mac_header;
2498 }
2499 
skb_reset_inner_mac_header(struct sk_buff * skb)2500 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2501 {
2502 	skb->inner_mac_header = skb->data - skb->head;
2503 }
2504 
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)2505 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2506 					    const int offset)
2507 {
2508 	skb_reset_inner_mac_header(skb);
2509 	skb->inner_mac_header += offset;
2510 }
skb_transport_header_was_set(const struct sk_buff * skb)2511 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2512 {
2513 	return skb->transport_header != (typeof(skb->transport_header))~0U;
2514 }
2515 
skb_transport_header(const struct sk_buff * skb)2516 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2517 {
2518 	return skb->head + skb->transport_header;
2519 }
2520 
skb_reset_transport_header(struct sk_buff * skb)2521 static inline void skb_reset_transport_header(struct sk_buff *skb)
2522 {
2523 	skb->transport_header = skb->data - skb->head;
2524 }
2525 
skb_set_transport_header(struct sk_buff * skb,const int offset)2526 static inline void skb_set_transport_header(struct sk_buff *skb,
2527 					    const int offset)
2528 {
2529 	skb_reset_transport_header(skb);
2530 	skb->transport_header += offset;
2531 }
2532 
skb_network_header(const struct sk_buff * skb)2533 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2534 {
2535 	return skb->head + skb->network_header;
2536 }
2537 
skb_reset_network_header(struct sk_buff * skb)2538 static inline void skb_reset_network_header(struct sk_buff *skb)
2539 {
2540 	skb->network_header = skb->data - skb->head;
2541 }
2542 
skb_set_network_header(struct sk_buff * skb,const int offset)2543 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2544 {
2545 	skb_reset_network_header(skb);
2546 	skb->network_header += offset;
2547 }
2548 
skb_mac_header(const struct sk_buff * skb)2549 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2550 {
2551 	return skb->head + skb->mac_header;
2552 }
2553 
skb_mac_offset(const struct sk_buff * skb)2554 static inline int skb_mac_offset(const struct sk_buff *skb)
2555 {
2556 	return skb_mac_header(skb) - skb->data;
2557 }
2558 
skb_mac_header_len(const struct sk_buff * skb)2559 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2560 {
2561 	return skb->network_header - skb->mac_header;
2562 }
2563 
skb_mac_header_was_set(const struct sk_buff * skb)2564 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2565 {
2566 	return skb->mac_header != (typeof(skb->mac_header))~0U;
2567 }
2568 
skb_unset_mac_header(struct sk_buff * skb)2569 static inline void skb_unset_mac_header(struct sk_buff *skb)
2570 {
2571 	skb->mac_header = (typeof(skb->mac_header))~0U;
2572 }
2573 
skb_reset_mac_header(struct sk_buff * skb)2574 static inline void skb_reset_mac_header(struct sk_buff *skb)
2575 {
2576 	skb->mac_header = skb->data - skb->head;
2577 }
2578 
skb_set_mac_header(struct sk_buff * skb,const int offset)2579 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2580 {
2581 	skb_reset_mac_header(skb);
2582 	skb->mac_header += offset;
2583 }
2584 
skb_pop_mac_header(struct sk_buff * skb)2585 static inline void skb_pop_mac_header(struct sk_buff *skb)
2586 {
2587 	skb->mac_header = skb->network_header;
2588 }
2589 
skb_probe_transport_header(struct sk_buff * skb)2590 static inline void skb_probe_transport_header(struct sk_buff *skb)
2591 {
2592 	struct flow_keys_basic keys;
2593 
2594 	if (skb_transport_header_was_set(skb))
2595 		return;
2596 
2597 	if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2598 					     NULL, 0, 0, 0, 0))
2599 		skb_set_transport_header(skb, keys.control.thoff);
2600 }
2601 
skb_mac_header_rebuild(struct sk_buff * skb)2602 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2603 {
2604 	if (skb_mac_header_was_set(skb)) {
2605 		const unsigned char *old_mac = skb_mac_header(skb);
2606 
2607 		skb_set_mac_header(skb, -skb->mac_len);
2608 		memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2609 	}
2610 }
2611 
skb_checksum_start_offset(const struct sk_buff * skb)2612 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2613 {
2614 	return skb->csum_start - skb_headroom(skb);
2615 }
2616 
skb_checksum_start(const struct sk_buff * skb)2617 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2618 {
2619 	return skb->head + skb->csum_start;
2620 }
2621 
skb_transport_offset(const struct sk_buff * skb)2622 static inline int skb_transport_offset(const struct sk_buff *skb)
2623 {
2624 	return skb_transport_header(skb) - skb->data;
2625 }
2626 
skb_network_header_len(const struct sk_buff * skb)2627 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2628 {
2629 	return skb->transport_header - skb->network_header;
2630 }
2631 
skb_inner_network_header_len(const struct sk_buff * skb)2632 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2633 {
2634 	return skb->inner_transport_header - skb->inner_network_header;
2635 }
2636 
skb_network_offset(const struct sk_buff * skb)2637 static inline int skb_network_offset(const struct sk_buff *skb)
2638 {
2639 	return skb_network_header(skb) - skb->data;
2640 }
2641 
skb_inner_network_offset(const struct sk_buff * skb)2642 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2643 {
2644 	return skb_inner_network_header(skb) - skb->data;
2645 }
2646 
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)2647 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2648 {
2649 	return pskb_may_pull(skb, skb_network_offset(skb) + len);
2650 }
2651 
2652 /*
2653  * CPUs often take a performance hit when accessing unaligned memory
2654  * locations. The actual performance hit varies, it can be small if the
2655  * hardware handles it or large if we have to take an exception and fix it
2656  * in software.
2657  *
2658  * Since an ethernet header is 14 bytes network drivers often end up with
2659  * the IP header at an unaligned offset. The IP header can be aligned by
2660  * shifting the start of the packet by 2 bytes. Drivers should do this
2661  * with:
2662  *
2663  * skb_reserve(skb, NET_IP_ALIGN);
2664  *
2665  * The downside to this alignment of the IP header is that the DMA is now
2666  * unaligned. On some architectures the cost of an unaligned DMA is high
2667  * and this cost outweighs the gains made by aligning the IP header.
2668  *
2669  * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2670  * to be overridden.
2671  */
2672 #ifndef NET_IP_ALIGN
2673 #define NET_IP_ALIGN	2
2674 #endif
2675 
2676 /*
2677  * The networking layer reserves some headroom in skb data (via
2678  * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2679  * the header has to grow. In the default case, if the header has to grow
2680  * 32 bytes or less we avoid the reallocation.
2681  *
2682  * Unfortunately this headroom changes the DMA alignment of the resulting
2683  * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2684  * on some architectures. An architecture can override this value,
2685  * perhaps setting it to a cacheline in size (since that will maintain
2686  * cacheline alignment of the DMA). It must be a power of 2.
2687  *
2688  * Various parts of the networking layer expect at least 32 bytes of
2689  * headroom, you should not reduce this.
2690  *
2691  * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2692  * to reduce average number of cache lines per packet.
2693  * get_rps_cpu() for example only access one 64 bytes aligned block :
2694  * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2695  */
2696 #ifndef NET_SKB_PAD
2697 #define NET_SKB_PAD	max(32, L1_CACHE_BYTES)
2698 #endif
2699 
2700 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2701 
__skb_set_length(struct sk_buff * skb,unsigned int len)2702 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2703 {
2704 	if (WARN_ON(skb_is_nonlinear(skb)))
2705 		return;
2706 	skb->len = len;
2707 	skb_set_tail_pointer(skb, len);
2708 }
2709 
__skb_trim(struct sk_buff * skb,unsigned int len)2710 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2711 {
2712 	__skb_set_length(skb, len);
2713 }
2714 
2715 void skb_trim(struct sk_buff *skb, unsigned int len);
2716 
__pskb_trim(struct sk_buff * skb,unsigned int len)2717 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2718 {
2719 	if (skb->data_len)
2720 		return ___pskb_trim(skb, len);
2721 	__skb_trim(skb, len);
2722 	return 0;
2723 }
2724 
pskb_trim(struct sk_buff * skb,unsigned int len)2725 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2726 {
2727 	return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2728 }
2729 
2730 /**
2731  *	pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2732  *	@skb: buffer to alter
2733  *	@len: new length
2734  *
2735  *	This is identical to pskb_trim except that the caller knows that
2736  *	the skb is not cloned so we should never get an error due to out-
2737  *	of-memory.
2738  */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)2739 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2740 {
2741 	int err = pskb_trim(skb, len);
2742 	BUG_ON(err);
2743 }
2744 
__skb_grow(struct sk_buff * skb,unsigned int len)2745 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2746 {
2747 	unsigned int diff = len - skb->len;
2748 
2749 	if (skb_tailroom(skb) < diff) {
2750 		int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2751 					   GFP_ATOMIC);
2752 		if (ret)
2753 			return ret;
2754 	}
2755 	__skb_set_length(skb, len);
2756 	return 0;
2757 }
2758 
2759 /**
2760  *	skb_orphan - orphan a buffer
2761  *	@skb: buffer to orphan
2762  *
2763  *	If a buffer currently has an owner then we call the owner's
2764  *	destructor function and make the @skb unowned. The buffer continues
2765  *	to exist but is no longer charged to its former owner.
2766  */
skb_orphan(struct sk_buff * skb)2767 static inline void skb_orphan(struct sk_buff *skb)
2768 {
2769 	if (skb->destructor) {
2770 		skb->destructor(skb);
2771 		skb->destructor = NULL;
2772 		skb->sk		= NULL;
2773 	} else {
2774 		BUG_ON(skb->sk);
2775 	}
2776 }
2777 
2778 /**
2779  *	skb_orphan_frags - orphan the frags contained in a buffer
2780  *	@skb: buffer to orphan frags from
2781  *	@gfp_mask: allocation mask for replacement pages
2782  *
2783  *	For each frag in the SKB which needs a destructor (i.e. has an
2784  *	owner) create a copy of that frag and release the original
2785  *	page by calling the destructor.
2786  */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)2787 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2788 {
2789 	if (likely(!skb_zcopy(skb)))
2790 		return 0;
2791 	if (!skb_zcopy_is_nouarg(skb) &&
2792 	    skb_uarg(skb)->callback == sock_zerocopy_callback)
2793 		return 0;
2794 	return skb_copy_ubufs(skb, gfp_mask);
2795 }
2796 
2797 /* 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)2798 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2799 {
2800 	if (likely(!skb_zcopy(skb)))
2801 		return 0;
2802 	return skb_copy_ubufs(skb, gfp_mask);
2803 }
2804 
2805 /**
2806  *	__skb_queue_purge - empty a list
2807  *	@list: list to empty
2808  *
2809  *	Delete all buffers on an &sk_buff list. Each buffer is removed from
2810  *	the list and one reference dropped. This function does not take the
2811  *	list lock and the caller must hold the relevant locks to use it.
2812  */
__skb_queue_purge(struct sk_buff_head * list)2813 static inline void __skb_queue_purge(struct sk_buff_head *list)
2814 {
2815 	struct sk_buff *skb;
2816 	while ((skb = __skb_dequeue(list)) != NULL)
2817 		kfree_skb(skb);
2818 }
2819 void skb_queue_purge(struct sk_buff_head *list);
2820 
2821 unsigned int skb_rbtree_purge(struct rb_root *root);
2822 
2823 void *netdev_alloc_frag(unsigned int fragsz);
2824 
2825 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2826 				   gfp_t gfp_mask);
2827 
2828 /**
2829  *	netdev_alloc_skb - allocate an skbuff for rx on a specific device
2830  *	@dev: network device to receive on
2831  *	@length: length to allocate
2832  *
2833  *	Allocate a new &sk_buff and assign it a usage count of one. The
2834  *	buffer has unspecified headroom built in. Users should allocate
2835  *	the headroom they think they need without accounting for the
2836  *	built in space. The built in space is used for optimisations.
2837  *
2838  *	%NULL is returned if there is no free memory. Although this function
2839  *	allocates memory it can be called from an interrupt.
2840  */
netdev_alloc_skb(struct net_device * dev,unsigned int length)2841 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2842 					       unsigned int length)
2843 {
2844 	return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2845 }
2846 
2847 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)2848 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2849 					      gfp_t gfp_mask)
2850 {
2851 	return __netdev_alloc_skb(NULL, length, gfp_mask);
2852 }
2853 
2854 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)2855 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2856 {
2857 	return netdev_alloc_skb(NULL, length);
2858 }
2859 
2860 
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)2861 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2862 		unsigned int length, gfp_t gfp)
2863 {
2864 	struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2865 
2866 	if (NET_IP_ALIGN && skb)
2867 		skb_reserve(skb, NET_IP_ALIGN);
2868 	return skb;
2869 }
2870 
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)2871 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2872 		unsigned int length)
2873 {
2874 	return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2875 }
2876 
skb_free_frag(void * addr)2877 static inline void skb_free_frag(void *addr)
2878 {
2879 	page_frag_free(addr);
2880 }
2881 
2882 void *napi_alloc_frag(unsigned int fragsz);
2883 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2884 				 unsigned int length, gfp_t gfp_mask);
napi_alloc_skb(struct napi_struct * napi,unsigned int length)2885 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2886 					     unsigned int length)
2887 {
2888 	return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2889 }
2890 void napi_consume_skb(struct sk_buff *skb, int budget);
2891 
2892 void __kfree_skb_flush(void);
2893 void __kfree_skb_defer(struct sk_buff *skb);
2894 
2895 /**
2896  * __dev_alloc_pages - allocate page for network Rx
2897  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2898  * @order: size of the allocation
2899  *
2900  * Allocate a new page.
2901  *
2902  * %NULL is returned if there is no free memory.
2903 */
__dev_alloc_pages(gfp_t gfp_mask,unsigned int order)2904 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2905 					     unsigned int order)
2906 {
2907 	/* This piece of code contains several assumptions.
2908 	 * 1.  This is for device Rx, therefor a cold page is preferred.
2909 	 * 2.  The expectation is the user wants a compound page.
2910 	 * 3.  If requesting a order 0 page it will not be compound
2911 	 *     due to the check to see if order has a value in prep_new_page
2912 	 * 4.  __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2913 	 *     code in gfp_to_alloc_flags that should be enforcing this.
2914 	 */
2915 	gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2916 
2917 	return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2918 }
2919 
dev_alloc_pages(unsigned int order)2920 static inline struct page *dev_alloc_pages(unsigned int order)
2921 {
2922 	return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2923 }
2924 
2925 /**
2926  * __dev_alloc_page - allocate a page for network Rx
2927  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2928  *
2929  * Allocate a new page.
2930  *
2931  * %NULL is returned if there is no free memory.
2932  */
__dev_alloc_page(gfp_t gfp_mask)2933 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2934 {
2935 	return __dev_alloc_pages(gfp_mask, 0);
2936 }
2937 
dev_alloc_page(void)2938 static inline struct page *dev_alloc_page(void)
2939 {
2940 	return dev_alloc_pages(0);
2941 }
2942 
2943 /**
2944  *	skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2945  *	@page: The page that was allocated from skb_alloc_page
2946  *	@skb: The skb that may need pfmemalloc set
2947  */
skb_propagate_pfmemalloc(struct page * page,struct sk_buff * skb)2948 static inline void skb_propagate_pfmemalloc(struct page *page,
2949 					     struct sk_buff *skb)
2950 {
2951 	if (page_is_pfmemalloc(page))
2952 		skb->pfmemalloc = true;
2953 }
2954 
2955 /**
2956  * skb_frag_off() - Returns the offset of a skb fragment
2957  * @frag: the paged fragment
2958  */
skb_frag_off(const skb_frag_t * frag)2959 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2960 {
2961 	return frag->bv_offset;
2962 }
2963 
2964 /**
2965  * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2966  * @frag: skb fragment
2967  * @delta: value to add
2968  */
skb_frag_off_add(skb_frag_t * frag,int delta)2969 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2970 {
2971 	frag->bv_offset += delta;
2972 }
2973 
2974 /**
2975  * skb_frag_off_set() - Sets the offset of a skb fragment
2976  * @frag: skb fragment
2977  * @offset: offset of fragment
2978  */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)2979 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2980 {
2981 	frag->bv_offset = offset;
2982 }
2983 
2984 /**
2985  * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2986  * @fragto: skb fragment where offset is set
2987  * @fragfrom: skb fragment offset is copied from
2988  */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)2989 static inline void skb_frag_off_copy(skb_frag_t *fragto,
2990 				     const skb_frag_t *fragfrom)
2991 {
2992 	fragto->bv_offset = fragfrom->bv_offset;
2993 }
2994 
2995 /**
2996  * skb_frag_page - retrieve the page referred to by a paged fragment
2997  * @frag: the paged fragment
2998  *
2999  * Returns the &struct page associated with @frag.
3000  */
skb_frag_page(const skb_frag_t * frag)3001 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3002 {
3003 	return frag->bv_page;
3004 }
3005 
3006 /**
3007  * __skb_frag_ref - take an addition reference on a paged fragment.
3008  * @frag: the paged fragment
3009  *
3010  * Takes an additional reference on the paged fragment @frag.
3011  */
__skb_frag_ref(skb_frag_t * frag)3012 static inline void __skb_frag_ref(skb_frag_t *frag)
3013 {
3014 	get_page(skb_frag_page(frag));
3015 }
3016 
3017 /**
3018  * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3019  * @skb: the buffer
3020  * @f: the fragment offset.
3021  *
3022  * Takes an additional reference on the @f'th paged fragment of @skb.
3023  */
skb_frag_ref(struct sk_buff * skb,int f)3024 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3025 {
3026 	__skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3027 }
3028 
3029 /**
3030  * __skb_frag_unref - release a reference on a paged fragment.
3031  * @frag: the paged fragment
3032  *
3033  * Releases a reference on the paged fragment @frag.
3034  */
__skb_frag_unref(skb_frag_t * frag)3035 static inline void __skb_frag_unref(skb_frag_t *frag)
3036 {
3037 	put_page(skb_frag_page(frag));
3038 }
3039 
3040 /**
3041  * skb_frag_unref - release a reference on a paged fragment of an skb.
3042  * @skb: the buffer
3043  * @f: the fragment offset
3044  *
3045  * Releases a reference on the @f'th paged fragment of @skb.
3046  */
skb_frag_unref(struct sk_buff * skb,int f)3047 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3048 {
3049 	__skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3050 }
3051 
3052 /**
3053  * skb_frag_address - gets the address of the data contained in a paged fragment
3054  * @frag: the paged fragment buffer
3055  *
3056  * Returns the address of the data within @frag. The page must already
3057  * be mapped.
3058  */
skb_frag_address(const skb_frag_t * frag)3059 static inline void *skb_frag_address(const skb_frag_t *frag)
3060 {
3061 	return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3062 }
3063 
3064 /**
3065  * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3066  * @frag: the paged fragment buffer
3067  *
3068  * Returns the address of the data within @frag. Checks that the page
3069  * is mapped and returns %NULL otherwise.
3070  */
skb_frag_address_safe(const skb_frag_t * frag)3071 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3072 {
3073 	void *ptr = page_address(skb_frag_page(frag));
3074 	if (unlikely(!ptr))
3075 		return NULL;
3076 
3077 	return ptr + skb_frag_off(frag);
3078 }
3079 
3080 /**
3081  * skb_frag_page_copy() - sets the page in a fragment from another fragment
3082  * @fragto: skb fragment where page is set
3083  * @fragfrom: skb fragment page is copied from
3084  */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3085 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3086 				      const skb_frag_t *fragfrom)
3087 {
3088 	fragto->bv_page = fragfrom->bv_page;
3089 }
3090 
3091 /**
3092  * __skb_frag_set_page - sets the page contained in a paged fragment
3093  * @frag: the paged fragment
3094  * @page: the page to set
3095  *
3096  * Sets the fragment @frag to contain @page.
3097  */
__skb_frag_set_page(skb_frag_t * frag,struct page * page)3098 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3099 {
3100 	frag->bv_page = page;
3101 }
3102 
3103 /**
3104  * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3105  * @skb: the buffer
3106  * @f: the fragment offset
3107  * @page: the page to set
3108  *
3109  * Sets the @f'th fragment of @skb to contain @page.
3110  */
skb_frag_set_page(struct sk_buff * skb,int f,struct page * page)3111 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3112 				     struct page *page)
3113 {
3114 	__skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3115 }
3116 
3117 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3118 
3119 /**
3120  * skb_frag_dma_map - maps a paged fragment via the DMA API
3121  * @dev: the device to map the fragment to
3122  * @frag: the paged fragment to map
3123  * @offset: the offset within the fragment (starting at the
3124  *          fragment's own offset)
3125  * @size: the number of bytes to map
3126  * @dir: the direction of the mapping (``PCI_DMA_*``)
3127  *
3128  * Maps the page associated with @frag to @device.
3129  */
skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3130 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3131 					  const skb_frag_t *frag,
3132 					  size_t offset, size_t size,
3133 					  enum dma_data_direction dir)
3134 {
3135 	return dma_map_page(dev, skb_frag_page(frag),
3136 			    skb_frag_off(frag) + offset, size, dir);
3137 }
3138 
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3139 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3140 					gfp_t gfp_mask)
3141 {
3142 	return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3143 }
3144 
3145 
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3146 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3147 						  gfp_t gfp_mask)
3148 {
3149 	return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3150 }
3151 
3152 
3153 /**
3154  *	skb_clone_writable - is the header of a clone writable
3155  *	@skb: buffer to check
3156  *	@len: length up to which to write
3157  *
3158  *	Returns true if modifying the header part of the cloned buffer
3159  *	does not requires the data to be copied.
3160  */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3161 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3162 {
3163 	return !skb_header_cloned(skb) &&
3164 	       skb_headroom(skb) + len <= skb->hdr_len;
3165 }
3166 
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3167 static inline int skb_try_make_writable(struct sk_buff *skb,
3168 					unsigned int write_len)
3169 {
3170 	return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3171 	       pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3172 }
3173 
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3174 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3175 			    int cloned)
3176 {
3177 	int delta = 0;
3178 
3179 	if (headroom > skb_headroom(skb))
3180 		delta = headroom - skb_headroom(skb);
3181 
3182 	if (delta || cloned)
3183 		return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3184 					GFP_ATOMIC);
3185 	return 0;
3186 }
3187 
3188 /**
3189  *	skb_cow - copy header of skb when it is required
3190  *	@skb: buffer to cow
3191  *	@headroom: needed headroom
3192  *
3193  *	If the skb passed lacks sufficient headroom or its data part
3194  *	is shared, data is reallocated. If reallocation fails, an error
3195  *	is returned and original skb is not changed.
3196  *
3197  *	The result is skb with writable area skb->head...skb->tail
3198  *	and at least @headroom of space at head.
3199  */
skb_cow(struct sk_buff * skb,unsigned int headroom)3200 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3201 {
3202 	return __skb_cow(skb, headroom, skb_cloned(skb));
3203 }
3204 
3205 /**
3206  *	skb_cow_head - skb_cow but only making the head writable
3207  *	@skb: buffer to cow
3208  *	@headroom: needed headroom
3209  *
3210  *	This function is identical to skb_cow except that we replace the
3211  *	skb_cloned check by skb_header_cloned.  It should be used when
3212  *	you only need to push on some header and do not need to modify
3213  *	the data.
3214  */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3215 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3216 {
3217 	return __skb_cow(skb, headroom, skb_header_cloned(skb));
3218 }
3219 
3220 /**
3221  *	skb_padto	- pad an skbuff up to a minimal size
3222  *	@skb: buffer to pad
3223  *	@len: minimal length
3224  *
3225  *	Pads up a buffer to ensure the trailing bytes exist and are
3226  *	blanked. If the buffer already contains sufficient data it
3227  *	is untouched. Otherwise it is extended. Returns zero on
3228  *	success. The skb is freed on error.
3229  */
skb_padto(struct sk_buff * skb,unsigned int len)3230 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3231 {
3232 	unsigned int size = skb->len;
3233 	if (likely(size >= len))
3234 		return 0;
3235 	return skb_pad(skb, len - size);
3236 }
3237 
3238 /**
3239  *	__skb_put_padto - increase size and pad an skbuff up to a minimal size
3240  *	@skb: buffer to pad
3241  *	@len: minimal length
3242  *	@free_on_error: free buffer on error
3243  *
3244  *	Pads up a buffer to ensure the trailing bytes exist and are
3245  *	blanked. If the buffer already contains sufficient data it
3246  *	is untouched. Otherwise it is extended. Returns zero on
3247  *	success. The skb is freed on error if @free_on_error is true.
3248  */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3249 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3250 					       unsigned int len,
3251 					       bool free_on_error)
3252 {
3253 	unsigned int size = skb->len;
3254 
3255 	if (unlikely(size < len)) {
3256 		len -= size;
3257 		if (__skb_pad(skb, len, free_on_error))
3258 			return -ENOMEM;
3259 		__skb_put(skb, len);
3260 	}
3261 	return 0;
3262 }
3263 
3264 /**
3265  *	skb_put_padto - increase size and pad an skbuff up to a minimal size
3266  *	@skb: buffer to pad
3267  *	@len: minimal length
3268  *
3269  *	Pads up a buffer to ensure the trailing bytes exist and are
3270  *	blanked. If the buffer already contains sufficient data it
3271  *	is untouched. Otherwise it is extended. Returns zero on
3272  *	success. The skb is freed on error.
3273  */
skb_put_padto(struct sk_buff * skb,unsigned int len)3274 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3275 {
3276 	return __skb_put_padto(skb, len, true);
3277 }
3278 
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3279 static inline int skb_add_data(struct sk_buff *skb,
3280 			       struct iov_iter *from, int copy)
3281 {
3282 	const int off = skb->len;
3283 
3284 	if (skb->ip_summed == CHECKSUM_NONE) {
3285 		__wsum csum = 0;
3286 		if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3287 					         &csum, from)) {
3288 			skb->csum = csum_block_add(skb->csum, csum, off);
3289 			return 0;
3290 		}
3291 	} else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3292 		return 0;
3293 
3294 	__skb_trim(skb, off);
3295 	return -EFAULT;
3296 }
3297 
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3298 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3299 				    const struct page *page, int off)
3300 {
3301 	if (skb_zcopy(skb))
3302 		return false;
3303 	if (i) {
3304 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3305 
3306 		return page == skb_frag_page(frag) &&
3307 		       off == skb_frag_off(frag) + skb_frag_size(frag);
3308 	}
3309 	return false;
3310 }
3311 
__skb_linearize(struct sk_buff * skb)3312 static inline int __skb_linearize(struct sk_buff *skb)
3313 {
3314 	return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3315 }
3316 
3317 /**
3318  *	skb_linearize - convert paged skb to linear one
3319  *	@skb: buffer to linarize
3320  *
3321  *	If there is no free memory -ENOMEM is returned, otherwise zero
3322  *	is returned and the old skb data released.
3323  */
skb_linearize(struct sk_buff * skb)3324 static inline int skb_linearize(struct sk_buff *skb)
3325 {
3326 	return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3327 }
3328 
3329 /**
3330  * skb_has_shared_frag - can any frag be overwritten
3331  * @skb: buffer to test
3332  *
3333  * Return true if the skb has at least one frag that might be modified
3334  * by an external entity (as in vmsplice()/sendfile())
3335  */
skb_has_shared_frag(const struct sk_buff * skb)3336 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3337 {
3338 	return skb_is_nonlinear(skb) &&
3339 	       skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3340 }
3341 
3342 /**
3343  *	skb_linearize_cow - make sure skb is linear and writable
3344  *	@skb: buffer to process
3345  *
3346  *	If there is no free memory -ENOMEM is returned, otherwise zero
3347  *	is returned and the old skb data released.
3348  */
skb_linearize_cow(struct sk_buff * skb)3349 static inline int skb_linearize_cow(struct sk_buff *skb)
3350 {
3351 	return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3352 	       __skb_linearize(skb) : 0;
3353 }
3354 
3355 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3356 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3357 		     unsigned int off)
3358 {
3359 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3360 		skb->csum = csum_block_sub(skb->csum,
3361 					   csum_partial(start, len, 0), off);
3362 	else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3363 		 skb_checksum_start_offset(skb) < 0)
3364 		skb->ip_summed = CHECKSUM_NONE;
3365 }
3366 
3367 /**
3368  *	skb_postpull_rcsum - update checksum for received skb after pull
3369  *	@skb: buffer to update
3370  *	@start: start of data before pull
3371  *	@len: length of data pulled
3372  *
3373  *	After doing a pull on a received packet, you need to call this to
3374  *	update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3375  *	CHECKSUM_NONE so that it can be recomputed from scratch.
3376  */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3377 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3378 				      const void *start, unsigned int len)
3379 {
3380 	__skb_postpull_rcsum(skb, start, len, 0);
3381 }
3382 
3383 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3384 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3385 		     unsigned int off)
3386 {
3387 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3388 		skb->csum = csum_block_add(skb->csum,
3389 					   csum_partial(start, len, 0), off);
3390 }
3391 
3392 /**
3393  *	skb_postpush_rcsum - update checksum for received skb after push
3394  *	@skb: buffer to update
3395  *	@start: start of data after push
3396  *	@len: length of data pushed
3397  *
3398  *	After doing a push on a received packet, you need to call this to
3399  *	update the CHECKSUM_COMPLETE checksum.
3400  */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3401 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3402 				      const void *start, unsigned int len)
3403 {
3404 	__skb_postpush_rcsum(skb, start, len, 0);
3405 }
3406 
3407 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3408 
3409 /**
3410  *	skb_push_rcsum - push skb and update receive checksum
3411  *	@skb: buffer to update
3412  *	@len: length of data pulled
3413  *
3414  *	This function performs an skb_push on the packet and updates
3415  *	the CHECKSUM_COMPLETE checksum.  It should be used on
3416  *	receive path processing instead of skb_push unless you know
3417  *	that the checksum difference is zero (e.g., a valid IP header)
3418  *	or you are setting ip_summed to CHECKSUM_NONE.
3419  */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3420 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3421 {
3422 	skb_push(skb, len);
3423 	skb_postpush_rcsum(skb, skb->data, len);
3424 	return skb->data;
3425 }
3426 
3427 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3428 /**
3429  *	pskb_trim_rcsum - trim received skb and update checksum
3430  *	@skb: buffer to trim
3431  *	@len: new length
3432  *
3433  *	This is exactly the same as pskb_trim except that it ensures the
3434  *	checksum of received packets are still valid after the operation.
3435  *	It can change skb pointers.
3436  */
3437 
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)3438 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3439 {
3440 	if (likely(len >= skb->len))
3441 		return 0;
3442 	return pskb_trim_rcsum_slow(skb, len);
3443 }
3444 
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)3445 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3446 {
3447 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3448 		skb->ip_summed = CHECKSUM_NONE;
3449 	__skb_trim(skb, len);
3450 	return 0;
3451 }
3452 
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)3453 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3454 {
3455 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3456 		skb->ip_summed = CHECKSUM_NONE;
3457 	return __skb_grow(skb, len);
3458 }
3459 
3460 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3461 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3462 #define skb_rb_last(root)  rb_to_skb(rb_last(root))
3463 #define skb_rb_next(skb)   rb_to_skb(rb_next(&(skb)->rbnode))
3464 #define skb_rb_prev(skb)   rb_to_skb(rb_prev(&(skb)->rbnode))
3465 
3466 #define skb_queue_walk(queue, skb) \
3467 		for (skb = (queue)->next;					\
3468 		     skb != (struct sk_buff *)(queue);				\
3469 		     skb = skb->next)
3470 
3471 #define skb_queue_walk_safe(queue, skb, tmp)					\
3472 		for (skb = (queue)->next, tmp = skb->next;			\
3473 		     skb != (struct sk_buff *)(queue);				\
3474 		     skb = tmp, tmp = skb->next)
3475 
3476 #define skb_queue_walk_from(queue, skb)						\
3477 		for (; skb != (struct sk_buff *)(queue);			\
3478 		     skb = skb->next)
3479 
3480 #define skb_rbtree_walk(skb, root)						\
3481 		for (skb = skb_rb_first(root); skb != NULL;			\
3482 		     skb = skb_rb_next(skb))
3483 
3484 #define skb_rbtree_walk_from(skb)						\
3485 		for (; skb != NULL;						\
3486 		     skb = skb_rb_next(skb))
3487 
3488 #define skb_rbtree_walk_from_safe(skb, tmp)					\
3489 		for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL);	\
3490 		     skb = tmp)
3491 
3492 #define skb_queue_walk_from_safe(queue, skb, tmp)				\
3493 		for (tmp = skb->next;						\
3494 		     skb != (struct sk_buff *)(queue);				\
3495 		     skb = tmp, tmp = skb->next)
3496 
3497 #define skb_queue_reverse_walk(queue, skb) \
3498 		for (skb = (queue)->prev;					\
3499 		     skb != (struct sk_buff *)(queue);				\
3500 		     skb = skb->prev)
3501 
3502 #define skb_queue_reverse_walk_safe(queue, skb, tmp)				\
3503 		for (skb = (queue)->prev, tmp = skb->prev;			\
3504 		     skb != (struct sk_buff *)(queue);				\
3505 		     skb = tmp, tmp = skb->prev)
3506 
3507 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp)			\
3508 		for (tmp = skb->prev;						\
3509 		     skb != (struct sk_buff *)(queue);				\
3510 		     skb = tmp, tmp = skb->prev)
3511 
skb_has_frag_list(const struct sk_buff * skb)3512 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3513 {
3514 	return skb_shinfo(skb)->frag_list != NULL;
3515 }
3516 
skb_frag_list_init(struct sk_buff * skb)3517 static inline void skb_frag_list_init(struct sk_buff *skb)
3518 {
3519 	skb_shinfo(skb)->frag_list = NULL;
3520 }
3521 
3522 #define skb_walk_frags(skb, iter)	\
3523 	for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3524 
3525 
3526 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3527 				int *err, long *timeo_p,
3528 				const struct sk_buff *skb);
3529 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3530 					  struct sk_buff_head *queue,
3531 					  unsigned int flags,
3532 					  int *off, int *err,
3533 					  struct sk_buff **last);
3534 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3535 					struct sk_buff_head *queue,
3536 					unsigned int flags, int *off, int *err,
3537 					struct sk_buff **last);
3538 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3539 				    struct sk_buff_head *sk_queue,
3540 				    unsigned int flags, int *off, int *err);
3541 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3542 				  int *err);
3543 __poll_t datagram_poll(struct file *file, struct socket *sock,
3544 			   struct poll_table_struct *wait);
3545 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3546 			   struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)3547 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3548 					struct msghdr *msg, int size)
3549 {
3550 	return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3551 }
3552 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3553 				   struct msghdr *msg);
3554 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3555 			   struct iov_iter *to, int len,
3556 			   struct ahash_request *hash);
3557 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3558 				 struct iov_iter *from, int len);
3559 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3560 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3561 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)3562 static inline void skb_free_datagram_locked(struct sock *sk,
3563 					    struct sk_buff *skb)
3564 {
3565 	__skb_free_datagram_locked(sk, skb, 0);
3566 }
3567 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3568 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3569 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3570 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3571 			      int len);
3572 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3573 		    struct pipe_inode_info *pipe, unsigned int len,
3574 		    unsigned int flags);
3575 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3576 			 int len);
3577 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3578 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3579 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3580 		 int len, int hlen);
3581 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3582 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3583 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3584 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3585 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3586 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3587 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3588 				 unsigned int offset);
3589 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3590 int skb_ensure_writable(struct sk_buff *skb, int write_len);
3591 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3592 int skb_vlan_pop(struct sk_buff *skb);
3593 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3594 int skb_eth_pop(struct sk_buff *skb);
3595 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3596 		 const unsigned char *src);
3597 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3598 		  int mac_len, bool ethernet);
3599 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3600 		 bool ethernet);
3601 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3602 int skb_mpls_dec_ttl(struct sk_buff *skb);
3603 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3604 			     gfp_t gfp);
3605 
memcpy_from_msg(void * data,struct msghdr * msg,int len)3606 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3607 {
3608 	return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3609 }
3610 
memcpy_to_msg(struct msghdr * msg,void * data,int len)3611 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3612 {
3613 	return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3614 }
3615 
3616 struct skb_checksum_ops {
3617 	__wsum (*update)(const void *mem, int len, __wsum wsum);
3618 	__wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3619 };
3620 
3621 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3622 
3623 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3624 		      __wsum csum, const struct skb_checksum_ops *ops);
3625 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3626 		    __wsum csum);
3627 
3628 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * data,int hlen,void * buffer)3629 __skb_header_pointer(const struct sk_buff *skb, int offset,
3630 		     int len, void *data, int hlen, void *buffer)
3631 {
3632 	if (hlen - offset >= len)
3633 		return data + offset;
3634 
3635 	if (!skb ||
3636 	    skb_copy_bits(skb, offset, buffer, len) < 0)
3637 		return NULL;
3638 
3639 	return buffer;
3640 }
3641 
3642 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)3643 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3644 {
3645 	return __skb_header_pointer(skb, offset, len, skb->data,
3646 				    skb_headlen(skb), buffer);
3647 }
3648 
3649 /**
3650  *	skb_needs_linearize - check if we need to linearize a given skb
3651  *			      depending on the given device features.
3652  *	@skb: socket buffer to check
3653  *	@features: net device features
3654  *
3655  *	Returns true if either:
3656  *	1. skb has frag_list and the device doesn't support FRAGLIST, or
3657  *	2. skb is fragmented and the device does not support SG.
3658  */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)3659 static inline bool skb_needs_linearize(struct sk_buff *skb,
3660 				       netdev_features_t features)
3661 {
3662 	return skb_is_nonlinear(skb) &&
3663 	       ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3664 		(skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3665 }
3666 
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)3667 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3668 					     void *to,
3669 					     const unsigned int len)
3670 {
3671 	memcpy(to, skb->data, len);
3672 }
3673 
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)3674 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3675 						    const int offset, void *to,
3676 						    const unsigned int len)
3677 {
3678 	memcpy(to, skb->data + offset, len);
3679 }
3680 
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)3681 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3682 					   const void *from,
3683 					   const unsigned int len)
3684 {
3685 	memcpy(skb->data, from, len);
3686 }
3687 
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)3688 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3689 						  const int offset,
3690 						  const void *from,
3691 						  const unsigned int len)
3692 {
3693 	memcpy(skb->data + offset, from, len);
3694 }
3695 
3696 void skb_init(void);
3697 
skb_get_ktime(const struct sk_buff * skb)3698 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3699 {
3700 	return skb->tstamp;
3701 }
3702 
3703 /**
3704  *	skb_get_timestamp - get timestamp from a skb
3705  *	@skb: skb to get stamp from
3706  *	@stamp: pointer to struct __kernel_old_timeval to store stamp in
3707  *
3708  *	Timestamps are stored in the skb as offsets to a base timestamp.
3709  *	This function converts the offset back to a struct timeval and stores
3710  *	it in stamp.
3711  */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)3712 static inline void skb_get_timestamp(const struct sk_buff *skb,
3713 				     struct __kernel_old_timeval *stamp)
3714 {
3715 	*stamp = ns_to_kernel_old_timeval(skb->tstamp);
3716 }
3717 
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)3718 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3719 					 struct __kernel_sock_timeval *stamp)
3720 {
3721 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3722 
3723 	stamp->tv_sec = ts.tv_sec;
3724 	stamp->tv_usec = ts.tv_nsec / 1000;
3725 }
3726 
skb_get_timestampns(const struct sk_buff * skb,struct __kernel_old_timespec * stamp)3727 static inline void skb_get_timestampns(const struct sk_buff *skb,
3728 				       struct __kernel_old_timespec *stamp)
3729 {
3730 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3731 
3732 	stamp->tv_sec = ts.tv_sec;
3733 	stamp->tv_nsec = ts.tv_nsec;
3734 }
3735 
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)3736 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3737 					   struct __kernel_timespec *stamp)
3738 {
3739 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3740 
3741 	stamp->tv_sec = ts.tv_sec;
3742 	stamp->tv_nsec = ts.tv_nsec;
3743 }
3744 
__net_timestamp(struct sk_buff * skb)3745 static inline void __net_timestamp(struct sk_buff *skb)
3746 {
3747 	skb->tstamp = ktime_get_real();
3748 }
3749 
net_timedelta(ktime_t t)3750 static inline ktime_t net_timedelta(ktime_t t)
3751 {
3752 	return ktime_sub(ktime_get_real(), t);
3753 }
3754 
net_invalid_timestamp(void)3755 static inline ktime_t net_invalid_timestamp(void)
3756 {
3757 	return 0;
3758 }
3759 
skb_metadata_len(const struct sk_buff * skb)3760 static inline u8 skb_metadata_len(const struct sk_buff *skb)
3761 {
3762 	return skb_shinfo(skb)->meta_len;
3763 }
3764 
skb_metadata_end(const struct sk_buff * skb)3765 static inline void *skb_metadata_end(const struct sk_buff *skb)
3766 {
3767 	return skb_mac_header(skb);
3768 }
3769 
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)3770 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3771 					  const struct sk_buff *skb_b,
3772 					  u8 meta_len)
3773 {
3774 	const void *a = skb_metadata_end(skb_a);
3775 	const void *b = skb_metadata_end(skb_b);
3776 	/* Using more efficient varaiant than plain call to memcmp(). */
3777 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3778 	u64 diffs = 0;
3779 
3780 	switch (meta_len) {
3781 #define __it(x, op) (x -= sizeof(u##op))
3782 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3783 	case 32: diffs |= __it_diff(a, b, 64);
3784 		fallthrough;
3785 	case 24: diffs |= __it_diff(a, b, 64);
3786 		fallthrough;
3787 	case 16: diffs |= __it_diff(a, b, 64);
3788 		fallthrough;
3789 	case  8: diffs |= __it_diff(a, b, 64);
3790 		break;
3791 	case 28: diffs |= __it_diff(a, b, 64);
3792 		fallthrough;
3793 	case 20: diffs |= __it_diff(a, b, 64);
3794 		fallthrough;
3795 	case 12: diffs |= __it_diff(a, b, 64);
3796 		fallthrough;
3797 	case  4: diffs |= __it_diff(a, b, 32);
3798 		break;
3799 	}
3800 	return diffs;
3801 #else
3802 	return memcmp(a - meta_len, b - meta_len, meta_len);
3803 #endif
3804 }
3805 
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)3806 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3807 					const struct sk_buff *skb_b)
3808 {
3809 	u8 len_a = skb_metadata_len(skb_a);
3810 	u8 len_b = skb_metadata_len(skb_b);
3811 
3812 	if (!(len_a | len_b))
3813 		return false;
3814 
3815 	return len_a != len_b ?
3816 	       true : __skb_metadata_differs(skb_a, skb_b, len_a);
3817 }
3818 
skb_metadata_set(struct sk_buff * skb,u8 meta_len)3819 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3820 {
3821 	skb_shinfo(skb)->meta_len = meta_len;
3822 }
3823 
skb_metadata_clear(struct sk_buff * skb)3824 static inline void skb_metadata_clear(struct sk_buff *skb)
3825 {
3826 	skb_metadata_set(skb, 0);
3827 }
3828 
3829 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3830 
3831 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3832 
3833 void skb_clone_tx_timestamp(struct sk_buff *skb);
3834 bool skb_defer_rx_timestamp(struct sk_buff *skb);
3835 
3836 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3837 
skb_clone_tx_timestamp(struct sk_buff * skb)3838 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3839 {
3840 }
3841 
skb_defer_rx_timestamp(struct sk_buff * skb)3842 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3843 {
3844 	return false;
3845 }
3846 
3847 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3848 
3849 /**
3850  * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3851  *
3852  * PHY drivers may accept clones of transmitted packets for
3853  * timestamping via their phy_driver.txtstamp method. These drivers
3854  * must call this function to return the skb back to the stack with a
3855  * timestamp.
3856  *
3857  * @skb: clone of the original outgoing packet
3858  * @hwtstamps: hardware time stamps
3859  *
3860  */
3861 void skb_complete_tx_timestamp(struct sk_buff *skb,
3862 			       struct skb_shared_hwtstamps *hwtstamps);
3863 
3864 void __skb_tstamp_tx(struct sk_buff *orig_skb,
3865 		     struct skb_shared_hwtstamps *hwtstamps,
3866 		     struct sock *sk, int tstype);
3867 
3868 /**
3869  * skb_tstamp_tx - queue clone of skb with send time stamps
3870  * @orig_skb:	the original outgoing packet
3871  * @hwtstamps:	hardware time stamps, may be NULL if not available
3872  *
3873  * If the skb has a socket associated, then this function clones the
3874  * skb (thus sharing the actual data and optional structures), stores
3875  * the optional hardware time stamping information (if non NULL) or
3876  * generates a software time stamp (otherwise), then queues the clone
3877  * to the error queue of the socket.  Errors are silently ignored.
3878  */
3879 void skb_tstamp_tx(struct sk_buff *orig_skb,
3880 		   struct skb_shared_hwtstamps *hwtstamps);
3881 
3882 /**
3883  * skb_tx_timestamp() - Driver hook for transmit timestamping
3884  *
3885  * Ethernet MAC Drivers should call this function in their hard_xmit()
3886  * function immediately before giving the sk_buff to the MAC hardware.
3887  *
3888  * Specifically, one should make absolutely sure that this function is
3889  * called before TX completion of this packet can trigger.  Otherwise
3890  * the packet could potentially already be freed.
3891  *
3892  * @skb: A socket buffer.
3893  */
skb_tx_timestamp(struct sk_buff * skb)3894 static inline void skb_tx_timestamp(struct sk_buff *skb)
3895 {
3896 	skb_clone_tx_timestamp(skb);
3897 	if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3898 		skb_tstamp_tx(skb, NULL);
3899 }
3900 
3901 /**
3902  * skb_complete_wifi_ack - deliver skb with wifi status
3903  *
3904  * @skb: the original outgoing packet
3905  * @acked: ack status
3906  *
3907  */
3908 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3909 
3910 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3911 __sum16 __skb_checksum_complete(struct sk_buff *skb);
3912 
skb_csum_unnecessary(const struct sk_buff * skb)3913 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3914 {
3915 	return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3916 		skb->csum_valid ||
3917 		(skb->ip_summed == CHECKSUM_PARTIAL &&
3918 		 skb_checksum_start_offset(skb) >= 0));
3919 }
3920 
3921 /**
3922  *	skb_checksum_complete - Calculate checksum of an entire packet
3923  *	@skb: packet to process
3924  *
3925  *	This function calculates the checksum over the entire packet plus
3926  *	the value of skb->csum.  The latter can be used to supply the
3927  *	checksum of a pseudo header as used by TCP/UDP.  It returns the
3928  *	checksum.
3929  *
3930  *	For protocols that contain complete checksums such as ICMP/TCP/UDP,
3931  *	this function can be used to verify that checksum on received
3932  *	packets.  In that case the function should return zero if the
3933  *	checksum is correct.  In particular, this function will return zero
3934  *	if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3935  *	hardware has already verified the correctness of the checksum.
3936  */
skb_checksum_complete(struct sk_buff * skb)3937 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3938 {
3939 	return skb_csum_unnecessary(skb) ?
3940 	       0 : __skb_checksum_complete(skb);
3941 }
3942 
__skb_decr_checksum_unnecessary(struct sk_buff * skb)3943 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3944 {
3945 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3946 		if (skb->csum_level == 0)
3947 			skb->ip_summed = CHECKSUM_NONE;
3948 		else
3949 			skb->csum_level--;
3950 	}
3951 }
3952 
__skb_incr_checksum_unnecessary(struct sk_buff * skb)3953 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3954 {
3955 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3956 		if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3957 			skb->csum_level++;
3958 	} else if (skb->ip_summed == CHECKSUM_NONE) {
3959 		skb->ip_summed = CHECKSUM_UNNECESSARY;
3960 		skb->csum_level = 0;
3961 	}
3962 }
3963 
__skb_reset_checksum_unnecessary(struct sk_buff * skb)3964 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
3965 {
3966 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3967 		skb->ip_summed = CHECKSUM_NONE;
3968 		skb->csum_level = 0;
3969 	}
3970 }
3971 
3972 /* Check if we need to perform checksum complete validation.
3973  *
3974  * Returns true if checksum complete is needed, false otherwise
3975  * (either checksum is unnecessary or zero checksum is allowed).
3976  */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)3977 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3978 						  bool zero_okay,
3979 						  __sum16 check)
3980 {
3981 	if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3982 		skb->csum_valid = 1;
3983 		__skb_decr_checksum_unnecessary(skb);
3984 		return false;
3985 	}
3986 
3987 	return true;
3988 }
3989 
3990 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
3991  * in checksum_init.
3992  */
3993 #define CHECKSUM_BREAK 76
3994 
3995 /* Unset checksum-complete
3996  *
3997  * Unset checksum complete can be done when packet is being modified
3998  * (uncompressed for instance) and checksum-complete value is
3999  * invalidated.
4000  */
skb_checksum_complete_unset(struct sk_buff * skb)4001 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4002 {
4003 	if (skb->ip_summed == CHECKSUM_COMPLETE)
4004 		skb->ip_summed = CHECKSUM_NONE;
4005 }
4006 
4007 /* Validate (init) checksum based on checksum complete.
4008  *
4009  * Return values:
4010  *   0: checksum is validated or try to in skb_checksum_complete. In the latter
4011  *	case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4012  *	checksum is stored in skb->csum for use in __skb_checksum_complete
4013  *   non-zero: value of invalid checksum
4014  *
4015  */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)4016 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4017 						       bool complete,
4018 						       __wsum psum)
4019 {
4020 	if (skb->ip_summed == CHECKSUM_COMPLETE) {
4021 		if (!csum_fold(csum_add(psum, skb->csum))) {
4022 			skb->csum_valid = 1;
4023 			return 0;
4024 		}
4025 	}
4026 
4027 	skb->csum = psum;
4028 
4029 	if (complete || skb->len <= CHECKSUM_BREAK) {
4030 		__sum16 csum;
4031 
4032 		csum = __skb_checksum_complete(skb);
4033 		skb->csum_valid = !csum;
4034 		return csum;
4035 	}
4036 
4037 	return 0;
4038 }
4039 
null_compute_pseudo(struct sk_buff * skb,int proto)4040 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4041 {
4042 	return 0;
4043 }
4044 
4045 /* Perform checksum validate (init). Note that this is a macro since we only
4046  * want to calculate the pseudo header which is an input function if necessary.
4047  * First we try to validate without any computation (checksum unnecessary) and
4048  * then calculate based on checksum complete calling the function to compute
4049  * pseudo header.
4050  *
4051  * Return values:
4052  *   0: checksum is validated or try to in skb_checksum_complete
4053  *   non-zero: value of invalid checksum
4054  */
4055 #define __skb_checksum_validate(skb, proto, complete,			\
4056 				zero_okay, check, compute_pseudo)	\
4057 ({									\
4058 	__sum16 __ret = 0;						\
4059 	skb->csum_valid = 0;						\
4060 	if (__skb_checksum_validate_needed(skb, zero_okay, check))	\
4061 		__ret = __skb_checksum_validate_complete(skb,		\
4062 				complete, compute_pseudo(skb, proto));	\
4063 	__ret;								\
4064 })
4065 
4066 #define skb_checksum_init(skb, proto, compute_pseudo)			\
4067 	__skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4068 
4069 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo)	\
4070 	__skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4071 
4072 #define skb_checksum_validate(skb, proto, compute_pseudo)		\
4073 	__skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4074 
4075 #define skb_checksum_validate_zero_check(skb, proto, check,		\
4076 					 compute_pseudo)		\
4077 	__skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4078 
4079 #define skb_checksum_simple_validate(skb)				\
4080 	__skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4081 
__skb_checksum_convert_check(struct sk_buff * skb)4082 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4083 {
4084 	return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4085 }
4086 
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4087 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4088 {
4089 	skb->csum = ~pseudo;
4090 	skb->ip_summed = CHECKSUM_COMPLETE;
4091 }
4092 
4093 #define skb_checksum_try_convert(skb, proto, compute_pseudo)	\
4094 do {									\
4095 	if (__skb_checksum_convert_check(skb))				\
4096 		__skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4097 } while (0)
4098 
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4099 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4100 					      u16 start, u16 offset)
4101 {
4102 	skb->ip_summed = CHECKSUM_PARTIAL;
4103 	skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4104 	skb->csum_offset = offset - start;
4105 }
4106 
4107 /* Update skbuf and packet to reflect the remote checksum offload operation.
4108  * When called, ptr indicates the starting point for skb->csum when
4109  * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4110  * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4111  */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4112 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4113 				       int start, int offset, bool nopartial)
4114 {
4115 	__wsum delta;
4116 
4117 	if (!nopartial) {
4118 		skb_remcsum_adjust_partial(skb, ptr, start, offset);
4119 		return;
4120 	}
4121 
4122 	 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4123 		__skb_checksum_complete(skb);
4124 		skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4125 	}
4126 
4127 	delta = remcsum_adjust(ptr, skb->csum, start, offset);
4128 
4129 	/* Adjust skb->csum since we changed the packet */
4130 	skb->csum = csum_add(skb->csum, delta);
4131 }
4132 
skb_nfct(const struct sk_buff * skb)4133 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4134 {
4135 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4136 	return (void *)(skb->_nfct & NFCT_PTRMASK);
4137 #else
4138 	return NULL;
4139 #endif
4140 }
4141 
skb_get_nfct(const struct sk_buff * skb)4142 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4143 {
4144 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4145 	return skb->_nfct;
4146 #else
4147 	return 0UL;
4148 #endif
4149 }
4150 
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4151 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4152 {
4153 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4154 	skb->_nfct = nfct;
4155 #endif
4156 }
4157 
4158 #ifdef CONFIG_SKB_EXTENSIONS
4159 enum skb_ext_id {
4160 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4161 	SKB_EXT_BRIDGE_NF,
4162 #endif
4163 #ifdef CONFIG_XFRM
4164 	SKB_EXT_SEC_PATH,
4165 #endif
4166 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4167 	TC_SKB_EXT,
4168 #endif
4169 #if IS_ENABLED(CONFIG_MPTCP)
4170 	SKB_EXT_MPTCP,
4171 #endif
4172 	SKB_EXT_NUM, /* must be last */
4173 };
4174 
4175 /**
4176  *	struct skb_ext - sk_buff extensions
4177  *	@refcnt: 1 on allocation, deallocated on 0
4178  *	@offset: offset to add to @data to obtain extension address
4179  *	@chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4180  *	@data: start of extension data, variable sized
4181  *
4182  *	Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4183  *	to use 'u8' types while allowing up to 2kb worth of extension data.
4184  */
4185 struct skb_ext {
4186 	refcount_t refcnt;
4187 	u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4188 	u8 chunks;		/* same */
4189 	char data[] __aligned(8);
4190 };
4191 
4192 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4193 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4194 		    struct skb_ext *ext);
4195 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4196 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4197 void __skb_ext_put(struct skb_ext *ext);
4198 
skb_ext_put(struct sk_buff * skb)4199 static inline void skb_ext_put(struct sk_buff *skb)
4200 {
4201 	if (skb->active_extensions)
4202 		__skb_ext_put(skb->extensions);
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,
4206 				  const struct sk_buff *src)
4207 {
4208 	dst->active_extensions = src->active_extensions;
4209 
4210 	if (src->active_extensions) {
4211 		struct skb_ext *ext = src->extensions;
4212 
4213 		refcount_inc(&ext->refcnt);
4214 		dst->extensions = ext;
4215 	}
4216 }
4217 
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4218 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4219 {
4220 	skb_ext_put(dst);
4221 	__skb_ext_copy(dst, src);
4222 }
4223 
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4224 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4225 {
4226 	return !!ext->offset[i];
4227 }
4228 
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4229 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4230 {
4231 	return skb->active_extensions & (1 << id);
4232 }
4233 
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4234 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4235 {
4236 	if (skb_ext_exist(skb, id))
4237 		__skb_ext_del(skb, id);
4238 }
4239 
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4240 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4241 {
4242 	if (skb_ext_exist(skb, id)) {
4243 		struct skb_ext *ext = skb->extensions;
4244 
4245 		return (void *)ext + (ext->offset[id] << 3);
4246 	}
4247 
4248 	return NULL;
4249 }
4250 
skb_ext_reset(struct sk_buff * skb)4251 static inline void skb_ext_reset(struct sk_buff *skb)
4252 {
4253 	if (unlikely(skb->active_extensions)) {
4254 		__skb_ext_put(skb->extensions);
4255 		skb->active_extensions = 0;
4256 	}
4257 }
4258 
skb_has_extensions(struct sk_buff * skb)4259 static inline bool skb_has_extensions(struct sk_buff *skb)
4260 {
4261 	return unlikely(skb->active_extensions);
4262 }
4263 #else
skb_ext_put(struct sk_buff * skb)4264 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4265 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4266 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4267 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)4268 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4269 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4270 #endif /* CONFIG_SKB_EXTENSIONS */
4271 
nf_reset_ct(struct sk_buff * skb)4272 static inline void nf_reset_ct(struct sk_buff *skb)
4273 {
4274 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4275 	nf_conntrack_put(skb_nfct(skb));
4276 	skb->_nfct = 0;
4277 #endif
4278 }
4279 
nf_reset_trace(struct sk_buff * skb)4280 static inline void nf_reset_trace(struct sk_buff *skb)
4281 {
4282 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4283 	skb->nf_trace = 0;
4284 #endif
4285 }
4286 
ipvs_reset(struct sk_buff * skb)4287 static inline void ipvs_reset(struct sk_buff *skb)
4288 {
4289 #if IS_ENABLED(CONFIG_IP_VS)
4290 	skb->ipvs_property = 0;
4291 #endif
4292 }
4293 
4294 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4295 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4296 			     bool copy)
4297 {
4298 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4299 	dst->_nfct = src->_nfct;
4300 	nf_conntrack_get(skb_nfct(src));
4301 #endif
4302 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4303 	if (copy)
4304 		dst->nf_trace = src->nf_trace;
4305 #endif
4306 }
4307 
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4308 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4309 {
4310 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4311 	nf_conntrack_put(skb_nfct(dst));
4312 #endif
4313 	__nf_copy(dst, src, true);
4314 }
4315 
4316 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4317 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4318 {
4319 	to->secmark = from->secmark;
4320 }
4321 
skb_init_secmark(struct sk_buff * skb)4322 static inline void skb_init_secmark(struct sk_buff *skb)
4323 {
4324 	skb->secmark = 0;
4325 }
4326 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4327 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4328 { }
4329 
skb_init_secmark(struct sk_buff * skb)4330 static inline void skb_init_secmark(struct sk_buff *skb)
4331 { }
4332 #endif
4333 
secpath_exists(const struct sk_buff * skb)4334 static inline int secpath_exists(const struct sk_buff *skb)
4335 {
4336 #ifdef CONFIG_XFRM
4337 	return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4338 #else
4339 	return 0;
4340 #endif
4341 }
4342 
skb_irq_freeable(const struct sk_buff * skb)4343 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4344 {
4345 	return !skb->destructor &&
4346 		!secpath_exists(skb) &&
4347 		!skb_nfct(skb) &&
4348 		!skb->_skb_refdst &&
4349 		!skb_has_frag_list(skb);
4350 }
4351 
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)4352 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4353 {
4354 	skb->queue_mapping = queue_mapping;
4355 }
4356 
skb_get_queue_mapping(const struct sk_buff * skb)4357 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4358 {
4359 	return skb->queue_mapping;
4360 }
4361 
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)4362 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4363 {
4364 	to->queue_mapping = from->queue_mapping;
4365 }
4366 
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)4367 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4368 {
4369 	skb->queue_mapping = rx_queue + 1;
4370 }
4371 
skb_get_rx_queue(const struct sk_buff * skb)4372 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4373 {
4374 	return skb->queue_mapping - 1;
4375 }
4376 
skb_rx_queue_recorded(const struct sk_buff * skb)4377 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4378 {
4379 	return skb->queue_mapping != 0;
4380 }
4381 
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)4382 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4383 {
4384 	skb->dst_pending_confirm = val;
4385 }
4386 
skb_get_dst_pending_confirm(const struct sk_buff * skb)4387 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4388 {
4389 	return skb->dst_pending_confirm != 0;
4390 }
4391 
skb_sec_path(const struct sk_buff * skb)4392 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4393 {
4394 #ifdef CONFIG_XFRM
4395 	return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4396 #else
4397 	return NULL;
4398 #endif
4399 }
4400 
4401 /* Keeps track of mac header offset relative to skb->head.
4402  * It is useful for TSO of Tunneling protocol. e.g. GRE.
4403  * For non-tunnel skb it points to skb_mac_header() and for
4404  * tunnel skb it points to outer mac header.
4405  * Keeps track of level of encapsulation of network headers.
4406  */
4407 struct skb_gso_cb {
4408 	union {
4409 		int	mac_offset;
4410 		int	data_offset;
4411 	};
4412 	int	encap_level;
4413 	__wsum	csum;
4414 	__u16	csum_start;
4415 };
4416 #define SKB_GSO_CB_OFFSET	32
4417 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4418 
skb_tnl_header_len(const struct sk_buff * inner_skb)4419 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4420 {
4421 	return (skb_mac_header(inner_skb) - inner_skb->head) -
4422 		SKB_GSO_CB(inner_skb)->mac_offset;
4423 }
4424 
gso_pskb_expand_head(struct sk_buff * skb,int extra)4425 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4426 {
4427 	int new_headroom, headroom;
4428 	int ret;
4429 
4430 	headroom = skb_headroom(skb);
4431 	ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4432 	if (ret)
4433 		return ret;
4434 
4435 	new_headroom = skb_headroom(skb);
4436 	SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4437 	return 0;
4438 }
4439 
gso_reset_checksum(struct sk_buff * skb,__wsum res)4440 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4441 {
4442 	/* Do not update partial checksums if remote checksum is enabled. */
4443 	if (skb->remcsum_offload)
4444 		return;
4445 
4446 	SKB_GSO_CB(skb)->csum = res;
4447 	SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4448 }
4449 
4450 /* Compute the checksum for a gso segment. First compute the checksum value
4451  * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4452  * then add in skb->csum (checksum from csum_start to end of packet).
4453  * skb->csum and csum_start are then updated to reflect the checksum of the
4454  * resultant packet starting from the transport header-- the resultant checksum
4455  * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4456  * header.
4457  */
gso_make_checksum(struct sk_buff * skb,__wsum res)4458 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4459 {
4460 	unsigned char *csum_start = skb_transport_header(skb);
4461 	int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4462 	__wsum partial = SKB_GSO_CB(skb)->csum;
4463 
4464 	SKB_GSO_CB(skb)->csum = res;
4465 	SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4466 
4467 	return csum_fold(csum_partial(csum_start, plen, partial));
4468 }
4469 
skb_is_gso(const struct sk_buff * skb)4470 static inline bool skb_is_gso(const struct sk_buff *skb)
4471 {
4472 	return skb_shinfo(skb)->gso_size;
4473 }
4474 
4475 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)4476 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4477 {
4478 	return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4479 }
4480 
4481 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)4482 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4483 {
4484 	return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4485 }
4486 
4487 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)4488 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4489 {
4490 	return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4491 }
4492 
skb_gso_reset(struct sk_buff * skb)4493 static inline void skb_gso_reset(struct sk_buff *skb)
4494 {
4495 	skb_shinfo(skb)->gso_size = 0;
4496 	skb_shinfo(skb)->gso_segs = 0;
4497 	skb_shinfo(skb)->gso_type = 0;
4498 }
4499 
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)4500 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4501 					 u16 increment)
4502 {
4503 	if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4504 		return;
4505 	shinfo->gso_size += increment;
4506 }
4507 
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)4508 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4509 					 u16 decrement)
4510 {
4511 	if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4512 		return;
4513 	shinfo->gso_size -= decrement;
4514 }
4515 
4516 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4517 
skb_warn_if_lro(const struct sk_buff * skb)4518 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4519 {
4520 	/* LRO sets gso_size but not gso_type, whereas if GSO is really
4521 	 * wanted then gso_type will be set. */
4522 	const struct skb_shared_info *shinfo = skb_shinfo(skb);
4523 
4524 	if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4525 	    unlikely(shinfo->gso_type == 0)) {
4526 		__skb_warn_lro_forwarding(skb);
4527 		return true;
4528 	}
4529 	return false;
4530 }
4531 
skb_forward_csum(struct sk_buff * skb)4532 static inline void skb_forward_csum(struct sk_buff *skb)
4533 {
4534 	/* Unfortunately we don't support this one.  Any brave souls? */
4535 	if (skb->ip_summed == CHECKSUM_COMPLETE)
4536 		skb->ip_summed = CHECKSUM_NONE;
4537 }
4538 
4539 /**
4540  * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4541  * @skb: skb to check
4542  *
4543  * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4544  * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4545  * use this helper, to document places where we make this assertion.
4546  */
skb_checksum_none_assert(const struct sk_buff * skb)4547 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4548 {
4549 #ifdef DEBUG
4550 	BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4551 #endif
4552 }
4553 
4554 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4555 
4556 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4557 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4558 				     unsigned int transport_len,
4559 				     __sum16(*skb_chkf)(struct sk_buff *skb));
4560 
4561 /**
4562  * skb_head_is_locked - Determine if the skb->head is locked down
4563  * @skb: skb to check
4564  *
4565  * The head on skbs build around a head frag can be removed if they are
4566  * not cloned.  This function returns true if the skb head is locked down
4567  * due to either being allocated via kmalloc, or by being a clone with
4568  * multiple references to the head.
4569  */
skb_head_is_locked(const struct sk_buff * skb)4570 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4571 {
4572 	return !skb->head_frag || skb_cloned(skb);
4573 }
4574 
4575 /* Local Checksum Offload.
4576  * Compute outer checksum based on the assumption that the
4577  * inner checksum will be offloaded later.
4578  * See Documentation/networking/checksum-offloads.rst for
4579  * explanation of how this works.
4580  * Fill in outer checksum adjustment (e.g. with sum of outer
4581  * pseudo-header) before calling.
4582  * Also ensure that inner checksum is in linear data area.
4583  */
lco_csum(struct sk_buff * skb)4584 static inline __wsum lco_csum(struct sk_buff *skb)
4585 {
4586 	unsigned char *csum_start = skb_checksum_start(skb);
4587 	unsigned char *l4_hdr = skb_transport_header(skb);
4588 	__wsum partial;
4589 
4590 	/* Start with complement of inner checksum adjustment */
4591 	partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4592 						    skb->csum_offset));
4593 
4594 	/* Add in checksum of our headers (incl. outer checksum
4595 	 * adjustment filled in by caller) and return result.
4596 	 */
4597 	return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4598 }
4599 
skb_is_redirected(const struct sk_buff * skb)4600 static inline bool skb_is_redirected(const struct sk_buff *skb)
4601 {
4602 #ifdef CONFIG_NET_REDIRECT
4603 	return skb->redirected;
4604 #else
4605 	return false;
4606 #endif
4607 }
4608 
skb_set_redirected(struct sk_buff * skb,bool from_ingress)4609 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4610 {
4611 #ifdef CONFIG_NET_REDIRECT
4612 	skb->redirected = 1;
4613 	skb->from_ingress = from_ingress;
4614 	if (skb->from_ingress)
4615 		skb->tstamp = 0;
4616 #endif
4617 }
4618 
skb_reset_redirect(struct sk_buff * skb)4619 static inline void skb_reset_redirect(struct sk_buff *skb)
4620 {
4621 #ifdef CONFIG_NET_REDIRECT
4622 	skb->redirected = 0;
4623 #endif
4624 }
4625 
skb_csum_is_sctp(struct sk_buff * skb)4626 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
4627 {
4628 	return skb->csum_not_inet;
4629 }
4630 
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)4631 static inline void skb_set_kcov_handle(struct sk_buff *skb,
4632 				       const u64 kcov_handle)
4633 {
4634 #ifdef CONFIG_KCOV
4635 	skb->kcov_handle = kcov_handle;
4636 #endif
4637 }
4638 
skb_get_kcov_handle(struct sk_buff * skb)4639 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
4640 {
4641 #ifdef CONFIG_KCOV
4642 	return skb->kcov_handle;
4643 #else
4644 	return 0;
4645 #endif
4646 }
4647 
4648 #endif	/* __KERNEL__ */
4649 #endif	/* _LINUX_SKBUFF_H */
4650