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 <linux/llist.h>
40 #include <net/flow.h>
41 #include <net/page_pool.h>
42 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
43 #include <linux/netfilter/nf_conntrack_common.h>
44 #endif
45 #include <net/net_debug.h>
46 #include <net/dropreason.h>
47 
48 /**
49  * DOC: skb checksums
50  *
51  * The interface for checksum offload between the stack and networking drivers
52  * is as follows...
53  *
54  * IP checksum related features
55  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
56  *
57  * Drivers advertise checksum offload capabilities in the features of a device.
58  * From the stack's point of view these are capabilities offered by the driver.
59  * A driver typically only advertises features that it is capable of offloading
60  * to its device.
61  *
62  * .. flat-table:: Checksum related device features
63  *   :widths: 1 10
64  *
65  *   * - %NETIF_F_HW_CSUM
66  *     - The driver (or its device) is able to compute one
67  *	 IP (one's complement) checksum for any combination
68  *	 of protocols or protocol layering. The checksum is
69  *	 computed and set in a packet per the CHECKSUM_PARTIAL
70  *	 interface (see below).
71  *
72  *   * - %NETIF_F_IP_CSUM
73  *     - Driver (device) is only able to checksum plain
74  *	 TCP or UDP packets over IPv4. These are specifically
75  *	 unencapsulated packets of the form IPv4|TCP or
76  *	 IPv4|UDP where the Protocol field in the IPv4 header
77  *	 is TCP or UDP. The IPv4 header may contain IP options.
78  *	 This feature cannot be set in features for a device
79  *	 with NETIF_F_HW_CSUM also set. This feature is being
80  *	 DEPRECATED (see below).
81  *
82  *   * - %NETIF_F_IPV6_CSUM
83  *     - Driver (device) is only able to checksum plain
84  *	 TCP or UDP packets over IPv6. These are specifically
85  *	 unencapsulated packets of the form IPv6|TCP or
86  *	 IPv6|UDP where the Next Header field in the IPv6
87  *	 header is either TCP or UDP. IPv6 extension headers
88  *	 are not supported with this feature. This feature
89  *	 cannot be set in features for a device with
90  *	 NETIF_F_HW_CSUM also set. This feature is being
91  *	 DEPRECATED (see below).
92  *
93  *   * - %NETIF_F_RXCSUM
94  *     - Driver (device) performs receive checksum offload.
95  *	 This flag is only used to disable the RX checksum
96  *	 feature for a device. The stack will accept receive
97  *	 checksum indication in packets received on a device
98  *	 regardless of whether NETIF_F_RXCSUM is set.
99  *
100  * Checksumming of received packets by device
101  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
102  *
103  * Indication of checksum verification is set in &sk_buff.ip_summed.
104  * Possible values are:
105  *
106  * - %CHECKSUM_NONE
107  *
108  *   Device did not checksum this packet e.g. due to lack of capabilities.
109  *   The packet contains full (though not verified) checksum in packet but
110  *   not in skb->csum. Thus, skb->csum is undefined in this case.
111  *
112  * - %CHECKSUM_UNNECESSARY
113  *
114  *   The hardware you're dealing with doesn't calculate the full checksum
115  *   (as in %CHECKSUM_COMPLETE), but it does parse headers and verify checksums
116  *   for specific protocols. For such packets it will set %CHECKSUM_UNNECESSARY
117  *   if their checksums are okay. &sk_buff.csum is still undefined in this case
118  *   though. A driver or device must never modify the checksum field in the
119  *   packet even if checksum is verified.
120  *
121  *   %CHECKSUM_UNNECESSARY is applicable to following protocols:
122  *
123  *     - TCP: IPv6 and IPv4.
124  *     - UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
125  *       zero UDP checksum for either IPv4 or IPv6, the networking stack
126  *       may perform further validation in this case.
127  *     - GRE: only if the checksum is present in the header.
128  *     - SCTP: indicates the CRC in SCTP header has been validated.
129  *     - FCOE: indicates the CRC in FC frame has been validated.
130  *
131  *   &sk_buff.csum_level indicates the number of consecutive checksums found in
132  *   the packet minus one that have been verified as %CHECKSUM_UNNECESSARY.
133  *   For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
134  *   and a device is able to verify the checksums for UDP (possibly zero),
135  *   GRE (checksum flag is set) and TCP, &sk_buff.csum_level would be set to
136  *   two. If the device were only able to verify the UDP checksum and not
137  *   GRE, either because it doesn't support GRE checksum or because GRE
138  *   checksum is bad, skb->csum_level would be set to zero (TCP checksum is
139  *   not considered in this case).
140  *
141  * - %CHECKSUM_COMPLETE
142  *
143  *   This is the most generic way. The device supplied checksum of the _whole_
144  *   packet as seen by netif_rx() and fills in &sk_buff.csum. This means the
145  *   hardware doesn't need to parse L3/L4 headers to implement this.
146  *
147  *   Notes:
148  *
149  *   - Even if device supports only some protocols, but is able to produce
150  *     skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
151  *   - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
152  *
153  * - %CHECKSUM_PARTIAL
154  *
155  *   A checksum is set up to be offloaded to a device as described in the
156  *   output description for CHECKSUM_PARTIAL. This may occur on a packet
157  *   received directly from another Linux OS, e.g., a virtualized Linux kernel
158  *   on the same host, or it may be set in the input path in GRO or remote
159  *   checksum offload. For the purposes of checksum verification, the checksum
160  *   referred to by skb->csum_start + skb->csum_offset and any preceding
161  *   checksums in the packet are considered verified. Any checksums in the
162  *   packet that are after the checksum being offloaded are not considered to
163  *   be verified.
164  *
165  * Checksumming on transmit for non-GSO
166  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
167  *
168  * The stack requests checksum offload in the &sk_buff.ip_summed for a packet.
169  * Values are:
170  *
171  * - %CHECKSUM_PARTIAL
172  *
173  *   The driver is required to checksum the packet as seen by hard_start_xmit()
174  *   from &sk_buff.csum_start up to the end, and to record/write the checksum at
175  *   offset &sk_buff.csum_start + &sk_buff.csum_offset.
176  *   A driver may verify that the
177  *   csum_start and csum_offset values are valid values given the length and
178  *   offset of the packet, but it should not attempt to validate that the
179  *   checksum refers to a legitimate transport layer checksum -- it is the
180  *   purview of the stack to validate that csum_start and csum_offset are set
181  *   correctly.
182  *
183  *   When the stack requests checksum offload for a packet, the driver MUST
184  *   ensure that the checksum is set correctly. A driver can either offload the
185  *   checksum calculation to the device, or call skb_checksum_help (in the case
186  *   that the device does not support offload for a particular checksum).
187  *
188  *   %NETIF_F_IP_CSUM and %NETIF_F_IPV6_CSUM are being deprecated in favor of
189  *   %NETIF_F_HW_CSUM. New devices should use %NETIF_F_HW_CSUM to indicate
190  *   checksum offload capability.
191  *   skb_csum_hwoffload_help() can be called to resolve %CHECKSUM_PARTIAL based
192  *   on network device checksumming capabilities: if a packet does not match
193  *   them, skb_checksum_help() or skb_crc32c_help() (depending on the value of
194  *   &sk_buff.csum_not_inet, see :ref:`crc`)
195  *   is called to resolve the checksum.
196  *
197  * - %CHECKSUM_NONE
198  *
199  *   The skb was already checksummed by the protocol, or a checksum is not
200  *   required.
201  *
202  * - %CHECKSUM_UNNECESSARY
203  *
204  *   This has the same meaning as CHECKSUM_NONE for checksum offload on
205  *   output.
206  *
207  * - %CHECKSUM_COMPLETE
208  *
209  *   Not used in checksum output. If a driver observes a packet with this value
210  *   set in skbuff, it should treat the packet as if %CHECKSUM_NONE were set.
211  *
212  * .. _crc:
213  *
214  * Non-IP checksum (CRC) offloads
215  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
216  *
217  * .. flat-table::
218  *   :widths: 1 10
219  *
220  *   * - %NETIF_F_SCTP_CRC
221  *     - This feature indicates that a device is capable of
222  *	 offloading the SCTP CRC in a packet. To perform this offload the stack
223  *	 will set csum_start and csum_offset accordingly, set ip_summed to
224  *	 %CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication
225  *	 in the skbuff that the %CHECKSUM_PARTIAL refers to CRC32c.
226  *	 A driver that supports both IP checksum offload and SCTP CRC32c offload
227  *	 must verify which offload is configured for a packet by testing the
228  *	 value of &sk_buff.csum_not_inet; skb_crc32c_csum_help() is provided to
229  *	 resolve %CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
230  *
231  *   * - %NETIF_F_FCOE_CRC
232  *     - This feature indicates that a device is capable of offloading the FCOE
233  *	 CRC in a packet. To perform this offload the stack will set ip_summed
234  *	 to %CHECKSUM_PARTIAL and set csum_start and csum_offset
235  *	 accordingly. Note that there is no indication in the skbuff that the
236  *	 %CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
237  *	 both IP checksum offload and FCOE CRC offload must verify which offload
238  *	 is configured for a packet, presumably by inspecting packet headers.
239  *
240  * Checksumming on output with GSO
241  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
242  *
243  * In the case of a GSO packet (skb_is_gso() is true), checksum offload
244  * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
245  * gso_type is %SKB_GSO_TCPV4 or %SKB_GSO_TCPV6, TCP checksum offload as
246  * part of the GSO operation is implied. If a checksum is being offloaded
247  * with GSO then ip_summed is %CHECKSUM_PARTIAL, and both csum_start and
248  * csum_offset are set to refer to the outermost checksum being offloaded
249  * (two offloaded checksums are possible with UDP encapsulation).
250  */
251 
252 /* Don't change this without changing skb_csum_unnecessary! */
253 #define CHECKSUM_NONE		0
254 #define CHECKSUM_UNNECESSARY	1
255 #define CHECKSUM_COMPLETE	2
256 #define CHECKSUM_PARTIAL	3
257 
258 /* Maximum value in skb->csum_level */
259 #define SKB_MAX_CSUM_LEVEL	3
260 
261 #define SKB_DATA_ALIGN(X)	ALIGN(X, SMP_CACHE_BYTES)
262 #define SKB_WITH_OVERHEAD(X)	\
263 	((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
264 #define SKB_MAX_ORDER(X, ORDER) \
265 	SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
266 #define SKB_MAX_HEAD(X)		(SKB_MAX_ORDER((X), 0))
267 #define SKB_MAX_ALLOC		(SKB_MAX_ORDER(0, 2))
268 
269 /* return minimum truesize of one skb containing X bytes of data */
270 #define SKB_TRUESIZE(X) ((X) +						\
271 			 SKB_DATA_ALIGN(sizeof(struct sk_buff)) +	\
272 			 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
273 
274 struct ahash_request;
275 struct net_device;
276 struct scatterlist;
277 struct pipe_inode_info;
278 struct iov_iter;
279 struct napi_struct;
280 struct bpf_prog;
281 union bpf_attr;
282 struct skb_ext;
283 
284 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
285 struct nf_bridge_info {
286 	enum {
287 		BRNF_PROTO_UNCHANGED,
288 		BRNF_PROTO_8021Q,
289 		BRNF_PROTO_PPPOE
290 	} orig_proto:8;
291 	u8			pkt_otherhost:1;
292 	u8			in_prerouting:1;
293 	u8			bridged_dnat:1;
294 	__u16			frag_max_size;
295 	struct net_device	*physindev;
296 
297 	/* always valid & non-NULL from FORWARD on, for physdev match */
298 	struct net_device	*physoutdev;
299 	union {
300 		/* prerouting: detect dnat in orig/reply direction */
301 		__be32          ipv4_daddr;
302 		struct in6_addr ipv6_daddr;
303 
304 		/* after prerouting + nat detected: store original source
305 		 * mac since neigh resolution overwrites it, only used while
306 		 * skb is out in neigh layer.
307 		 */
308 		char neigh_header[8];
309 	};
310 };
311 #endif
312 
313 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
314 /* Chain in tc_skb_ext will be used to share the tc chain with
315  * ovs recirc_id. It will be set to the current chain by tc
316  * and read by ovs to recirc_id.
317  */
318 struct tc_skb_ext {
319 	__u32 chain;
320 	__u16 mru;
321 	__u16 zone;
322 	u8 post_ct:1;
323 	u8 post_ct_snat:1;
324 	u8 post_ct_dnat:1;
325 };
326 #endif
327 
328 struct sk_buff_head {
329 	/* These two members must be first to match sk_buff. */
330 	struct_group_tagged(sk_buff_list, list,
331 		struct sk_buff	*next;
332 		struct sk_buff	*prev;
333 	);
334 
335 	__u32		qlen;
336 	spinlock_t	lock;
337 };
338 
339 struct sk_buff;
340 
341 /* To allow 64K frame to be packed as single skb without frag_list we
342  * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
343  * buffers which do not start on a page boundary.
344  *
345  * Since GRO uses frags we allocate at least 16 regardless of page
346  * size.
347  */
348 #if (65536/PAGE_SIZE + 1) < 16
349 #define MAX_SKB_FRAGS 16UL
350 #else
351 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
352 #endif
353 extern int sysctl_max_skb_frags;
354 
355 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
356  * segment using its current segmentation instead.
357  */
358 #define GSO_BY_FRAGS	0xFFFF
359 
360 typedef struct bio_vec skb_frag_t;
361 
362 /**
363  * skb_frag_size() - Returns the size of a skb fragment
364  * @frag: skb fragment
365  */
skb_frag_size(const skb_frag_t * frag)366 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
367 {
368 	return frag->bv_len;
369 }
370 
371 /**
372  * skb_frag_size_set() - Sets the size of a skb fragment
373  * @frag: skb fragment
374  * @size: size of fragment
375  */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)376 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
377 {
378 	frag->bv_len = size;
379 }
380 
381 /**
382  * skb_frag_size_add() - Increments the size of a skb fragment by @delta
383  * @frag: skb fragment
384  * @delta: value to add
385  */
skb_frag_size_add(skb_frag_t * frag,int delta)386 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
387 {
388 	frag->bv_len += delta;
389 }
390 
391 /**
392  * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
393  * @frag: skb fragment
394  * @delta: value to subtract
395  */
skb_frag_size_sub(skb_frag_t * frag,int delta)396 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
397 {
398 	frag->bv_len -= delta;
399 }
400 
401 /**
402  * skb_frag_must_loop - Test if %p is a high memory page
403  * @p: fragment's page
404  */
skb_frag_must_loop(struct page * p)405 static inline bool skb_frag_must_loop(struct page *p)
406 {
407 #if defined(CONFIG_HIGHMEM)
408 	if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
409 		return true;
410 #endif
411 	return false;
412 }
413 
414 /**
415  *	skb_frag_foreach_page - loop over pages in a fragment
416  *
417  *	@f:		skb frag to operate on
418  *	@f_off:		offset from start of f->bv_page
419  *	@f_len:		length from f_off to loop over
420  *	@p:		(temp var) current page
421  *	@p_off:		(temp var) offset from start of current page,
422  *	                           non-zero only on first page.
423  *	@p_len:		(temp var) length in current page,
424  *				   < PAGE_SIZE only on first and last page.
425  *	@copied:	(temp var) length so far, excluding current p_len.
426  *
427  *	A fragment can hold a compound page, in which case per-page
428  *	operations, notably kmap_atomic, must be called for each
429  *	regular page.
430  */
431 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied)	\
432 	for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT),		\
433 	     p_off = (f_off) & (PAGE_SIZE - 1),				\
434 	     p_len = skb_frag_must_loop(p) ?				\
435 	     min_t(u32, f_len, PAGE_SIZE - p_off) : f_len,		\
436 	     copied = 0;						\
437 	     copied < f_len;						\
438 	     copied += p_len, p++, p_off = 0,				\
439 	     p_len = min_t(u32, f_len - copied, PAGE_SIZE))		\
440 
441 #define HAVE_HW_TIME_STAMP
442 
443 /**
444  * struct skb_shared_hwtstamps - hardware time stamps
445  * @hwtstamp:		hardware time stamp transformed into duration
446  *			since arbitrary point in time
447  * @netdev_data:	address/cookie of network device driver used as
448  *			reference to actual hardware time stamp
449  *
450  * Software time stamps generated by ktime_get_real() are stored in
451  * skb->tstamp.
452  *
453  * hwtstamps can only be compared against other hwtstamps from
454  * the same device.
455  *
456  * This structure is attached to packets as part of the
457  * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
458  */
459 struct skb_shared_hwtstamps {
460 	union {
461 		ktime_t	hwtstamp;
462 		void *netdev_data;
463 	};
464 };
465 
466 /* Definitions for tx_flags in struct skb_shared_info */
467 enum {
468 	/* generate hardware time stamp */
469 	SKBTX_HW_TSTAMP = 1 << 0,
470 
471 	/* generate software time stamp when queueing packet to NIC */
472 	SKBTX_SW_TSTAMP = 1 << 1,
473 
474 	/* device driver is going to provide hardware time stamp */
475 	SKBTX_IN_PROGRESS = 1 << 2,
476 
477 	/* generate hardware time stamp based on cycles if supported */
478 	SKBTX_HW_TSTAMP_USE_CYCLES = 1 << 3,
479 
480 	/* generate wifi status information (where possible) */
481 	SKBTX_WIFI_STATUS = 1 << 4,
482 
483 	/* determine hardware time stamp based on time or cycles */
484 	SKBTX_HW_TSTAMP_NETDEV = 1 << 5,
485 
486 	/* generate software time stamp when entering packet scheduling */
487 	SKBTX_SCHED_TSTAMP = 1 << 6,
488 };
489 
490 #define SKBTX_ANY_SW_TSTAMP	(SKBTX_SW_TSTAMP    | \
491 				 SKBTX_SCHED_TSTAMP)
492 #define SKBTX_ANY_TSTAMP	(SKBTX_HW_TSTAMP | \
493 				 SKBTX_HW_TSTAMP_USE_CYCLES | \
494 				 SKBTX_ANY_SW_TSTAMP)
495 
496 /* Definitions for flags in struct skb_shared_info */
497 enum {
498 	/* use zcopy routines */
499 	SKBFL_ZEROCOPY_ENABLE = BIT(0),
500 
501 	/* This indicates at least one fragment might be overwritten
502 	 * (as in vmsplice(), sendfile() ...)
503 	 * If we need to compute a TX checksum, we'll need to copy
504 	 * all frags to avoid possible bad checksum
505 	 */
506 	SKBFL_SHARED_FRAG = BIT(1),
507 
508 	/* segment contains only zerocopy data and should not be
509 	 * charged to the kernel memory.
510 	 */
511 	SKBFL_PURE_ZEROCOPY = BIT(2),
512 
513 	SKBFL_DONT_ORPHAN = BIT(3),
514 
515 	/* page references are managed by the ubuf_info, so it's safe to
516 	 * use frags only up until ubuf_info is released
517 	 */
518 	SKBFL_MANAGED_FRAG_REFS = BIT(4),
519 };
520 
521 #define SKBFL_ZEROCOPY_FRAG	(SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
522 #define SKBFL_ALL_ZEROCOPY	(SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \
523 				 SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS)
524 
525 /*
526  * The callback notifies userspace to release buffers when skb DMA is done in
527  * lower device, the skb last reference should be 0 when calling this.
528  * The zerocopy_success argument is true if zero copy transmit occurred,
529  * false on data copy or out of memory error caused by data copy attempt.
530  * The ctx field is used to track device context.
531  * The desc field is used to track userspace buffer index.
532  */
533 struct ubuf_info {
534 	void (*callback)(struct sk_buff *, struct ubuf_info *,
535 			 bool zerocopy_success);
536 	refcount_t refcnt;
537 	u8 flags;
538 };
539 
540 struct ubuf_info_msgzc {
541 	struct ubuf_info ubuf;
542 
543 	union {
544 		struct {
545 			unsigned long desc;
546 			void *ctx;
547 		};
548 		struct {
549 			u32 id;
550 			u16 len;
551 			u16 zerocopy:1;
552 			u32 bytelen;
553 		};
554 	};
555 
556 	struct mmpin {
557 		struct user_struct *user;
558 		unsigned int num_pg;
559 	} mmp;
560 };
561 
562 #define skb_uarg(SKB)	((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
563 #define uarg_to_msgzc(ubuf_ptr)	container_of((ubuf_ptr), struct ubuf_info_msgzc, \
564 					     ubuf)
565 
566 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
567 void mm_unaccount_pinned_pages(struct mmpin *mmp);
568 
569 /* This data is invariant across clones and lives at
570  * the end of the header data, ie. at skb->end.
571  */
572 struct skb_shared_info {
573 	__u8		flags;
574 	__u8		meta_len;
575 	__u8		nr_frags;
576 	__u8		tx_flags;
577 	unsigned short	gso_size;
578 	/* Warning: this field is not always filled in (UFO)! */
579 	unsigned short	gso_segs;
580 	struct sk_buff	*frag_list;
581 	struct skb_shared_hwtstamps hwtstamps;
582 	unsigned int	gso_type;
583 	u32		tskey;
584 
585 	/*
586 	 * Warning : all fields before dataref are cleared in __alloc_skb()
587 	 */
588 	atomic_t	dataref;
589 	unsigned int	xdp_frags_size;
590 
591 	/* Intermediate layers must ensure that destructor_arg
592 	 * remains valid until skb destructor */
593 	void *		destructor_arg;
594 
595 	/* must be last field, see pskb_expand_head() */
596 	skb_frag_t	frags[MAX_SKB_FRAGS];
597 };
598 
599 /**
600  * DOC: dataref and headerless skbs
601  *
602  * Transport layers send out clones of payload skbs they hold for
603  * retransmissions. To allow lower layers of the stack to prepend their headers
604  * we split &skb_shared_info.dataref into two halves.
605  * The lower 16 bits count the overall number of references.
606  * The higher 16 bits indicate how many of the references are payload-only.
607  * skb_header_cloned() checks if skb is allowed to add / write the headers.
608  *
609  * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr
610  * (via __skb_header_release()). Any clone created from marked skb will get
611  * &sk_buff.hdr_len populated with the available headroom.
612  * If there's the only clone in existence it's able to modify the headroom
613  * at will. The sequence of calls inside the transport layer is::
614  *
615  *  <alloc skb>
616  *  skb_reserve()
617  *  __skb_header_release()
618  *  skb_clone()
619  *  // send the clone down the stack
620  *
621  * This is not a very generic construct and it depends on the transport layers
622  * doing the right thing. In practice there's usually only one payload-only skb.
623  * Having multiple payload-only skbs with different lengths of hdr_len is not
624  * possible. The payload-only skbs should never leave their owner.
625  */
626 #define SKB_DATAREF_SHIFT 16
627 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
628 
629 
630 enum {
631 	SKB_FCLONE_UNAVAILABLE,	/* skb has no fclone (from head_cache) */
632 	SKB_FCLONE_ORIG,	/* orig skb (from fclone_cache) */
633 	SKB_FCLONE_CLONE,	/* companion fclone skb (from fclone_cache) */
634 };
635 
636 enum {
637 	SKB_GSO_TCPV4 = 1 << 0,
638 
639 	/* This indicates the skb is from an untrusted source. */
640 	SKB_GSO_DODGY = 1 << 1,
641 
642 	/* This indicates the tcp segment has CWR set. */
643 	SKB_GSO_TCP_ECN = 1 << 2,
644 
645 	SKB_GSO_TCP_FIXEDID = 1 << 3,
646 
647 	SKB_GSO_TCPV6 = 1 << 4,
648 
649 	SKB_GSO_FCOE = 1 << 5,
650 
651 	SKB_GSO_GRE = 1 << 6,
652 
653 	SKB_GSO_GRE_CSUM = 1 << 7,
654 
655 	SKB_GSO_IPXIP4 = 1 << 8,
656 
657 	SKB_GSO_IPXIP6 = 1 << 9,
658 
659 	SKB_GSO_UDP_TUNNEL = 1 << 10,
660 
661 	SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
662 
663 	SKB_GSO_PARTIAL = 1 << 12,
664 
665 	SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
666 
667 	SKB_GSO_SCTP = 1 << 14,
668 
669 	SKB_GSO_ESP = 1 << 15,
670 
671 	SKB_GSO_UDP = 1 << 16,
672 
673 	SKB_GSO_UDP_L4 = 1 << 17,
674 
675 	SKB_GSO_FRAGLIST = 1 << 18,
676 };
677 
678 #if BITS_PER_LONG > 32
679 #define NET_SKBUFF_DATA_USES_OFFSET 1
680 #endif
681 
682 #ifdef NET_SKBUFF_DATA_USES_OFFSET
683 typedef unsigned int sk_buff_data_t;
684 #else
685 typedef unsigned char *sk_buff_data_t;
686 #endif
687 
688 /**
689  * DOC: Basic sk_buff geometry
690  *
691  * struct sk_buff itself is a metadata structure and does not hold any packet
692  * data. All the data is held in associated buffers.
693  *
694  * &sk_buff.head points to the main "head" buffer. The head buffer is divided
695  * into two parts:
696  *
697  *  - data buffer, containing headers and sometimes payload;
698  *    this is the part of the skb operated on by the common helpers
699  *    such as skb_put() or skb_pull();
700  *  - shared info (struct skb_shared_info) which holds an array of pointers
701  *    to read-only data in the (page, offset, length) format.
702  *
703  * Optionally &skb_shared_info.frag_list may point to another skb.
704  *
705  * Basic diagram may look like this::
706  *
707  *                                  ---------------
708  *                                 | sk_buff       |
709  *                                  ---------------
710  *     ,---------------------------  + head
711  *    /          ,-----------------  + data
712  *   /          /      ,-----------  + tail
713  *  |          |      |            , + end
714  *  |          |      |           |
715  *  v          v      v           v
716  *   -----------------------------------------------
717  *  | headroom | data |  tailroom | skb_shared_info |
718  *   -----------------------------------------------
719  *                                 + [page frag]
720  *                                 + [page frag]
721  *                                 + [page frag]
722  *                                 + [page frag]       ---------
723  *                                 + frag_list    --> | sk_buff |
724  *                                                     ---------
725  *
726  */
727 
728 /**
729  *	struct sk_buff - socket buffer
730  *	@next: Next buffer in list
731  *	@prev: Previous buffer in list
732  *	@tstamp: Time we arrived/left
733  *	@skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
734  *		for retransmit timer
735  *	@rbnode: RB tree node, alternative to next/prev for netem/tcp
736  *	@list: queue head
737  *	@ll_node: anchor in an llist (eg socket defer_list)
738  *	@sk: Socket we are owned by
739  *	@ip_defrag_offset: (aka @sk) alternate use of @sk, used in
740  *		fragmentation management
741  *	@dev: Device we arrived on/are leaving by
742  *	@dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
743  *	@cb: Control buffer. Free for use by every layer. Put private vars here
744  *	@_skb_refdst: destination entry (with norefcount bit)
745  *	@sp: the security path, used for xfrm
746  *	@len: Length of actual data
747  *	@data_len: Data length
748  *	@mac_len: Length of link layer header
749  *	@hdr_len: writable header length of cloned skb
750  *	@csum: Checksum (must include start/offset pair)
751  *	@csum_start: Offset from skb->head where checksumming should start
752  *	@csum_offset: Offset from csum_start where checksum should be stored
753  *	@priority: Packet queueing priority
754  *	@ignore_df: allow local fragmentation
755  *	@cloned: Head may be cloned (check refcnt to be sure)
756  *	@ip_summed: Driver fed us an IP checksum
757  *	@nohdr: Payload reference only, must not modify header
758  *	@pkt_type: Packet class
759  *	@fclone: skbuff clone status
760  *	@ipvs_property: skbuff is owned by ipvs
761  *	@inner_protocol_type: whether the inner protocol is
762  *		ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
763  *	@remcsum_offload: remote checksum offload is enabled
764  *	@offload_fwd_mark: Packet was L2-forwarded in hardware
765  *	@offload_l3_fwd_mark: Packet was L3-forwarded in hardware
766  *	@tc_skip_classify: do not classify packet. set by IFB device
767  *	@tc_at_ingress: used within tc_classify to distinguish in/egress
768  *	@redirected: packet was redirected by packet classifier
769  *	@from_ingress: packet was redirected from the ingress path
770  *	@nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
771  *	@peeked: this packet has been seen already, so stats have been
772  *		done for it, don't do them again
773  *	@nf_trace: netfilter packet trace flag
774  *	@protocol: Packet protocol from driver
775  *	@destructor: Destruct function
776  *	@tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
777  *	@_sk_redir: socket redirection information for skmsg
778  *	@_nfct: Associated connection, if any (with nfctinfo bits)
779  *	@nf_bridge: Saved data about a bridged frame - see br_netfilter.c
780  *	@skb_iif: ifindex of device we arrived on
781  *	@tc_index: Traffic control index
782  *	@hash: the packet hash
783  *	@queue_mapping: Queue mapping for multiqueue devices
784  *	@head_frag: skb was allocated from page fragments,
785  *		not allocated by kmalloc() or vmalloc().
786  *	@pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
787  *	@pp_recycle: mark the packet for recycling instead of freeing (implies
788  *		page_pool support on driver)
789  *	@active_extensions: active extensions (skb_ext_id types)
790  *	@ndisc_nodetype: router type (from link layer)
791  *	@ooo_okay: allow the mapping of a socket to a queue to be changed
792  *	@l4_hash: indicate hash is a canonical 4-tuple hash over transport
793  *		ports.
794  *	@sw_hash: indicates hash was computed in software stack
795  *	@wifi_acked_valid: wifi_acked was set
796  *	@wifi_acked: whether frame was acked on wifi or not
797  *	@no_fcs:  Request NIC to treat last 4 bytes as Ethernet FCS
798  *	@encapsulation: indicates the inner headers in the skbuff are valid
799  *	@encap_hdr_csum: software checksum is needed
800  *	@csum_valid: checksum is already valid
801  *	@csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
802  *	@csum_complete_sw: checksum was completed by software
803  *	@csum_level: indicates the number of consecutive checksums found in
804  *		the packet minus one that have been verified as
805  *		CHECKSUM_UNNECESSARY (max 3)
806  *	@scm_io_uring: SKB holds io_uring registered files
807  *	@dst_pending_confirm: need to confirm neighbour
808  *	@decrypted: Decrypted SKB
809  *	@slow_gro: state present at GRO time, slower prepare step required
810  *	@mono_delivery_time: When set, skb->tstamp has the
811  *		delivery_time in mono clock base (i.e. EDT).  Otherwise, the
812  *		skb->tstamp has the (rcv) timestamp at ingress and
813  *		delivery_time at egress.
814  *	@napi_id: id of the NAPI struct this skb came from
815  *	@sender_cpu: (aka @napi_id) source CPU in XPS
816  *	@alloc_cpu: CPU which did the skb allocation.
817  *	@secmark: security marking
818  *	@mark: Generic packet mark
819  *	@reserved_tailroom: (aka @mark) number of bytes of free space available
820  *		at the tail of an sk_buff
821  *	@vlan_present: VLAN tag is present
822  *	@vlan_proto: vlan encapsulation protocol
823  *	@vlan_tci: vlan tag control information
824  *	@inner_protocol: Protocol (encapsulation)
825  *	@inner_ipproto: (aka @inner_protocol) stores ipproto when
826  *		skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
827  *	@inner_transport_header: Inner transport layer header (encapsulation)
828  *	@inner_network_header: Network layer header (encapsulation)
829  *	@inner_mac_header: Link layer header (encapsulation)
830  *	@transport_header: Transport layer header
831  *	@network_header: Network layer header
832  *	@mac_header: Link layer header
833  *	@kcov_handle: KCOV remote handle for remote coverage collection
834  *	@tail: Tail pointer
835  *	@end: End pointer
836  *	@head: Head of buffer
837  *	@data: Data head pointer
838  *	@truesize: Buffer size
839  *	@users: User count - see {datagram,tcp}.c
840  *	@extensions: allocated extensions, valid if active_extensions is nonzero
841  */
842 
843 struct sk_buff {
844 	union {
845 		struct {
846 			/* These two members must be first to match sk_buff_head. */
847 			struct sk_buff		*next;
848 			struct sk_buff		*prev;
849 
850 			union {
851 				struct net_device	*dev;
852 				/* Some protocols might use this space to store information,
853 				 * while device pointer would be NULL.
854 				 * UDP receive path is one user.
855 				 */
856 				unsigned long		dev_scratch;
857 			};
858 		};
859 		struct rb_node		rbnode; /* used in netem, ip4 defrag, and tcp stack */
860 		struct list_head	list;
861 		struct llist_node	ll_node;
862 	};
863 
864 	union {
865 		struct sock		*sk;
866 		int			ip_defrag_offset;
867 	};
868 
869 	union {
870 		ktime_t		tstamp;
871 		u64		skb_mstamp_ns; /* earliest departure time */
872 	};
873 	/*
874 	 * This is the control buffer. It is free to use for every
875 	 * layer. Please put your private variables there. If you
876 	 * want to keep them across layers you have to do a skb_clone()
877 	 * first. This is owned by whoever has the skb queued ATM.
878 	 */
879 	char			cb[48] __aligned(8);
880 
881 	union {
882 		struct {
883 			unsigned long	_skb_refdst;
884 			void		(*destructor)(struct sk_buff *skb);
885 		};
886 		struct list_head	tcp_tsorted_anchor;
887 #ifdef CONFIG_NET_SOCK_MSG
888 		unsigned long		_sk_redir;
889 #endif
890 	};
891 
892 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
893 	unsigned long		 _nfct;
894 #endif
895 	unsigned int		len,
896 				data_len;
897 	__u16			mac_len,
898 				hdr_len;
899 
900 	/* Following fields are _not_ copied in __copy_skb_header()
901 	 * Note that queue_mapping is here mostly to fill a hole.
902 	 */
903 	__u16			queue_mapping;
904 
905 /* if you move cloned around you also must adapt those constants */
906 #ifdef __BIG_ENDIAN_BITFIELD
907 #define CLONED_MASK	(1 << 7)
908 #else
909 #define CLONED_MASK	1
910 #endif
911 #define CLONED_OFFSET		offsetof(struct sk_buff, __cloned_offset)
912 
913 	/* private: */
914 	__u8			__cloned_offset[0];
915 	/* public: */
916 	__u8			cloned:1,
917 				nohdr:1,
918 				fclone:2,
919 				peeked:1,
920 				head_frag:1,
921 				pfmemalloc:1,
922 				pp_recycle:1; /* page_pool recycle indicator */
923 #ifdef CONFIG_SKB_EXTENSIONS
924 	__u8			active_extensions;
925 #endif
926 
927 	/* Fields enclosed in headers group are copied
928 	 * using a single memcpy() in __copy_skb_header()
929 	 */
930 	struct_group(headers,
931 
932 	/* private: */
933 	__u8			__pkt_type_offset[0];
934 	/* public: */
935 	__u8			pkt_type:3; /* see PKT_TYPE_MAX */
936 	__u8			ignore_df:1;
937 	__u8			nf_trace:1;
938 	__u8			ip_summed:2;
939 	__u8			ooo_okay:1;
940 
941 	__u8			l4_hash:1;
942 	__u8			sw_hash:1;
943 	__u8			wifi_acked_valid:1;
944 	__u8			wifi_acked:1;
945 	__u8			no_fcs:1;
946 	/* Indicates the inner headers are valid in the skbuff. */
947 	__u8			encapsulation:1;
948 	__u8			encap_hdr_csum:1;
949 	__u8			csum_valid:1;
950 
951 	/* private: */
952 	__u8			__pkt_vlan_present_offset[0];
953 	/* public: */
954 	__u8			vlan_present:1;	/* See PKT_VLAN_PRESENT_BIT */
955 	__u8			csum_complete_sw:1;
956 	__u8			csum_level:2;
957 	__u8			dst_pending_confirm:1;
958 	__u8			mono_delivery_time:1;	/* See SKB_MONO_DELIVERY_TIME_MASK */
959 #ifdef CONFIG_NET_CLS_ACT
960 	__u8			tc_skip_classify:1;
961 	__u8			tc_at_ingress:1;	/* See TC_AT_INGRESS_MASK */
962 #endif
963 #ifdef CONFIG_IPV6_NDISC_NODETYPE
964 	__u8			ndisc_nodetype:2;
965 #endif
966 
967 	__u8			ipvs_property:1;
968 	__u8			inner_protocol_type:1;
969 	__u8			remcsum_offload:1;
970 #ifdef CONFIG_NET_SWITCHDEV
971 	__u8			offload_fwd_mark:1;
972 	__u8			offload_l3_fwd_mark:1;
973 #endif
974 	__u8			redirected:1;
975 #ifdef CONFIG_NET_REDIRECT
976 	__u8			from_ingress:1;
977 #endif
978 #ifdef CONFIG_NETFILTER_SKIP_EGRESS
979 	__u8			nf_skip_egress:1;
980 #endif
981 #ifdef CONFIG_TLS_DEVICE
982 	__u8			decrypted:1;
983 #endif
984 	__u8			slow_gro:1;
985 	__u8			csum_not_inet:1;
986 	__u8			scm_io_uring:1;
987 
988 #ifdef CONFIG_NET_SCHED
989 	__u16			tc_index;	/* traffic control index */
990 #endif
991 
992 	union {
993 		__wsum		csum;
994 		struct {
995 			__u16	csum_start;
996 			__u16	csum_offset;
997 		};
998 	};
999 	__u32			priority;
1000 	int			skb_iif;
1001 	__u32			hash;
1002 	__be16			vlan_proto;
1003 	__u16			vlan_tci;
1004 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
1005 	union {
1006 		unsigned int	napi_id;
1007 		unsigned int	sender_cpu;
1008 	};
1009 #endif
1010 	u16			alloc_cpu;
1011 #ifdef CONFIG_NETWORK_SECMARK
1012 	__u32		secmark;
1013 #endif
1014 
1015 	union {
1016 		__u32		mark;
1017 		__u32		reserved_tailroom;
1018 	};
1019 
1020 	union {
1021 		__be16		inner_protocol;
1022 		__u8		inner_ipproto;
1023 	};
1024 
1025 	__u16			inner_transport_header;
1026 	__u16			inner_network_header;
1027 	__u16			inner_mac_header;
1028 
1029 	__be16			protocol;
1030 	__u16			transport_header;
1031 	__u16			network_header;
1032 	__u16			mac_header;
1033 
1034 #ifdef CONFIG_KCOV
1035 	u64			kcov_handle;
1036 #endif
1037 
1038 	); /* end headers group */
1039 
1040 	/* These elements must be at the end, see alloc_skb() for details.  */
1041 	sk_buff_data_t		tail;
1042 	sk_buff_data_t		end;
1043 	unsigned char		*head,
1044 				*data;
1045 	unsigned int		truesize;
1046 	refcount_t		users;
1047 
1048 #ifdef CONFIG_SKB_EXTENSIONS
1049 	/* only useable after checking ->active_extensions != 0 */
1050 	struct skb_ext		*extensions;
1051 #endif
1052 };
1053 
1054 /* if you move pkt_type around you also must adapt those constants */
1055 #ifdef __BIG_ENDIAN_BITFIELD
1056 #define PKT_TYPE_MAX	(7 << 5)
1057 #else
1058 #define PKT_TYPE_MAX	7
1059 #endif
1060 #define PKT_TYPE_OFFSET		offsetof(struct sk_buff, __pkt_type_offset)
1061 
1062 /* if you move pkt_vlan_present, tc_at_ingress, or mono_delivery_time
1063  * around, you also must adapt these constants.
1064  */
1065 #ifdef __BIG_ENDIAN_BITFIELD
1066 #define PKT_VLAN_PRESENT_BIT	7
1067 #define TC_AT_INGRESS_MASK		(1 << 0)
1068 #define SKB_MONO_DELIVERY_TIME_MASK	(1 << 2)
1069 #else
1070 #define PKT_VLAN_PRESENT_BIT	0
1071 #define TC_AT_INGRESS_MASK		(1 << 7)
1072 #define SKB_MONO_DELIVERY_TIME_MASK	(1 << 5)
1073 #endif
1074 #define PKT_VLAN_PRESENT_OFFSET	offsetof(struct sk_buff, __pkt_vlan_present_offset)
1075 
1076 #ifdef __KERNEL__
1077 /*
1078  *	Handling routines are only of interest to the kernel
1079  */
1080 
1081 #define SKB_ALLOC_FCLONE	0x01
1082 #define SKB_ALLOC_RX		0x02
1083 #define SKB_ALLOC_NAPI		0x04
1084 
1085 /**
1086  * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
1087  * @skb: buffer
1088  */
skb_pfmemalloc(const struct sk_buff * skb)1089 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
1090 {
1091 	return unlikely(skb->pfmemalloc);
1092 }
1093 
1094 /*
1095  * skb might have a dst pointer attached, refcounted or not.
1096  * _skb_refdst low order bit is set if refcount was _not_ taken
1097  */
1098 #define SKB_DST_NOREF	1UL
1099 #define SKB_DST_PTRMASK	~(SKB_DST_NOREF)
1100 
1101 /**
1102  * skb_dst - returns skb dst_entry
1103  * @skb: buffer
1104  *
1105  * Returns skb dst_entry, regardless of reference taken or not.
1106  */
skb_dst(const struct sk_buff * skb)1107 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1108 {
1109 	/* If refdst was not refcounted, check we still are in a
1110 	 * rcu_read_lock section
1111 	 */
1112 	WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1113 		!rcu_read_lock_held() &&
1114 		!rcu_read_lock_bh_held());
1115 	return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
1116 }
1117 
1118 /**
1119  * skb_dst_set - sets skb dst
1120  * @skb: buffer
1121  * @dst: dst entry
1122  *
1123  * Sets skb dst, assuming a reference was taken on dst and should
1124  * be released by skb_dst_drop()
1125  */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)1126 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1127 {
1128 	skb->slow_gro |= !!dst;
1129 	skb->_skb_refdst = (unsigned long)dst;
1130 }
1131 
1132 /**
1133  * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1134  * @skb: buffer
1135  * @dst: dst entry
1136  *
1137  * Sets skb dst, assuming a reference was not taken on dst.
1138  * If dst entry is cached, we do not take reference and dst_release
1139  * will be avoided by refdst_drop. If dst entry is not cached, we take
1140  * reference, so that last dst_release can destroy the dst immediately.
1141  */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)1142 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1143 {
1144 	WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1145 	skb->slow_gro |= !!dst;
1146 	skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1147 }
1148 
1149 /**
1150  * skb_dst_is_noref - Test if skb dst isn't refcounted
1151  * @skb: buffer
1152  */
skb_dst_is_noref(const struct sk_buff * skb)1153 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1154 {
1155 	return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1156 }
1157 
1158 /**
1159  * skb_rtable - Returns the skb &rtable
1160  * @skb: buffer
1161  */
skb_rtable(const struct sk_buff * skb)1162 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1163 {
1164 	return (struct rtable *)skb_dst(skb);
1165 }
1166 
1167 /* For mangling skb->pkt_type from user space side from applications
1168  * such as nft, tc, etc, we only allow a conservative subset of
1169  * possible pkt_types to be set.
1170 */
skb_pkt_type_ok(u32 ptype)1171 static inline bool skb_pkt_type_ok(u32 ptype)
1172 {
1173 	return ptype <= PACKET_OTHERHOST;
1174 }
1175 
1176 /**
1177  * skb_napi_id - Returns the skb's NAPI id
1178  * @skb: buffer
1179  */
skb_napi_id(const struct sk_buff * skb)1180 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1181 {
1182 #ifdef CONFIG_NET_RX_BUSY_POLL
1183 	return skb->napi_id;
1184 #else
1185 	return 0;
1186 #endif
1187 }
1188 
1189 /**
1190  * skb_unref - decrement the skb's reference count
1191  * @skb: buffer
1192  *
1193  * Returns true if we can free the skb.
1194  */
skb_unref(struct sk_buff * skb)1195 static inline bool skb_unref(struct sk_buff *skb)
1196 {
1197 	if (unlikely(!skb))
1198 		return false;
1199 	if (likely(refcount_read(&skb->users) == 1))
1200 		smp_rmb();
1201 	else if (likely(!refcount_dec_and_test(&skb->users)))
1202 		return false;
1203 
1204 	return true;
1205 }
1206 
1207 void __fix_address
1208 kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason);
1209 
1210 /**
1211  *	kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1212  *	@skb: buffer to free
1213  */
kfree_skb(struct sk_buff * skb)1214 static inline void kfree_skb(struct sk_buff *skb)
1215 {
1216 	kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1217 }
1218 
1219 void skb_release_head_state(struct sk_buff *skb);
1220 void kfree_skb_list_reason(struct sk_buff *segs,
1221 			   enum skb_drop_reason reason);
1222 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1223 void skb_tx_error(struct sk_buff *skb);
1224 
kfree_skb_list(struct sk_buff * segs)1225 static inline void kfree_skb_list(struct sk_buff *segs)
1226 {
1227 	kfree_skb_list_reason(segs, SKB_DROP_REASON_NOT_SPECIFIED);
1228 }
1229 
1230 #ifdef CONFIG_TRACEPOINTS
1231 void consume_skb(struct sk_buff *skb);
1232 #else
consume_skb(struct sk_buff * skb)1233 static inline void consume_skb(struct sk_buff *skb)
1234 {
1235 	return kfree_skb(skb);
1236 }
1237 #endif
1238 
1239 void __consume_stateless_skb(struct sk_buff *skb);
1240 void  __kfree_skb(struct sk_buff *skb);
1241 extern struct kmem_cache *skbuff_head_cache;
1242 
1243 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1244 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1245 		      bool *fragstolen, int *delta_truesize);
1246 
1247 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1248 			    int node);
1249 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1250 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1251 struct sk_buff *build_skb_around(struct sk_buff *skb,
1252 				 void *data, unsigned int frag_size);
1253 void skb_attempt_defer_free(struct sk_buff *skb);
1254 
1255 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1256 
1257 /**
1258  * alloc_skb - allocate a network buffer
1259  * @size: size to allocate
1260  * @priority: allocation mask
1261  *
1262  * This function is a convenient wrapper around __alloc_skb().
1263  */
alloc_skb(unsigned int size,gfp_t priority)1264 static inline struct sk_buff *alloc_skb(unsigned int size,
1265 					gfp_t priority)
1266 {
1267 	return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1268 }
1269 
1270 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1271 				     unsigned long data_len,
1272 				     int max_page_order,
1273 				     int *errcode,
1274 				     gfp_t gfp_mask);
1275 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1276 
1277 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1278 struct sk_buff_fclones {
1279 	struct sk_buff	skb1;
1280 
1281 	struct sk_buff	skb2;
1282 
1283 	refcount_t	fclone_ref;
1284 };
1285 
1286 /**
1287  *	skb_fclone_busy - check if fclone is busy
1288  *	@sk: socket
1289  *	@skb: buffer
1290  *
1291  * Returns true if skb is a fast clone, and its clone is not freed.
1292  * Some drivers call skb_orphan() in their ndo_start_xmit(),
1293  * so we also check that this didnt happen.
1294  */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1295 static inline bool skb_fclone_busy(const struct sock *sk,
1296 				   const struct sk_buff *skb)
1297 {
1298 	const struct sk_buff_fclones *fclones;
1299 
1300 	fclones = container_of(skb, struct sk_buff_fclones, skb1);
1301 
1302 	return skb->fclone == SKB_FCLONE_ORIG &&
1303 	       refcount_read(&fclones->fclone_ref) > 1 &&
1304 	       READ_ONCE(fclones->skb2.sk) == sk;
1305 }
1306 
1307 /**
1308  * alloc_skb_fclone - allocate a network buffer from fclone cache
1309  * @size: size to allocate
1310  * @priority: allocation mask
1311  *
1312  * This function is a convenient wrapper around __alloc_skb().
1313  */
alloc_skb_fclone(unsigned int size,gfp_t priority)1314 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1315 					       gfp_t priority)
1316 {
1317 	return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1318 }
1319 
1320 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1321 void skb_headers_offset_update(struct sk_buff *skb, int off);
1322 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1323 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1324 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1325 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1326 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1327 				   gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1328 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1329 					  gfp_t gfp_mask)
1330 {
1331 	return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1332 }
1333 
1334 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1335 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1336 				     unsigned int headroom);
1337 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1338 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1339 				int newtailroom, gfp_t priority);
1340 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1341 				     int offset, int len);
1342 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1343 			      int offset, int len);
1344 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1345 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1346 
1347 /**
1348  *	skb_pad			-	zero pad the tail of an skb
1349  *	@skb: buffer to pad
1350  *	@pad: space to pad
1351  *
1352  *	Ensure that a buffer is followed by a padding area that is zero
1353  *	filled. Used by network drivers which may DMA or transfer data
1354  *	beyond the buffer end onto the wire.
1355  *
1356  *	May return error in out of memory cases. The skb is freed on error.
1357  */
skb_pad(struct sk_buff * skb,int pad)1358 static inline int skb_pad(struct sk_buff *skb, int pad)
1359 {
1360 	return __skb_pad(skb, pad, true);
1361 }
1362 #define dev_kfree_skb(a)	consume_skb(a)
1363 
1364 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1365 			 int offset, size_t size);
1366 
1367 struct skb_seq_state {
1368 	__u32		lower_offset;
1369 	__u32		upper_offset;
1370 	__u32		frag_idx;
1371 	__u32		stepped_offset;
1372 	struct sk_buff	*root_skb;
1373 	struct sk_buff	*cur_skb;
1374 	__u8		*frag_data;
1375 	__u32		frag_off;
1376 };
1377 
1378 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1379 			  unsigned int to, struct skb_seq_state *st);
1380 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1381 			  struct skb_seq_state *st);
1382 void skb_abort_seq_read(struct skb_seq_state *st);
1383 
1384 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1385 			   unsigned int to, struct ts_config *config);
1386 
1387 /*
1388  * Packet hash types specify the type of hash in skb_set_hash.
1389  *
1390  * Hash types refer to the protocol layer addresses which are used to
1391  * construct a packet's hash. The hashes are used to differentiate or identify
1392  * flows of the protocol layer for the hash type. Hash types are either
1393  * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1394  *
1395  * Properties of hashes:
1396  *
1397  * 1) Two packets in different flows have different hash values
1398  * 2) Two packets in the same flow should have the same hash value
1399  *
1400  * A hash at a higher layer is considered to be more specific. A driver should
1401  * set the most specific hash possible.
1402  *
1403  * A driver cannot indicate a more specific hash than the layer at which a hash
1404  * was computed. For instance an L3 hash cannot be set as an L4 hash.
1405  *
1406  * A driver may indicate a hash level which is less specific than the
1407  * actual layer the hash was computed on. For instance, a hash computed
1408  * at L4 may be considered an L3 hash. This should only be done if the
1409  * driver can't unambiguously determine that the HW computed the hash at
1410  * the higher layer. Note that the "should" in the second property above
1411  * permits this.
1412  */
1413 enum pkt_hash_types {
1414 	PKT_HASH_TYPE_NONE,	/* Undefined type */
1415 	PKT_HASH_TYPE_L2,	/* Input: src_MAC, dest_MAC */
1416 	PKT_HASH_TYPE_L3,	/* Input: src_IP, dst_IP */
1417 	PKT_HASH_TYPE_L4,	/* Input: src_IP, dst_IP, src_port, dst_port */
1418 };
1419 
skb_clear_hash(struct sk_buff * skb)1420 static inline void skb_clear_hash(struct sk_buff *skb)
1421 {
1422 	skb->hash = 0;
1423 	skb->sw_hash = 0;
1424 	skb->l4_hash = 0;
1425 }
1426 
skb_clear_hash_if_not_l4(struct sk_buff * skb)1427 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1428 {
1429 	if (!skb->l4_hash)
1430 		skb_clear_hash(skb);
1431 }
1432 
1433 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1434 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1435 {
1436 	skb->l4_hash = is_l4;
1437 	skb->sw_hash = is_sw;
1438 	skb->hash = hash;
1439 }
1440 
1441 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1442 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1443 {
1444 	/* Used by drivers to set hash from HW */
1445 	__skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1446 }
1447 
1448 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1449 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1450 {
1451 	__skb_set_hash(skb, hash, true, is_l4);
1452 }
1453 
1454 void __skb_get_hash(struct sk_buff *skb);
1455 u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1456 u32 skb_get_poff(const struct sk_buff *skb);
1457 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1458 		   const struct flow_keys_basic *keys, int hlen);
1459 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1460 			    const void *data, int hlen_proto);
1461 
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto)1462 static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1463 					int thoff, u8 ip_proto)
1464 {
1465 	return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1466 }
1467 
1468 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1469 			     const struct flow_dissector_key *key,
1470 			     unsigned int key_count);
1471 
1472 struct bpf_flow_dissector;
1473 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1474 		     __be16 proto, int nhoff, int hlen, unsigned int flags);
1475 
1476 bool __skb_flow_dissect(const struct net *net,
1477 			const struct sk_buff *skb,
1478 			struct flow_dissector *flow_dissector,
1479 			void *target_container, const void *data,
1480 			__be16 proto, int nhoff, int hlen, unsigned int flags);
1481 
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1482 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1483 				    struct flow_dissector *flow_dissector,
1484 				    void *target_container, unsigned int flags)
1485 {
1486 	return __skb_flow_dissect(NULL, skb, flow_dissector,
1487 				  target_container, NULL, 0, 0, 0, flags);
1488 }
1489 
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1490 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1491 					      struct flow_keys *flow,
1492 					      unsigned int flags)
1493 {
1494 	memset(flow, 0, sizeof(*flow));
1495 	return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1496 				  flow, NULL, 0, 0, 0, flags);
1497 }
1498 
1499 static inline bool
skb_flow_dissect_flow_keys_basic(const struct net * net,const struct sk_buff * skb,struct flow_keys_basic * flow,const void * data,__be16 proto,int nhoff,int hlen,unsigned int flags)1500 skb_flow_dissect_flow_keys_basic(const struct net *net,
1501 				 const struct sk_buff *skb,
1502 				 struct flow_keys_basic *flow,
1503 				 const void *data, __be16 proto,
1504 				 int nhoff, int hlen, unsigned int flags)
1505 {
1506 	memset(flow, 0, sizeof(*flow));
1507 	return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1508 				  data, proto, nhoff, hlen, flags);
1509 }
1510 
1511 void skb_flow_dissect_meta(const struct sk_buff *skb,
1512 			   struct flow_dissector *flow_dissector,
1513 			   void *target_container);
1514 
1515 /* Gets a skb connection tracking info, ctinfo map should be a
1516  * map of mapsize to translate enum ip_conntrack_info states
1517  * to user states.
1518  */
1519 void
1520 skb_flow_dissect_ct(const struct sk_buff *skb,
1521 		    struct flow_dissector *flow_dissector,
1522 		    void *target_container,
1523 		    u16 *ctinfo_map, size_t mapsize,
1524 		    bool post_ct, u16 zone);
1525 void
1526 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1527 			     struct flow_dissector *flow_dissector,
1528 			     void *target_container);
1529 
1530 void skb_flow_dissect_hash(const struct sk_buff *skb,
1531 			   struct flow_dissector *flow_dissector,
1532 			   void *target_container);
1533 
skb_get_hash(struct sk_buff * skb)1534 static inline __u32 skb_get_hash(struct sk_buff *skb)
1535 {
1536 	if (!skb->l4_hash && !skb->sw_hash)
1537 		__skb_get_hash(skb);
1538 
1539 	return skb->hash;
1540 }
1541 
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1542 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1543 {
1544 	if (!skb->l4_hash && !skb->sw_hash) {
1545 		struct flow_keys keys;
1546 		__u32 hash = __get_hash_from_flowi6(fl6, &keys);
1547 
1548 		__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1549 	}
1550 
1551 	return skb->hash;
1552 }
1553 
1554 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1555 			   const siphash_key_t *perturb);
1556 
skb_get_hash_raw(const struct sk_buff * skb)1557 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1558 {
1559 	return skb->hash;
1560 }
1561 
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1562 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1563 {
1564 	to->hash = from->hash;
1565 	to->sw_hash = from->sw_hash;
1566 	to->l4_hash = from->l4_hash;
1567 };
1568 
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1569 static inline void skb_copy_decrypted(struct sk_buff *to,
1570 				      const struct sk_buff *from)
1571 {
1572 #ifdef CONFIG_TLS_DEVICE
1573 	to->decrypted = from->decrypted;
1574 #endif
1575 }
1576 
1577 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1578 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1579 {
1580 	return skb->head + skb->end;
1581 }
1582 
skb_end_offset(const struct sk_buff * skb)1583 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1584 {
1585 	return skb->end;
1586 }
1587 
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1588 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1589 {
1590 	skb->end = offset;
1591 }
1592 #else
skb_end_pointer(const struct sk_buff * skb)1593 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1594 {
1595 	return skb->end;
1596 }
1597 
skb_end_offset(const struct sk_buff * skb)1598 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1599 {
1600 	return skb->end - skb->head;
1601 }
1602 
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1603 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1604 {
1605 	skb->end = skb->head + offset;
1606 }
1607 #endif
1608 
1609 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
1610 				       struct ubuf_info *uarg);
1611 
1612 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
1613 
1614 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
1615 			   bool success);
1616 
1617 int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk,
1618 			    struct sk_buff *skb, struct iov_iter *from,
1619 			    size_t length);
1620 
skb_zerocopy_iter_dgram(struct sk_buff * skb,struct msghdr * msg,int len)1621 static inline int skb_zerocopy_iter_dgram(struct sk_buff *skb,
1622 					  struct msghdr *msg, int len)
1623 {
1624 	return __zerocopy_sg_from_iter(msg, skb->sk, skb, &msg->msg_iter, len);
1625 }
1626 
1627 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
1628 			     struct msghdr *msg, int len,
1629 			     struct ubuf_info *uarg);
1630 
1631 /* Internal */
1632 #define skb_shinfo(SKB)	((struct skb_shared_info *)(skb_end_pointer(SKB)))
1633 
skb_hwtstamps(struct sk_buff * skb)1634 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1635 {
1636 	return &skb_shinfo(skb)->hwtstamps;
1637 }
1638 
skb_zcopy(struct sk_buff * skb)1639 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1640 {
1641 	bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1642 
1643 	return is_zcopy ? skb_uarg(skb) : NULL;
1644 }
1645 
skb_zcopy_pure(const struct sk_buff * skb)1646 static inline bool skb_zcopy_pure(const struct sk_buff *skb)
1647 {
1648 	return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY;
1649 }
1650 
skb_zcopy_managed(const struct sk_buff * skb)1651 static inline bool skb_zcopy_managed(const struct sk_buff *skb)
1652 {
1653 	return skb_shinfo(skb)->flags & SKBFL_MANAGED_FRAG_REFS;
1654 }
1655 
skb_pure_zcopy_same(const struct sk_buff * skb1,const struct sk_buff * skb2)1656 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1,
1657 				       const struct sk_buff *skb2)
1658 {
1659 	return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2);
1660 }
1661 
net_zcopy_get(struct ubuf_info * uarg)1662 static inline void net_zcopy_get(struct ubuf_info *uarg)
1663 {
1664 	refcount_inc(&uarg->refcnt);
1665 }
1666 
skb_zcopy_init(struct sk_buff * skb,struct ubuf_info * uarg)1667 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1668 {
1669 	skb_shinfo(skb)->destructor_arg = uarg;
1670 	skb_shinfo(skb)->flags |= uarg->flags;
1671 }
1672 
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1673 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1674 				 bool *have_ref)
1675 {
1676 	if (skb && uarg && !skb_zcopy(skb)) {
1677 		if (unlikely(have_ref && *have_ref))
1678 			*have_ref = false;
1679 		else
1680 			net_zcopy_get(uarg);
1681 		skb_zcopy_init(skb, uarg);
1682 	}
1683 }
1684 
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1685 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1686 {
1687 	skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1688 	skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1689 }
1690 
skb_zcopy_is_nouarg(struct sk_buff * skb)1691 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1692 {
1693 	return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1694 }
1695 
skb_zcopy_get_nouarg(struct sk_buff * skb)1696 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1697 {
1698 	return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1699 }
1700 
net_zcopy_put(struct ubuf_info * uarg)1701 static inline void net_zcopy_put(struct ubuf_info *uarg)
1702 {
1703 	if (uarg)
1704 		uarg->callback(NULL, uarg, true);
1705 }
1706 
net_zcopy_put_abort(struct ubuf_info * uarg,bool have_uref)1707 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1708 {
1709 	if (uarg) {
1710 		if (uarg->callback == msg_zerocopy_callback)
1711 			msg_zerocopy_put_abort(uarg, have_uref);
1712 		else if (have_uref)
1713 			net_zcopy_put(uarg);
1714 	}
1715 }
1716 
1717 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy_success)1718 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1719 {
1720 	struct ubuf_info *uarg = skb_zcopy(skb);
1721 
1722 	if (uarg) {
1723 		if (!skb_zcopy_is_nouarg(skb))
1724 			uarg->callback(skb, uarg, zerocopy_success);
1725 
1726 		skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY;
1727 	}
1728 }
1729 
1730 void __skb_zcopy_downgrade_managed(struct sk_buff *skb);
1731 
skb_zcopy_downgrade_managed(struct sk_buff * skb)1732 static inline void skb_zcopy_downgrade_managed(struct sk_buff *skb)
1733 {
1734 	if (unlikely(skb_zcopy_managed(skb)))
1735 		__skb_zcopy_downgrade_managed(skb);
1736 }
1737 
skb_mark_not_on_list(struct sk_buff * skb)1738 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1739 {
1740 	skb->next = NULL;
1741 }
1742 
1743 /* Iterate through singly-linked GSO fragments of an skb. */
1744 #define skb_list_walk_safe(first, skb, next_skb)                               \
1745 	for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb);  \
1746 	     (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1747 
skb_list_del_init(struct sk_buff * skb)1748 static inline void skb_list_del_init(struct sk_buff *skb)
1749 {
1750 	__list_del_entry(&skb->list);
1751 	skb_mark_not_on_list(skb);
1752 }
1753 
1754 /**
1755  *	skb_queue_empty - check if a queue is empty
1756  *	@list: queue head
1757  *
1758  *	Returns true if the queue is empty, false otherwise.
1759  */
skb_queue_empty(const struct sk_buff_head * list)1760 static inline int skb_queue_empty(const struct sk_buff_head *list)
1761 {
1762 	return list->next == (const struct sk_buff *) list;
1763 }
1764 
1765 /**
1766  *	skb_queue_empty_lockless - check if a queue is empty
1767  *	@list: queue head
1768  *
1769  *	Returns true if the queue is empty, false otherwise.
1770  *	This variant can be used in lockless contexts.
1771  */
skb_queue_empty_lockless(const struct sk_buff_head * list)1772 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1773 {
1774 	return READ_ONCE(list->next) == (const struct sk_buff *) list;
1775 }
1776 
1777 
1778 /**
1779  *	skb_queue_is_last - check if skb is the last entry in the queue
1780  *	@list: queue head
1781  *	@skb: buffer
1782  *
1783  *	Returns true if @skb is the last buffer on the list.
1784  */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1785 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1786 				     const struct sk_buff *skb)
1787 {
1788 	return skb->next == (const struct sk_buff *) list;
1789 }
1790 
1791 /**
1792  *	skb_queue_is_first - check if skb is the first entry in the queue
1793  *	@list: queue head
1794  *	@skb: buffer
1795  *
1796  *	Returns true if @skb is the first buffer on the list.
1797  */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1798 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1799 				      const struct sk_buff *skb)
1800 {
1801 	return skb->prev == (const struct sk_buff *) list;
1802 }
1803 
1804 /**
1805  *	skb_queue_next - return the next packet in the queue
1806  *	@list: queue head
1807  *	@skb: current buffer
1808  *
1809  *	Return the next packet in @list after @skb.  It is only valid to
1810  *	call this if skb_queue_is_last() evaluates to false.
1811  */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1812 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1813 					     const struct sk_buff *skb)
1814 {
1815 	/* This BUG_ON may seem severe, but if we just return then we
1816 	 * are going to dereference garbage.
1817 	 */
1818 	BUG_ON(skb_queue_is_last(list, skb));
1819 	return skb->next;
1820 }
1821 
1822 /**
1823  *	skb_queue_prev - return the prev packet in the queue
1824  *	@list: queue head
1825  *	@skb: current buffer
1826  *
1827  *	Return the prev packet in @list before @skb.  It is only valid to
1828  *	call this if skb_queue_is_first() evaluates to false.
1829  */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1830 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1831 					     const struct sk_buff *skb)
1832 {
1833 	/* This BUG_ON may seem severe, but if we just return then we
1834 	 * are going to dereference garbage.
1835 	 */
1836 	BUG_ON(skb_queue_is_first(list, skb));
1837 	return skb->prev;
1838 }
1839 
1840 /**
1841  *	skb_get - reference buffer
1842  *	@skb: buffer to reference
1843  *
1844  *	Makes another reference to a socket buffer and returns a pointer
1845  *	to the buffer.
1846  */
skb_get(struct sk_buff * skb)1847 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1848 {
1849 	refcount_inc(&skb->users);
1850 	return skb;
1851 }
1852 
1853 /*
1854  * If users == 1, we are the only owner and can avoid redundant atomic changes.
1855  */
1856 
1857 /**
1858  *	skb_cloned - is the buffer a clone
1859  *	@skb: buffer to check
1860  *
1861  *	Returns true if the buffer was generated with skb_clone() and is
1862  *	one of multiple shared copies of the buffer. Cloned buffers are
1863  *	shared data so must not be written to under normal circumstances.
1864  */
skb_cloned(const struct sk_buff * skb)1865 static inline int skb_cloned(const struct sk_buff *skb)
1866 {
1867 	return skb->cloned &&
1868 	       (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1869 }
1870 
skb_unclone(struct sk_buff * skb,gfp_t pri)1871 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1872 {
1873 	might_sleep_if(gfpflags_allow_blocking(pri));
1874 
1875 	if (skb_cloned(skb))
1876 		return pskb_expand_head(skb, 0, 0, pri);
1877 
1878 	return 0;
1879 }
1880 
1881 /* This variant of skb_unclone() makes sure skb->truesize
1882  * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1883  *
1884  * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
1885  * when various debugging features are in place.
1886  */
1887 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
skb_unclone_keeptruesize(struct sk_buff * skb,gfp_t pri)1888 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
1889 {
1890 	might_sleep_if(gfpflags_allow_blocking(pri));
1891 
1892 	if (skb_cloned(skb))
1893 		return __skb_unclone_keeptruesize(skb, pri);
1894 	return 0;
1895 }
1896 
1897 /**
1898  *	skb_header_cloned - is the header a clone
1899  *	@skb: buffer to check
1900  *
1901  *	Returns true if modifying the header part of the buffer requires
1902  *	the data to be copied.
1903  */
skb_header_cloned(const struct sk_buff * skb)1904 static inline int skb_header_cloned(const struct sk_buff *skb)
1905 {
1906 	int dataref;
1907 
1908 	if (!skb->cloned)
1909 		return 0;
1910 
1911 	dataref = atomic_read(&skb_shinfo(skb)->dataref);
1912 	dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1913 	return dataref != 1;
1914 }
1915 
skb_header_unclone(struct sk_buff * skb,gfp_t pri)1916 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1917 {
1918 	might_sleep_if(gfpflags_allow_blocking(pri));
1919 
1920 	if (skb_header_cloned(skb))
1921 		return pskb_expand_head(skb, 0, 0, pri);
1922 
1923 	return 0;
1924 }
1925 
1926 /**
1927  * __skb_header_release() - allow clones to use the headroom
1928  * @skb: buffer to operate on
1929  *
1930  * See "DOC: dataref and headerless skbs".
1931  */
__skb_header_release(struct sk_buff * skb)1932 static inline void __skb_header_release(struct sk_buff *skb)
1933 {
1934 	skb->nohdr = 1;
1935 	atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1936 }
1937 
1938 
1939 /**
1940  *	skb_shared - is the buffer shared
1941  *	@skb: buffer to check
1942  *
1943  *	Returns true if more than one person has a reference to this
1944  *	buffer.
1945  */
skb_shared(const struct sk_buff * skb)1946 static inline int skb_shared(const struct sk_buff *skb)
1947 {
1948 	return refcount_read(&skb->users) != 1;
1949 }
1950 
1951 /**
1952  *	skb_share_check - check if buffer is shared and if so clone it
1953  *	@skb: buffer to check
1954  *	@pri: priority for memory allocation
1955  *
1956  *	If the buffer is shared the buffer is cloned and the old copy
1957  *	drops a reference. A new clone with a single reference is returned.
1958  *	If the buffer is not shared the original buffer is returned. When
1959  *	being called from interrupt status or with spinlocks held pri must
1960  *	be GFP_ATOMIC.
1961  *
1962  *	NULL is returned on a memory allocation failure.
1963  */
skb_share_check(struct sk_buff * skb,gfp_t pri)1964 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1965 {
1966 	might_sleep_if(gfpflags_allow_blocking(pri));
1967 	if (skb_shared(skb)) {
1968 		struct sk_buff *nskb = skb_clone(skb, pri);
1969 
1970 		if (likely(nskb))
1971 			consume_skb(skb);
1972 		else
1973 			kfree_skb(skb);
1974 		skb = nskb;
1975 	}
1976 	return skb;
1977 }
1978 
1979 /*
1980  *	Copy shared buffers into a new sk_buff. We effectively do COW on
1981  *	packets to handle cases where we have a local reader and forward
1982  *	and a couple of other messy ones. The normal one is tcpdumping
1983  *	a packet thats being forwarded.
1984  */
1985 
1986 /**
1987  *	skb_unshare - make a copy of a shared buffer
1988  *	@skb: buffer to check
1989  *	@pri: priority for memory allocation
1990  *
1991  *	If the socket buffer is a clone then this function creates a new
1992  *	copy of the data, drops a reference count on the old copy and returns
1993  *	the new copy with the reference count at 1. If the buffer is not a clone
1994  *	the original buffer is returned. When called with a spinlock held or
1995  *	from interrupt state @pri must be %GFP_ATOMIC
1996  *
1997  *	%NULL is returned on a memory allocation failure.
1998  */
skb_unshare(struct sk_buff * skb,gfp_t pri)1999 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
2000 					  gfp_t pri)
2001 {
2002 	might_sleep_if(gfpflags_allow_blocking(pri));
2003 	if (skb_cloned(skb)) {
2004 		struct sk_buff *nskb = skb_copy(skb, pri);
2005 
2006 		/* Free our shared copy */
2007 		if (likely(nskb))
2008 			consume_skb(skb);
2009 		else
2010 			kfree_skb(skb);
2011 		skb = nskb;
2012 	}
2013 	return skb;
2014 }
2015 
2016 /**
2017  *	skb_peek - peek at the head of an &sk_buff_head
2018  *	@list_: list to peek at
2019  *
2020  *	Peek an &sk_buff. Unlike most other operations you _MUST_
2021  *	be careful with this one. A peek leaves the buffer on the
2022  *	list and someone else may run off with it. You must hold
2023  *	the appropriate locks or have a private queue to do this.
2024  *
2025  *	Returns %NULL for an empty list or a pointer to the head element.
2026  *	The reference count is not incremented and the reference is therefore
2027  *	volatile. Use with caution.
2028  */
skb_peek(const struct sk_buff_head * list_)2029 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
2030 {
2031 	struct sk_buff *skb = list_->next;
2032 
2033 	if (skb == (struct sk_buff *)list_)
2034 		skb = NULL;
2035 	return skb;
2036 }
2037 
2038 /**
2039  *	__skb_peek - peek at the head of a non-empty &sk_buff_head
2040  *	@list_: list to peek at
2041  *
2042  *	Like skb_peek(), but the caller knows that the list is not empty.
2043  */
__skb_peek(const struct sk_buff_head * list_)2044 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
2045 {
2046 	return list_->next;
2047 }
2048 
2049 /**
2050  *	skb_peek_next - peek skb following the given one from a queue
2051  *	@skb: skb to start from
2052  *	@list_: list to peek at
2053  *
2054  *	Returns %NULL when the end of the list is met or a pointer to the
2055  *	next element. The reference count is not incremented and the
2056  *	reference is therefore volatile. Use with caution.
2057  */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)2058 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
2059 		const struct sk_buff_head *list_)
2060 {
2061 	struct sk_buff *next = skb->next;
2062 
2063 	if (next == (struct sk_buff *)list_)
2064 		next = NULL;
2065 	return next;
2066 }
2067 
2068 /**
2069  *	skb_peek_tail - peek at the tail of an &sk_buff_head
2070  *	@list_: list to peek at
2071  *
2072  *	Peek an &sk_buff. Unlike most other operations you _MUST_
2073  *	be careful with this one. A peek leaves the buffer on the
2074  *	list and someone else may run off with it. You must hold
2075  *	the appropriate locks or have a private queue to do this.
2076  *
2077  *	Returns %NULL for an empty list or a pointer to the tail element.
2078  *	The reference count is not incremented and the reference is therefore
2079  *	volatile. Use with caution.
2080  */
skb_peek_tail(const struct sk_buff_head * list_)2081 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
2082 {
2083 	struct sk_buff *skb = READ_ONCE(list_->prev);
2084 
2085 	if (skb == (struct sk_buff *)list_)
2086 		skb = NULL;
2087 	return skb;
2088 
2089 }
2090 
2091 /**
2092  *	skb_queue_len	- get queue length
2093  *	@list_: list to measure
2094  *
2095  *	Return the length of an &sk_buff queue.
2096  */
skb_queue_len(const struct sk_buff_head * list_)2097 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
2098 {
2099 	return list_->qlen;
2100 }
2101 
2102 /**
2103  *	skb_queue_len_lockless	- get queue length
2104  *	@list_: list to measure
2105  *
2106  *	Return the length of an &sk_buff queue.
2107  *	This variant can be used in lockless contexts.
2108  */
skb_queue_len_lockless(const struct sk_buff_head * list_)2109 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
2110 {
2111 	return READ_ONCE(list_->qlen);
2112 }
2113 
2114 /**
2115  *	__skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
2116  *	@list: queue to initialize
2117  *
2118  *	This initializes only the list and queue length aspects of
2119  *	an sk_buff_head object.  This allows to initialize the list
2120  *	aspects of an sk_buff_head without reinitializing things like
2121  *	the spinlock.  It can also be used for on-stack sk_buff_head
2122  *	objects where the spinlock is known to not be used.
2123  */
__skb_queue_head_init(struct sk_buff_head * list)2124 static inline void __skb_queue_head_init(struct sk_buff_head *list)
2125 {
2126 	list->prev = list->next = (struct sk_buff *)list;
2127 	list->qlen = 0;
2128 }
2129 
2130 /*
2131  * This function creates a split out lock class for each invocation;
2132  * this is needed for now since a whole lot of users of the skb-queue
2133  * infrastructure in drivers have different locking usage (in hardirq)
2134  * than the networking core (in softirq only). In the long run either the
2135  * network layer or drivers should need annotation to consolidate the
2136  * main types of usage into 3 classes.
2137  */
skb_queue_head_init(struct sk_buff_head * list)2138 static inline void skb_queue_head_init(struct sk_buff_head *list)
2139 {
2140 	spin_lock_init(&list->lock);
2141 	__skb_queue_head_init(list);
2142 }
2143 
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)2144 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
2145 		struct lock_class_key *class)
2146 {
2147 	skb_queue_head_init(list);
2148 	lockdep_set_class(&list->lock, class);
2149 }
2150 
2151 /*
2152  *	Insert an sk_buff on a list.
2153  *
2154  *	The "__skb_xxxx()" functions are the non-atomic ones that
2155  *	can only be called with interrupts disabled.
2156  */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)2157 static inline void __skb_insert(struct sk_buff *newsk,
2158 				struct sk_buff *prev, struct sk_buff *next,
2159 				struct sk_buff_head *list)
2160 {
2161 	/* See skb_queue_empty_lockless() and skb_peek_tail()
2162 	 * for the opposite READ_ONCE()
2163 	 */
2164 	WRITE_ONCE(newsk->next, next);
2165 	WRITE_ONCE(newsk->prev, prev);
2166 	WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk);
2167 	WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk);
2168 	WRITE_ONCE(list->qlen, list->qlen + 1);
2169 }
2170 
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)2171 static inline void __skb_queue_splice(const struct sk_buff_head *list,
2172 				      struct sk_buff *prev,
2173 				      struct sk_buff *next)
2174 {
2175 	struct sk_buff *first = list->next;
2176 	struct sk_buff *last = list->prev;
2177 
2178 	WRITE_ONCE(first->prev, prev);
2179 	WRITE_ONCE(prev->next, first);
2180 
2181 	WRITE_ONCE(last->next, next);
2182 	WRITE_ONCE(next->prev, last);
2183 }
2184 
2185 /**
2186  *	skb_queue_splice - join two skb lists, this is designed for stacks
2187  *	@list: the new list to add
2188  *	@head: the place to add it in the first list
2189  */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)2190 static inline void skb_queue_splice(const struct sk_buff_head *list,
2191 				    struct sk_buff_head *head)
2192 {
2193 	if (!skb_queue_empty(list)) {
2194 		__skb_queue_splice(list, (struct sk_buff *) head, head->next);
2195 		head->qlen += list->qlen;
2196 	}
2197 }
2198 
2199 /**
2200  *	skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2201  *	@list: the new list to add
2202  *	@head: the place to add it in the first list
2203  *
2204  *	The list at @list is reinitialised
2205  */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)2206 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2207 					 struct sk_buff_head *head)
2208 {
2209 	if (!skb_queue_empty(list)) {
2210 		__skb_queue_splice(list, (struct sk_buff *) head, head->next);
2211 		head->qlen += list->qlen;
2212 		__skb_queue_head_init(list);
2213 	}
2214 }
2215 
2216 /**
2217  *	skb_queue_splice_tail - join two skb lists, each list being a queue
2218  *	@list: the new list to add
2219  *	@head: the place to add it in the first list
2220  */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)2221 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2222 					 struct sk_buff_head *head)
2223 {
2224 	if (!skb_queue_empty(list)) {
2225 		__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2226 		head->qlen += list->qlen;
2227 	}
2228 }
2229 
2230 /**
2231  *	skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2232  *	@list: the new list to add
2233  *	@head: the place to add it in the first list
2234  *
2235  *	Each of the lists is a queue.
2236  *	The list at @list is reinitialised
2237  */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)2238 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2239 					      struct sk_buff_head *head)
2240 {
2241 	if (!skb_queue_empty(list)) {
2242 		__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2243 		head->qlen += list->qlen;
2244 		__skb_queue_head_init(list);
2245 	}
2246 }
2247 
2248 /**
2249  *	__skb_queue_after - queue a buffer at the list head
2250  *	@list: list to use
2251  *	@prev: place after this buffer
2252  *	@newsk: buffer to queue
2253  *
2254  *	Queue a buffer int the middle of a list. This function takes no locks
2255  *	and you must therefore hold required locks before calling it.
2256  *
2257  *	A buffer cannot be placed on two lists at the same time.
2258  */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)2259 static inline void __skb_queue_after(struct sk_buff_head *list,
2260 				     struct sk_buff *prev,
2261 				     struct sk_buff *newsk)
2262 {
2263 	__skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list);
2264 }
2265 
2266 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2267 		struct sk_buff_head *list);
2268 
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)2269 static inline void __skb_queue_before(struct sk_buff_head *list,
2270 				      struct sk_buff *next,
2271 				      struct sk_buff *newsk)
2272 {
2273 	__skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list);
2274 }
2275 
2276 /**
2277  *	__skb_queue_head - queue a buffer at the list head
2278  *	@list: list to use
2279  *	@newsk: buffer to queue
2280  *
2281  *	Queue a buffer at the start of a list. This function takes no locks
2282  *	and you must therefore hold required locks before calling it.
2283  *
2284  *	A buffer cannot be placed on two lists at the same time.
2285  */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)2286 static inline void __skb_queue_head(struct sk_buff_head *list,
2287 				    struct sk_buff *newsk)
2288 {
2289 	__skb_queue_after(list, (struct sk_buff *)list, newsk);
2290 }
2291 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2292 
2293 /**
2294  *	__skb_queue_tail - queue a buffer at the list tail
2295  *	@list: list to use
2296  *	@newsk: buffer to queue
2297  *
2298  *	Queue a buffer at the end of a list. This function takes no locks
2299  *	and you must therefore hold required locks before calling it.
2300  *
2301  *	A buffer cannot be placed on two lists at the same time.
2302  */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2303 static inline void __skb_queue_tail(struct sk_buff_head *list,
2304 				   struct sk_buff *newsk)
2305 {
2306 	__skb_queue_before(list, (struct sk_buff *)list, newsk);
2307 }
2308 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2309 
2310 /*
2311  * remove sk_buff from list. _Must_ be called atomically, and with
2312  * the list known..
2313  */
2314 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2315 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2316 {
2317 	struct sk_buff *next, *prev;
2318 
2319 	WRITE_ONCE(list->qlen, list->qlen - 1);
2320 	next	   = skb->next;
2321 	prev	   = skb->prev;
2322 	skb->next  = skb->prev = NULL;
2323 	WRITE_ONCE(next->prev, prev);
2324 	WRITE_ONCE(prev->next, next);
2325 }
2326 
2327 /**
2328  *	__skb_dequeue - remove from the head of the queue
2329  *	@list: list to dequeue from
2330  *
2331  *	Remove the head of the list. This function does not take any locks
2332  *	so must be used with appropriate locks held only. The head item is
2333  *	returned or %NULL if the list is empty.
2334  */
__skb_dequeue(struct sk_buff_head * list)2335 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2336 {
2337 	struct sk_buff *skb = skb_peek(list);
2338 	if (skb)
2339 		__skb_unlink(skb, list);
2340 	return skb;
2341 }
2342 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2343 
2344 /**
2345  *	__skb_dequeue_tail - remove from the tail of the queue
2346  *	@list: list to dequeue from
2347  *
2348  *	Remove the tail of the list. This function does not take any locks
2349  *	so must be used with appropriate locks held only. The tail item is
2350  *	returned or %NULL if the list is empty.
2351  */
__skb_dequeue_tail(struct sk_buff_head * list)2352 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2353 {
2354 	struct sk_buff *skb = skb_peek_tail(list);
2355 	if (skb)
2356 		__skb_unlink(skb, list);
2357 	return skb;
2358 }
2359 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2360 
2361 
skb_is_nonlinear(const struct sk_buff * skb)2362 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2363 {
2364 	return skb->data_len;
2365 }
2366 
skb_headlen(const struct sk_buff * skb)2367 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2368 {
2369 	return skb->len - skb->data_len;
2370 }
2371 
__skb_pagelen(const struct sk_buff * skb)2372 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2373 {
2374 	unsigned int i, len = 0;
2375 
2376 	for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2377 		len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2378 	return len;
2379 }
2380 
skb_pagelen(const struct sk_buff * skb)2381 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2382 {
2383 	return skb_headlen(skb) + __skb_pagelen(skb);
2384 }
2385 
__skb_fill_page_desc_noacc(struct skb_shared_info * shinfo,int i,struct page * page,int off,int size)2386 static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo,
2387 					      int i, struct page *page,
2388 					      int off, int size)
2389 {
2390 	skb_frag_t *frag = &shinfo->frags[i];
2391 
2392 	/*
2393 	 * Propagate page pfmemalloc to the skb if we can. The problem is
2394 	 * that not all callers have unique ownership of the page but rely
2395 	 * on page_is_pfmemalloc doing the right thing(tm).
2396 	 */
2397 	frag->bv_page		  = page;
2398 	frag->bv_offset		  = off;
2399 	skb_frag_size_set(frag, size);
2400 }
2401 
2402 /**
2403  * skb_len_add - adds a number to len fields of skb
2404  * @skb: buffer to add len to
2405  * @delta: number of bytes to add
2406  */
skb_len_add(struct sk_buff * skb,int delta)2407 static inline void skb_len_add(struct sk_buff *skb, int delta)
2408 {
2409 	skb->len += delta;
2410 	skb->data_len += delta;
2411 	skb->truesize += delta;
2412 }
2413 
2414 /**
2415  * __skb_fill_page_desc - initialise a paged fragment in an skb
2416  * @skb: buffer containing fragment to be initialised
2417  * @i: paged fragment index to initialise
2418  * @page: the page to use for this fragment
2419  * @off: the offset to the data with @page
2420  * @size: the length of the data
2421  *
2422  * Initialises the @i'th fragment of @skb to point to &size bytes at
2423  * offset @off within @page.
2424  *
2425  * Does not take any additional reference on the fragment.
2426  */
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2427 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2428 					struct page *page, int off, int size)
2429 {
2430 	__skb_fill_page_desc_noacc(skb_shinfo(skb), i, page, off, size);
2431 	page = compound_head(page);
2432 	if (page_is_pfmemalloc(page))
2433 		skb->pfmemalloc	= true;
2434 }
2435 
2436 /**
2437  * skb_fill_page_desc - initialise a paged fragment in an skb
2438  * @skb: buffer containing fragment to be initialised
2439  * @i: paged fragment index to initialise
2440  * @page: the page to use for this fragment
2441  * @off: the offset to the data with @page
2442  * @size: the length of the data
2443  *
2444  * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2445  * @skb to point to @size bytes at offset @off within @page. In
2446  * addition updates @skb such that @i is the last fragment.
2447  *
2448  * Does not take any additional reference on the fragment.
2449  */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2450 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2451 				      struct page *page, int off, int size)
2452 {
2453 	__skb_fill_page_desc(skb, i, page, off, size);
2454 	skb_shinfo(skb)->nr_frags = i + 1;
2455 }
2456 
2457 /**
2458  * skb_fill_page_desc_noacc - initialise a paged fragment in an skb
2459  * @skb: buffer containing fragment to be initialised
2460  * @i: paged fragment index to initialise
2461  * @page: the page to use for this fragment
2462  * @off: the offset to the data with @page
2463  * @size: the length of the data
2464  *
2465  * Variant of skb_fill_page_desc() which does not deal with
2466  * pfmemalloc, if page is not owned by us.
2467  */
skb_fill_page_desc_noacc(struct sk_buff * skb,int i,struct page * page,int off,int size)2468 static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i,
2469 					    struct page *page, int off,
2470 					    int size)
2471 {
2472 	struct skb_shared_info *shinfo = skb_shinfo(skb);
2473 
2474 	__skb_fill_page_desc_noacc(shinfo, i, page, off, size);
2475 	shinfo->nr_frags = i + 1;
2476 }
2477 
2478 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2479 		     int size, unsigned int truesize);
2480 
2481 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2482 			  unsigned int truesize);
2483 
2484 #define SKB_LINEAR_ASSERT(skb)  BUG_ON(skb_is_nonlinear(skb))
2485 
2486 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2487 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2488 {
2489 	return skb->head + skb->tail;
2490 }
2491 
skb_reset_tail_pointer(struct sk_buff * skb)2492 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2493 {
2494 	skb->tail = skb->data - skb->head;
2495 }
2496 
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2497 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2498 {
2499 	skb_reset_tail_pointer(skb);
2500 	skb->tail += offset;
2501 }
2502 
2503 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2504 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2505 {
2506 	return skb->tail;
2507 }
2508 
skb_reset_tail_pointer(struct sk_buff * skb)2509 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2510 {
2511 	skb->tail = skb->data;
2512 }
2513 
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2514 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2515 {
2516 	skb->tail = skb->data + offset;
2517 }
2518 
2519 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2520 
skb_assert_len(struct sk_buff * skb)2521 static inline void skb_assert_len(struct sk_buff *skb)
2522 {
2523 #ifdef CONFIG_DEBUG_NET
2524 	if (WARN_ONCE(!skb->len, "%s\n", __func__))
2525 		DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2526 #endif /* CONFIG_DEBUG_NET */
2527 }
2528 
2529 /*
2530  *	Add data to an sk_buff
2531  */
2532 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2533 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2534 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2535 {
2536 	void *tmp = skb_tail_pointer(skb);
2537 	SKB_LINEAR_ASSERT(skb);
2538 	skb->tail += len;
2539 	skb->len  += len;
2540 	return tmp;
2541 }
2542 
__skb_put_zero(struct sk_buff * skb,unsigned int len)2543 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2544 {
2545 	void *tmp = __skb_put(skb, len);
2546 
2547 	memset(tmp, 0, len);
2548 	return tmp;
2549 }
2550 
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2551 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2552 				   unsigned int len)
2553 {
2554 	void *tmp = __skb_put(skb, len);
2555 
2556 	memcpy(tmp, data, len);
2557 	return tmp;
2558 }
2559 
__skb_put_u8(struct sk_buff * skb,u8 val)2560 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2561 {
2562 	*(u8 *)__skb_put(skb, 1) = val;
2563 }
2564 
skb_put_zero(struct sk_buff * skb,unsigned int len)2565 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2566 {
2567 	void *tmp = skb_put(skb, len);
2568 
2569 	memset(tmp, 0, len);
2570 
2571 	return tmp;
2572 }
2573 
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2574 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2575 				 unsigned int len)
2576 {
2577 	void *tmp = skb_put(skb, len);
2578 
2579 	memcpy(tmp, data, len);
2580 
2581 	return tmp;
2582 }
2583 
skb_put_u8(struct sk_buff * skb,u8 val)2584 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2585 {
2586 	*(u8 *)skb_put(skb, 1) = val;
2587 }
2588 
2589 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2590 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2591 {
2592 	skb->data -= len;
2593 	skb->len  += len;
2594 	return skb->data;
2595 }
2596 
2597 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2598 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2599 {
2600 	skb->len -= len;
2601 	if (unlikely(skb->len < skb->data_len)) {
2602 #if defined(CONFIG_DEBUG_NET)
2603 		skb->len += len;
2604 		pr_err("__skb_pull(len=%u)\n", len);
2605 		skb_dump(KERN_ERR, skb, false);
2606 #endif
2607 		BUG();
2608 	}
2609 	return skb->data += len;
2610 }
2611 
skb_pull_inline(struct sk_buff * skb,unsigned int len)2612 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2613 {
2614 	return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2615 }
2616 
2617 void *skb_pull_data(struct sk_buff *skb, size_t len);
2618 
2619 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2620 
pskb_may_pull(struct sk_buff * skb,unsigned int len)2621 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2622 {
2623 	if (likely(len <= skb_headlen(skb)))
2624 		return true;
2625 	if (unlikely(len > skb->len))
2626 		return false;
2627 	return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2628 }
2629 
pskb_pull(struct sk_buff * skb,unsigned int len)2630 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2631 {
2632 	if (!pskb_may_pull(skb, len))
2633 		return NULL;
2634 
2635 	skb->len -= len;
2636 	return skb->data += len;
2637 }
2638 
2639 void skb_condense(struct sk_buff *skb);
2640 
2641 /**
2642  *	skb_headroom - bytes at buffer head
2643  *	@skb: buffer to check
2644  *
2645  *	Return the number of bytes of free space at the head of an &sk_buff.
2646  */
skb_headroom(const struct sk_buff * skb)2647 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2648 {
2649 	return skb->data - skb->head;
2650 }
2651 
2652 /**
2653  *	skb_tailroom - bytes at buffer end
2654  *	@skb: buffer to check
2655  *
2656  *	Return the number of bytes of free space at the tail of an sk_buff
2657  */
skb_tailroom(const struct sk_buff * skb)2658 static inline int skb_tailroom(const struct sk_buff *skb)
2659 {
2660 	return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2661 }
2662 
2663 /**
2664  *	skb_availroom - bytes at buffer end
2665  *	@skb: buffer to check
2666  *
2667  *	Return the number of bytes of free space at the tail of an sk_buff
2668  *	allocated by sk_stream_alloc()
2669  */
skb_availroom(const struct sk_buff * skb)2670 static inline int skb_availroom(const struct sk_buff *skb)
2671 {
2672 	if (skb_is_nonlinear(skb))
2673 		return 0;
2674 
2675 	return skb->end - skb->tail - skb->reserved_tailroom;
2676 }
2677 
2678 /**
2679  *	skb_reserve - adjust headroom
2680  *	@skb: buffer to alter
2681  *	@len: bytes to move
2682  *
2683  *	Increase the headroom of an empty &sk_buff by reducing the tail
2684  *	room. This is only allowed for an empty buffer.
2685  */
skb_reserve(struct sk_buff * skb,int len)2686 static inline void skb_reserve(struct sk_buff *skb, int len)
2687 {
2688 	skb->data += len;
2689 	skb->tail += len;
2690 }
2691 
2692 /**
2693  *	skb_tailroom_reserve - adjust reserved_tailroom
2694  *	@skb: buffer to alter
2695  *	@mtu: maximum amount of headlen permitted
2696  *	@needed_tailroom: minimum amount of reserved_tailroom
2697  *
2698  *	Set reserved_tailroom so that headlen can be as large as possible but
2699  *	not larger than mtu and tailroom cannot be smaller than
2700  *	needed_tailroom.
2701  *	The required headroom should already have been reserved before using
2702  *	this function.
2703  */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2704 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2705 					unsigned int needed_tailroom)
2706 {
2707 	SKB_LINEAR_ASSERT(skb);
2708 	if (mtu < skb_tailroom(skb) - needed_tailroom)
2709 		/* use at most mtu */
2710 		skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2711 	else
2712 		/* use up to all available space */
2713 		skb->reserved_tailroom = needed_tailroom;
2714 }
2715 
2716 #define ENCAP_TYPE_ETHER	0
2717 #define ENCAP_TYPE_IPPROTO	1
2718 
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2719 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2720 					  __be16 protocol)
2721 {
2722 	skb->inner_protocol = protocol;
2723 	skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2724 }
2725 
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2726 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2727 					 __u8 ipproto)
2728 {
2729 	skb->inner_ipproto = ipproto;
2730 	skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2731 }
2732 
skb_reset_inner_headers(struct sk_buff * skb)2733 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2734 {
2735 	skb->inner_mac_header = skb->mac_header;
2736 	skb->inner_network_header = skb->network_header;
2737 	skb->inner_transport_header = skb->transport_header;
2738 }
2739 
skb_reset_mac_len(struct sk_buff * skb)2740 static inline void skb_reset_mac_len(struct sk_buff *skb)
2741 {
2742 	skb->mac_len = skb->network_header - skb->mac_header;
2743 }
2744 
skb_inner_transport_header(const struct sk_buff * skb)2745 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2746 							*skb)
2747 {
2748 	return skb->head + skb->inner_transport_header;
2749 }
2750 
skb_inner_transport_offset(const struct sk_buff * skb)2751 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2752 {
2753 	return skb_inner_transport_header(skb) - skb->data;
2754 }
2755 
skb_reset_inner_transport_header(struct sk_buff * skb)2756 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2757 {
2758 	skb->inner_transport_header = skb->data - skb->head;
2759 }
2760 
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2761 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2762 						   const int offset)
2763 {
2764 	skb_reset_inner_transport_header(skb);
2765 	skb->inner_transport_header += offset;
2766 }
2767 
skb_inner_network_header(const struct sk_buff * skb)2768 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2769 {
2770 	return skb->head + skb->inner_network_header;
2771 }
2772 
skb_reset_inner_network_header(struct sk_buff * skb)2773 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2774 {
2775 	skb->inner_network_header = skb->data - skb->head;
2776 }
2777 
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2778 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2779 						const int offset)
2780 {
2781 	skb_reset_inner_network_header(skb);
2782 	skb->inner_network_header += offset;
2783 }
2784 
skb_inner_mac_header(const struct sk_buff * skb)2785 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2786 {
2787 	return skb->head + skb->inner_mac_header;
2788 }
2789 
skb_reset_inner_mac_header(struct sk_buff * skb)2790 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2791 {
2792 	skb->inner_mac_header = skb->data - skb->head;
2793 }
2794 
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)2795 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2796 					    const int offset)
2797 {
2798 	skb_reset_inner_mac_header(skb);
2799 	skb->inner_mac_header += offset;
2800 }
skb_transport_header_was_set(const struct sk_buff * skb)2801 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2802 {
2803 	return skb->transport_header != (typeof(skb->transport_header))~0U;
2804 }
2805 
skb_transport_header(const struct sk_buff * skb)2806 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2807 {
2808 	DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
2809 	return skb->head + skb->transport_header;
2810 }
2811 
skb_reset_transport_header(struct sk_buff * skb)2812 static inline void skb_reset_transport_header(struct sk_buff *skb)
2813 {
2814 	skb->transport_header = skb->data - skb->head;
2815 }
2816 
skb_set_transport_header(struct sk_buff * skb,const int offset)2817 static inline void skb_set_transport_header(struct sk_buff *skb,
2818 					    const int offset)
2819 {
2820 	skb_reset_transport_header(skb);
2821 	skb->transport_header += offset;
2822 }
2823 
skb_network_header(const struct sk_buff * skb)2824 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2825 {
2826 	return skb->head + skb->network_header;
2827 }
2828 
skb_reset_network_header(struct sk_buff * skb)2829 static inline void skb_reset_network_header(struct sk_buff *skb)
2830 {
2831 	skb->network_header = skb->data - skb->head;
2832 }
2833 
skb_set_network_header(struct sk_buff * skb,const int offset)2834 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2835 {
2836 	skb_reset_network_header(skb);
2837 	skb->network_header += offset;
2838 }
2839 
skb_mac_header_was_set(const struct sk_buff * skb)2840 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2841 {
2842 	return skb->mac_header != (typeof(skb->mac_header))~0U;
2843 }
2844 
skb_mac_header(const struct sk_buff * skb)2845 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2846 {
2847 	DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
2848 	return skb->head + skb->mac_header;
2849 }
2850 
skb_mac_offset(const struct sk_buff * skb)2851 static inline int skb_mac_offset(const struct sk_buff *skb)
2852 {
2853 	return skb_mac_header(skb) - skb->data;
2854 }
2855 
skb_mac_header_len(const struct sk_buff * skb)2856 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2857 {
2858 	DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
2859 	return skb->network_header - skb->mac_header;
2860 }
2861 
skb_unset_mac_header(struct sk_buff * skb)2862 static inline void skb_unset_mac_header(struct sk_buff *skb)
2863 {
2864 	skb->mac_header = (typeof(skb->mac_header))~0U;
2865 }
2866 
skb_reset_mac_header(struct sk_buff * skb)2867 static inline void skb_reset_mac_header(struct sk_buff *skb)
2868 {
2869 	skb->mac_header = skb->data - skb->head;
2870 }
2871 
skb_set_mac_header(struct sk_buff * skb,const int offset)2872 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2873 {
2874 	skb_reset_mac_header(skb);
2875 	skb->mac_header += offset;
2876 }
2877 
skb_pop_mac_header(struct sk_buff * skb)2878 static inline void skb_pop_mac_header(struct sk_buff *skb)
2879 {
2880 	skb->mac_header = skb->network_header;
2881 }
2882 
skb_probe_transport_header(struct sk_buff * skb)2883 static inline void skb_probe_transport_header(struct sk_buff *skb)
2884 {
2885 	struct flow_keys_basic keys;
2886 
2887 	if (skb_transport_header_was_set(skb))
2888 		return;
2889 
2890 	if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2891 					     NULL, 0, 0, 0, 0))
2892 		skb_set_transport_header(skb, keys.control.thoff);
2893 }
2894 
skb_mac_header_rebuild(struct sk_buff * skb)2895 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2896 {
2897 	if (skb_mac_header_was_set(skb)) {
2898 		const unsigned char *old_mac = skb_mac_header(skb);
2899 
2900 		skb_set_mac_header(skb, -skb->mac_len);
2901 		memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2902 	}
2903 }
2904 
skb_checksum_start_offset(const struct sk_buff * skb)2905 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2906 {
2907 	return skb->csum_start - skb_headroom(skb);
2908 }
2909 
skb_checksum_start(const struct sk_buff * skb)2910 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2911 {
2912 	return skb->head + skb->csum_start;
2913 }
2914 
skb_transport_offset(const struct sk_buff * skb)2915 static inline int skb_transport_offset(const struct sk_buff *skb)
2916 {
2917 	return skb_transport_header(skb) - skb->data;
2918 }
2919 
skb_network_header_len(const struct sk_buff * skb)2920 static inline u32 skb_network_header_len(const struct sk_buff *skb)
2921 {
2922 	return skb->transport_header - skb->network_header;
2923 }
2924 
skb_inner_network_header_len(const struct sk_buff * skb)2925 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2926 {
2927 	return skb->inner_transport_header - skb->inner_network_header;
2928 }
2929 
skb_network_offset(const struct sk_buff * skb)2930 static inline int skb_network_offset(const struct sk_buff *skb)
2931 {
2932 	return skb_network_header(skb) - skb->data;
2933 }
2934 
skb_inner_network_offset(const struct sk_buff * skb)2935 static inline int skb_inner_network_offset(const struct sk_buff *skb)
2936 {
2937 	return skb_inner_network_header(skb) - skb->data;
2938 }
2939 
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)2940 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2941 {
2942 	return pskb_may_pull(skb, skb_network_offset(skb) + len);
2943 }
2944 
2945 /*
2946  * CPUs often take a performance hit when accessing unaligned memory
2947  * locations. The actual performance hit varies, it can be small if the
2948  * hardware handles it or large if we have to take an exception and fix it
2949  * in software.
2950  *
2951  * Since an ethernet header is 14 bytes network drivers often end up with
2952  * the IP header at an unaligned offset. The IP header can be aligned by
2953  * shifting the start of the packet by 2 bytes. Drivers should do this
2954  * with:
2955  *
2956  * skb_reserve(skb, NET_IP_ALIGN);
2957  *
2958  * The downside to this alignment of the IP header is that the DMA is now
2959  * unaligned. On some architectures the cost of an unaligned DMA is high
2960  * and this cost outweighs the gains made by aligning the IP header.
2961  *
2962  * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2963  * to be overridden.
2964  */
2965 #ifndef NET_IP_ALIGN
2966 #define NET_IP_ALIGN	2
2967 #endif
2968 
2969 /*
2970  * The networking layer reserves some headroom in skb data (via
2971  * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2972  * the header has to grow. In the default case, if the header has to grow
2973  * 32 bytes or less we avoid the reallocation.
2974  *
2975  * Unfortunately this headroom changes the DMA alignment of the resulting
2976  * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2977  * on some architectures. An architecture can override this value,
2978  * perhaps setting it to a cacheline in size (since that will maintain
2979  * cacheline alignment of the DMA). It must be a power of 2.
2980  *
2981  * Various parts of the networking layer expect at least 32 bytes of
2982  * headroom, you should not reduce this.
2983  *
2984  * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2985  * to reduce average number of cache lines per packet.
2986  * get_rps_cpu() for example only access one 64 bytes aligned block :
2987  * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2988  */
2989 #ifndef NET_SKB_PAD
2990 #define NET_SKB_PAD	max(32, L1_CACHE_BYTES)
2991 #endif
2992 
2993 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2994 
__skb_set_length(struct sk_buff * skb,unsigned int len)2995 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2996 {
2997 	if (WARN_ON(skb_is_nonlinear(skb)))
2998 		return;
2999 	skb->len = len;
3000 	skb_set_tail_pointer(skb, len);
3001 }
3002 
__skb_trim(struct sk_buff * skb,unsigned int len)3003 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
3004 {
3005 	__skb_set_length(skb, len);
3006 }
3007 
3008 void skb_trim(struct sk_buff *skb, unsigned int len);
3009 
__pskb_trim(struct sk_buff * skb,unsigned int len)3010 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
3011 {
3012 	if (skb->data_len)
3013 		return ___pskb_trim(skb, len);
3014 	__skb_trim(skb, len);
3015 	return 0;
3016 }
3017 
pskb_trim(struct sk_buff * skb,unsigned int len)3018 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
3019 {
3020 	return (len < skb->len) ? __pskb_trim(skb, len) : 0;
3021 }
3022 
3023 /**
3024  *	pskb_trim_unique - remove end from a paged unique (not cloned) buffer
3025  *	@skb: buffer to alter
3026  *	@len: new length
3027  *
3028  *	This is identical to pskb_trim except that the caller knows that
3029  *	the skb is not cloned so we should never get an error due to out-
3030  *	of-memory.
3031  */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)3032 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
3033 {
3034 	int err = pskb_trim(skb, len);
3035 	BUG_ON(err);
3036 }
3037 
__skb_grow(struct sk_buff * skb,unsigned int len)3038 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
3039 {
3040 	unsigned int diff = len - skb->len;
3041 
3042 	if (skb_tailroom(skb) < diff) {
3043 		int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
3044 					   GFP_ATOMIC);
3045 		if (ret)
3046 			return ret;
3047 	}
3048 	__skb_set_length(skb, len);
3049 	return 0;
3050 }
3051 
3052 /**
3053  *	skb_orphan - orphan a buffer
3054  *	@skb: buffer to orphan
3055  *
3056  *	If a buffer currently has an owner then we call the owner's
3057  *	destructor function and make the @skb unowned. The buffer continues
3058  *	to exist but is no longer charged to its former owner.
3059  */
skb_orphan(struct sk_buff * skb)3060 static inline void skb_orphan(struct sk_buff *skb)
3061 {
3062 	if (skb->destructor) {
3063 		skb->destructor(skb);
3064 		skb->destructor = NULL;
3065 		skb->sk		= NULL;
3066 	} else {
3067 		BUG_ON(skb->sk);
3068 	}
3069 }
3070 
3071 /**
3072  *	skb_orphan_frags - orphan the frags contained in a buffer
3073  *	@skb: buffer to orphan frags from
3074  *	@gfp_mask: allocation mask for replacement pages
3075  *
3076  *	For each frag in the SKB which needs a destructor (i.e. has an
3077  *	owner) create a copy of that frag and release the original
3078  *	page by calling the destructor.
3079  */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)3080 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
3081 {
3082 	if (likely(!skb_zcopy(skb)))
3083 		return 0;
3084 	if (skb_shinfo(skb)->flags & SKBFL_DONT_ORPHAN)
3085 		return 0;
3086 	return skb_copy_ubufs(skb, gfp_mask);
3087 }
3088 
3089 /* 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)3090 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
3091 {
3092 	if (likely(!skb_zcopy(skb)))
3093 		return 0;
3094 	return skb_copy_ubufs(skb, gfp_mask);
3095 }
3096 
3097 /**
3098  *	__skb_queue_purge - empty a list
3099  *	@list: list to empty
3100  *
3101  *	Delete all buffers on an &sk_buff list. Each buffer is removed from
3102  *	the list and one reference dropped. This function does not take the
3103  *	list lock and the caller must hold the relevant locks to use it.
3104  */
__skb_queue_purge(struct sk_buff_head * list)3105 static inline void __skb_queue_purge(struct sk_buff_head *list)
3106 {
3107 	struct sk_buff *skb;
3108 	while ((skb = __skb_dequeue(list)) != NULL)
3109 		kfree_skb(skb);
3110 }
3111 void skb_queue_purge(struct sk_buff_head *list);
3112 
3113 unsigned int skb_rbtree_purge(struct rb_root *root);
3114 
3115 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3116 
3117 /**
3118  * netdev_alloc_frag - allocate a page fragment
3119  * @fragsz: fragment size
3120  *
3121  * Allocates a frag from a page for receive buffer.
3122  * Uses GFP_ATOMIC allocations.
3123  */
netdev_alloc_frag(unsigned int fragsz)3124 static inline void *netdev_alloc_frag(unsigned int fragsz)
3125 {
3126 	return __netdev_alloc_frag_align(fragsz, ~0u);
3127 }
3128 
netdev_alloc_frag_align(unsigned int fragsz,unsigned int align)3129 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
3130 					    unsigned int align)
3131 {
3132 	WARN_ON_ONCE(!is_power_of_2(align));
3133 	return __netdev_alloc_frag_align(fragsz, -align);
3134 }
3135 
3136 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
3137 				   gfp_t gfp_mask);
3138 
3139 /**
3140  *	netdev_alloc_skb - allocate an skbuff for rx on a specific device
3141  *	@dev: network device to receive on
3142  *	@length: length to allocate
3143  *
3144  *	Allocate a new &sk_buff and assign it a usage count of one. The
3145  *	buffer has unspecified headroom built in. Users should allocate
3146  *	the headroom they think they need without accounting for the
3147  *	built in space. The built in space is used for optimisations.
3148  *
3149  *	%NULL is returned if there is no free memory. Although this function
3150  *	allocates memory it can be called from an interrupt.
3151  */
netdev_alloc_skb(struct net_device * dev,unsigned int length)3152 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
3153 					       unsigned int length)
3154 {
3155 	return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
3156 }
3157 
3158 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)3159 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
3160 					      gfp_t gfp_mask)
3161 {
3162 	return __netdev_alloc_skb(NULL, length, gfp_mask);
3163 }
3164 
3165 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)3166 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
3167 {
3168 	return netdev_alloc_skb(NULL, length);
3169 }
3170 
3171 
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)3172 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
3173 		unsigned int length, gfp_t gfp)
3174 {
3175 	struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
3176 
3177 	if (NET_IP_ALIGN && skb)
3178 		skb_reserve(skb, NET_IP_ALIGN);
3179 	return skb;
3180 }
3181 
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)3182 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
3183 		unsigned int length)
3184 {
3185 	return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
3186 }
3187 
skb_free_frag(void * addr)3188 static inline void skb_free_frag(void *addr)
3189 {
3190 	page_frag_free(addr);
3191 }
3192 
3193 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3194 
napi_alloc_frag(unsigned int fragsz)3195 static inline void *napi_alloc_frag(unsigned int fragsz)
3196 {
3197 	return __napi_alloc_frag_align(fragsz, ~0u);
3198 }
3199 
napi_alloc_frag_align(unsigned int fragsz,unsigned int align)3200 static inline void *napi_alloc_frag_align(unsigned int fragsz,
3201 					  unsigned int align)
3202 {
3203 	WARN_ON_ONCE(!is_power_of_2(align));
3204 	return __napi_alloc_frag_align(fragsz, -align);
3205 }
3206 
3207 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
3208 				 unsigned int length, gfp_t gfp_mask);
napi_alloc_skb(struct napi_struct * napi,unsigned int length)3209 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
3210 					     unsigned int length)
3211 {
3212 	return __napi_alloc_skb(napi, length, GFP_ATOMIC);
3213 }
3214 void napi_consume_skb(struct sk_buff *skb, int budget);
3215 
3216 void napi_skb_free_stolen_head(struct sk_buff *skb);
3217 void __kfree_skb_defer(struct sk_buff *skb);
3218 
3219 /**
3220  * __dev_alloc_pages - allocate page for network Rx
3221  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3222  * @order: size of the allocation
3223  *
3224  * Allocate a new page.
3225  *
3226  * %NULL is returned if there is no free memory.
3227 */
__dev_alloc_pages(gfp_t gfp_mask,unsigned int order)3228 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
3229 					     unsigned int order)
3230 {
3231 	/* This piece of code contains several assumptions.
3232 	 * 1.  This is for device Rx, therefor a cold page is preferred.
3233 	 * 2.  The expectation is the user wants a compound page.
3234 	 * 3.  If requesting a order 0 page it will not be compound
3235 	 *     due to the check to see if order has a value in prep_new_page
3236 	 * 4.  __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3237 	 *     code in gfp_to_alloc_flags that should be enforcing this.
3238 	 */
3239 	gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3240 
3241 	return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
3242 }
3243 
dev_alloc_pages(unsigned int order)3244 static inline struct page *dev_alloc_pages(unsigned int order)
3245 {
3246 	return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
3247 }
3248 
3249 /**
3250  * __dev_alloc_page - allocate a page for network Rx
3251  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3252  *
3253  * Allocate a new page.
3254  *
3255  * %NULL is returned if there is no free memory.
3256  */
__dev_alloc_page(gfp_t gfp_mask)3257 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
3258 {
3259 	return __dev_alloc_pages(gfp_mask, 0);
3260 }
3261 
dev_alloc_page(void)3262 static inline struct page *dev_alloc_page(void)
3263 {
3264 	return dev_alloc_pages(0);
3265 }
3266 
3267 /**
3268  * dev_page_is_reusable - check whether a page can be reused for network Rx
3269  * @page: the page to test
3270  *
3271  * A page shouldn't be considered for reusing/recycling if it was allocated
3272  * under memory pressure or at a distant memory node.
3273  *
3274  * Returns false if this page should be returned to page allocator, true
3275  * otherwise.
3276  */
dev_page_is_reusable(const struct page * page)3277 static inline bool dev_page_is_reusable(const struct page *page)
3278 {
3279 	return likely(page_to_nid(page) == numa_mem_id() &&
3280 		      !page_is_pfmemalloc(page));
3281 }
3282 
3283 /**
3284  *	skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3285  *	@page: The page that was allocated from skb_alloc_page
3286  *	@skb: The skb that may need pfmemalloc set
3287  */
skb_propagate_pfmemalloc(const struct page * page,struct sk_buff * skb)3288 static inline void skb_propagate_pfmemalloc(const struct page *page,
3289 					    struct sk_buff *skb)
3290 {
3291 	if (page_is_pfmemalloc(page))
3292 		skb->pfmemalloc = true;
3293 }
3294 
3295 /**
3296  * skb_frag_off() - Returns the offset of a skb fragment
3297  * @frag: the paged fragment
3298  */
skb_frag_off(const skb_frag_t * frag)3299 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3300 {
3301 	return frag->bv_offset;
3302 }
3303 
3304 /**
3305  * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3306  * @frag: skb fragment
3307  * @delta: value to add
3308  */
skb_frag_off_add(skb_frag_t * frag,int delta)3309 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3310 {
3311 	frag->bv_offset += delta;
3312 }
3313 
3314 /**
3315  * skb_frag_off_set() - Sets the offset of a skb fragment
3316  * @frag: skb fragment
3317  * @offset: offset of fragment
3318  */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)3319 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3320 {
3321 	frag->bv_offset = offset;
3322 }
3323 
3324 /**
3325  * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3326  * @fragto: skb fragment where offset is set
3327  * @fragfrom: skb fragment offset is copied from
3328  */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3329 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3330 				     const skb_frag_t *fragfrom)
3331 {
3332 	fragto->bv_offset = fragfrom->bv_offset;
3333 }
3334 
3335 /**
3336  * skb_frag_page - retrieve the page referred to by a paged fragment
3337  * @frag: the paged fragment
3338  *
3339  * Returns the &struct page associated with @frag.
3340  */
skb_frag_page(const skb_frag_t * frag)3341 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3342 {
3343 	return frag->bv_page;
3344 }
3345 
3346 /**
3347  * __skb_frag_ref - take an addition reference on a paged fragment.
3348  * @frag: the paged fragment
3349  *
3350  * Takes an additional reference on the paged fragment @frag.
3351  */
__skb_frag_ref(skb_frag_t * frag)3352 static inline void __skb_frag_ref(skb_frag_t *frag)
3353 {
3354 	get_page(skb_frag_page(frag));
3355 }
3356 
3357 /**
3358  * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3359  * @skb: the buffer
3360  * @f: the fragment offset.
3361  *
3362  * Takes an additional reference on the @f'th paged fragment of @skb.
3363  */
skb_frag_ref(struct sk_buff * skb,int f)3364 static inline void skb_frag_ref(struct sk_buff *skb, int f)
3365 {
3366 	__skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3367 }
3368 
3369 /**
3370  * __skb_frag_unref - release a reference on a paged fragment.
3371  * @frag: the paged fragment
3372  * @recycle: recycle the page if allocated via page_pool
3373  *
3374  * Releases a reference on the paged fragment @frag
3375  * or recycles the page via the page_pool API.
3376  */
__skb_frag_unref(skb_frag_t * frag,bool recycle)3377 static inline void __skb_frag_unref(skb_frag_t *frag, bool recycle)
3378 {
3379 	struct page *page = skb_frag_page(frag);
3380 
3381 #ifdef CONFIG_PAGE_POOL
3382 	if (recycle && page_pool_return_skb_page(page))
3383 		return;
3384 #endif
3385 	put_page(page);
3386 }
3387 
3388 /**
3389  * skb_frag_unref - release a reference on a paged fragment of an skb.
3390  * @skb: the buffer
3391  * @f: the fragment offset
3392  *
3393  * Releases a reference on the @f'th paged fragment of @skb.
3394  */
skb_frag_unref(struct sk_buff * skb,int f)3395 static inline void skb_frag_unref(struct sk_buff *skb, int f)
3396 {
3397 	struct skb_shared_info *shinfo = skb_shinfo(skb);
3398 
3399 	if (!skb_zcopy_managed(skb))
3400 		__skb_frag_unref(&shinfo->frags[f], skb->pp_recycle);
3401 }
3402 
3403 /**
3404  * skb_frag_address - gets the address of the data contained in a paged fragment
3405  * @frag: the paged fragment buffer
3406  *
3407  * Returns the address of the data within @frag. The page must already
3408  * be mapped.
3409  */
skb_frag_address(const skb_frag_t * frag)3410 static inline void *skb_frag_address(const skb_frag_t *frag)
3411 {
3412 	return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3413 }
3414 
3415 /**
3416  * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3417  * @frag: the paged fragment buffer
3418  *
3419  * Returns the address of the data within @frag. Checks that the page
3420  * is mapped and returns %NULL otherwise.
3421  */
skb_frag_address_safe(const skb_frag_t * frag)3422 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3423 {
3424 	void *ptr = page_address(skb_frag_page(frag));
3425 	if (unlikely(!ptr))
3426 		return NULL;
3427 
3428 	return ptr + skb_frag_off(frag);
3429 }
3430 
3431 /**
3432  * skb_frag_page_copy() - sets the page in a fragment from another fragment
3433  * @fragto: skb fragment where page is set
3434  * @fragfrom: skb fragment page is copied from
3435  */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3436 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3437 				      const skb_frag_t *fragfrom)
3438 {
3439 	fragto->bv_page = fragfrom->bv_page;
3440 }
3441 
3442 /**
3443  * __skb_frag_set_page - sets the page contained in a paged fragment
3444  * @frag: the paged fragment
3445  * @page: the page to set
3446  *
3447  * Sets the fragment @frag to contain @page.
3448  */
__skb_frag_set_page(skb_frag_t * frag,struct page * page)3449 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3450 {
3451 	frag->bv_page = page;
3452 }
3453 
3454 /**
3455  * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3456  * @skb: the buffer
3457  * @f: the fragment offset
3458  * @page: the page to set
3459  *
3460  * Sets the @f'th fragment of @skb to contain @page.
3461  */
skb_frag_set_page(struct sk_buff * skb,int f,struct page * page)3462 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3463 				     struct page *page)
3464 {
3465 	__skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3466 }
3467 
3468 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3469 
3470 /**
3471  * skb_frag_dma_map - maps a paged fragment via the DMA API
3472  * @dev: the device to map the fragment to
3473  * @frag: the paged fragment to map
3474  * @offset: the offset within the fragment (starting at the
3475  *          fragment's own offset)
3476  * @size: the number of bytes to map
3477  * @dir: the direction of the mapping (``PCI_DMA_*``)
3478  *
3479  * Maps the page associated with @frag to @device.
3480  */
skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3481 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3482 					  const skb_frag_t *frag,
3483 					  size_t offset, size_t size,
3484 					  enum dma_data_direction dir)
3485 {
3486 	return dma_map_page(dev, skb_frag_page(frag),
3487 			    skb_frag_off(frag) + offset, size, dir);
3488 }
3489 
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3490 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3491 					gfp_t gfp_mask)
3492 {
3493 	return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3494 }
3495 
3496 
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3497 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3498 						  gfp_t gfp_mask)
3499 {
3500 	return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3501 }
3502 
3503 
3504 /**
3505  *	skb_clone_writable - is the header of a clone writable
3506  *	@skb: buffer to check
3507  *	@len: length up to which to write
3508  *
3509  *	Returns true if modifying the header part of the cloned buffer
3510  *	does not requires the data to be copied.
3511  */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3512 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3513 {
3514 	return !skb_header_cloned(skb) &&
3515 	       skb_headroom(skb) + len <= skb->hdr_len;
3516 }
3517 
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3518 static inline int skb_try_make_writable(struct sk_buff *skb,
3519 					unsigned int write_len)
3520 {
3521 	return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3522 	       pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3523 }
3524 
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3525 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3526 			    int cloned)
3527 {
3528 	int delta = 0;
3529 
3530 	if (headroom > skb_headroom(skb))
3531 		delta = headroom - skb_headroom(skb);
3532 
3533 	if (delta || cloned)
3534 		return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3535 					GFP_ATOMIC);
3536 	return 0;
3537 }
3538 
3539 /**
3540  *	skb_cow - copy header of skb when it is required
3541  *	@skb: buffer to cow
3542  *	@headroom: needed headroom
3543  *
3544  *	If the skb passed lacks sufficient headroom or its data part
3545  *	is shared, data is reallocated. If reallocation fails, an error
3546  *	is returned and original skb is not changed.
3547  *
3548  *	The result is skb with writable area skb->head...skb->tail
3549  *	and at least @headroom of space at head.
3550  */
skb_cow(struct sk_buff * skb,unsigned int headroom)3551 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3552 {
3553 	return __skb_cow(skb, headroom, skb_cloned(skb));
3554 }
3555 
3556 /**
3557  *	skb_cow_head - skb_cow but only making the head writable
3558  *	@skb: buffer to cow
3559  *	@headroom: needed headroom
3560  *
3561  *	This function is identical to skb_cow except that we replace the
3562  *	skb_cloned check by skb_header_cloned.  It should be used when
3563  *	you only need to push on some header and do not need to modify
3564  *	the data.
3565  */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3566 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3567 {
3568 	return __skb_cow(skb, headroom, skb_header_cloned(skb));
3569 }
3570 
3571 /**
3572  *	skb_padto	- pad an skbuff up to a minimal size
3573  *	@skb: buffer to pad
3574  *	@len: minimal length
3575  *
3576  *	Pads up a buffer to ensure the trailing bytes exist and are
3577  *	blanked. If the buffer already contains sufficient data it
3578  *	is untouched. Otherwise it is extended. Returns zero on
3579  *	success. The skb is freed on error.
3580  */
skb_padto(struct sk_buff * skb,unsigned int len)3581 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3582 {
3583 	unsigned int size = skb->len;
3584 	if (likely(size >= len))
3585 		return 0;
3586 	return skb_pad(skb, len - size);
3587 }
3588 
3589 /**
3590  *	__skb_put_padto - increase size and pad an skbuff up to a minimal size
3591  *	@skb: buffer to pad
3592  *	@len: minimal length
3593  *	@free_on_error: free buffer on error
3594  *
3595  *	Pads up a buffer to ensure the trailing bytes exist and are
3596  *	blanked. If the buffer already contains sufficient data it
3597  *	is untouched. Otherwise it is extended. Returns zero on
3598  *	success. The skb is freed on error if @free_on_error is true.
3599  */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3600 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3601 					       unsigned int len,
3602 					       bool free_on_error)
3603 {
3604 	unsigned int size = skb->len;
3605 
3606 	if (unlikely(size < len)) {
3607 		len -= size;
3608 		if (__skb_pad(skb, len, free_on_error))
3609 			return -ENOMEM;
3610 		__skb_put(skb, len);
3611 	}
3612 	return 0;
3613 }
3614 
3615 /**
3616  *	skb_put_padto - increase size and pad an skbuff up to a minimal size
3617  *	@skb: buffer to pad
3618  *	@len: minimal length
3619  *
3620  *	Pads up a buffer to ensure the trailing bytes exist and are
3621  *	blanked. If the buffer already contains sufficient data it
3622  *	is untouched. Otherwise it is extended. Returns zero on
3623  *	success. The skb is freed on error.
3624  */
skb_put_padto(struct sk_buff * skb,unsigned int len)3625 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3626 {
3627 	return __skb_put_padto(skb, len, true);
3628 }
3629 
skb_add_data(struct sk_buff * skb,struct iov_iter * from,int copy)3630 static inline int skb_add_data(struct sk_buff *skb,
3631 			       struct iov_iter *from, int copy)
3632 {
3633 	const int off = skb->len;
3634 
3635 	if (skb->ip_summed == CHECKSUM_NONE) {
3636 		__wsum csum = 0;
3637 		if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3638 					         &csum, from)) {
3639 			skb->csum = csum_block_add(skb->csum, csum, off);
3640 			return 0;
3641 		}
3642 	} else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3643 		return 0;
3644 
3645 	__skb_trim(skb, off);
3646 	return -EFAULT;
3647 }
3648 
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3649 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3650 				    const struct page *page, int off)
3651 {
3652 	if (skb_zcopy(skb))
3653 		return false;
3654 	if (i) {
3655 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3656 
3657 		return page == skb_frag_page(frag) &&
3658 		       off == skb_frag_off(frag) + skb_frag_size(frag);
3659 	}
3660 	return false;
3661 }
3662 
__skb_linearize(struct sk_buff * skb)3663 static inline int __skb_linearize(struct sk_buff *skb)
3664 {
3665 	return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3666 }
3667 
3668 /**
3669  *	skb_linearize - convert paged skb to linear one
3670  *	@skb: buffer to linarize
3671  *
3672  *	If there is no free memory -ENOMEM is returned, otherwise zero
3673  *	is returned and the old skb data released.
3674  */
skb_linearize(struct sk_buff * skb)3675 static inline int skb_linearize(struct sk_buff *skb)
3676 {
3677 	return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3678 }
3679 
3680 /**
3681  * skb_has_shared_frag - can any frag be overwritten
3682  * @skb: buffer to test
3683  *
3684  * Return true if the skb has at least one frag that might be modified
3685  * by an external entity (as in vmsplice()/sendfile())
3686  */
skb_has_shared_frag(const struct sk_buff * skb)3687 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3688 {
3689 	return skb_is_nonlinear(skb) &&
3690 	       skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3691 }
3692 
3693 /**
3694  *	skb_linearize_cow - make sure skb is linear and writable
3695  *	@skb: buffer to process
3696  *
3697  *	If there is no free memory -ENOMEM is returned, otherwise zero
3698  *	is returned and the old skb data released.
3699  */
skb_linearize_cow(struct sk_buff * skb)3700 static inline int skb_linearize_cow(struct sk_buff *skb)
3701 {
3702 	return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3703 	       __skb_linearize(skb) : 0;
3704 }
3705 
3706 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3707 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3708 		     unsigned int off)
3709 {
3710 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3711 		skb->csum = csum_block_sub(skb->csum,
3712 					   csum_partial(start, len, 0), off);
3713 	else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3714 		 skb_checksum_start_offset(skb) < 0)
3715 		skb->ip_summed = CHECKSUM_NONE;
3716 }
3717 
3718 /**
3719  *	skb_postpull_rcsum - update checksum for received skb after pull
3720  *	@skb: buffer to update
3721  *	@start: start of data before pull
3722  *	@len: length of data pulled
3723  *
3724  *	After doing a pull on a received packet, you need to call this to
3725  *	update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3726  *	CHECKSUM_NONE so that it can be recomputed from scratch.
3727  */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3728 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3729 				      const void *start, unsigned int len)
3730 {
3731 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3732 		skb->csum = wsum_negate(csum_partial(start, len,
3733 						     wsum_negate(skb->csum)));
3734 	else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3735 		 skb_checksum_start_offset(skb) < 0)
3736 		skb->ip_summed = CHECKSUM_NONE;
3737 }
3738 
3739 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3740 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3741 		     unsigned int off)
3742 {
3743 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3744 		skb->csum = csum_block_add(skb->csum,
3745 					   csum_partial(start, len, 0), off);
3746 }
3747 
3748 /**
3749  *	skb_postpush_rcsum - update checksum for received skb after push
3750  *	@skb: buffer to update
3751  *	@start: start of data after push
3752  *	@len: length of data pushed
3753  *
3754  *	After doing a push on a received packet, you need to call this to
3755  *	update the CHECKSUM_COMPLETE checksum.
3756  */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3757 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3758 				      const void *start, unsigned int len)
3759 {
3760 	__skb_postpush_rcsum(skb, start, len, 0);
3761 }
3762 
3763 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3764 
3765 /**
3766  *	skb_push_rcsum - push skb and update receive checksum
3767  *	@skb: buffer to update
3768  *	@len: length of data pulled
3769  *
3770  *	This function performs an skb_push on the packet and updates
3771  *	the CHECKSUM_COMPLETE checksum.  It should be used on
3772  *	receive path processing instead of skb_push unless you know
3773  *	that the checksum difference is zero (e.g., a valid IP header)
3774  *	or you are setting ip_summed to CHECKSUM_NONE.
3775  */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3776 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3777 {
3778 	skb_push(skb, len);
3779 	skb_postpush_rcsum(skb, skb->data, len);
3780 	return skb->data;
3781 }
3782 
3783 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3784 /**
3785  *	pskb_trim_rcsum - trim received skb and update checksum
3786  *	@skb: buffer to trim
3787  *	@len: new length
3788  *
3789  *	This is exactly the same as pskb_trim except that it ensures the
3790  *	checksum of received packets are still valid after the operation.
3791  *	It can change skb pointers.
3792  */
3793 
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)3794 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3795 {
3796 	if (likely(len >= skb->len))
3797 		return 0;
3798 	return pskb_trim_rcsum_slow(skb, len);
3799 }
3800 
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)3801 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3802 {
3803 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3804 		skb->ip_summed = CHECKSUM_NONE;
3805 	__skb_trim(skb, len);
3806 	return 0;
3807 }
3808 
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)3809 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3810 {
3811 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3812 		skb->ip_summed = CHECKSUM_NONE;
3813 	return __skb_grow(skb, len);
3814 }
3815 
3816 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3817 #define skb_rb_first(root) rb_to_skb(rb_first(root))
3818 #define skb_rb_last(root)  rb_to_skb(rb_last(root))
3819 #define skb_rb_next(skb)   rb_to_skb(rb_next(&(skb)->rbnode))
3820 #define skb_rb_prev(skb)   rb_to_skb(rb_prev(&(skb)->rbnode))
3821 
3822 #define skb_queue_walk(queue, skb) \
3823 		for (skb = (queue)->next;					\
3824 		     skb != (struct sk_buff *)(queue);				\
3825 		     skb = skb->next)
3826 
3827 #define skb_queue_walk_safe(queue, skb, tmp)					\
3828 		for (skb = (queue)->next, tmp = skb->next;			\
3829 		     skb != (struct sk_buff *)(queue);				\
3830 		     skb = tmp, tmp = skb->next)
3831 
3832 #define skb_queue_walk_from(queue, skb)						\
3833 		for (; skb != (struct sk_buff *)(queue);			\
3834 		     skb = skb->next)
3835 
3836 #define skb_rbtree_walk(skb, root)						\
3837 		for (skb = skb_rb_first(root); skb != NULL;			\
3838 		     skb = skb_rb_next(skb))
3839 
3840 #define skb_rbtree_walk_from(skb)						\
3841 		for (; skb != NULL;						\
3842 		     skb = skb_rb_next(skb))
3843 
3844 #define skb_rbtree_walk_from_safe(skb, tmp)					\
3845 		for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL);	\
3846 		     skb = tmp)
3847 
3848 #define skb_queue_walk_from_safe(queue, skb, tmp)				\
3849 		for (tmp = skb->next;						\
3850 		     skb != (struct sk_buff *)(queue);				\
3851 		     skb = tmp, tmp = skb->next)
3852 
3853 #define skb_queue_reverse_walk(queue, skb) \
3854 		for (skb = (queue)->prev;					\
3855 		     skb != (struct sk_buff *)(queue);				\
3856 		     skb = skb->prev)
3857 
3858 #define skb_queue_reverse_walk_safe(queue, skb, tmp)				\
3859 		for (skb = (queue)->prev, tmp = skb->prev;			\
3860 		     skb != (struct sk_buff *)(queue);				\
3861 		     skb = tmp, tmp = skb->prev)
3862 
3863 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp)			\
3864 		for (tmp = skb->prev;						\
3865 		     skb != (struct sk_buff *)(queue);				\
3866 		     skb = tmp, tmp = skb->prev)
3867 
skb_has_frag_list(const struct sk_buff * skb)3868 static inline bool skb_has_frag_list(const struct sk_buff *skb)
3869 {
3870 	return skb_shinfo(skb)->frag_list != NULL;
3871 }
3872 
skb_frag_list_init(struct sk_buff * skb)3873 static inline void skb_frag_list_init(struct sk_buff *skb)
3874 {
3875 	skb_shinfo(skb)->frag_list = NULL;
3876 }
3877 
3878 #define skb_walk_frags(skb, iter)	\
3879 	for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3880 
3881 
3882 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3883 				int *err, long *timeo_p,
3884 				const struct sk_buff *skb);
3885 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3886 					  struct sk_buff_head *queue,
3887 					  unsigned int flags,
3888 					  int *off, int *err,
3889 					  struct sk_buff **last);
3890 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3891 					struct sk_buff_head *queue,
3892 					unsigned int flags, int *off, int *err,
3893 					struct sk_buff **last);
3894 struct sk_buff *__skb_recv_datagram(struct sock *sk,
3895 				    struct sk_buff_head *sk_queue,
3896 				    unsigned int flags, int *off, int *err);
3897 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err);
3898 __poll_t datagram_poll(struct file *file, struct socket *sock,
3899 			   struct poll_table_struct *wait);
3900 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3901 			   struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)3902 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3903 					struct msghdr *msg, int size)
3904 {
3905 	return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3906 }
3907 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3908 				   struct msghdr *msg);
3909 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3910 			   struct iov_iter *to, int len,
3911 			   struct ahash_request *hash);
3912 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3913 				 struct iov_iter *from, int len);
3914 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3915 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3916 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)3917 static inline void skb_free_datagram_locked(struct sock *sk,
3918 					    struct sk_buff *skb)
3919 {
3920 	__skb_free_datagram_locked(sk, skb, 0);
3921 }
3922 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3923 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3924 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3925 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3926 			      int len);
3927 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3928 		    struct pipe_inode_info *pipe, unsigned int len,
3929 		    unsigned int flags);
3930 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3931 			 int len);
3932 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
3933 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3934 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3935 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3936 		 int len, int hlen);
3937 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3938 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3939 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3940 bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3941 bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3942 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3943 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3944 				 unsigned int offset);
3945 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3946 int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len);
3947 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3948 int skb_vlan_pop(struct sk_buff *skb);
3949 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
3950 int skb_eth_pop(struct sk_buff *skb);
3951 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3952 		 const unsigned char *src);
3953 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3954 		  int mac_len, bool ethernet);
3955 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3956 		 bool ethernet);
3957 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3958 int skb_mpls_dec_ttl(struct sk_buff *skb);
3959 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3960 			     gfp_t gfp);
3961 
memcpy_from_msg(void * data,struct msghdr * msg,int len)3962 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3963 {
3964 	return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3965 }
3966 
memcpy_to_msg(struct msghdr * msg,void * data,int len)3967 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3968 {
3969 	return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3970 }
3971 
3972 struct skb_checksum_ops {
3973 	__wsum (*update)(const void *mem, int len, __wsum wsum);
3974 	__wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3975 };
3976 
3977 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3978 
3979 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3980 		      __wsum csum, const struct skb_checksum_ops *ops);
3981 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3982 		    __wsum csum);
3983 
3984 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,const void * data,int hlen,void * buffer)3985 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
3986 		     const void *data, int hlen, void *buffer)
3987 {
3988 	if (likely(hlen - offset >= len))
3989 		return (void *)data + offset;
3990 
3991 	if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
3992 		return NULL;
3993 
3994 	return buffer;
3995 }
3996 
3997 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)3998 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3999 {
4000 	return __skb_header_pointer(skb, offset, len, skb->data,
4001 				    skb_headlen(skb), buffer);
4002 }
4003 
4004 /**
4005  *	skb_needs_linearize - check if we need to linearize a given skb
4006  *			      depending on the given device features.
4007  *	@skb: socket buffer to check
4008  *	@features: net device features
4009  *
4010  *	Returns true if either:
4011  *	1. skb has frag_list and the device doesn't support FRAGLIST, or
4012  *	2. skb is fragmented and the device does not support SG.
4013  */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)4014 static inline bool skb_needs_linearize(struct sk_buff *skb,
4015 				       netdev_features_t features)
4016 {
4017 	return skb_is_nonlinear(skb) &&
4018 	       ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
4019 		(skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
4020 }
4021 
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)4022 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
4023 					     void *to,
4024 					     const unsigned int len)
4025 {
4026 	memcpy(to, skb->data, len);
4027 }
4028 
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)4029 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
4030 						    const int offset, void *to,
4031 						    const unsigned int len)
4032 {
4033 	memcpy(to, skb->data + offset, len);
4034 }
4035 
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)4036 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
4037 					   const void *from,
4038 					   const unsigned int len)
4039 {
4040 	memcpy(skb->data, from, len);
4041 }
4042 
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)4043 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
4044 						  const int offset,
4045 						  const void *from,
4046 						  const unsigned int len)
4047 {
4048 	memcpy(skb->data + offset, from, len);
4049 }
4050 
4051 void skb_init(void);
4052 
skb_get_ktime(const struct sk_buff * skb)4053 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
4054 {
4055 	return skb->tstamp;
4056 }
4057 
4058 /**
4059  *	skb_get_timestamp - get timestamp from a skb
4060  *	@skb: skb to get stamp from
4061  *	@stamp: pointer to struct __kernel_old_timeval to store stamp in
4062  *
4063  *	Timestamps are stored in the skb as offsets to a base timestamp.
4064  *	This function converts the offset back to a struct timeval and stores
4065  *	it in stamp.
4066  */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)4067 static inline void skb_get_timestamp(const struct sk_buff *skb,
4068 				     struct __kernel_old_timeval *stamp)
4069 {
4070 	*stamp = ns_to_kernel_old_timeval(skb->tstamp);
4071 }
4072 
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)4073 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
4074 					 struct __kernel_sock_timeval *stamp)
4075 {
4076 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4077 
4078 	stamp->tv_sec = ts.tv_sec;
4079 	stamp->tv_usec = ts.tv_nsec / 1000;
4080 }
4081 
skb_get_timestampns(const struct sk_buff * skb,struct __kernel_old_timespec * stamp)4082 static inline void skb_get_timestampns(const struct sk_buff *skb,
4083 				       struct __kernel_old_timespec *stamp)
4084 {
4085 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4086 
4087 	stamp->tv_sec = ts.tv_sec;
4088 	stamp->tv_nsec = ts.tv_nsec;
4089 }
4090 
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)4091 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
4092 					   struct __kernel_timespec *stamp)
4093 {
4094 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4095 
4096 	stamp->tv_sec = ts.tv_sec;
4097 	stamp->tv_nsec = ts.tv_nsec;
4098 }
4099 
__net_timestamp(struct sk_buff * skb)4100 static inline void __net_timestamp(struct sk_buff *skb)
4101 {
4102 	skb->tstamp = ktime_get_real();
4103 	skb->mono_delivery_time = 0;
4104 }
4105 
net_timedelta(ktime_t t)4106 static inline ktime_t net_timedelta(ktime_t t)
4107 {
4108 	return ktime_sub(ktime_get_real(), t);
4109 }
4110 
skb_set_delivery_time(struct sk_buff * skb,ktime_t kt,bool mono)4111 static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt,
4112 					 bool mono)
4113 {
4114 	skb->tstamp = kt;
4115 	skb->mono_delivery_time = kt && mono;
4116 }
4117 
4118 DECLARE_STATIC_KEY_FALSE(netstamp_needed_key);
4119 
4120 /* It is used in the ingress path to clear the delivery_time.
4121  * If needed, set the skb->tstamp to the (rcv) timestamp.
4122  */
skb_clear_delivery_time(struct sk_buff * skb)4123 static inline void skb_clear_delivery_time(struct sk_buff *skb)
4124 {
4125 	if (skb->mono_delivery_time) {
4126 		skb->mono_delivery_time = 0;
4127 		if (static_branch_unlikely(&netstamp_needed_key))
4128 			skb->tstamp = ktime_get_real();
4129 		else
4130 			skb->tstamp = 0;
4131 	}
4132 }
4133 
skb_clear_tstamp(struct sk_buff * skb)4134 static inline void skb_clear_tstamp(struct sk_buff *skb)
4135 {
4136 	if (skb->mono_delivery_time)
4137 		return;
4138 
4139 	skb->tstamp = 0;
4140 }
4141 
skb_tstamp(const struct sk_buff * skb)4142 static inline ktime_t skb_tstamp(const struct sk_buff *skb)
4143 {
4144 	if (skb->mono_delivery_time)
4145 		return 0;
4146 
4147 	return skb->tstamp;
4148 }
4149 
skb_tstamp_cond(const struct sk_buff * skb,bool cond)4150 static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond)
4151 {
4152 	if (!skb->mono_delivery_time && skb->tstamp)
4153 		return skb->tstamp;
4154 
4155 	if (static_branch_unlikely(&netstamp_needed_key) || cond)
4156 		return ktime_get_real();
4157 
4158 	return 0;
4159 }
4160 
skb_metadata_len(const struct sk_buff * skb)4161 static inline u8 skb_metadata_len(const struct sk_buff *skb)
4162 {
4163 	return skb_shinfo(skb)->meta_len;
4164 }
4165 
skb_metadata_end(const struct sk_buff * skb)4166 static inline void *skb_metadata_end(const struct sk_buff *skb)
4167 {
4168 	return skb_mac_header(skb);
4169 }
4170 
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)4171 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
4172 					  const struct sk_buff *skb_b,
4173 					  u8 meta_len)
4174 {
4175 	const void *a = skb_metadata_end(skb_a);
4176 	const void *b = skb_metadata_end(skb_b);
4177 	/* Using more efficient varaiant than plain call to memcmp(). */
4178 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
4179 	u64 diffs = 0;
4180 
4181 	switch (meta_len) {
4182 #define __it(x, op) (x -= sizeof(u##op))
4183 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
4184 	case 32: diffs |= __it_diff(a, b, 64);
4185 		fallthrough;
4186 	case 24: diffs |= __it_diff(a, b, 64);
4187 		fallthrough;
4188 	case 16: diffs |= __it_diff(a, b, 64);
4189 		fallthrough;
4190 	case  8: diffs |= __it_diff(a, b, 64);
4191 		break;
4192 	case 28: diffs |= __it_diff(a, b, 64);
4193 		fallthrough;
4194 	case 20: diffs |= __it_diff(a, b, 64);
4195 		fallthrough;
4196 	case 12: diffs |= __it_diff(a, b, 64);
4197 		fallthrough;
4198 	case  4: diffs |= __it_diff(a, b, 32);
4199 		break;
4200 	}
4201 	return diffs;
4202 #else
4203 	return memcmp(a - meta_len, b - meta_len, meta_len);
4204 #endif
4205 }
4206 
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)4207 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
4208 					const struct sk_buff *skb_b)
4209 {
4210 	u8 len_a = skb_metadata_len(skb_a);
4211 	u8 len_b = skb_metadata_len(skb_b);
4212 
4213 	if (!(len_a | len_b))
4214 		return false;
4215 
4216 	return len_a != len_b ?
4217 	       true : __skb_metadata_differs(skb_a, skb_b, len_a);
4218 }
4219 
skb_metadata_set(struct sk_buff * skb,u8 meta_len)4220 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
4221 {
4222 	skb_shinfo(skb)->meta_len = meta_len;
4223 }
4224 
skb_metadata_clear(struct sk_buff * skb)4225 static inline void skb_metadata_clear(struct sk_buff *skb)
4226 {
4227 	skb_metadata_set(skb, 0);
4228 }
4229 
4230 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
4231 
4232 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
4233 
4234 void skb_clone_tx_timestamp(struct sk_buff *skb);
4235 bool skb_defer_rx_timestamp(struct sk_buff *skb);
4236 
4237 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
4238 
skb_clone_tx_timestamp(struct sk_buff * skb)4239 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
4240 {
4241 }
4242 
skb_defer_rx_timestamp(struct sk_buff * skb)4243 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
4244 {
4245 	return false;
4246 }
4247 
4248 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
4249 
4250 /**
4251  * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
4252  *
4253  * PHY drivers may accept clones of transmitted packets for
4254  * timestamping via their phy_driver.txtstamp method. These drivers
4255  * must call this function to return the skb back to the stack with a
4256  * timestamp.
4257  *
4258  * @skb: clone of the original outgoing packet
4259  * @hwtstamps: hardware time stamps
4260  *
4261  */
4262 void skb_complete_tx_timestamp(struct sk_buff *skb,
4263 			       struct skb_shared_hwtstamps *hwtstamps);
4264 
4265 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4266 		     struct skb_shared_hwtstamps *hwtstamps,
4267 		     struct sock *sk, int tstype);
4268 
4269 /**
4270  * skb_tstamp_tx - queue clone of skb with send time stamps
4271  * @orig_skb:	the original outgoing packet
4272  * @hwtstamps:	hardware time stamps, may be NULL if not available
4273  *
4274  * If the skb has a socket associated, then this function clones the
4275  * skb (thus sharing the actual data and optional structures), stores
4276  * the optional hardware time stamping information (if non NULL) or
4277  * generates a software time stamp (otherwise), then queues the clone
4278  * to the error queue of the socket.  Errors are silently ignored.
4279  */
4280 void skb_tstamp_tx(struct sk_buff *orig_skb,
4281 		   struct skb_shared_hwtstamps *hwtstamps);
4282 
4283 /**
4284  * skb_tx_timestamp() - Driver hook for transmit timestamping
4285  *
4286  * Ethernet MAC Drivers should call this function in their hard_xmit()
4287  * function immediately before giving the sk_buff to the MAC hardware.
4288  *
4289  * Specifically, one should make absolutely sure that this function is
4290  * called before TX completion of this packet can trigger.  Otherwise
4291  * the packet could potentially already be freed.
4292  *
4293  * @skb: A socket buffer.
4294  */
skb_tx_timestamp(struct sk_buff * skb)4295 static inline void skb_tx_timestamp(struct sk_buff *skb)
4296 {
4297 	skb_clone_tx_timestamp(skb);
4298 	if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
4299 		skb_tstamp_tx(skb, NULL);
4300 }
4301 
4302 /**
4303  * skb_complete_wifi_ack - deliver skb with wifi status
4304  *
4305  * @skb: the original outgoing packet
4306  * @acked: ack status
4307  *
4308  */
4309 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4310 
4311 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4312 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4313 
skb_csum_unnecessary(const struct sk_buff * skb)4314 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4315 {
4316 	return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4317 		skb->csum_valid ||
4318 		(skb->ip_summed == CHECKSUM_PARTIAL &&
4319 		 skb_checksum_start_offset(skb) >= 0));
4320 }
4321 
4322 /**
4323  *	skb_checksum_complete - Calculate checksum of an entire packet
4324  *	@skb: packet to process
4325  *
4326  *	This function calculates the checksum over the entire packet plus
4327  *	the value of skb->csum.  The latter can be used to supply the
4328  *	checksum of a pseudo header as used by TCP/UDP.  It returns the
4329  *	checksum.
4330  *
4331  *	For protocols that contain complete checksums such as ICMP/TCP/UDP,
4332  *	this function can be used to verify that checksum on received
4333  *	packets.  In that case the function should return zero if the
4334  *	checksum is correct.  In particular, this function will return zero
4335  *	if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4336  *	hardware has already verified the correctness of the checksum.
4337  */
skb_checksum_complete(struct sk_buff * skb)4338 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4339 {
4340 	return skb_csum_unnecessary(skb) ?
4341 	       0 : __skb_checksum_complete(skb);
4342 }
4343 
__skb_decr_checksum_unnecessary(struct sk_buff * skb)4344 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4345 {
4346 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4347 		if (skb->csum_level == 0)
4348 			skb->ip_summed = CHECKSUM_NONE;
4349 		else
4350 			skb->csum_level--;
4351 	}
4352 }
4353 
__skb_incr_checksum_unnecessary(struct sk_buff * skb)4354 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4355 {
4356 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4357 		if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4358 			skb->csum_level++;
4359 	} else if (skb->ip_summed == CHECKSUM_NONE) {
4360 		skb->ip_summed = CHECKSUM_UNNECESSARY;
4361 		skb->csum_level = 0;
4362 	}
4363 }
4364 
__skb_reset_checksum_unnecessary(struct sk_buff * skb)4365 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4366 {
4367 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4368 		skb->ip_summed = CHECKSUM_NONE;
4369 		skb->csum_level = 0;
4370 	}
4371 }
4372 
4373 /* Check if we need to perform checksum complete validation.
4374  *
4375  * Returns true if checksum complete is needed, false otherwise
4376  * (either checksum is unnecessary or zero checksum is allowed).
4377  */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)4378 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4379 						  bool zero_okay,
4380 						  __sum16 check)
4381 {
4382 	if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4383 		skb->csum_valid = 1;
4384 		__skb_decr_checksum_unnecessary(skb);
4385 		return false;
4386 	}
4387 
4388 	return true;
4389 }
4390 
4391 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4392  * in checksum_init.
4393  */
4394 #define CHECKSUM_BREAK 76
4395 
4396 /* Unset checksum-complete
4397  *
4398  * Unset checksum complete can be done when packet is being modified
4399  * (uncompressed for instance) and checksum-complete value is
4400  * invalidated.
4401  */
skb_checksum_complete_unset(struct sk_buff * skb)4402 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4403 {
4404 	if (skb->ip_summed == CHECKSUM_COMPLETE)
4405 		skb->ip_summed = CHECKSUM_NONE;
4406 }
4407 
4408 /* Validate (init) checksum based on checksum complete.
4409  *
4410  * Return values:
4411  *   0: checksum is validated or try to in skb_checksum_complete. In the latter
4412  *	case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4413  *	checksum is stored in skb->csum for use in __skb_checksum_complete
4414  *   non-zero: value of invalid checksum
4415  *
4416  */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)4417 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4418 						       bool complete,
4419 						       __wsum psum)
4420 {
4421 	if (skb->ip_summed == CHECKSUM_COMPLETE) {
4422 		if (!csum_fold(csum_add(psum, skb->csum))) {
4423 			skb->csum_valid = 1;
4424 			return 0;
4425 		}
4426 	}
4427 
4428 	skb->csum = psum;
4429 
4430 	if (complete || skb->len <= CHECKSUM_BREAK) {
4431 		__sum16 csum;
4432 
4433 		csum = __skb_checksum_complete(skb);
4434 		skb->csum_valid = !csum;
4435 		return csum;
4436 	}
4437 
4438 	return 0;
4439 }
4440 
null_compute_pseudo(struct sk_buff * skb,int proto)4441 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4442 {
4443 	return 0;
4444 }
4445 
4446 /* Perform checksum validate (init). Note that this is a macro since we only
4447  * want to calculate the pseudo header which is an input function if necessary.
4448  * First we try to validate without any computation (checksum unnecessary) and
4449  * then calculate based on checksum complete calling the function to compute
4450  * pseudo header.
4451  *
4452  * Return values:
4453  *   0: checksum is validated or try to in skb_checksum_complete
4454  *   non-zero: value of invalid checksum
4455  */
4456 #define __skb_checksum_validate(skb, proto, complete,			\
4457 				zero_okay, check, compute_pseudo)	\
4458 ({									\
4459 	__sum16 __ret = 0;						\
4460 	skb->csum_valid = 0;						\
4461 	if (__skb_checksum_validate_needed(skb, zero_okay, check))	\
4462 		__ret = __skb_checksum_validate_complete(skb,		\
4463 				complete, compute_pseudo(skb, proto));	\
4464 	__ret;								\
4465 })
4466 
4467 #define skb_checksum_init(skb, proto, compute_pseudo)			\
4468 	__skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4469 
4470 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo)	\
4471 	__skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4472 
4473 #define skb_checksum_validate(skb, proto, compute_pseudo)		\
4474 	__skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4475 
4476 #define skb_checksum_validate_zero_check(skb, proto, check,		\
4477 					 compute_pseudo)		\
4478 	__skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4479 
4480 #define skb_checksum_simple_validate(skb)				\
4481 	__skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4482 
__skb_checksum_convert_check(struct sk_buff * skb)4483 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4484 {
4485 	return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4486 }
4487 
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4488 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4489 {
4490 	skb->csum = ~pseudo;
4491 	skb->ip_summed = CHECKSUM_COMPLETE;
4492 }
4493 
4494 #define skb_checksum_try_convert(skb, proto, compute_pseudo)	\
4495 do {									\
4496 	if (__skb_checksum_convert_check(skb))				\
4497 		__skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4498 } while (0)
4499 
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4500 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4501 					      u16 start, u16 offset)
4502 {
4503 	skb->ip_summed = CHECKSUM_PARTIAL;
4504 	skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4505 	skb->csum_offset = offset - start;
4506 }
4507 
4508 /* Update skbuf and packet to reflect the remote checksum offload operation.
4509  * When called, ptr indicates the starting point for skb->csum when
4510  * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4511  * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4512  */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4513 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4514 				       int start, int offset, bool nopartial)
4515 {
4516 	__wsum delta;
4517 
4518 	if (!nopartial) {
4519 		skb_remcsum_adjust_partial(skb, ptr, start, offset);
4520 		return;
4521 	}
4522 
4523 	if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4524 		__skb_checksum_complete(skb);
4525 		skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4526 	}
4527 
4528 	delta = remcsum_adjust(ptr, skb->csum, start, offset);
4529 
4530 	/* Adjust skb->csum since we changed the packet */
4531 	skb->csum = csum_add(skb->csum, delta);
4532 }
4533 
skb_nfct(const struct sk_buff * skb)4534 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4535 {
4536 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4537 	return (void *)(skb->_nfct & NFCT_PTRMASK);
4538 #else
4539 	return NULL;
4540 #endif
4541 }
4542 
skb_get_nfct(const struct sk_buff * skb)4543 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4544 {
4545 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4546 	return skb->_nfct;
4547 #else
4548 	return 0UL;
4549 #endif
4550 }
4551 
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4552 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4553 {
4554 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4555 	skb->slow_gro |= !!nfct;
4556 	skb->_nfct = nfct;
4557 #endif
4558 }
4559 
4560 #ifdef CONFIG_SKB_EXTENSIONS
4561 enum skb_ext_id {
4562 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4563 	SKB_EXT_BRIDGE_NF,
4564 #endif
4565 #ifdef CONFIG_XFRM
4566 	SKB_EXT_SEC_PATH,
4567 #endif
4568 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4569 	TC_SKB_EXT,
4570 #endif
4571 #if IS_ENABLED(CONFIG_MPTCP)
4572 	SKB_EXT_MPTCP,
4573 #endif
4574 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
4575 	SKB_EXT_MCTP,
4576 #endif
4577 	SKB_EXT_NUM, /* must be last */
4578 };
4579 
4580 /**
4581  *	struct skb_ext - sk_buff extensions
4582  *	@refcnt: 1 on allocation, deallocated on 0
4583  *	@offset: offset to add to @data to obtain extension address
4584  *	@chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4585  *	@data: start of extension data, variable sized
4586  *
4587  *	Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4588  *	to use 'u8' types while allowing up to 2kb worth of extension data.
4589  */
4590 struct skb_ext {
4591 	refcount_t refcnt;
4592 	u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4593 	u8 chunks;		/* same */
4594 	char data[] __aligned(8);
4595 };
4596 
4597 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4598 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4599 		    struct skb_ext *ext);
4600 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4601 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4602 void __skb_ext_put(struct skb_ext *ext);
4603 
skb_ext_put(struct sk_buff * skb)4604 static inline void skb_ext_put(struct sk_buff *skb)
4605 {
4606 	if (skb->active_extensions)
4607 		__skb_ext_put(skb->extensions);
4608 }
4609 
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4610 static inline void __skb_ext_copy(struct sk_buff *dst,
4611 				  const struct sk_buff *src)
4612 {
4613 	dst->active_extensions = src->active_extensions;
4614 
4615 	if (src->active_extensions) {
4616 		struct skb_ext *ext = src->extensions;
4617 
4618 		refcount_inc(&ext->refcnt);
4619 		dst->extensions = ext;
4620 	}
4621 }
4622 
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4623 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4624 {
4625 	skb_ext_put(dst);
4626 	__skb_ext_copy(dst, src);
4627 }
4628 
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4629 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4630 {
4631 	return !!ext->offset[i];
4632 }
4633 
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4634 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4635 {
4636 	return skb->active_extensions & (1 << id);
4637 }
4638 
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4639 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4640 {
4641 	if (skb_ext_exist(skb, id))
4642 		__skb_ext_del(skb, id);
4643 }
4644 
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4645 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4646 {
4647 	if (skb_ext_exist(skb, id)) {
4648 		struct skb_ext *ext = skb->extensions;
4649 
4650 		return (void *)ext + (ext->offset[id] << 3);
4651 	}
4652 
4653 	return NULL;
4654 }
4655 
skb_ext_reset(struct sk_buff * skb)4656 static inline void skb_ext_reset(struct sk_buff *skb)
4657 {
4658 	if (unlikely(skb->active_extensions)) {
4659 		__skb_ext_put(skb->extensions);
4660 		skb->active_extensions = 0;
4661 	}
4662 }
4663 
skb_has_extensions(struct sk_buff * skb)4664 static inline bool skb_has_extensions(struct sk_buff *skb)
4665 {
4666 	return unlikely(skb->active_extensions);
4667 }
4668 #else
skb_ext_put(struct sk_buff * skb)4669 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4670 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4671 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4672 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)4673 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4674 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4675 #endif /* CONFIG_SKB_EXTENSIONS */
4676 
nf_reset_ct(struct sk_buff * skb)4677 static inline void nf_reset_ct(struct sk_buff *skb)
4678 {
4679 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4680 	nf_conntrack_put(skb_nfct(skb));
4681 	skb->_nfct = 0;
4682 #endif
4683 }
4684 
nf_reset_trace(struct sk_buff * skb)4685 static inline void nf_reset_trace(struct sk_buff *skb)
4686 {
4687 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4688 	skb->nf_trace = 0;
4689 #endif
4690 }
4691 
ipvs_reset(struct sk_buff * skb)4692 static inline void ipvs_reset(struct sk_buff *skb)
4693 {
4694 #if IS_ENABLED(CONFIG_IP_VS)
4695 	skb->ipvs_property = 0;
4696 #endif
4697 }
4698 
4699 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4700 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4701 			     bool copy)
4702 {
4703 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4704 	dst->_nfct = src->_nfct;
4705 	nf_conntrack_get(skb_nfct(src));
4706 #endif
4707 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4708 	if (copy)
4709 		dst->nf_trace = src->nf_trace;
4710 #endif
4711 }
4712 
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4713 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4714 {
4715 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4716 	nf_conntrack_put(skb_nfct(dst));
4717 #endif
4718 	dst->slow_gro = src->slow_gro;
4719 	__nf_copy(dst, src, true);
4720 }
4721 
4722 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4723 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4724 {
4725 	to->secmark = from->secmark;
4726 }
4727 
skb_init_secmark(struct sk_buff * skb)4728 static inline void skb_init_secmark(struct sk_buff *skb)
4729 {
4730 	skb->secmark = 0;
4731 }
4732 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4733 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4734 { }
4735 
skb_init_secmark(struct sk_buff * skb)4736 static inline void skb_init_secmark(struct sk_buff *skb)
4737 { }
4738 #endif
4739 
secpath_exists(const struct sk_buff * skb)4740 static inline int secpath_exists(const struct sk_buff *skb)
4741 {
4742 #ifdef CONFIG_XFRM
4743 	return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4744 #else
4745 	return 0;
4746 #endif
4747 }
4748 
skb_irq_freeable(const struct sk_buff * skb)4749 static inline bool skb_irq_freeable(const struct sk_buff *skb)
4750 {
4751 	return !skb->destructor &&
4752 		!secpath_exists(skb) &&
4753 		!skb_nfct(skb) &&
4754 		!skb->_skb_refdst &&
4755 		!skb_has_frag_list(skb);
4756 }
4757 
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)4758 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4759 {
4760 	skb->queue_mapping = queue_mapping;
4761 }
4762 
skb_get_queue_mapping(const struct sk_buff * skb)4763 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4764 {
4765 	return skb->queue_mapping;
4766 }
4767 
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)4768 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4769 {
4770 	to->queue_mapping = from->queue_mapping;
4771 }
4772 
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)4773 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4774 {
4775 	skb->queue_mapping = rx_queue + 1;
4776 }
4777 
skb_get_rx_queue(const struct sk_buff * skb)4778 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4779 {
4780 	return skb->queue_mapping - 1;
4781 }
4782 
skb_rx_queue_recorded(const struct sk_buff * skb)4783 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4784 {
4785 	return skb->queue_mapping != 0;
4786 }
4787 
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)4788 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4789 {
4790 	skb->dst_pending_confirm = val;
4791 }
4792 
skb_get_dst_pending_confirm(const struct sk_buff * skb)4793 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4794 {
4795 	return skb->dst_pending_confirm != 0;
4796 }
4797 
skb_sec_path(const struct sk_buff * skb)4798 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
4799 {
4800 #ifdef CONFIG_XFRM
4801 	return skb_ext_find(skb, SKB_EXT_SEC_PATH);
4802 #else
4803 	return NULL;
4804 #endif
4805 }
4806 
4807 /* Keeps track of mac header offset relative to skb->head.
4808  * It is useful for TSO of Tunneling protocol. e.g. GRE.
4809  * For non-tunnel skb it points to skb_mac_header() and for
4810  * tunnel skb it points to outer mac header.
4811  * Keeps track of level of encapsulation of network headers.
4812  */
4813 struct skb_gso_cb {
4814 	union {
4815 		int	mac_offset;
4816 		int	data_offset;
4817 	};
4818 	int	encap_level;
4819 	__wsum	csum;
4820 	__u16	csum_start;
4821 };
4822 #define SKB_GSO_CB_OFFSET	32
4823 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
4824 
skb_tnl_header_len(const struct sk_buff * inner_skb)4825 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4826 {
4827 	return (skb_mac_header(inner_skb) - inner_skb->head) -
4828 		SKB_GSO_CB(inner_skb)->mac_offset;
4829 }
4830 
gso_pskb_expand_head(struct sk_buff * skb,int extra)4831 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4832 {
4833 	int new_headroom, headroom;
4834 	int ret;
4835 
4836 	headroom = skb_headroom(skb);
4837 	ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4838 	if (ret)
4839 		return ret;
4840 
4841 	new_headroom = skb_headroom(skb);
4842 	SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4843 	return 0;
4844 }
4845 
gso_reset_checksum(struct sk_buff * skb,__wsum res)4846 static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4847 {
4848 	/* Do not update partial checksums if remote checksum is enabled. */
4849 	if (skb->remcsum_offload)
4850 		return;
4851 
4852 	SKB_GSO_CB(skb)->csum = res;
4853 	SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4854 }
4855 
4856 /* Compute the checksum for a gso segment. First compute the checksum value
4857  * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4858  * then add in skb->csum (checksum from csum_start to end of packet).
4859  * skb->csum and csum_start are then updated to reflect the checksum of the
4860  * resultant packet starting from the transport header-- the resultant checksum
4861  * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4862  * header.
4863  */
gso_make_checksum(struct sk_buff * skb,__wsum res)4864 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4865 {
4866 	unsigned char *csum_start = skb_transport_header(skb);
4867 	int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4868 	__wsum partial = SKB_GSO_CB(skb)->csum;
4869 
4870 	SKB_GSO_CB(skb)->csum = res;
4871 	SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4872 
4873 	return csum_fold(csum_partial(csum_start, plen, partial));
4874 }
4875 
skb_is_gso(const struct sk_buff * skb)4876 static inline bool skb_is_gso(const struct sk_buff *skb)
4877 {
4878 	return skb_shinfo(skb)->gso_size;
4879 }
4880 
4881 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)4882 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4883 {
4884 	return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4885 }
4886 
4887 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)4888 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4889 {
4890 	return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4891 }
4892 
4893 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)4894 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4895 {
4896 	return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4897 }
4898 
skb_gso_reset(struct sk_buff * skb)4899 static inline void skb_gso_reset(struct sk_buff *skb)
4900 {
4901 	skb_shinfo(skb)->gso_size = 0;
4902 	skb_shinfo(skb)->gso_segs = 0;
4903 	skb_shinfo(skb)->gso_type = 0;
4904 }
4905 
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)4906 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4907 					 u16 increment)
4908 {
4909 	if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4910 		return;
4911 	shinfo->gso_size += increment;
4912 }
4913 
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)4914 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4915 					 u16 decrement)
4916 {
4917 	if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4918 		return;
4919 	shinfo->gso_size -= decrement;
4920 }
4921 
4922 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4923 
skb_warn_if_lro(const struct sk_buff * skb)4924 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4925 {
4926 	/* LRO sets gso_size but not gso_type, whereas if GSO is really
4927 	 * wanted then gso_type will be set. */
4928 	const struct skb_shared_info *shinfo = skb_shinfo(skb);
4929 
4930 	if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4931 	    unlikely(shinfo->gso_type == 0)) {
4932 		__skb_warn_lro_forwarding(skb);
4933 		return true;
4934 	}
4935 	return false;
4936 }
4937 
skb_forward_csum(struct sk_buff * skb)4938 static inline void skb_forward_csum(struct sk_buff *skb)
4939 {
4940 	/* Unfortunately we don't support this one.  Any brave souls? */
4941 	if (skb->ip_summed == CHECKSUM_COMPLETE)
4942 		skb->ip_summed = CHECKSUM_NONE;
4943 }
4944 
4945 /**
4946  * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4947  * @skb: skb to check
4948  *
4949  * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4950  * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4951  * use this helper, to document places where we make this assertion.
4952  */
skb_checksum_none_assert(const struct sk_buff * skb)4953 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4954 {
4955 	DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE);
4956 }
4957 
4958 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4959 
4960 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4961 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4962 				     unsigned int transport_len,
4963 				     __sum16(*skb_chkf)(struct sk_buff *skb));
4964 
4965 /**
4966  * skb_head_is_locked - Determine if the skb->head is locked down
4967  * @skb: skb to check
4968  *
4969  * The head on skbs build around a head frag can be removed if they are
4970  * not cloned.  This function returns true if the skb head is locked down
4971  * due to either being allocated via kmalloc, or by being a clone with
4972  * multiple references to the head.
4973  */
skb_head_is_locked(const struct sk_buff * skb)4974 static inline bool skb_head_is_locked(const struct sk_buff *skb)
4975 {
4976 	return !skb->head_frag || skb_cloned(skb);
4977 }
4978 
4979 /* Local Checksum Offload.
4980  * Compute outer checksum based on the assumption that the
4981  * inner checksum will be offloaded later.
4982  * See Documentation/networking/checksum-offloads.rst for
4983  * explanation of how this works.
4984  * Fill in outer checksum adjustment (e.g. with sum of outer
4985  * pseudo-header) before calling.
4986  * Also ensure that inner checksum is in linear data area.
4987  */
lco_csum(struct sk_buff * skb)4988 static inline __wsum lco_csum(struct sk_buff *skb)
4989 {
4990 	unsigned char *csum_start = skb_checksum_start(skb);
4991 	unsigned char *l4_hdr = skb_transport_header(skb);
4992 	__wsum partial;
4993 
4994 	/* Start with complement of inner checksum adjustment */
4995 	partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4996 						    skb->csum_offset));
4997 
4998 	/* Add in checksum of our headers (incl. outer checksum
4999 	 * adjustment filled in by caller) and return result.
5000 	 */
5001 	return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
5002 }
5003 
skb_is_redirected(const struct sk_buff * skb)5004 static inline bool skb_is_redirected(const struct sk_buff *skb)
5005 {
5006 	return skb->redirected;
5007 }
5008 
skb_set_redirected(struct sk_buff * skb,bool from_ingress)5009 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
5010 {
5011 	skb->redirected = 1;
5012 #ifdef CONFIG_NET_REDIRECT
5013 	skb->from_ingress = from_ingress;
5014 	if (skb->from_ingress)
5015 		skb_clear_tstamp(skb);
5016 #endif
5017 }
5018 
skb_reset_redirect(struct sk_buff * skb)5019 static inline void skb_reset_redirect(struct sk_buff *skb)
5020 {
5021 	skb->redirected = 0;
5022 }
5023 
skb_csum_is_sctp(struct sk_buff * skb)5024 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
5025 {
5026 	return skb->csum_not_inet;
5027 }
5028 
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)5029 static inline void skb_set_kcov_handle(struct sk_buff *skb,
5030 				       const u64 kcov_handle)
5031 {
5032 #ifdef CONFIG_KCOV
5033 	skb->kcov_handle = kcov_handle;
5034 #endif
5035 }
5036 
skb_get_kcov_handle(struct sk_buff * skb)5037 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
5038 {
5039 #ifdef CONFIG_KCOV
5040 	return skb->kcov_handle;
5041 #else
5042 	return 0;
5043 #endif
5044 }
5045 
5046 #ifdef CONFIG_PAGE_POOL
skb_mark_for_recycle(struct sk_buff * skb)5047 static inline void skb_mark_for_recycle(struct sk_buff *skb)
5048 {
5049 	skb->pp_recycle = 1;
5050 }
5051 #endif
5052 
skb_pp_recycle(struct sk_buff * skb,void * data)5053 static inline bool skb_pp_recycle(struct sk_buff *skb, void *data)
5054 {
5055 	if (!IS_ENABLED(CONFIG_PAGE_POOL) || !skb->pp_recycle)
5056 		return false;
5057 	return page_pool_return_skb_page(virt_to_page(data));
5058 }
5059 
5060 #endif	/* __KERNEL__ */
5061 #endif	/* _LINUX_SKBUFF_H */
5062