xref: /linux/include/linux/skbuff.h (revision 1a9239bb4253f9076b5b4b2a1a4e8d7defd77a95)
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 <net/checksum.h>
27 #include <linux/rcupdate.h>
28 #include <linux/dma-mapping.h>
29 #include <linux/netdev_features.h>
30 #include <net/flow_dissector.h>
31 #include <linux/in6.h>
32 #include <linux/if_packet.h>
33 #include <linux/llist.h>
34 #include <linux/page_frag_cache.h>
35 #include <net/flow.h>
36 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
37 #include <linux/netfilter/nf_conntrack_common.h>
38 #endif
39 #include <net/net_debug.h>
40 #include <net/dropreason-core.h>
41 #include <net/netmem.h>
42 
43 /**
44  * DOC: skb checksums
45  *
46  * The interface for checksum offload between the stack and networking drivers
47  * is as follows...
48  *
49  * IP checksum related features
50  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
51  *
52  * Drivers advertise checksum offload capabilities in the features of a device.
53  * From the stack's point of view these are capabilities offered by the driver.
54  * A driver typically only advertises features that it is capable of offloading
55  * to its device.
56  *
57  * .. flat-table:: Checksum related device features
58  *   :widths: 1 10
59  *
60  *   * - %NETIF_F_HW_CSUM
61  *     - The driver (or its device) is able to compute one
62  *	 IP (one's complement) checksum for any combination
63  *	 of protocols or protocol layering. The checksum is
64  *	 computed and set in a packet per the CHECKSUM_PARTIAL
65  *	 interface (see below).
66  *
67  *   * - %NETIF_F_IP_CSUM
68  *     - Driver (device) is only able to checksum plain
69  *	 TCP or UDP packets over IPv4. These are specifically
70  *	 unencapsulated packets of the form IPv4|TCP or
71  *	 IPv4|UDP where the Protocol field in the IPv4 header
72  *	 is TCP or UDP. The IPv4 header may contain IP options.
73  *	 This feature cannot be set in features for a device
74  *	 with NETIF_F_HW_CSUM also set. This feature is being
75  *	 DEPRECATED (see below).
76  *
77  *   * - %NETIF_F_IPV6_CSUM
78  *     - Driver (device) is only able to checksum plain
79  *	 TCP or UDP packets over IPv6. These are specifically
80  *	 unencapsulated packets of the form IPv6|TCP or
81  *	 IPv6|UDP where the Next Header field in the IPv6
82  *	 header is either TCP or UDP. IPv6 extension headers
83  *	 are not supported with this feature. This feature
84  *	 cannot be set in features for a device with
85  *	 NETIF_F_HW_CSUM also set. This feature is being
86  *	 DEPRECATED (see below).
87  *
88  *   * - %NETIF_F_RXCSUM
89  *     - Driver (device) performs receive checksum offload.
90  *	 This flag is only used to disable the RX checksum
91  *	 feature for a device. The stack will accept receive
92  *	 checksum indication in packets received on a device
93  *	 regardless of whether NETIF_F_RXCSUM is set.
94  *
95  * Checksumming of received packets by device
96  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
97  *
98  * Indication of checksum verification is set in &sk_buff.ip_summed.
99  * Possible values are:
100  *
101  * - %CHECKSUM_NONE
102  *
103  *   Device did not checksum this packet e.g. due to lack of capabilities.
104  *   The packet contains full (though not verified) checksum in packet but
105  *   not in skb->csum. Thus, skb->csum is undefined in this case.
106  *
107  * - %CHECKSUM_UNNECESSARY
108  *
109  *   The hardware you're dealing with doesn't calculate the full checksum
110  *   (as in %CHECKSUM_COMPLETE), but it does parse headers and verify checksums
111  *   for specific protocols. For such packets it will set %CHECKSUM_UNNECESSARY
112  *   if their checksums are okay. &sk_buff.csum is still undefined in this case
113  *   though. A driver or device must never modify the checksum field in the
114  *   packet even if checksum is verified.
115  *
116  *   %CHECKSUM_UNNECESSARY is applicable to following protocols:
117  *
118  *     - TCP: IPv6 and IPv4.
119  *     - UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
120  *       zero UDP checksum for either IPv4 or IPv6, the networking stack
121  *       may perform further validation in this case.
122  *     - GRE: only if the checksum is present in the header.
123  *     - SCTP: indicates the CRC in SCTP header has been validated.
124  *     - FCOE: indicates the CRC in FC frame has been validated.
125  *
126  *   &sk_buff.csum_level indicates the number of consecutive checksums found in
127  *   the packet minus one that have been verified as %CHECKSUM_UNNECESSARY.
128  *   For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
129  *   and a device is able to verify the checksums for UDP (possibly zero),
130  *   GRE (checksum flag is set) and TCP, &sk_buff.csum_level would be set to
131  *   two. If the device were only able to verify the UDP checksum and not
132  *   GRE, either because it doesn't support GRE checksum or because GRE
133  *   checksum is bad, skb->csum_level would be set to zero (TCP checksum is
134  *   not considered in this case).
135  *
136  * - %CHECKSUM_COMPLETE
137  *
138  *   This is the most generic way. The device supplied checksum of the _whole_
139  *   packet as seen by netif_rx() and fills in &sk_buff.csum. This means the
140  *   hardware doesn't need to parse L3/L4 headers to implement this.
141  *
142  *   Notes:
143  *
144  *   - Even if device supports only some protocols, but is able to produce
145  *     skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
146  *   - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
147  *
148  * - %CHECKSUM_PARTIAL
149  *
150  *   A checksum is set up to be offloaded to a device as described in the
151  *   output description for CHECKSUM_PARTIAL. This may occur on a packet
152  *   received directly from another Linux OS, e.g., a virtualized Linux kernel
153  *   on the same host, or it may be set in the input path in GRO or remote
154  *   checksum offload. For the purposes of checksum verification, the checksum
155  *   referred to by skb->csum_start + skb->csum_offset and any preceding
156  *   checksums in the packet are considered verified. Any checksums in the
157  *   packet that are after the checksum being offloaded are not considered to
158  *   be verified.
159  *
160  * Checksumming on transmit for non-GSO
161  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
162  *
163  * The stack requests checksum offload in the &sk_buff.ip_summed for a packet.
164  * Values are:
165  *
166  * - %CHECKSUM_PARTIAL
167  *
168  *   The driver is required to checksum the packet as seen by hard_start_xmit()
169  *   from &sk_buff.csum_start up to the end, and to record/write the checksum at
170  *   offset &sk_buff.csum_start + &sk_buff.csum_offset.
171  *   A driver may verify that the
172  *   csum_start and csum_offset values are valid values given the length and
173  *   offset of the packet, but it should not attempt to validate that the
174  *   checksum refers to a legitimate transport layer checksum -- it is the
175  *   purview of the stack to validate that csum_start and csum_offset are set
176  *   correctly.
177  *
178  *   When the stack requests checksum offload for a packet, the driver MUST
179  *   ensure that the checksum is set correctly. A driver can either offload the
180  *   checksum calculation to the device, or call skb_checksum_help (in the case
181  *   that the device does not support offload for a particular checksum).
182  *
183  *   %NETIF_F_IP_CSUM and %NETIF_F_IPV6_CSUM are being deprecated in favor of
184  *   %NETIF_F_HW_CSUM. New devices should use %NETIF_F_HW_CSUM to indicate
185  *   checksum offload capability.
186  *   skb_csum_hwoffload_help() can be called to resolve %CHECKSUM_PARTIAL based
187  *   on network device checksumming capabilities: if a packet does not match
188  *   them, skb_checksum_help() or skb_crc32c_help() (depending on the value of
189  *   &sk_buff.csum_not_inet, see :ref:`crc`)
190  *   is called to resolve the checksum.
191  *
192  * - %CHECKSUM_NONE
193  *
194  *   The skb was already checksummed by the protocol, or a checksum is not
195  *   required.
196  *
197  * - %CHECKSUM_UNNECESSARY
198  *
199  *   This has the same meaning as CHECKSUM_NONE for checksum offload on
200  *   output.
201  *
202  * - %CHECKSUM_COMPLETE
203  *
204  *   Not used in checksum output. If a driver observes a packet with this value
205  *   set in skbuff, it should treat the packet as if %CHECKSUM_NONE were set.
206  *
207  * .. _crc:
208  *
209  * Non-IP checksum (CRC) offloads
210  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
211  *
212  * .. flat-table::
213  *   :widths: 1 10
214  *
215  *   * - %NETIF_F_SCTP_CRC
216  *     - This feature indicates that a device is capable of
217  *	 offloading the SCTP CRC in a packet. To perform this offload the stack
218  *	 will set csum_start and csum_offset accordingly, set ip_summed to
219  *	 %CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication
220  *	 in the skbuff that the %CHECKSUM_PARTIAL refers to CRC32c.
221  *	 A driver that supports both IP checksum offload and SCTP CRC32c offload
222  *	 must verify which offload is configured for a packet by testing the
223  *	 value of &sk_buff.csum_not_inet; skb_crc32c_csum_help() is provided to
224  *	 resolve %CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
225  *
226  *   * - %NETIF_F_FCOE_CRC
227  *     - This feature indicates that a device is capable of offloading the FCOE
228  *	 CRC in a packet. To perform this offload the stack will set ip_summed
229  *	 to %CHECKSUM_PARTIAL and set csum_start and csum_offset
230  *	 accordingly. Note that there is no indication in the skbuff that the
231  *	 %CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
232  *	 both IP checksum offload and FCOE CRC offload must verify which offload
233  *	 is configured for a packet, presumably by inspecting packet headers.
234  *
235  * Checksumming on output with GSO
236  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
237  *
238  * In the case of a GSO packet (skb_is_gso() is true), checksum offload
239  * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
240  * gso_type is %SKB_GSO_TCPV4 or %SKB_GSO_TCPV6, TCP checksum offload as
241  * part of the GSO operation is implied. If a checksum is being offloaded
242  * with GSO then ip_summed is %CHECKSUM_PARTIAL, and both csum_start and
243  * csum_offset are set to refer to the outermost checksum being offloaded
244  * (two offloaded checksums are possible with UDP encapsulation).
245  */
246 
247 /* Don't change this without changing skb_csum_unnecessary! */
248 #define CHECKSUM_NONE		0
249 #define CHECKSUM_UNNECESSARY	1
250 #define CHECKSUM_COMPLETE	2
251 #define CHECKSUM_PARTIAL	3
252 
253 /* Maximum value in skb->csum_level */
254 #define SKB_MAX_CSUM_LEVEL	3
255 
256 #define SKB_DATA_ALIGN(X)	ALIGN(X, SMP_CACHE_BYTES)
257 #define SKB_WITH_OVERHEAD(X)	\
258 	((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
259 
260 /* For X bytes available in skb->head, what is the minimal
261  * allocation needed, knowing struct skb_shared_info needs
262  * to be aligned.
263  */
264 #define SKB_HEAD_ALIGN(X) (SKB_DATA_ALIGN(X) + \
265 	SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
266 
267 #define SKB_MAX_ORDER(X, ORDER) \
268 	SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
269 #define SKB_MAX_HEAD(X)		(SKB_MAX_ORDER((X), 0))
270 #define SKB_MAX_ALLOC		(SKB_MAX_ORDER(0, 2))
271 
272 /* return minimum truesize of one skb containing X bytes of data */
273 #define SKB_TRUESIZE(X) ((X) +						\
274 			 SKB_DATA_ALIGN(sizeof(struct sk_buff)) +	\
275 			 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
276 
277 struct ahash_request;
278 struct net_device;
279 struct scatterlist;
280 struct pipe_inode_info;
281 struct iov_iter;
282 struct napi_struct;
283 struct bpf_prog;
284 union bpf_attr;
285 struct skb_ext;
286 struct ts_config;
287 
288 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
289 struct nf_bridge_info {
290 	enum {
291 		BRNF_PROTO_UNCHANGED,
292 		BRNF_PROTO_8021Q,
293 		BRNF_PROTO_PPPOE
294 	} orig_proto:8;
295 	u8			pkt_otherhost:1;
296 	u8			in_prerouting:1;
297 	u8			bridged_dnat:1;
298 	u8			sabotage_in_done:1;
299 	__u16			frag_max_size;
300 	int			physinif;
301 
302 	/* always valid & non-NULL from FORWARD on, for physdev match */
303 	struct net_device	*physoutdev;
304 	union {
305 		/* prerouting: detect dnat in orig/reply direction */
306 		__be32          ipv4_daddr;
307 		struct in6_addr ipv6_daddr;
308 
309 		/* after prerouting + nat detected: store original source
310 		 * mac since neigh resolution overwrites it, only used while
311 		 * skb is out in neigh layer.
312 		 */
313 		char neigh_header[8];
314 	};
315 };
316 #endif
317 
318 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
319 /* Chain in tc_skb_ext will be used to share the tc chain with
320  * ovs recirc_id. It will be set to the current chain by tc
321  * and read by ovs to recirc_id.
322  */
323 struct tc_skb_ext {
324 	union {
325 		u64 act_miss_cookie;
326 		__u32 chain;
327 	};
328 	__u16 mru;
329 	__u16 zone;
330 	u8 post_ct:1;
331 	u8 post_ct_snat:1;
332 	u8 post_ct_dnat:1;
333 	u8 act_miss:1; /* Set if act_miss_cookie is used */
334 	u8 l2_miss:1; /* Set by bridge upon FDB or MDB miss */
335 };
336 #endif
337 
338 struct sk_buff_head {
339 	/* These two members must be first to match sk_buff. */
340 	struct_group_tagged(sk_buff_list, list,
341 		struct sk_buff	*next;
342 		struct sk_buff	*prev;
343 	);
344 
345 	__u32		qlen;
346 	spinlock_t	lock;
347 };
348 
349 struct sk_buff;
350 
351 #ifndef CONFIG_MAX_SKB_FRAGS
352 # define CONFIG_MAX_SKB_FRAGS 17
353 #endif
354 
355 #define MAX_SKB_FRAGS CONFIG_MAX_SKB_FRAGS
356 
357 /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
358  * segment using its current segmentation instead.
359  */
360 #define GSO_BY_FRAGS	0xFFFF
361 
362 typedef struct skb_frag {
363 	netmem_ref netmem;
364 	unsigned int len;
365 	unsigned int offset;
366 } skb_frag_t;
367 
368 /**
369  * skb_frag_size() - Returns the size of a skb fragment
370  * @frag: skb fragment
371  */
skb_frag_size(const skb_frag_t * frag)372 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
373 {
374 	return frag->len;
375 }
376 
377 /**
378  * skb_frag_size_set() - Sets the size of a skb fragment
379  * @frag: skb fragment
380  * @size: size of fragment
381  */
skb_frag_size_set(skb_frag_t * frag,unsigned int size)382 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
383 {
384 	frag->len = size;
385 }
386 
387 /**
388  * skb_frag_size_add() - Increments the size of a skb fragment by @delta
389  * @frag: skb fragment
390  * @delta: value to add
391  */
skb_frag_size_add(skb_frag_t * frag,int delta)392 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
393 {
394 	frag->len += delta;
395 }
396 
397 /**
398  * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
399  * @frag: skb fragment
400  * @delta: value to subtract
401  */
skb_frag_size_sub(skb_frag_t * frag,int delta)402 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
403 {
404 	frag->len -= delta;
405 }
406 
407 /**
408  * skb_frag_must_loop - Test if %p is a high memory page
409  * @p: fragment's page
410  */
skb_frag_must_loop(struct page * p)411 static inline bool skb_frag_must_loop(struct page *p)
412 {
413 #if defined(CONFIG_HIGHMEM)
414 	if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p))
415 		return true;
416 #endif
417 	return false;
418 }
419 
420 /**
421  *	skb_frag_foreach_page - loop over pages in a fragment
422  *
423  *	@f:		skb frag to operate on
424  *	@f_off:		offset from start of f->netmem
425  *	@f_len:		length from f_off to loop over
426  *	@p:		(temp var) current page
427  *	@p_off:		(temp var) offset from start of current page,
428  *	                           non-zero only on first page.
429  *	@p_len:		(temp var) length in current page,
430  *				   < PAGE_SIZE only on first and last page.
431  *	@copied:	(temp var) length so far, excluding current p_len.
432  *
433  *	A fragment can hold a compound page, in which case per-page
434  *	operations, notably kmap_atomic, must be called for each
435  *	regular page.
436  */
437 #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied)	\
438 	for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT),		\
439 	     p_off = (f_off) & (PAGE_SIZE - 1),				\
440 	     p_len = skb_frag_must_loop(p) ?				\
441 	     min_t(u32, f_len, PAGE_SIZE - p_off) : f_len,		\
442 	     copied = 0;						\
443 	     copied < f_len;						\
444 	     copied += p_len, p++, p_off = 0,				\
445 	     p_len = min_t(u32, f_len - copied, PAGE_SIZE))		\
446 
447 /**
448  * struct skb_shared_hwtstamps - hardware time stamps
449  * @hwtstamp:		hardware time stamp transformed into duration
450  *			since arbitrary point in time
451  * @netdev_data:	address/cookie of network device driver used as
452  *			reference to actual hardware time stamp
453  *
454  * Software time stamps generated by ktime_get_real() are stored in
455  * skb->tstamp.
456  *
457  * hwtstamps can only be compared against other hwtstamps from
458  * the same device.
459  *
460  * This structure is attached to packets as part of the
461  * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
462  */
463 struct skb_shared_hwtstamps {
464 	union {
465 		ktime_t	hwtstamp;
466 		void *netdev_data;
467 	};
468 };
469 
470 /* Definitions for tx_flags in struct skb_shared_info */
471 enum {
472 	/* generate hardware time stamp */
473 	SKBTX_HW_TSTAMP_NOBPF = 1 << 0,
474 
475 	/* generate software time stamp when queueing packet to NIC */
476 	SKBTX_SW_TSTAMP = 1 << 1,
477 
478 	/* device driver is going to provide hardware time stamp */
479 	SKBTX_IN_PROGRESS = 1 << 2,
480 
481 	/* generate software time stamp on packet tx completion */
482 	SKBTX_COMPLETION_TSTAMP = 1 << 3,
483 
484 	/* generate wifi status information (where possible) */
485 	SKBTX_WIFI_STATUS = 1 << 4,
486 
487 	/* determine hardware time stamp based on time or cycles */
488 	SKBTX_HW_TSTAMP_NETDEV = 1 << 5,
489 
490 	/* generate software time stamp when entering packet scheduling */
491 	SKBTX_SCHED_TSTAMP = 1 << 6,
492 
493 	/* used for bpf extension when a bpf program is loaded */
494 	SKBTX_BPF = 1 << 7,
495 };
496 
497 #define SKBTX_HW_TSTAMP		(SKBTX_HW_TSTAMP_NOBPF | SKBTX_BPF)
498 
499 #define SKBTX_ANY_SW_TSTAMP	(SKBTX_SW_TSTAMP    | \
500 				 SKBTX_SCHED_TSTAMP | \
501 				 SKBTX_BPF          | \
502 				 SKBTX_COMPLETION_TSTAMP)
503 #define SKBTX_ANY_TSTAMP	(SKBTX_HW_TSTAMP | \
504 				 SKBTX_ANY_SW_TSTAMP)
505 
506 /* Definitions for flags in struct skb_shared_info */
507 enum {
508 	/* use zcopy routines */
509 	SKBFL_ZEROCOPY_ENABLE = BIT(0),
510 
511 	/* This indicates at least one fragment might be overwritten
512 	 * (as in vmsplice(), sendfile() ...)
513 	 * If we need to compute a TX checksum, we'll need to copy
514 	 * all frags to avoid possible bad checksum
515 	 */
516 	SKBFL_SHARED_FRAG = BIT(1),
517 
518 	/* segment contains only zerocopy data and should not be
519 	 * charged to the kernel memory.
520 	 */
521 	SKBFL_PURE_ZEROCOPY = BIT(2),
522 
523 	SKBFL_DONT_ORPHAN = BIT(3),
524 
525 	/* page references are managed by the ubuf_info, so it's safe to
526 	 * use frags only up until ubuf_info is released
527 	 */
528 	SKBFL_MANAGED_FRAG_REFS = BIT(4),
529 };
530 
531 #define SKBFL_ZEROCOPY_FRAG	(SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG)
532 #define SKBFL_ALL_ZEROCOPY	(SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \
533 				 SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS)
534 
535 struct ubuf_info_ops {
536 	void (*complete)(struct sk_buff *, struct ubuf_info *,
537 			 bool zerocopy_success);
538 	/* has to be compatible with skb_zcopy_set() */
539 	int (*link_skb)(struct sk_buff *skb, struct ubuf_info *uarg);
540 };
541 
542 /*
543  * The callback notifies userspace to release buffers when skb DMA is done in
544  * lower device, the skb last reference should be 0 when calling this.
545  * The zerocopy_success argument is true if zero copy transmit occurred,
546  * false on data copy or out of memory error caused by data copy attempt.
547  * The ctx field is used to track device context.
548  * The desc field is used to track userspace buffer index.
549  */
550 struct ubuf_info {
551 	const struct ubuf_info_ops *ops;
552 	refcount_t refcnt;
553 	u8 flags;
554 };
555 
556 struct ubuf_info_msgzc {
557 	struct ubuf_info ubuf;
558 
559 	union {
560 		struct {
561 			unsigned long desc;
562 			void *ctx;
563 		};
564 		struct {
565 			u32 id;
566 			u16 len;
567 			u16 zerocopy:1;
568 			u32 bytelen;
569 		};
570 	};
571 
572 	struct mmpin {
573 		struct user_struct *user;
574 		unsigned int num_pg;
575 	} mmp;
576 };
577 
578 #define skb_uarg(SKB)	((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
579 #define uarg_to_msgzc(ubuf_ptr)	container_of((ubuf_ptr), struct ubuf_info_msgzc, \
580 					     ubuf)
581 
582 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
583 void mm_unaccount_pinned_pages(struct mmpin *mmp);
584 
585 /* Preserve some data across TX submission and completion.
586  *
587  * Note, this state is stored in the driver. Extending the layout
588  * might need some special care.
589  */
590 struct xsk_tx_metadata_compl {
591 	__u64 *tx_timestamp;
592 };
593 
594 /* This data is invariant across clones and lives at
595  * the end of the header data, ie. at skb->end.
596  */
597 struct skb_shared_info {
598 	__u8		flags;
599 	__u8		meta_len;
600 	__u8		nr_frags;
601 	__u8		tx_flags;
602 	unsigned short	gso_size;
603 	/* Warning: this field is not always filled in (UFO)! */
604 	unsigned short	gso_segs;
605 	struct sk_buff	*frag_list;
606 	union {
607 		struct skb_shared_hwtstamps hwtstamps;
608 		struct xsk_tx_metadata_compl xsk_meta;
609 	};
610 	unsigned int	gso_type;
611 	u32		tskey;
612 
613 	/*
614 	 * Warning : all fields before dataref are cleared in __alloc_skb()
615 	 */
616 	atomic_t	dataref;
617 
618 	union {
619 		struct {
620 			u32		xdp_frags_size;
621 			u32		xdp_frags_truesize;
622 		};
623 
624 		/*
625 		 * Intermediate layers must ensure that destructor_arg
626 		 * remains valid until skb destructor.
627 		 */
628 		void		*destructor_arg;
629 	};
630 
631 	/* must be last field, see pskb_expand_head() */
632 	skb_frag_t	frags[MAX_SKB_FRAGS];
633 };
634 
635 /**
636  * DOC: dataref and headerless skbs
637  *
638  * Transport layers send out clones of payload skbs they hold for
639  * retransmissions. To allow lower layers of the stack to prepend their headers
640  * we split &skb_shared_info.dataref into two halves.
641  * The lower 16 bits count the overall number of references.
642  * The higher 16 bits indicate how many of the references are payload-only.
643  * skb_header_cloned() checks if skb is allowed to add / write the headers.
644  *
645  * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr
646  * (via __skb_header_release()). Any clone created from marked skb will get
647  * &sk_buff.hdr_len populated with the available headroom.
648  * If there's the only clone in existence it's able to modify the headroom
649  * at will. The sequence of calls inside the transport layer is::
650  *
651  *  <alloc skb>
652  *  skb_reserve()
653  *  __skb_header_release()
654  *  skb_clone()
655  *  // send the clone down the stack
656  *
657  * This is not a very generic construct and it depends on the transport layers
658  * doing the right thing. In practice there's usually only one payload-only skb.
659  * Having multiple payload-only skbs with different lengths of hdr_len is not
660  * possible. The payload-only skbs should never leave their owner.
661  */
662 #define SKB_DATAREF_SHIFT 16
663 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
664 
665 
666 enum {
667 	SKB_FCLONE_UNAVAILABLE,	/* skb has no fclone (from head_cache) */
668 	SKB_FCLONE_ORIG,	/* orig skb (from fclone_cache) */
669 	SKB_FCLONE_CLONE,	/* companion fclone skb (from fclone_cache) */
670 };
671 
672 enum {
673 	SKB_GSO_TCPV4 = 1 << 0,
674 
675 	/* This indicates the skb is from an untrusted source. */
676 	SKB_GSO_DODGY = 1 << 1,
677 
678 	/* This indicates the tcp segment has CWR set. */
679 	SKB_GSO_TCP_ECN = 1 << 2,
680 
681 	SKB_GSO_TCP_FIXEDID = 1 << 3,
682 
683 	SKB_GSO_TCPV6 = 1 << 4,
684 
685 	SKB_GSO_FCOE = 1 << 5,
686 
687 	SKB_GSO_GRE = 1 << 6,
688 
689 	SKB_GSO_GRE_CSUM = 1 << 7,
690 
691 	SKB_GSO_IPXIP4 = 1 << 8,
692 
693 	SKB_GSO_IPXIP6 = 1 << 9,
694 
695 	SKB_GSO_UDP_TUNNEL = 1 << 10,
696 
697 	SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
698 
699 	SKB_GSO_PARTIAL = 1 << 12,
700 
701 	SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
702 
703 	SKB_GSO_SCTP = 1 << 14,
704 
705 	SKB_GSO_ESP = 1 << 15,
706 
707 	SKB_GSO_UDP = 1 << 16,
708 
709 	SKB_GSO_UDP_L4 = 1 << 17,
710 
711 	SKB_GSO_FRAGLIST = 1 << 18,
712 
713 	SKB_GSO_TCP_ACCECN = 1 << 19,
714 };
715 
716 #if BITS_PER_LONG > 32
717 #define NET_SKBUFF_DATA_USES_OFFSET 1
718 #endif
719 
720 #ifdef NET_SKBUFF_DATA_USES_OFFSET
721 typedef unsigned int sk_buff_data_t;
722 #else
723 typedef unsigned char *sk_buff_data_t;
724 #endif
725 
726 enum skb_tstamp_type {
727 	SKB_CLOCK_REALTIME,
728 	SKB_CLOCK_MONOTONIC,
729 	SKB_CLOCK_TAI,
730 	__SKB_CLOCK_MAX = SKB_CLOCK_TAI,
731 };
732 
733 /**
734  * DOC: Basic sk_buff geometry
735  *
736  * struct sk_buff itself is a metadata structure and does not hold any packet
737  * data. All the data is held in associated buffers.
738  *
739  * &sk_buff.head points to the main "head" buffer. The head buffer is divided
740  * into two parts:
741  *
742  *  - data buffer, containing headers and sometimes payload;
743  *    this is the part of the skb operated on by the common helpers
744  *    such as skb_put() or skb_pull();
745  *  - shared info (struct skb_shared_info) which holds an array of pointers
746  *    to read-only data in the (page, offset, length) format.
747  *
748  * Optionally &skb_shared_info.frag_list may point to another skb.
749  *
750  * Basic diagram may look like this::
751  *
752  *                                  ---------------
753  *                                 | sk_buff       |
754  *                                  ---------------
755  *     ,---------------------------  + head
756  *    /          ,-----------------  + data
757  *   /          /      ,-----------  + tail
758  *  |          |      |            , + end
759  *  |          |      |           |
760  *  v          v      v           v
761  *   -----------------------------------------------
762  *  | headroom | data |  tailroom | skb_shared_info |
763  *   -----------------------------------------------
764  *                                 + [page frag]
765  *                                 + [page frag]
766  *                                 + [page frag]
767  *                                 + [page frag]       ---------
768  *                                 + frag_list    --> | sk_buff |
769  *                                                     ---------
770  *
771  */
772 
773 /**
774  *	struct sk_buff - socket buffer
775  *	@next: Next buffer in list
776  *	@prev: Previous buffer in list
777  *	@tstamp: Time we arrived/left
778  *	@skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
779  *		for retransmit timer
780  *	@rbnode: RB tree node, alternative to next/prev for netem/tcp
781  *	@list: queue head
782  *	@ll_node: anchor in an llist (eg socket defer_list)
783  *	@sk: Socket we are owned by
784  *	@dev: Device we arrived on/are leaving by
785  *	@dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
786  *	@cb: Control buffer. Free for use by every layer. Put private vars here
787  *	@_skb_refdst: destination entry (with norefcount bit)
788  *	@len: Length of actual data
789  *	@data_len: Data length
790  *	@mac_len: Length of link layer header
791  *	@hdr_len: writable header length of cloned skb
792  *	@csum: Checksum (must include start/offset pair)
793  *	@csum_start: Offset from skb->head where checksumming should start
794  *	@csum_offset: Offset from csum_start where checksum should be stored
795  *	@priority: Packet queueing priority
796  *	@ignore_df: allow local fragmentation
797  *	@cloned: Head may be cloned (check refcnt to be sure)
798  *	@ip_summed: Driver fed us an IP checksum
799  *	@nohdr: Payload reference only, must not modify header
800  *	@pkt_type: Packet class
801  *	@fclone: skbuff clone status
802  *	@ipvs_property: skbuff is owned by ipvs
803  *	@inner_protocol_type: whether the inner protocol is
804  *		ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
805  *	@remcsum_offload: remote checksum offload is enabled
806  *	@offload_fwd_mark: Packet was L2-forwarded in hardware
807  *	@offload_l3_fwd_mark: Packet was L3-forwarded in hardware
808  *	@tc_skip_classify: do not classify packet. set by IFB device
809  *	@tc_at_ingress: used within tc_classify to distinguish in/egress
810  *	@redirected: packet was redirected by packet classifier
811  *	@from_ingress: packet was redirected from the ingress path
812  *	@nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
813  *	@peeked: this packet has been seen already, so stats have been
814  *		done for it, don't do them again
815  *	@nf_trace: netfilter packet trace flag
816  *	@protocol: Packet protocol from driver
817  *	@destructor: Destruct function
818  *	@tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
819  *	@_sk_redir: socket redirection information for skmsg
820  *	@_nfct: Associated connection, if any (with nfctinfo bits)
821  *	@skb_iif: ifindex of device we arrived on
822  *	@tc_index: Traffic control index
823  *	@hash: the packet hash
824  *	@queue_mapping: Queue mapping for multiqueue devices
825  *	@head_frag: skb was allocated from page fragments,
826  *		not allocated by kmalloc() or vmalloc().
827  *	@pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
828  *	@pp_recycle: mark the packet for recycling instead of freeing (implies
829  *		page_pool support on driver)
830  *	@active_extensions: active extensions (skb_ext_id types)
831  *	@ndisc_nodetype: router type (from link layer)
832  *	@ooo_okay: allow the mapping of a socket to a queue to be changed
833  *	@l4_hash: indicate hash is a canonical 4-tuple hash over transport
834  *		ports.
835  *	@sw_hash: indicates hash was computed in software stack
836  *	@wifi_acked_valid: wifi_acked was set
837  *	@wifi_acked: whether frame was acked on wifi or not
838  *	@no_fcs:  Request NIC to treat last 4 bytes as Ethernet FCS
839  *	@encapsulation: indicates the inner headers in the skbuff are valid
840  *	@encap_hdr_csum: software checksum is needed
841  *	@csum_valid: checksum is already valid
842  *	@csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
843  *	@csum_complete_sw: checksum was completed by software
844  *	@csum_level: indicates the number of consecutive checksums found in
845  *		the packet minus one that have been verified as
846  *		CHECKSUM_UNNECESSARY (max 3)
847  *	@unreadable: indicates that at least 1 of the fragments in this skb is
848  *		unreadable.
849  *	@dst_pending_confirm: need to confirm neighbour
850  *	@decrypted: Decrypted SKB
851  *	@slow_gro: state present at GRO time, slower prepare step required
852  *	@tstamp_type: When set, skb->tstamp has the
853  *		delivery_time clock base of skb->tstamp.
854  *	@napi_id: id of the NAPI struct this skb came from
855  *	@sender_cpu: (aka @napi_id) source CPU in XPS
856  *	@alloc_cpu: CPU which did the skb allocation.
857  *	@secmark: security marking
858  *	@mark: Generic packet mark
859  *	@reserved_tailroom: (aka @mark) number of bytes of free space available
860  *		at the tail of an sk_buff
861  *	@vlan_all: vlan fields (proto & tci)
862  *	@vlan_proto: vlan encapsulation protocol
863  *	@vlan_tci: vlan tag control information
864  *	@inner_protocol: Protocol (encapsulation)
865  *	@inner_ipproto: (aka @inner_protocol) stores ipproto when
866  *		skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
867  *	@inner_transport_header: Inner transport layer header (encapsulation)
868  *	@inner_network_header: Network layer header (encapsulation)
869  *	@inner_mac_header: Link layer header (encapsulation)
870  *	@transport_header: Transport layer header
871  *	@network_header: Network layer header
872  *	@mac_header: Link layer header
873  *	@kcov_handle: KCOV remote handle for remote coverage collection
874  *	@tail: Tail pointer
875  *	@end: End pointer
876  *	@head: Head of buffer
877  *	@data: Data head pointer
878  *	@truesize: Buffer size
879  *	@users: User count - see {datagram,tcp}.c
880  *	@extensions: allocated extensions, valid if active_extensions is nonzero
881  */
882 
883 struct sk_buff {
884 	union {
885 		struct {
886 			/* These two members must be first to match sk_buff_head. */
887 			struct sk_buff		*next;
888 			struct sk_buff		*prev;
889 
890 			union {
891 				struct net_device	*dev;
892 				/* Some protocols might use this space to store information,
893 				 * while device pointer would be NULL.
894 				 * UDP receive path is one user.
895 				 */
896 				unsigned long		dev_scratch;
897 			};
898 		};
899 		struct rb_node		rbnode; /* used in netem, ip4 defrag, and tcp stack */
900 		struct list_head	list;
901 		struct llist_node	ll_node;
902 	};
903 
904 	struct sock		*sk;
905 
906 	union {
907 		ktime_t		tstamp;
908 		u64		skb_mstamp_ns; /* earliest departure time */
909 	};
910 	/*
911 	 * This is the control buffer. It is free to use for every
912 	 * layer. Please put your private variables there. If you
913 	 * want to keep them across layers you have to do a skb_clone()
914 	 * first. This is owned by whoever has the skb queued ATM.
915 	 */
916 	char			cb[48] __aligned(8);
917 
918 	union {
919 		struct {
920 			unsigned long	_skb_refdst;
921 			void		(*destructor)(struct sk_buff *skb);
922 		};
923 		struct list_head	tcp_tsorted_anchor;
924 #ifdef CONFIG_NET_SOCK_MSG
925 		unsigned long		_sk_redir;
926 #endif
927 	};
928 
929 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
930 	unsigned long		 _nfct;
931 #endif
932 	unsigned int		len,
933 				data_len;
934 	__u16			mac_len,
935 				hdr_len;
936 
937 	/* Following fields are _not_ copied in __copy_skb_header()
938 	 * Note that queue_mapping is here mostly to fill a hole.
939 	 */
940 	__u16			queue_mapping;
941 
942 /* if you move cloned around you also must adapt those constants */
943 #ifdef __BIG_ENDIAN_BITFIELD
944 #define CLONED_MASK	(1 << 7)
945 #else
946 #define CLONED_MASK	1
947 #endif
948 #define CLONED_OFFSET		offsetof(struct sk_buff, __cloned_offset)
949 
950 	/* private: */
951 	__u8			__cloned_offset[0];
952 	/* public: */
953 	__u8			cloned:1,
954 				nohdr:1,
955 				fclone:2,
956 				peeked:1,
957 				head_frag:1,
958 				pfmemalloc:1,
959 				pp_recycle:1; /* page_pool recycle indicator */
960 #ifdef CONFIG_SKB_EXTENSIONS
961 	__u8			active_extensions;
962 #endif
963 
964 	/* Fields enclosed in headers group are copied
965 	 * using a single memcpy() in __copy_skb_header()
966 	 */
967 	struct_group(headers,
968 
969 	/* private: */
970 	__u8			__pkt_type_offset[0];
971 	/* public: */
972 	__u8			pkt_type:3; /* see PKT_TYPE_MAX */
973 	__u8			ignore_df:1;
974 	__u8			dst_pending_confirm:1;
975 	__u8			ip_summed:2;
976 	__u8			ooo_okay:1;
977 
978 	/* private: */
979 	__u8			__mono_tc_offset[0];
980 	/* public: */
981 	__u8			tstamp_type:2;	/* See skb_tstamp_type */
982 #ifdef CONFIG_NET_XGRESS
983 	__u8			tc_at_ingress:1;	/* See TC_AT_INGRESS_MASK */
984 	__u8			tc_skip_classify:1;
985 #endif
986 	__u8			remcsum_offload:1;
987 	__u8			csum_complete_sw:1;
988 	__u8			csum_level:2;
989 	__u8			inner_protocol_type:1;
990 
991 	__u8			l4_hash:1;
992 	__u8			sw_hash:1;
993 #ifdef CONFIG_WIRELESS
994 	__u8			wifi_acked_valid:1;
995 	__u8			wifi_acked:1;
996 #endif
997 	__u8			no_fcs:1;
998 	/* Indicates the inner headers are valid in the skbuff. */
999 	__u8			encapsulation:1;
1000 	__u8			encap_hdr_csum:1;
1001 	__u8			csum_valid:1;
1002 #ifdef CONFIG_IPV6_NDISC_NODETYPE
1003 	__u8			ndisc_nodetype:2;
1004 #endif
1005 
1006 #if IS_ENABLED(CONFIG_IP_VS)
1007 	__u8			ipvs_property:1;
1008 #endif
1009 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
1010 	__u8			nf_trace:1;
1011 #endif
1012 #ifdef CONFIG_NET_SWITCHDEV
1013 	__u8			offload_fwd_mark:1;
1014 	__u8			offload_l3_fwd_mark:1;
1015 #endif
1016 	__u8			redirected:1;
1017 #ifdef CONFIG_NET_REDIRECT
1018 	__u8			from_ingress:1;
1019 #endif
1020 #ifdef CONFIG_NETFILTER_SKIP_EGRESS
1021 	__u8			nf_skip_egress:1;
1022 #endif
1023 #ifdef CONFIG_SKB_DECRYPTED
1024 	__u8			decrypted:1;
1025 #endif
1026 	__u8			slow_gro:1;
1027 #if IS_ENABLED(CONFIG_IP_SCTP)
1028 	__u8			csum_not_inet:1;
1029 #endif
1030 	__u8			unreadable:1;
1031 #if defined(CONFIG_NET_SCHED) || defined(CONFIG_NET_XGRESS)
1032 	__u16			tc_index;	/* traffic control index */
1033 #endif
1034 
1035 	u16			alloc_cpu;
1036 
1037 	union {
1038 		__wsum		csum;
1039 		struct {
1040 			__u16	csum_start;
1041 			__u16	csum_offset;
1042 		};
1043 	};
1044 	__u32			priority;
1045 	int			skb_iif;
1046 	__u32			hash;
1047 	union {
1048 		u32		vlan_all;
1049 		struct {
1050 			__be16	vlan_proto;
1051 			__u16	vlan_tci;
1052 		};
1053 	};
1054 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
1055 	union {
1056 		unsigned int	napi_id;
1057 		unsigned int	sender_cpu;
1058 	};
1059 #endif
1060 #ifdef CONFIG_NETWORK_SECMARK
1061 	__u32		secmark;
1062 #endif
1063 
1064 	union {
1065 		__u32		mark;
1066 		__u32		reserved_tailroom;
1067 	};
1068 
1069 	union {
1070 		__be16		inner_protocol;
1071 		__u8		inner_ipproto;
1072 	};
1073 
1074 	__u16			inner_transport_header;
1075 	__u16			inner_network_header;
1076 	__u16			inner_mac_header;
1077 
1078 	__be16			protocol;
1079 	__u16			transport_header;
1080 	__u16			network_header;
1081 	__u16			mac_header;
1082 
1083 #ifdef CONFIG_KCOV
1084 	u64			kcov_handle;
1085 #endif
1086 
1087 	); /* end headers group */
1088 
1089 	/* These elements must be at the end, see alloc_skb() for details.  */
1090 	sk_buff_data_t		tail;
1091 	sk_buff_data_t		end;
1092 	unsigned char		*head,
1093 				*data;
1094 	unsigned int		truesize;
1095 	refcount_t		users;
1096 
1097 #ifdef CONFIG_SKB_EXTENSIONS
1098 	/* only usable after checking ->active_extensions != 0 */
1099 	struct skb_ext		*extensions;
1100 #endif
1101 };
1102 
1103 /* if you move pkt_type around you also must adapt those constants */
1104 #ifdef __BIG_ENDIAN_BITFIELD
1105 #define PKT_TYPE_MAX	(7 << 5)
1106 #else
1107 #define PKT_TYPE_MAX	7
1108 #endif
1109 #define PKT_TYPE_OFFSET		offsetof(struct sk_buff, __pkt_type_offset)
1110 
1111 /* if you move tc_at_ingress or tstamp_type
1112  * around, you also must adapt these constants.
1113  */
1114 #ifdef __BIG_ENDIAN_BITFIELD
1115 #define SKB_TSTAMP_TYPE_MASK		(3 << 6)
1116 #define SKB_TSTAMP_TYPE_RSHIFT		(6)
1117 #define TC_AT_INGRESS_MASK		(1 << 5)
1118 #else
1119 #define SKB_TSTAMP_TYPE_MASK		(3)
1120 #define TC_AT_INGRESS_MASK		(1 << 2)
1121 #endif
1122 #define SKB_BF_MONO_TC_OFFSET		offsetof(struct sk_buff, __mono_tc_offset)
1123 
1124 #ifdef __KERNEL__
1125 /*
1126  *	Handling routines are only of interest to the kernel
1127  */
1128 
1129 #define SKB_ALLOC_FCLONE	0x01
1130 #define SKB_ALLOC_RX		0x02
1131 #define SKB_ALLOC_NAPI		0x04
1132 
1133 /**
1134  * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
1135  * @skb: buffer
1136  */
skb_pfmemalloc(const struct sk_buff * skb)1137 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
1138 {
1139 	return unlikely(skb->pfmemalloc);
1140 }
1141 
1142 /*
1143  * skb might have a dst pointer attached, refcounted or not.
1144  * _skb_refdst low order bit is set if refcount was _not_ taken
1145  */
1146 #define SKB_DST_NOREF	1UL
1147 #define SKB_DST_PTRMASK	~(SKB_DST_NOREF)
1148 
1149 /**
1150  * skb_dst - returns skb dst_entry
1151  * @skb: buffer
1152  *
1153  * Returns: skb dst_entry, regardless of reference taken or not.
1154  */
skb_dst(const struct sk_buff * skb)1155 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
1156 {
1157 	/* If refdst was not refcounted, check we still are in a
1158 	 * rcu_read_lock section
1159 	 */
1160 	WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
1161 		!rcu_read_lock_held() &&
1162 		!rcu_read_lock_bh_held());
1163 	return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
1164 }
1165 
1166 /**
1167  * skb_dst_set - sets skb dst
1168  * @skb: buffer
1169  * @dst: dst entry
1170  *
1171  * Sets skb dst, assuming a reference was taken on dst and should
1172  * be released by skb_dst_drop()
1173  */
skb_dst_set(struct sk_buff * skb,struct dst_entry * dst)1174 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
1175 {
1176 	skb->slow_gro |= !!dst;
1177 	skb->_skb_refdst = (unsigned long)dst;
1178 }
1179 
1180 /**
1181  * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
1182  * @skb: buffer
1183  * @dst: dst entry
1184  *
1185  * Sets skb dst, assuming a reference was not taken on dst.
1186  * If dst entry is cached, we do not take reference and dst_release
1187  * will be avoided by refdst_drop. If dst entry is not cached, we take
1188  * reference, so that last dst_release can destroy the dst immediately.
1189  */
skb_dst_set_noref(struct sk_buff * skb,struct dst_entry * dst)1190 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
1191 {
1192 	WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1193 	skb->slow_gro |= !!dst;
1194 	skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
1195 }
1196 
1197 /**
1198  * skb_dst_is_noref - Test if skb dst isn't refcounted
1199  * @skb: buffer
1200  */
skb_dst_is_noref(const struct sk_buff * skb)1201 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1202 {
1203 	return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
1204 }
1205 
1206 /* For mangling skb->pkt_type from user space side from applications
1207  * such as nft, tc, etc, we only allow a conservative subset of
1208  * possible pkt_types to be set.
1209 */
skb_pkt_type_ok(u32 ptype)1210 static inline bool skb_pkt_type_ok(u32 ptype)
1211 {
1212 	return ptype <= PACKET_OTHERHOST;
1213 }
1214 
1215 /**
1216  * skb_napi_id - Returns the skb's NAPI id
1217  * @skb: buffer
1218  */
skb_napi_id(const struct sk_buff * skb)1219 static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1220 {
1221 #ifdef CONFIG_NET_RX_BUSY_POLL
1222 	return skb->napi_id;
1223 #else
1224 	return 0;
1225 #endif
1226 }
1227 
skb_wifi_acked_valid(const struct sk_buff * skb)1228 static inline bool skb_wifi_acked_valid(const struct sk_buff *skb)
1229 {
1230 #ifdef CONFIG_WIRELESS
1231 	return skb->wifi_acked_valid;
1232 #else
1233 	return 0;
1234 #endif
1235 }
1236 
1237 /**
1238  * skb_unref - decrement the skb's reference count
1239  * @skb: buffer
1240  *
1241  * Returns: true if we can free the skb.
1242  */
skb_unref(struct sk_buff * skb)1243 static inline bool skb_unref(struct sk_buff *skb)
1244 {
1245 	if (unlikely(!skb))
1246 		return false;
1247 	if (!IS_ENABLED(CONFIG_DEBUG_NET) && likely(refcount_read(&skb->users) == 1))
1248 		smp_rmb();
1249 	else if (likely(!refcount_dec_and_test(&skb->users)))
1250 		return false;
1251 
1252 	return true;
1253 }
1254 
skb_data_unref(const struct sk_buff * skb,struct skb_shared_info * shinfo)1255 static inline bool skb_data_unref(const struct sk_buff *skb,
1256 				  struct skb_shared_info *shinfo)
1257 {
1258 	int bias;
1259 
1260 	if (!skb->cloned)
1261 		return true;
1262 
1263 	bias = skb->nohdr ? (1 << SKB_DATAREF_SHIFT) + 1 : 1;
1264 
1265 	if (atomic_read(&shinfo->dataref) == bias)
1266 		smp_rmb();
1267 	else if (atomic_sub_return(bias, &shinfo->dataref))
1268 		return false;
1269 
1270 	return true;
1271 }
1272 
1273 void __fix_address sk_skb_reason_drop(struct sock *sk, struct sk_buff *skb,
1274 				      enum skb_drop_reason reason);
1275 
1276 static inline void
kfree_skb_reason(struct sk_buff * skb,enum skb_drop_reason reason)1277 kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason)
1278 {
1279 	sk_skb_reason_drop(NULL, skb, reason);
1280 }
1281 
1282 /**
1283  *	kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason
1284  *	@skb: buffer to free
1285  */
kfree_skb(struct sk_buff * skb)1286 static inline void kfree_skb(struct sk_buff *skb)
1287 {
1288 	kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
1289 }
1290 
1291 void skb_release_head_state(struct sk_buff *skb);
1292 void kfree_skb_list_reason(struct sk_buff *segs,
1293 			   enum skb_drop_reason reason);
1294 void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
1295 void skb_tx_error(struct sk_buff *skb);
1296 
kfree_skb_list(struct sk_buff * segs)1297 static inline void kfree_skb_list(struct sk_buff *segs)
1298 {
1299 	kfree_skb_list_reason(segs, SKB_DROP_REASON_NOT_SPECIFIED);
1300 }
1301 
1302 #ifdef CONFIG_TRACEPOINTS
1303 void consume_skb(struct sk_buff *skb);
1304 #else
consume_skb(struct sk_buff * skb)1305 static inline void consume_skb(struct sk_buff *skb)
1306 {
1307 	return kfree_skb(skb);
1308 }
1309 #endif
1310 
1311 void __consume_stateless_skb(struct sk_buff *skb);
1312 void  __kfree_skb(struct sk_buff *skb);
1313 
1314 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1315 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1316 		      bool *fragstolen, int *delta_truesize);
1317 
1318 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1319 			    int node);
1320 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1321 struct sk_buff *build_skb(void *data, unsigned int frag_size);
1322 struct sk_buff *build_skb_around(struct sk_buff *skb,
1323 				 void *data, unsigned int frag_size);
1324 void skb_attempt_defer_free(struct sk_buff *skb);
1325 
1326 u32 napi_skb_cache_get_bulk(void **skbs, u32 n);
1327 struct sk_buff *napi_build_skb(void *data, unsigned int frag_size);
1328 struct sk_buff *slab_build_skb(void *data);
1329 
1330 /**
1331  * alloc_skb - allocate a network buffer
1332  * @size: size to allocate
1333  * @priority: allocation mask
1334  *
1335  * This function is a convenient wrapper around __alloc_skb().
1336  */
alloc_skb(unsigned int size,gfp_t priority)1337 static inline struct sk_buff *alloc_skb(unsigned int size,
1338 					gfp_t priority)
1339 {
1340 	return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1341 }
1342 
1343 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1344 				     unsigned long data_len,
1345 				     int max_page_order,
1346 				     int *errcode,
1347 				     gfp_t gfp_mask);
1348 struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
1349 
1350 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
1351 struct sk_buff_fclones {
1352 	struct sk_buff	skb1;
1353 
1354 	struct sk_buff	skb2;
1355 
1356 	refcount_t	fclone_ref;
1357 };
1358 
1359 /**
1360  *	skb_fclone_busy - check if fclone is busy
1361  *	@sk: socket
1362  *	@skb: buffer
1363  *
1364  * Returns: true if skb is a fast clone, and its clone is not freed.
1365  * Some drivers call skb_orphan() in their ndo_start_xmit(),
1366  * so we also check that didn't happen.
1367  */
skb_fclone_busy(const struct sock * sk,const struct sk_buff * skb)1368 static inline bool skb_fclone_busy(const struct sock *sk,
1369 				   const struct sk_buff *skb)
1370 {
1371 	const struct sk_buff_fclones *fclones;
1372 
1373 	fclones = container_of(skb, struct sk_buff_fclones, skb1);
1374 
1375 	return skb->fclone == SKB_FCLONE_ORIG &&
1376 	       refcount_read(&fclones->fclone_ref) > 1 &&
1377 	       READ_ONCE(fclones->skb2.sk) == sk;
1378 }
1379 
1380 /**
1381  * alloc_skb_fclone - allocate a network buffer from fclone cache
1382  * @size: size to allocate
1383  * @priority: allocation mask
1384  *
1385  * This function is a convenient wrapper around __alloc_skb().
1386  */
alloc_skb_fclone(unsigned int size,gfp_t priority)1387 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1388 					       gfp_t priority)
1389 {
1390 	return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1391 }
1392 
1393 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1394 void skb_headers_offset_update(struct sk_buff *skb, int off);
1395 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1396 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1397 void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1398 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1399 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1400 				   gfp_t gfp_mask, bool fclone);
__pskb_copy(struct sk_buff * skb,int headroom,gfp_t gfp_mask)1401 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1402 					  gfp_t gfp_mask)
1403 {
1404 	return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1405 }
1406 
1407 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1408 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1409 				     unsigned int headroom);
1410 struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom);
1411 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1412 				int newtailroom, gfp_t priority);
1413 int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1414 				     int offset, int len);
1415 int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1416 			      int offset, int len);
1417 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1418 int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1419 
1420 /**
1421  *	skb_pad			-	zero pad the tail of an skb
1422  *	@skb: buffer to pad
1423  *	@pad: space to pad
1424  *
1425  *	Ensure that a buffer is followed by a padding area that is zero
1426  *	filled. Used by network drivers which may DMA or transfer data
1427  *	beyond the buffer end onto the wire.
1428  *
1429  *	May return error in out of memory cases. The skb is freed on error.
1430  */
skb_pad(struct sk_buff * skb,int pad)1431 static inline int skb_pad(struct sk_buff *skb, int pad)
1432 {
1433 	return __skb_pad(skb, pad, true);
1434 }
1435 #define dev_kfree_skb(a)	consume_skb(a)
1436 
1437 int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1438 			 int offset, size_t size, size_t max_frags);
1439 
1440 struct skb_seq_state {
1441 	__u32		lower_offset;
1442 	__u32		upper_offset;
1443 	__u32		frag_idx;
1444 	__u32		stepped_offset;
1445 	struct sk_buff	*root_skb;
1446 	struct sk_buff	*cur_skb;
1447 	__u8		*frag_data;
1448 	__u32		frag_off;
1449 };
1450 
1451 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1452 			  unsigned int to, struct skb_seq_state *st);
1453 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1454 			  struct skb_seq_state *st);
1455 void skb_abort_seq_read(struct skb_seq_state *st);
1456 int skb_copy_seq_read(struct skb_seq_state *st, int offset, void *to, int len);
1457 
1458 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1459 			   unsigned int to, struct ts_config *config);
1460 
1461 /*
1462  * Packet hash types specify the type of hash in skb_set_hash.
1463  *
1464  * Hash types refer to the protocol layer addresses which are used to
1465  * construct a packet's hash. The hashes are used to differentiate or identify
1466  * flows of the protocol layer for the hash type. Hash types are either
1467  * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1468  *
1469  * Properties of hashes:
1470  *
1471  * 1) Two packets in different flows have different hash values
1472  * 2) Two packets in the same flow should have the same hash value
1473  *
1474  * A hash at a higher layer is considered to be more specific. A driver should
1475  * set the most specific hash possible.
1476  *
1477  * A driver cannot indicate a more specific hash than the layer at which a hash
1478  * was computed. For instance an L3 hash cannot be set as an L4 hash.
1479  *
1480  * A driver may indicate a hash level which is less specific than the
1481  * actual layer the hash was computed on. For instance, a hash computed
1482  * at L4 may be considered an L3 hash. This should only be done if the
1483  * driver can't unambiguously determine that the HW computed the hash at
1484  * the higher layer. Note that the "should" in the second property above
1485  * permits this.
1486  */
1487 enum pkt_hash_types {
1488 	PKT_HASH_TYPE_NONE,	/* Undefined type */
1489 	PKT_HASH_TYPE_L2,	/* Input: src_MAC, dest_MAC */
1490 	PKT_HASH_TYPE_L3,	/* Input: src_IP, dst_IP */
1491 	PKT_HASH_TYPE_L4,	/* Input: src_IP, dst_IP, src_port, dst_port */
1492 };
1493 
skb_clear_hash(struct sk_buff * skb)1494 static inline void skb_clear_hash(struct sk_buff *skb)
1495 {
1496 	skb->hash = 0;
1497 	skb->sw_hash = 0;
1498 	skb->l4_hash = 0;
1499 }
1500 
skb_clear_hash_if_not_l4(struct sk_buff * skb)1501 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1502 {
1503 	if (!skb->l4_hash)
1504 		skb_clear_hash(skb);
1505 }
1506 
1507 static inline void
__skb_set_hash(struct sk_buff * skb,__u32 hash,bool is_sw,bool is_l4)1508 __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1509 {
1510 	skb->l4_hash = is_l4;
1511 	skb->sw_hash = is_sw;
1512 	skb->hash = hash;
1513 }
1514 
1515 static inline void
skb_set_hash(struct sk_buff * skb,__u32 hash,enum pkt_hash_types type)1516 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1517 {
1518 	/* Used by drivers to set hash from HW */
1519 	__skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1520 }
1521 
1522 static inline void
__skb_set_sw_hash(struct sk_buff * skb,__u32 hash,bool is_l4)1523 __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1524 {
1525 	__skb_set_hash(skb, hash, true, is_l4);
1526 }
1527 
1528 u32 __skb_get_hash_symmetric_net(const struct net *net, const struct sk_buff *skb);
1529 
__skb_get_hash_symmetric(const struct sk_buff * skb)1530 static inline u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1531 {
1532 	return __skb_get_hash_symmetric_net(NULL, skb);
1533 }
1534 
1535 void __skb_get_hash_net(const struct net *net, struct sk_buff *skb);
1536 u32 skb_get_poff(const struct sk_buff *skb);
1537 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1538 		   const struct flow_keys_basic *keys, int hlen);
1539 __be32 skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1540 			  const void *data, int hlen_proto);
1541 
1542 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1543 			     const struct flow_dissector_key *key,
1544 			     unsigned int key_count);
1545 
1546 struct bpf_flow_dissector;
1547 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1548 		     __be16 proto, int nhoff, int hlen, unsigned int flags);
1549 
1550 bool __skb_flow_dissect(const struct net *net,
1551 			const struct sk_buff *skb,
1552 			struct flow_dissector *flow_dissector,
1553 			void *target_container, const void *data,
1554 			__be16 proto, int nhoff, int hlen, unsigned int flags);
1555 
skb_flow_dissect(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,unsigned int flags)1556 static inline bool skb_flow_dissect(const struct sk_buff *skb,
1557 				    struct flow_dissector *flow_dissector,
1558 				    void *target_container, unsigned int flags)
1559 {
1560 	return __skb_flow_dissect(NULL, skb, flow_dissector,
1561 				  target_container, NULL, 0, 0, 0, flags);
1562 }
1563 
skb_flow_dissect_flow_keys(const struct sk_buff * skb,struct flow_keys * flow,unsigned int flags)1564 static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1565 					      struct flow_keys *flow,
1566 					      unsigned int flags)
1567 {
1568 	memset(flow, 0, sizeof(*flow));
1569 	return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1570 				  flow, NULL, 0, 0, 0, flags);
1571 }
1572 
1573 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)1574 skb_flow_dissect_flow_keys_basic(const struct net *net,
1575 				 const struct sk_buff *skb,
1576 				 struct flow_keys_basic *flow,
1577 				 const void *data, __be16 proto,
1578 				 int nhoff, int hlen, unsigned int flags)
1579 {
1580 	memset(flow, 0, sizeof(*flow));
1581 	return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
1582 				  data, proto, nhoff, hlen, flags);
1583 }
1584 
1585 void skb_flow_dissect_meta(const struct sk_buff *skb,
1586 			   struct flow_dissector *flow_dissector,
1587 			   void *target_container);
1588 
1589 /* Gets a skb connection tracking info, ctinfo map should be a
1590  * map of mapsize to translate enum ip_conntrack_info states
1591  * to user states.
1592  */
1593 void
1594 skb_flow_dissect_ct(const struct sk_buff *skb,
1595 		    struct flow_dissector *flow_dissector,
1596 		    void *target_container,
1597 		    u16 *ctinfo_map, size_t mapsize,
1598 		    bool post_ct, u16 zone);
1599 void
1600 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1601 			     struct flow_dissector *flow_dissector,
1602 			     void *target_container);
1603 
1604 void skb_flow_dissect_hash(const struct sk_buff *skb,
1605 			   struct flow_dissector *flow_dissector,
1606 			   void *target_container);
1607 
skb_get_hash_net(const struct net * net,struct sk_buff * skb)1608 static inline __u32 skb_get_hash_net(const struct net *net, struct sk_buff *skb)
1609 {
1610 	if (!skb->l4_hash && !skb->sw_hash)
1611 		__skb_get_hash_net(net, skb);
1612 
1613 	return skb->hash;
1614 }
1615 
skb_get_hash(struct sk_buff * skb)1616 static inline __u32 skb_get_hash(struct sk_buff *skb)
1617 {
1618 	if (!skb->l4_hash && !skb->sw_hash)
1619 		__skb_get_hash_net(NULL, skb);
1620 
1621 	return skb->hash;
1622 }
1623 
skb_get_hash_flowi6(struct sk_buff * skb,const struct flowi6 * fl6)1624 static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1625 {
1626 	if (!skb->l4_hash && !skb->sw_hash) {
1627 		struct flow_keys keys;
1628 		__u32 hash = __get_hash_from_flowi6(fl6, &keys);
1629 
1630 		__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1631 	}
1632 
1633 	return skb->hash;
1634 }
1635 
1636 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1637 			   const siphash_key_t *perturb);
1638 
skb_get_hash_raw(const struct sk_buff * skb)1639 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1640 {
1641 	return skb->hash;
1642 }
1643 
skb_copy_hash(struct sk_buff * to,const struct sk_buff * from)1644 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1645 {
1646 	to->hash = from->hash;
1647 	to->sw_hash = from->sw_hash;
1648 	to->l4_hash = from->l4_hash;
1649 };
1650 
skb_cmp_decrypted(const struct sk_buff * skb1,const struct sk_buff * skb2)1651 static inline int skb_cmp_decrypted(const struct sk_buff *skb1,
1652 				    const struct sk_buff *skb2)
1653 {
1654 #ifdef CONFIG_SKB_DECRYPTED
1655 	return skb2->decrypted - skb1->decrypted;
1656 #else
1657 	return 0;
1658 #endif
1659 }
1660 
skb_is_decrypted(const struct sk_buff * skb)1661 static inline bool skb_is_decrypted(const struct sk_buff *skb)
1662 {
1663 #ifdef CONFIG_SKB_DECRYPTED
1664 	return skb->decrypted;
1665 #else
1666 	return false;
1667 #endif
1668 }
1669 
skb_copy_decrypted(struct sk_buff * to,const struct sk_buff * from)1670 static inline void skb_copy_decrypted(struct sk_buff *to,
1671 				      const struct sk_buff *from)
1672 {
1673 #ifdef CONFIG_SKB_DECRYPTED
1674 	to->decrypted = from->decrypted;
1675 #endif
1676 }
1677 
1678 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_end_pointer(const struct sk_buff * skb)1679 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1680 {
1681 	return skb->head + skb->end;
1682 }
1683 
skb_end_offset(const struct sk_buff * skb)1684 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1685 {
1686 	return skb->end;
1687 }
1688 
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1689 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1690 {
1691 	skb->end = offset;
1692 }
1693 #else
skb_end_pointer(const struct sk_buff * skb)1694 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1695 {
1696 	return skb->end;
1697 }
1698 
skb_end_offset(const struct sk_buff * skb)1699 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1700 {
1701 	return skb->end - skb->head;
1702 }
1703 
skb_set_end_offset(struct sk_buff * skb,unsigned int offset)1704 static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset)
1705 {
1706 	skb->end = skb->head + offset;
1707 }
1708 #endif
1709 
1710 extern const struct ubuf_info_ops msg_zerocopy_ubuf_ops;
1711 
1712 struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
1713 				       struct ubuf_info *uarg);
1714 
1715 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
1716 
1717 int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk,
1718 			    struct sk_buff *skb, struct iov_iter *from,
1719 			    size_t length);
1720 
1721 int zerocopy_fill_skb_from_iter(struct sk_buff *skb,
1722 				struct iov_iter *from, size_t length);
1723 
skb_zerocopy_iter_dgram(struct sk_buff * skb,struct msghdr * msg,int len)1724 static inline int skb_zerocopy_iter_dgram(struct sk_buff *skb,
1725 					  struct msghdr *msg, int len)
1726 {
1727 	return __zerocopy_sg_from_iter(msg, skb->sk, skb, &msg->msg_iter, len);
1728 }
1729 
1730 int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
1731 			     struct msghdr *msg, int len,
1732 			     struct ubuf_info *uarg);
1733 
1734 /* Internal */
1735 #define skb_shinfo(SKB)	((struct skb_shared_info *)(skb_end_pointer(SKB)))
1736 
skb_hwtstamps(struct sk_buff * skb)1737 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1738 {
1739 	return &skb_shinfo(skb)->hwtstamps;
1740 }
1741 
skb_zcopy(struct sk_buff * skb)1742 static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1743 {
1744 	bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE;
1745 
1746 	return is_zcopy ? skb_uarg(skb) : NULL;
1747 }
1748 
skb_zcopy_pure(const struct sk_buff * skb)1749 static inline bool skb_zcopy_pure(const struct sk_buff *skb)
1750 {
1751 	return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY;
1752 }
1753 
skb_zcopy_managed(const struct sk_buff * skb)1754 static inline bool skb_zcopy_managed(const struct sk_buff *skb)
1755 {
1756 	return skb_shinfo(skb)->flags & SKBFL_MANAGED_FRAG_REFS;
1757 }
1758 
skb_pure_zcopy_same(const struct sk_buff * skb1,const struct sk_buff * skb2)1759 static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1,
1760 				       const struct sk_buff *skb2)
1761 {
1762 	return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2);
1763 }
1764 
net_zcopy_get(struct ubuf_info * uarg)1765 static inline void net_zcopy_get(struct ubuf_info *uarg)
1766 {
1767 	refcount_inc(&uarg->refcnt);
1768 }
1769 
skb_zcopy_init(struct sk_buff * skb,struct ubuf_info * uarg)1770 static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg)
1771 {
1772 	skb_shinfo(skb)->destructor_arg = uarg;
1773 	skb_shinfo(skb)->flags |= uarg->flags;
1774 }
1775 
skb_zcopy_set(struct sk_buff * skb,struct ubuf_info * uarg,bool * have_ref)1776 static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1777 				 bool *have_ref)
1778 {
1779 	if (skb && uarg && !skb_zcopy(skb)) {
1780 		if (unlikely(have_ref && *have_ref))
1781 			*have_ref = false;
1782 		else
1783 			net_zcopy_get(uarg);
1784 		skb_zcopy_init(skb, uarg);
1785 	}
1786 }
1787 
skb_zcopy_set_nouarg(struct sk_buff * skb,void * val)1788 static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1789 {
1790 	skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1791 	skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG;
1792 }
1793 
skb_zcopy_is_nouarg(struct sk_buff * skb)1794 static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1795 {
1796 	return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1797 }
1798 
skb_zcopy_get_nouarg(struct sk_buff * skb)1799 static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1800 {
1801 	return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1802 }
1803 
net_zcopy_put(struct ubuf_info * uarg)1804 static inline void net_zcopy_put(struct ubuf_info *uarg)
1805 {
1806 	if (uarg)
1807 		uarg->ops->complete(NULL, uarg, true);
1808 }
1809 
net_zcopy_put_abort(struct ubuf_info * uarg,bool have_uref)1810 static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref)
1811 {
1812 	if (uarg) {
1813 		if (uarg->ops == &msg_zerocopy_ubuf_ops)
1814 			msg_zerocopy_put_abort(uarg, have_uref);
1815 		else if (have_uref)
1816 			net_zcopy_put(uarg);
1817 	}
1818 }
1819 
1820 /* Release a reference on a zerocopy structure */
skb_zcopy_clear(struct sk_buff * skb,bool zerocopy_success)1821 static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success)
1822 {
1823 	struct ubuf_info *uarg = skb_zcopy(skb);
1824 
1825 	if (uarg) {
1826 		if (!skb_zcopy_is_nouarg(skb))
1827 			uarg->ops->complete(skb, uarg, zerocopy_success);
1828 
1829 		skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY;
1830 	}
1831 }
1832 
1833 void __skb_zcopy_downgrade_managed(struct sk_buff *skb);
1834 
skb_zcopy_downgrade_managed(struct sk_buff * skb)1835 static inline void skb_zcopy_downgrade_managed(struct sk_buff *skb)
1836 {
1837 	if (unlikely(skb_zcopy_managed(skb)))
1838 		__skb_zcopy_downgrade_managed(skb);
1839 }
1840 
1841 /* Return true if frags in this skb are readable by the host. */
skb_frags_readable(const struct sk_buff * skb)1842 static inline bool skb_frags_readable(const struct sk_buff *skb)
1843 {
1844 	return !skb->unreadable;
1845 }
1846 
skb_mark_not_on_list(struct sk_buff * skb)1847 static inline void skb_mark_not_on_list(struct sk_buff *skb)
1848 {
1849 	skb->next = NULL;
1850 }
1851 
skb_poison_list(struct sk_buff * skb)1852 static inline void skb_poison_list(struct sk_buff *skb)
1853 {
1854 #ifdef CONFIG_DEBUG_NET
1855 	skb->next = SKB_LIST_POISON_NEXT;
1856 #endif
1857 }
1858 
1859 /* Iterate through singly-linked GSO fragments of an skb. */
1860 #define skb_list_walk_safe(first, skb, next_skb)                               \
1861 	for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb);  \
1862 	     (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
1863 
skb_list_del_init(struct sk_buff * skb)1864 static inline void skb_list_del_init(struct sk_buff *skb)
1865 {
1866 	__list_del_entry(&skb->list);
1867 	skb_mark_not_on_list(skb);
1868 }
1869 
1870 /**
1871  *	skb_queue_empty - check if a queue is empty
1872  *	@list: queue head
1873  *
1874  *	Returns true if the queue is empty, false otherwise.
1875  */
skb_queue_empty(const struct sk_buff_head * list)1876 static inline int skb_queue_empty(const struct sk_buff_head *list)
1877 {
1878 	return list->next == (const struct sk_buff *) list;
1879 }
1880 
1881 /**
1882  *	skb_queue_empty_lockless - check if a queue is empty
1883  *	@list: queue head
1884  *
1885  *	Returns true if the queue is empty, false otherwise.
1886  *	This variant can be used in lockless contexts.
1887  */
skb_queue_empty_lockless(const struct sk_buff_head * list)1888 static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1889 {
1890 	return READ_ONCE(list->next) == (const struct sk_buff *) list;
1891 }
1892 
1893 
1894 /**
1895  *	skb_queue_is_last - check if skb is the last entry in the queue
1896  *	@list: queue head
1897  *	@skb: buffer
1898  *
1899  *	Returns true if @skb is the last buffer on the list.
1900  */
skb_queue_is_last(const struct sk_buff_head * list,const struct sk_buff * skb)1901 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1902 				     const struct sk_buff *skb)
1903 {
1904 	return skb->next == (const struct sk_buff *) list;
1905 }
1906 
1907 /**
1908  *	skb_queue_is_first - check if skb is the first entry in the queue
1909  *	@list: queue head
1910  *	@skb: buffer
1911  *
1912  *	Returns true if @skb is the first buffer on the list.
1913  */
skb_queue_is_first(const struct sk_buff_head * list,const struct sk_buff * skb)1914 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1915 				      const struct sk_buff *skb)
1916 {
1917 	return skb->prev == (const struct sk_buff *) list;
1918 }
1919 
1920 /**
1921  *	skb_queue_next - return the next packet in the queue
1922  *	@list: queue head
1923  *	@skb: current buffer
1924  *
1925  *	Return the next packet in @list after @skb.  It is only valid to
1926  *	call this if skb_queue_is_last() evaluates to false.
1927  */
skb_queue_next(const struct sk_buff_head * list,const struct sk_buff * skb)1928 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1929 					     const struct sk_buff *skb)
1930 {
1931 	/* This BUG_ON may seem severe, but if we just return then we
1932 	 * are going to dereference garbage.
1933 	 */
1934 	BUG_ON(skb_queue_is_last(list, skb));
1935 	return skb->next;
1936 }
1937 
1938 /**
1939  *	skb_queue_prev - return the prev packet in the queue
1940  *	@list: queue head
1941  *	@skb: current buffer
1942  *
1943  *	Return the prev packet in @list before @skb.  It is only valid to
1944  *	call this if skb_queue_is_first() evaluates to false.
1945  */
skb_queue_prev(const struct sk_buff_head * list,const struct sk_buff * skb)1946 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1947 					     const struct sk_buff *skb)
1948 {
1949 	/* This BUG_ON may seem severe, but if we just return then we
1950 	 * are going to dereference garbage.
1951 	 */
1952 	BUG_ON(skb_queue_is_first(list, skb));
1953 	return skb->prev;
1954 }
1955 
1956 /**
1957  *	skb_get - reference buffer
1958  *	@skb: buffer to reference
1959  *
1960  *	Makes another reference to a socket buffer and returns a pointer
1961  *	to the buffer.
1962  */
skb_get(struct sk_buff * skb)1963 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1964 {
1965 	refcount_inc(&skb->users);
1966 	return skb;
1967 }
1968 
1969 /*
1970  * If users == 1, we are the only owner and can avoid redundant atomic changes.
1971  */
1972 
1973 /**
1974  *	skb_cloned - is the buffer a clone
1975  *	@skb: buffer to check
1976  *
1977  *	Returns true if the buffer was generated with skb_clone() and is
1978  *	one of multiple shared copies of the buffer. Cloned buffers are
1979  *	shared data so must not be written to under normal circumstances.
1980  */
skb_cloned(const struct sk_buff * skb)1981 static inline int skb_cloned(const struct sk_buff *skb)
1982 {
1983 	return skb->cloned &&
1984 	       (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1985 }
1986 
skb_unclone(struct sk_buff * skb,gfp_t pri)1987 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1988 {
1989 	might_sleep_if(gfpflags_allow_blocking(pri));
1990 
1991 	if (skb_cloned(skb))
1992 		return pskb_expand_head(skb, 0, 0, pri);
1993 
1994 	return 0;
1995 }
1996 
1997 /* This variant of skb_unclone() makes sure skb->truesize
1998  * and skb_end_offset() are not changed, whenever a new skb->head is needed.
1999  *
2000  * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X))
2001  * when various debugging features are in place.
2002  */
2003 int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri);
skb_unclone_keeptruesize(struct sk_buff * skb,gfp_t pri)2004 static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri)
2005 {
2006 	might_sleep_if(gfpflags_allow_blocking(pri));
2007 
2008 	if (skb_cloned(skb))
2009 		return __skb_unclone_keeptruesize(skb, pri);
2010 	return 0;
2011 }
2012 
2013 /**
2014  *	skb_header_cloned - is the header a clone
2015  *	@skb: buffer to check
2016  *
2017  *	Returns true if modifying the header part of the buffer requires
2018  *	the data to be copied.
2019  */
skb_header_cloned(const struct sk_buff * skb)2020 static inline int skb_header_cloned(const struct sk_buff *skb)
2021 {
2022 	int dataref;
2023 
2024 	if (!skb->cloned)
2025 		return 0;
2026 
2027 	dataref = atomic_read(&skb_shinfo(skb)->dataref);
2028 	dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
2029 	return dataref != 1;
2030 }
2031 
skb_header_unclone(struct sk_buff * skb,gfp_t pri)2032 static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
2033 {
2034 	might_sleep_if(gfpflags_allow_blocking(pri));
2035 
2036 	if (skb_header_cloned(skb))
2037 		return pskb_expand_head(skb, 0, 0, pri);
2038 
2039 	return 0;
2040 }
2041 
2042 /**
2043  * __skb_header_release() - allow clones to use the headroom
2044  * @skb: buffer to operate on
2045  *
2046  * See "DOC: dataref and headerless skbs".
2047  */
__skb_header_release(struct sk_buff * skb)2048 static inline void __skb_header_release(struct sk_buff *skb)
2049 {
2050 	skb->nohdr = 1;
2051 	atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
2052 }
2053 
2054 
2055 /**
2056  *	skb_shared - is the buffer shared
2057  *	@skb: buffer to check
2058  *
2059  *	Returns true if more than one person has a reference to this
2060  *	buffer.
2061  */
skb_shared(const struct sk_buff * skb)2062 static inline int skb_shared(const struct sk_buff *skb)
2063 {
2064 	return refcount_read(&skb->users) != 1;
2065 }
2066 
2067 /**
2068  *	skb_share_check - check if buffer is shared and if so clone it
2069  *	@skb: buffer to check
2070  *	@pri: priority for memory allocation
2071  *
2072  *	If the buffer is shared the buffer is cloned and the old copy
2073  *	drops a reference. A new clone with a single reference is returned.
2074  *	If the buffer is not shared the original buffer is returned. When
2075  *	being called from interrupt status or with spinlocks held pri must
2076  *	be GFP_ATOMIC.
2077  *
2078  *	NULL is returned on a memory allocation failure.
2079  */
skb_share_check(struct sk_buff * skb,gfp_t pri)2080 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
2081 {
2082 	might_sleep_if(gfpflags_allow_blocking(pri));
2083 	if (skb_shared(skb)) {
2084 		struct sk_buff *nskb = skb_clone(skb, pri);
2085 
2086 		if (likely(nskb))
2087 			consume_skb(skb);
2088 		else
2089 			kfree_skb(skb);
2090 		skb = nskb;
2091 	}
2092 	return skb;
2093 }
2094 
2095 /*
2096  *	Copy shared buffers into a new sk_buff. We effectively do COW on
2097  *	packets to handle cases where we have a local reader and forward
2098  *	and a couple of other messy ones. The normal one is tcpdumping
2099  *	a packet that's being forwarded.
2100  */
2101 
2102 /**
2103  *	skb_unshare - make a copy of a shared buffer
2104  *	@skb: buffer to check
2105  *	@pri: priority for memory allocation
2106  *
2107  *	If the socket buffer is a clone then this function creates a new
2108  *	copy of the data, drops a reference count on the old copy and returns
2109  *	the new copy with the reference count at 1. If the buffer is not a clone
2110  *	the original buffer is returned. When called with a spinlock held or
2111  *	from interrupt state @pri must be %GFP_ATOMIC
2112  *
2113  *	%NULL is returned on a memory allocation failure.
2114  */
skb_unshare(struct sk_buff * skb,gfp_t pri)2115 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
2116 					  gfp_t pri)
2117 {
2118 	might_sleep_if(gfpflags_allow_blocking(pri));
2119 	if (skb_cloned(skb)) {
2120 		struct sk_buff *nskb = skb_copy(skb, pri);
2121 
2122 		/* Free our shared copy */
2123 		if (likely(nskb))
2124 			consume_skb(skb);
2125 		else
2126 			kfree_skb(skb);
2127 		skb = nskb;
2128 	}
2129 	return skb;
2130 }
2131 
2132 /**
2133  *	skb_peek - peek at the head of an &sk_buff_head
2134  *	@list_: list to peek at
2135  *
2136  *	Peek an &sk_buff. Unlike most other operations you _MUST_
2137  *	be careful with this one. A peek leaves the buffer on the
2138  *	list and someone else may run off with it. You must hold
2139  *	the appropriate locks or have a private queue to do this.
2140  *
2141  *	Returns %NULL for an empty list or a pointer to the head element.
2142  *	The reference count is not incremented and the reference is therefore
2143  *	volatile. Use with caution.
2144  */
skb_peek(const struct sk_buff_head * list_)2145 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
2146 {
2147 	struct sk_buff *skb = list_->next;
2148 
2149 	if (skb == (struct sk_buff *)list_)
2150 		skb = NULL;
2151 	return skb;
2152 }
2153 
2154 /**
2155  *	__skb_peek - peek at the head of a non-empty &sk_buff_head
2156  *	@list_: list to peek at
2157  *
2158  *	Like skb_peek(), but the caller knows that the list is not empty.
2159  */
__skb_peek(const struct sk_buff_head * list_)2160 static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
2161 {
2162 	return list_->next;
2163 }
2164 
2165 /**
2166  *	skb_peek_next - peek skb following the given one from a queue
2167  *	@skb: skb to start from
2168  *	@list_: list to peek at
2169  *
2170  *	Returns %NULL when the end of the list is met or a pointer to the
2171  *	next element. The reference count is not incremented and the
2172  *	reference is therefore volatile. Use with caution.
2173  */
skb_peek_next(struct sk_buff * skb,const struct sk_buff_head * list_)2174 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
2175 		const struct sk_buff_head *list_)
2176 {
2177 	struct sk_buff *next = skb->next;
2178 
2179 	if (next == (struct sk_buff *)list_)
2180 		next = NULL;
2181 	return next;
2182 }
2183 
2184 /**
2185  *	skb_peek_tail - peek at the tail of an &sk_buff_head
2186  *	@list_: list to peek at
2187  *
2188  *	Peek an &sk_buff. Unlike most other operations you _MUST_
2189  *	be careful with this one. A peek leaves the buffer on the
2190  *	list and someone else may run off with it. You must hold
2191  *	the appropriate locks or have a private queue to do this.
2192  *
2193  *	Returns %NULL for an empty list or a pointer to the tail element.
2194  *	The reference count is not incremented and the reference is therefore
2195  *	volatile. Use with caution.
2196  */
skb_peek_tail(const struct sk_buff_head * list_)2197 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
2198 {
2199 	struct sk_buff *skb = READ_ONCE(list_->prev);
2200 
2201 	if (skb == (struct sk_buff *)list_)
2202 		skb = NULL;
2203 	return skb;
2204 
2205 }
2206 
2207 /**
2208  *	skb_queue_len	- get queue length
2209  *	@list_: list to measure
2210  *
2211  *	Return the length of an &sk_buff queue.
2212  */
skb_queue_len(const struct sk_buff_head * list_)2213 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
2214 {
2215 	return list_->qlen;
2216 }
2217 
2218 /**
2219  *	skb_queue_len_lockless	- get queue length
2220  *	@list_: list to measure
2221  *
2222  *	Return the length of an &sk_buff queue.
2223  *	This variant can be used in lockless contexts.
2224  */
skb_queue_len_lockless(const struct sk_buff_head * list_)2225 static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
2226 {
2227 	return READ_ONCE(list_->qlen);
2228 }
2229 
2230 /**
2231  *	__skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
2232  *	@list: queue to initialize
2233  *
2234  *	This initializes only the list and queue length aspects of
2235  *	an sk_buff_head object.  This allows to initialize the list
2236  *	aspects of an sk_buff_head without reinitializing things like
2237  *	the spinlock.  It can also be used for on-stack sk_buff_head
2238  *	objects where the spinlock is known to not be used.
2239  */
__skb_queue_head_init(struct sk_buff_head * list)2240 static inline void __skb_queue_head_init(struct sk_buff_head *list)
2241 {
2242 	list->prev = list->next = (struct sk_buff *)list;
2243 	list->qlen = 0;
2244 }
2245 
2246 /*
2247  * This function creates a split out lock class for each invocation;
2248  * this is needed for now since a whole lot of users of the skb-queue
2249  * infrastructure in drivers have different locking usage (in hardirq)
2250  * than the networking core (in softirq only). In the long run either the
2251  * network layer or drivers should need annotation to consolidate the
2252  * main types of usage into 3 classes.
2253  */
skb_queue_head_init(struct sk_buff_head * list)2254 static inline void skb_queue_head_init(struct sk_buff_head *list)
2255 {
2256 	spin_lock_init(&list->lock);
2257 	__skb_queue_head_init(list);
2258 }
2259 
skb_queue_head_init_class(struct sk_buff_head * list,struct lock_class_key * class)2260 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
2261 		struct lock_class_key *class)
2262 {
2263 	skb_queue_head_init(list);
2264 	lockdep_set_class(&list->lock, class);
2265 }
2266 
2267 /*
2268  *	Insert an sk_buff on a list.
2269  *
2270  *	The "__skb_xxxx()" functions are the non-atomic ones that
2271  *	can only be called with interrupts disabled.
2272  */
__skb_insert(struct sk_buff * newsk,struct sk_buff * prev,struct sk_buff * next,struct sk_buff_head * list)2273 static inline void __skb_insert(struct sk_buff *newsk,
2274 				struct sk_buff *prev, struct sk_buff *next,
2275 				struct sk_buff_head *list)
2276 {
2277 	/* See skb_queue_empty_lockless() and skb_peek_tail()
2278 	 * for the opposite READ_ONCE()
2279 	 */
2280 	WRITE_ONCE(newsk->next, next);
2281 	WRITE_ONCE(newsk->prev, prev);
2282 	WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk);
2283 	WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk);
2284 	WRITE_ONCE(list->qlen, list->qlen + 1);
2285 }
2286 
__skb_queue_splice(const struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * next)2287 static inline void __skb_queue_splice(const struct sk_buff_head *list,
2288 				      struct sk_buff *prev,
2289 				      struct sk_buff *next)
2290 {
2291 	struct sk_buff *first = list->next;
2292 	struct sk_buff *last = list->prev;
2293 
2294 	WRITE_ONCE(first->prev, prev);
2295 	WRITE_ONCE(prev->next, first);
2296 
2297 	WRITE_ONCE(last->next, next);
2298 	WRITE_ONCE(next->prev, last);
2299 }
2300 
2301 /**
2302  *	skb_queue_splice - join two skb lists, this is designed for stacks
2303  *	@list: the new list to add
2304  *	@head: the place to add it in the first list
2305  */
skb_queue_splice(const struct sk_buff_head * list,struct sk_buff_head * head)2306 static inline void skb_queue_splice(const struct sk_buff_head *list,
2307 				    struct sk_buff_head *head)
2308 {
2309 	if (!skb_queue_empty(list)) {
2310 		__skb_queue_splice(list, (struct sk_buff *) head, head->next);
2311 		head->qlen += list->qlen;
2312 	}
2313 }
2314 
2315 /**
2316  *	skb_queue_splice_init - join two skb lists and reinitialise the emptied list
2317  *	@list: the new list to add
2318  *	@head: the place to add it in the first list
2319  *
2320  *	The list at @list is reinitialised
2321  */
skb_queue_splice_init(struct sk_buff_head * list,struct sk_buff_head * head)2322 static inline void skb_queue_splice_init(struct sk_buff_head *list,
2323 					 struct sk_buff_head *head)
2324 {
2325 	if (!skb_queue_empty(list)) {
2326 		__skb_queue_splice(list, (struct sk_buff *) head, head->next);
2327 		head->qlen += list->qlen;
2328 		__skb_queue_head_init(list);
2329 	}
2330 }
2331 
2332 /**
2333  *	skb_queue_splice_tail - join two skb lists, each list being a queue
2334  *	@list: the new list to add
2335  *	@head: the place to add it in the first list
2336  */
skb_queue_splice_tail(const struct sk_buff_head * list,struct sk_buff_head * head)2337 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
2338 					 struct sk_buff_head *head)
2339 {
2340 	if (!skb_queue_empty(list)) {
2341 		__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2342 		head->qlen += list->qlen;
2343 	}
2344 }
2345 
2346 /**
2347  *	skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
2348  *	@list: the new list to add
2349  *	@head: the place to add it in the first list
2350  *
2351  *	Each of the lists is a queue.
2352  *	The list at @list is reinitialised
2353  */
skb_queue_splice_tail_init(struct sk_buff_head * list,struct sk_buff_head * head)2354 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
2355 					      struct sk_buff_head *head)
2356 {
2357 	if (!skb_queue_empty(list)) {
2358 		__skb_queue_splice(list, head->prev, (struct sk_buff *) head);
2359 		head->qlen += list->qlen;
2360 		__skb_queue_head_init(list);
2361 	}
2362 }
2363 
2364 /**
2365  *	__skb_queue_after - queue a buffer at the list head
2366  *	@list: list to use
2367  *	@prev: place after this buffer
2368  *	@newsk: buffer to queue
2369  *
2370  *	Queue a buffer int the middle of a list. This function takes no locks
2371  *	and you must therefore hold required locks before calling it.
2372  *
2373  *	A buffer cannot be placed on two lists at the same time.
2374  */
__skb_queue_after(struct sk_buff_head * list,struct sk_buff * prev,struct sk_buff * newsk)2375 static inline void __skb_queue_after(struct sk_buff_head *list,
2376 				     struct sk_buff *prev,
2377 				     struct sk_buff *newsk)
2378 {
2379 	__skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list);
2380 }
2381 
2382 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2383 		struct sk_buff_head *list);
2384 
__skb_queue_before(struct sk_buff_head * list,struct sk_buff * next,struct sk_buff * newsk)2385 static inline void __skb_queue_before(struct sk_buff_head *list,
2386 				      struct sk_buff *next,
2387 				      struct sk_buff *newsk)
2388 {
2389 	__skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list);
2390 }
2391 
2392 /**
2393  *	__skb_queue_head - queue a buffer at the list head
2394  *	@list: list to use
2395  *	@newsk: buffer to queue
2396  *
2397  *	Queue a buffer at the start of a list. This function takes no locks
2398  *	and you must therefore hold required locks before calling it.
2399  *
2400  *	A buffer cannot be placed on two lists at the same time.
2401  */
__skb_queue_head(struct sk_buff_head * list,struct sk_buff * newsk)2402 static inline void __skb_queue_head(struct sk_buff_head *list,
2403 				    struct sk_buff *newsk)
2404 {
2405 	__skb_queue_after(list, (struct sk_buff *)list, newsk);
2406 }
2407 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
2408 
2409 /**
2410  *	__skb_queue_tail - queue a buffer at the list tail
2411  *	@list: list to use
2412  *	@newsk: buffer to queue
2413  *
2414  *	Queue a buffer at the end of a list. This function takes no locks
2415  *	and you must therefore hold required locks before calling it.
2416  *
2417  *	A buffer cannot be placed on two lists at the same time.
2418  */
__skb_queue_tail(struct sk_buff_head * list,struct sk_buff * newsk)2419 static inline void __skb_queue_tail(struct sk_buff_head *list,
2420 				   struct sk_buff *newsk)
2421 {
2422 	__skb_queue_before(list, (struct sk_buff *)list, newsk);
2423 }
2424 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
2425 
2426 /*
2427  * remove sk_buff from list. _Must_ be called atomically, and with
2428  * the list known..
2429  */
2430 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
__skb_unlink(struct sk_buff * skb,struct sk_buff_head * list)2431 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2432 {
2433 	struct sk_buff *next, *prev;
2434 
2435 	WRITE_ONCE(list->qlen, list->qlen - 1);
2436 	next	   = skb->next;
2437 	prev	   = skb->prev;
2438 	skb->next  = skb->prev = NULL;
2439 	WRITE_ONCE(next->prev, prev);
2440 	WRITE_ONCE(prev->next, next);
2441 }
2442 
2443 /**
2444  *	__skb_dequeue - remove from the head of the queue
2445  *	@list: list to dequeue from
2446  *
2447  *	Remove the head of the list. This function does not take any locks
2448  *	so must be used with appropriate locks held only. The head item is
2449  *	returned or %NULL if the list is empty.
2450  */
__skb_dequeue(struct sk_buff_head * list)2451 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2452 {
2453 	struct sk_buff *skb = skb_peek(list);
2454 	if (skb)
2455 		__skb_unlink(skb, list);
2456 	return skb;
2457 }
2458 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
2459 
2460 /**
2461  *	__skb_dequeue_tail - remove from the tail of the queue
2462  *	@list: list to dequeue from
2463  *
2464  *	Remove the tail of the list. This function does not take any locks
2465  *	so must be used with appropriate locks held only. The tail item is
2466  *	returned or %NULL if the list is empty.
2467  */
__skb_dequeue_tail(struct sk_buff_head * list)2468 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2469 {
2470 	struct sk_buff *skb = skb_peek_tail(list);
2471 	if (skb)
2472 		__skb_unlink(skb, list);
2473 	return skb;
2474 }
2475 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
2476 
2477 
skb_is_nonlinear(const struct sk_buff * skb)2478 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2479 {
2480 	return skb->data_len;
2481 }
2482 
skb_headlen(const struct sk_buff * skb)2483 static inline unsigned int skb_headlen(const struct sk_buff *skb)
2484 {
2485 	return skb->len - skb->data_len;
2486 }
2487 
__skb_pagelen(const struct sk_buff * skb)2488 static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2489 {
2490 	unsigned int i, len = 0;
2491 
2492 	for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2493 		len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2494 	return len;
2495 }
2496 
skb_pagelen(const struct sk_buff * skb)2497 static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2498 {
2499 	return skb_headlen(skb) + __skb_pagelen(skb);
2500 }
2501 
skb_frag_fill_netmem_desc(skb_frag_t * frag,netmem_ref netmem,int off,int size)2502 static inline void skb_frag_fill_netmem_desc(skb_frag_t *frag,
2503 					     netmem_ref netmem, int off,
2504 					     int size)
2505 {
2506 	frag->netmem = netmem;
2507 	frag->offset = off;
2508 	skb_frag_size_set(frag, size);
2509 }
2510 
skb_frag_fill_page_desc(skb_frag_t * frag,struct page * page,int off,int size)2511 static inline void skb_frag_fill_page_desc(skb_frag_t *frag,
2512 					   struct page *page,
2513 					   int off, int size)
2514 {
2515 	skb_frag_fill_netmem_desc(frag, page_to_netmem(page), off, size);
2516 }
2517 
__skb_fill_netmem_desc_noacc(struct skb_shared_info * shinfo,int i,netmem_ref netmem,int off,int size)2518 static inline void __skb_fill_netmem_desc_noacc(struct skb_shared_info *shinfo,
2519 						int i, netmem_ref netmem,
2520 						int off, int size)
2521 {
2522 	skb_frag_t *frag = &shinfo->frags[i];
2523 
2524 	skb_frag_fill_netmem_desc(frag, netmem, off, size);
2525 }
2526 
__skb_fill_page_desc_noacc(struct skb_shared_info * shinfo,int i,struct page * page,int off,int size)2527 static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo,
2528 					      int i, struct page *page,
2529 					      int off, int size)
2530 {
2531 	__skb_fill_netmem_desc_noacc(shinfo, i, page_to_netmem(page), off,
2532 				     size);
2533 }
2534 
2535 /**
2536  * skb_len_add - adds a number to len fields of skb
2537  * @skb: buffer to add len to
2538  * @delta: number of bytes to add
2539  */
skb_len_add(struct sk_buff * skb,int delta)2540 static inline void skb_len_add(struct sk_buff *skb, int delta)
2541 {
2542 	skb->len += delta;
2543 	skb->data_len += delta;
2544 	skb->truesize += delta;
2545 }
2546 
2547 /**
2548  * __skb_fill_netmem_desc - initialise a fragment in an skb
2549  * @skb: buffer containing fragment to be initialised
2550  * @i: fragment index to initialise
2551  * @netmem: the netmem to use for this fragment
2552  * @off: the offset to the data with @page
2553  * @size: the length of the data
2554  *
2555  * Initialises the @i'th fragment of @skb to point to &size bytes at
2556  * offset @off within @page.
2557  *
2558  * Does not take any additional reference on the fragment.
2559  */
__skb_fill_netmem_desc(struct sk_buff * skb,int i,netmem_ref netmem,int off,int size)2560 static inline void __skb_fill_netmem_desc(struct sk_buff *skb, int i,
2561 					  netmem_ref netmem, int off, int size)
2562 {
2563 	struct page *page;
2564 
2565 	__skb_fill_netmem_desc_noacc(skb_shinfo(skb), i, netmem, off, size);
2566 
2567 	if (netmem_is_net_iov(netmem)) {
2568 		skb->unreadable = true;
2569 		return;
2570 	}
2571 
2572 	page = netmem_to_page(netmem);
2573 
2574 	/* Propagate page pfmemalloc to the skb if we can. The problem is
2575 	 * that not all callers have unique ownership of the page but rely
2576 	 * on page_is_pfmemalloc doing the right thing(tm).
2577 	 */
2578 	page = compound_head(page);
2579 	if (page_is_pfmemalloc(page))
2580 		skb->pfmemalloc = true;
2581 }
2582 
__skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2583 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2584 					struct page *page, int off, int size)
2585 {
2586 	__skb_fill_netmem_desc(skb, i, page_to_netmem(page), off, size);
2587 }
2588 
skb_fill_netmem_desc(struct sk_buff * skb,int i,netmem_ref netmem,int off,int size)2589 static inline void skb_fill_netmem_desc(struct sk_buff *skb, int i,
2590 					netmem_ref netmem, int off, int size)
2591 {
2592 	__skb_fill_netmem_desc(skb, i, netmem, off, size);
2593 	skb_shinfo(skb)->nr_frags = i + 1;
2594 }
2595 
2596 /**
2597  * skb_fill_page_desc - initialise a paged fragment in an skb
2598  * @skb: buffer containing fragment to be initialised
2599  * @i: paged fragment index to initialise
2600  * @page: the page to use for this fragment
2601  * @off: the offset to the data with @page
2602  * @size: the length of the data
2603  *
2604  * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2605  * @skb to point to @size bytes at offset @off within @page. In
2606  * addition updates @skb such that @i is the last fragment.
2607  *
2608  * Does not take any additional reference on the fragment.
2609  */
skb_fill_page_desc(struct sk_buff * skb,int i,struct page * page,int off,int size)2610 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2611 				      struct page *page, int off, int size)
2612 {
2613 	skb_fill_netmem_desc(skb, i, page_to_netmem(page), off, size);
2614 }
2615 
2616 /**
2617  * skb_fill_page_desc_noacc - initialise a paged fragment in an skb
2618  * @skb: buffer containing fragment to be initialised
2619  * @i: paged fragment index to initialise
2620  * @page: the page to use for this fragment
2621  * @off: the offset to the data with @page
2622  * @size: the length of the data
2623  *
2624  * Variant of skb_fill_page_desc() which does not deal with
2625  * pfmemalloc, if page is not owned by us.
2626  */
skb_fill_page_desc_noacc(struct sk_buff * skb,int i,struct page * page,int off,int size)2627 static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i,
2628 					    struct page *page, int off,
2629 					    int size)
2630 {
2631 	struct skb_shared_info *shinfo = skb_shinfo(skb);
2632 
2633 	__skb_fill_page_desc_noacc(shinfo, i, page, off, size);
2634 	shinfo->nr_frags = i + 1;
2635 }
2636 
2637 void skb_add_rx_frag_netmem(struct sk_buff *skb, int i, netmem_ref netmem,
2638 			    int off, int size, unsigned int truesize);
2639 
skb_add_rx_frag(struct sk_buff * skb,int i,struct page * page,int off,int size,unsigned int truesize)2640 static inline void skb_add_rx_frag(struct sk_buff *skb, int i,
2641 				   struct page *page, int off, int size,
2642 				   unsigned int truesize)
2643 {
2644 	skb_add_rx_frag_netmem(skb, i, page_to_netmem(page), off, size,
2645 			       truesize);
2646 }
2647 
2648 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2649 			  unsigned int truesize);
2650 
2651 #define SKB_LINEAR_ASSERT(skb)  BUG_ON(skb_is_nonlinear(skb))
2652 
2653 #ifdef NET_SKBUFF_DATA_USES_OFFSET
skb_tail_pointer(const struct sk_buff * skb)2654 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2655 {
2656 	return skb->head + skb->tail;
2657 }
2658 
skb_reset_tail_pointer(struct sk_buff * skb)2659 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2660 {
2661 	skb->tail = skb->data - skb->head;
2662 }
2663 
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2664 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2665 {
2666 	skb_reset_tail_pointer(skb);
2667 	skb->tail += offset;
2668 }
2669 
2670 #else /* NET_SKBUFF_DATA_USES_OFFSET */
skb_tail_pointer(const struct sk_buff * skb)2671 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2672 {
2673 	return skb->tail;
2674 }
2675 
skb_reset_tail_pointer(struct sk_buff * skb)2676 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2677 {
2678 	skb->tail = skb->data;
2679 }
2680 
skb_set_tail_pointer(struct sk_buff * skb,const int offset)2681 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2682 {
2683 	skb->tail = skb->data + offset;
2684 }
2685 
2686 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
2687 
skb_assert_len(struct sk_buff * skb)2688 static inline void skb_assert_len(struct sk_buff *skb)
2689 {
2690 #ifdef CONFIG_DEBUG_NET
2691 	if (WARN_ONCE(!skb->len, "%s\n", __func__))
2692 		DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false);
2693 #endif /* CONFIG_DEBUG_NET */
2694 }
2695 
2696 #if defined(CONFIG_FAIL_SKB_REALLOC)
2697 void skb_might_realloc(struct sk_buff *skb);
2698 #else
skb_might_realloc(struct sk_buff * skb)2699 static inline void skb_might_realloc(struct sk_buff *skb) {}
2700 #endif
2701 
2702 /*
2703  *	Add data to an sk_buff
2704  */
2705 void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2706 void *skb_put(struct sk_buff *skb, unsigned int len);
__skb_put(struct sk_buff * skb,unsigned int len)2707 static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2708 {
2709 	void *tmp = skb_tail_pointer(skb);
2710 	SKB_LINEAR_ASSERT(skb);
2711 	skb->tail += len;
2712 	skb->len  += len;
2713 	return tmp;
2714 }
2715 
__skb_put_zero(struct sk_buff * skb,unsigned int len)2716 static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2717 {
2718 	void *tmp = __skb_put(skb, len);
2719 
2720 	memset(tmp, 0, len);
2721 	return tmp;
2722 }
2723 
__skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2724 static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2725 				   unsigned int len)
2726 {
2727 	void *tmp = __skb_put(skb, len);
2728 
2729 	memcpy(tmp, data, len);
2730 	return tmp;
2731 }
2732 
__skb_put_u8(struct sk_buff * skb,u8 val)2733 static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2734 {
2735 	*(u8 *)__skb_put(skb, 1) = val;
2736 }
2737 
skb_put_zero(struct sk_buff * skb,unsigned int len)2738 static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2739 {
2740 	void *tmp = skb_put(skb, len);
2741 
2742 	memset(tmp, 0, len);
2743 
2744 	return tmp;
2745 }
2746 
skb_put_data(struct sk_buff * skb,const void * data,unsigned int len)2747 static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2748 				 unsigned int len)
2749 {
2750 	void *tmp = skb_put(skb, len);
2751 
2752 	memcpy(tmp, data, len);
2753 
2754 	return tmp;
2755 }
2756 
skb_put_u8(struct sk_buff * skb,u8 val)2757 static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2758 {
2759 	*(u8 *)skb_put(skb, 1) = val;
2760 }
2761 
2762 void *skb_push(struct sk_buff *skb, unsigned int len);
__skb_push(struct sk_buff * skb,unsigned int len)2763 static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2764 {
2765 	DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
2766 
2767 	skb->data -= len;
2768 	skb->len  += len;
2769 	return skb->data;
2770 }
2771 
2772 void *skb_pull(struct sk_buff *skb, unsigned int len);
__skb_pull(struct sk_buff * skb,unsigned int len)2773 static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2774 {
2775 	DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
2776 
2777 	skb->len -= len;
2778 	if (unlikely(skb->len < skb->data_len)) {
2779 #if defined(CONFIG_DEBUG_NET)
2780 		skb->len += len;
2781 		pr_err("__skb_pull(len=%u)\n", len);
2782 		skb_dump(KERN_ERR, skb, false);
2783 #endif
2784 		BUG();
2785 	}
2786 	return skb->data += len;
2787 }
2788 
skb_pull_inline(struct sk_buff * skb,unsigned int len)2789 static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2790 {
2791 	return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2792 }
2793 
2794 void *skb_pull_data(struct sk_buff *skb, size_t len);
2795 
2796 void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2797 
2798 static inline enum skb_drop_reason
pskb_may_pull_reason(struct sk_buff * skb,unsigned int len)2799 pskb_may_pull_reason(struct sk_buff *skb, unsigned int len)
2800 {
2801 	DEBUG_NET_WARN_ON_ONCE(len > INT_MAX);
2802 	skb_might_realloc(skb);
2803 
2804 	if (likely(len <= skb_headlen(skb)))
2805 		return SKB_NOT_DROPPED_YET;
2806 
2807 	if (unlikely(len > skb->len))
2808 		return SKB_DROP_REASON_PKT_TOO_SMALL;
2809 
2810 	if (unlikely(!__pskb_pull_tail(skb, len - skb_headlen(skb))))
2811 		return SKB_DROP_REASON_NOMEM;
2812 
2813 	return SKB_NOT_DROPPED_YET;
2814 }
2815 
pskb_may_pull(struct sk_buff * skb,unsigned int len)2816 static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
2817 {
2818 	return pskb_may_pull_reason(skb, len) == SKB_NOT_DROPPED_YET;
2819 }
2820 
pskb_pull(struct sk_buff * skb,unsigned int len)2821 static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2822 {
2823 	if (!pskb_may_pull(skb, len))
2824 		return NULL;
2825 
2826 	skb->len -= len;
2827 	return skb->data += len;
2828 }
2829 
2830 void skb_condense(struct sk_buff *skb);
2831 
2832 /**
2833  *	skb_headroom - bytes at buffer head
2834  *	@skb: buffer to check
2835  *
2836  *	Return the number of bytes of free space at the head of an &sk_buff.
2837  */
skb_headroom(const struct sk_buff * skb)2838 static inline unsigned int skb_headroom(const struct sk_buff *skb)
2839 {
2840 	return skb->data - skb->head;
2841 }
2842 
2843 /**
2844  *	skb_tailroom - bytes at buffer end
2845  *	@skb: buffer to check
2846  *
2847  *	Return the number of bytes of free space at the tail of an sk_buff
2848  */
skb_tailroom(const struct sk_buff * skb)2849 static inline int skb_tailroom(const struct sk_buff *skb)
2850 {
2851 	return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2852 }
2853 
2854 /**
2855  *	skb_availroom - bytes at buffer end
2856  *	@skb: buffer to check
2857  *
2858  *	Return the number of bytes of free space at the tail of an sk_buff
2859  *	allocated by sk_stream_alloc()
2860  */
skb_availroom(const struct sk_buff * skb)2861 static inline int skb_availroom(const struct sk_buff *skb)
2862 {
2863 	if (skb_is_nonlinear(skb))
2864 		return 0;
2865 
2866 	return skb->end - skb->tail - skb->reserved_tailroom;
2867 }
2868 
2869 /**
2870  *	skb_reserve - adjust headroom
2871  *	@skb: buffer to alter
2872  *	@len: bytes to move
2873  *
2874  *	Increase the headroom of an empty &sk_buff by reducing the tail
2875  *	room. This is only allowed for an empty buffer.
2876  */
skb_reserve(struct sk_buff * skb,int len)2877 static inline void skb_reserve(struct sk_buff *skb, int len)
2878 {
2879 	skb->data += len;
2880 	skb->tail += len;
2881 }
2882 
2883 /**
2884  *	skb_tailroom_reserve - adjust reserved_tailroom
2885  *	@skb: buffer to alter
2886  *	@mtu: maximum amount of headlen permitted
2887  *	@needed_tailroom: minimum amount of reserved_tailroom
2888  *
2889  *	Set reserved_tailroom so that headlen can be as large as possible but
2890  *	not larger than mtu and tailroom cannot be smaller than
2891  *	needed_tailroom.
2892  *	The required headroom should already have been reserved before using
2893  *	this function.
2894  */
skb_tailroom_reserve(struct sk_buff * skb,unsigned int mtu,unsigned int needed_tailroom)2895 static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2896 					unsigned int needed_tailroom)
2897 {
2898 	SKB_LINEAR_ASSERT(skb);
2899 	if (mtu < skb_tailroom(skb) - needed_tailroom)
2900 		/* use at most mtu */
2901 		skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2902 	else
2903 		/* use up to all available space */
2904 		skb->reserved_tailroom = needed_tailroom;
2905 }
2906 
2907 #define ENCAP_TYPE_ETHER	0
2908 #define ENCAP_TYPE_IPPROTO	1
2909 
skb_set_inner_protocol(struct sk_buff * skb,__be16 protocol)2910 static inline void skb_set_inner_protocol(struct sk_buff *skb,
2911 					  __be16 protocol)
2912 {
2913 	skb->inner_protocol = protocol;
2914 	skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2915 }
2916 
skb_set_inner_ipproto(struct sk_buff * skb,__u8 ipproto)2917 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2918 					 __u8 ipproto)
2919 {
2920 	skb->inner_ipproto = ipproto;
2921 	skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2922 }
2923 
skb_reset_inner_headers(struct sk_buff * skb)2924 static inline void skb_reset_inner_headers(struct sk_buff *skb)
2925 {
2926 	skb->inner_mac_header = skb->mac_header;
2927 	skb->inner_network_header = skb->network_header;
2928 	skb->inner_transport_header = skb->transport_header;
2929 }
2930 
skb_mac_header_was_set(const struct sk_buff * skb)2931 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2932 {
2933 	return skb->mac_header != (typeof(skb->mac_header))~0U;
2934 }
2935 
skb_reset_mac_len(struct sk_buff * skb)2936 static inline void skb_reset_mac_len(struct sk_buff *skb)
2937 {
2938 	if (!skb_mac_header_was_set(skb)) {
2939 		DEBUG_NET_WARN_ON_ONCE(1);
2940 		skb->mac_len = 0;
2941 	} else {
2942 		skb->mac_len = skb->network_header - skb->mac_header;
2943 	}
2944 }
2945 
skb_inner_transport_header(const struct sk_buff * skb)2946 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2947 							*skb)
2948 {
2949 	return skb->head + skb->inner_transport_header;
2950 }
2951 
skb_inner_transport_offset(const struct sk_buff * skb)2952 static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2953 {
2954 	return skb_inner_transport_header(skb) - skb->data;
2955 }
2956 
skb_reset_inner_transport_header(struct sk_buff * skb)2957 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2958 {
2959 	long offset = skb->data - skb->head;
2960 
2961 	DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->inner_transport_header))offset);
2962 	skb->inner_transport_header = offset;
2963 }
2964 
skb_set_inner_transport_header(struct sk_buff * skb,const int offset)2965 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2966 						   const int offset)
2967 {
2968 	skb_reset_inner_transport_header(skb);
2969 	skb->inner_transport_header += offset;
2970 }
2971 
skb_inner_network_header(const struct sk_buff * skb)2972 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2973 {
2974 	return skb->head + skb->inner_network_header;
2975 }
2976 
skb_reset_inner_network_header(struct sk_buff * skb)2977 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2978 {
2979 	long offset = skb->data - skb->head;
2980 
2981 	DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->inner_network_header))offset);
2982 	skb->inner_network_header = offset;
2983 }
2984 
skb_set_inner_network_header(struct sk_buff * skb,const int offset)2985 static inline void skb_set_inner_network_header(struct sk_buff *skb,
2986 						const int offset)
2987 {
2988 	skb_reset_inner_network_header(skb);
2989 	skb->inner_network_header += offset;
2990 }
2991 
skb_inner_network_header_was_set(const struct sk_buff * skb)2992 static inline bool skb_inner_network_header_was_set(const struct sk_buff *skb)
2993 {
2994 	return skb->inner_network_header > 0;
2995 }
2996 
skb_inner_mac_header(const struct sk_buff * skb)2997 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2998 {
2999 	return skb->head + skb->inner_mac_header;
3000 }
3001 
skb_reset_inner_mac_header(struct sk_buff * skb)3002 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
3003 {
3004 	long offset = skb->data - skb->head;
3005 
3006 	DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->inner_mac_header))offset);
3007 	skb->inner_mac_header = offset;
3008 }
3009 
skb_set_inner_mac_header(struct sk_buff * skb,const int offset)3010 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
3011 					    const int offset)
3012 {
3013 	skb_reset_inner_mac_header(skb);
3014 	skb->inner_mac_header += offset;
3015 }
skb_transport_header_was_set(const struct sk_buff * skb)3016 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
3017 {
3018 	return skb->transport_header != (typeof(skb->transport_header))~0U;
3019 }
3020 
skb_transport_header(const struct sk_buff * skb)3021 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
3022 {
3023 	DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
3024 	return skb->head + skb->transport_header;
3025 }
3026 
skb_reset_transport_header(struct sk_buff * skb)3027 static inline void skb_reset_transport_header(struct sk_buff *skb)
3028 {
3029 	long offset = skb->data - skb->head;
3030 
3031 	DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->transport_header))offset);
3032 	skb->transport_header = offset;
3033 }
3034 
skb_set_transport_header(struct sk_buff * skb,const int offset)3035 static inline void skb_set_transport_header(struct sk_buff *skb,
3036 					    const int offset)
3037 {
3038 	skb_reset_transport_header(skb);
3039 	skb->transport_header += offset;
3040 }
3041 
skb_network_header(const struct sk_buff * skb)3042 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
3043 {
3044 	return skb->head + skb->network_header;
3045 }
3046 
skb_reset_network_header(struct sk_buff * skb)3047 static inline void skb_reset_network_header(struct sk_buff *skb)
3048 {
3049 	long offset = skb->data - skb->head;
3050 
3051 	DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->network_header))offset);
3052 	skb->network_header = offset;
3053 }
3054 
skb_set_network_header(struct sk_buff * skb,const int offset)3055 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
3056 {
3057 	skb_reset_network_header(skb);
3058 	skb->network_header += offset;
3059 }
3060 
skb_mac_header(const struct sk_buff * skb)3061 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
3062 {
3063 	DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
3064 	return skb->head + skb->mac_header;
3065 }
3066 
skb_mac_offset(const struct sk_buff * skb)3067 static inline int skb_mac_offset(const struct sk_buff *skb)
3068 {
3069 	return skb_mac_header(skb) - skb->data;
3070 }
3071 
skb_mac_header_len(const struct sk_buff * skb)3072 static inline u32 skb_mac_header_len(const struct sk_buff *skb)
3073 {
3074 	DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb));
3075 	return skb->network_header - skb->mac_header;
3076 }
3077 
skb_unset_mac_header(struct sk_buff * skb)3078 static inline void skb_unset_mac_header(struct sk_buff *skb)
3079 {
3080 	skb->mac_header = (typeof(skb->mac_header))~0U;
3081 }
3082 
skb_reset_mac_header(struct sk_buff * skb)3083 static inline void skb_reset_mac_header(struct sk_buff *skb)
3084 {
3085 	long offset = skb->data - skb->head;
3086 
3087 	DEBUG_NET_WARN_ON_ONCE(offset != (typeof(skb->mac_header))offset);
3088 	skb->mac_header = offset;
3089 }
3090 
skb_set_mac_header(struct sk_buff * skb,const int offset)3091 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
3092 {
3093 	skb_reset_mac_header(skb);
3094 	skb->mac_header += offset;
3095 }
3096 
skb_pop_mac_header(struct sk_buff * skb)3097 static inline void skb_pop_mac_header(struct sk_buff *skb)
3098 {
3099 	skb->mac_header = skb->network_header;
3100 }
3101 
skb_probe_transport_header(struct sk_buff * skb)3102 static inline void skb_probe_transport_header(struct sk_buff *skb)
3103 {
3104 	struct flow_keys_basic keys;
3105 
3106 	if (skb_transport_header_was_set(skb))
3107 		return;
3108 
3109 	if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
3110 					     NULL, 0, 0, 0, 0))
3111 		skb_set_transport_header(skb, keys.control.thoff);
3112 }
3113 
skb_mac_header_rebuild(struct sk_buff * skb)3114 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
3115 {
3116 	if (skb_mac_header_was_set(skb)) {
3117 		const unsigned char *old_mac = skb_mac_header(skb);
3118 
3119 		skb_set_mac_header(skb, -skb->mac_len);
3120 		memmove(skb_mac_header(skb), old_mac, skb->mac_len);
3121 	}
3122 }
3123 
3124 /* Move the full mac header up to current network_header.
3125  * Leaves skb->data pointing at offset skb->mac_len into the mac_header.
3126  * Must be provided the complete mac header length.
3127  */
skb_mac_header_rebuild_full(struct sk_buff * skb,u32 full_mac_len)3128 static inline void skb_mac_header_rebuild_full(struct sk_buff *skb, u32 full_mac_len)
3129 {
3130 	if (skb_mac_header_was_set(skb)) {
3131 		const unsigned char *old_mac = skb_mac_header(skb);
3132 
3133 		skb_set_mac_header(skb, -full_mac_len);
3134 		memmove(skb_mac_header(skb), old_mac, full_mac_len);
3135 		__skb_push(skb, full_mac_len - skb->mac_len);
3136 	}
3137 }
3138 
skb_checksum_start_offset(const struct sk_buff * skb)3139 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
3140 {
3141 	return skb->csum_start - skb_headroom(skb);
3142 }
3143 
skb_checksum_start(const struct sk_buff * skb)3144 static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
3145 {
3146 	return skb->head + skb->csum_start;
3147 }
3148 
skb_transport_offset(const struct sk_buff * skb)3149 static inline int skb_transport_offset(const struct sk_buff *skb)
3150 {
3151 	return skb_transport_header(skb) - skb->data;
3152 }
3153 
skb_network_header_len(const struct sk_buff * skb)3154 static inline u32 skb_network_header_len(const struct sk_buff *skb)
3155 {
3156 	DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb));
3157 	return skb->transport_header - skb->network_header;
3158 }
3159 
skb_inner_network_header_len(const struct sk_buff * skb)3160 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
3161 {
3162 	return skb->inner_transport_header - skb->inner_network_header;
3163 }
3164 
skb_network_offset(const struct sk_buff * skb)3165 static inline int skb_network_offset(const struct sk_buff *skb)
3166 {
3167 	return skb_network_header(skb) - skb->data;
3168 }
3169 
skb_inner_network_offset(const struct sk_buff * skb)3170 static inline int skb_inner_network_offset(const struct sk_buff *skb)
3171 {
3172 	return skb_inner_network_header(skb) - skb->data;
3173 }
3174 
3175 static inline enum skb_drop_reason
pskb_network_may_pull_reason(struct sk_buff * skb,unsigned int len)3176 pskb_network_may_pull_reason(struct sk_buff *skb, unsigned int len)
3177 {
3178 	return pskb_may_pull_reason(skb, skb_network_offset(skb) + len);
3179 }
3180 
pskb_network_may_pull(struct sk_buff * skb,unsigned int len)3181 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
3182 {
3183 	return pskb_network_may_pull_reason(skb, len) == SKB_NOT_DROPPED_YET;
3184 }
3185 
3186 /*
3187  * CPUs often take a performance hit when accessing unaligned memory
3188  * locations. The actual performance hit varies, it can be small if the
3189  * hardware handles it or large if we have to take an exception and fix it
3190  * in software.
3191  *
3192  * Since an ethernet header is 14 bytes network drivers often end up with
3193  * the IP header at an unaligned offset. The IP header can be aligned by
3194  * shifting the start of the packet by 2 bytes. Drivers should do this
3195  * with:
3196  *
3197  * skb_reserve(skb, NET_IP_ALIGN);
3198  *
3199  * The downside to this alignment of the IP header is that the DMA is now
3200  * unaligned. On some architectures the cost of an unaligned DMA is high
3201  * and this cost outweighs the gains made by aligning the IP header.
3202  *
3203  * Since this trade off varies between architectures, we allow NET_IP_ALIGN
3204  * to be overridden.
3205  */
3206 #ifndef NET_IP_ALIGN
3207 #define NET_IP_ALIGN	2
3208 #endif
3209 
3210 /*
3211  * The networking layer reserves some headroom in skb data (via
3212  * dev_alloc_skb). This is used to avoid having to reallocate skb data when
3213  * the header has to grow. In the default case, if the header has to grow
3214  * 32 bytes or less we avoid the reallocation.
3215  *
3216  * Unfortunately this headroom changes the DMA alignment of the resulting
3217  * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
3218  * on some architectures. An architecture can override this value,
3219  * perhaps setting it to a cacheline in size (since that will maintain
3220  * cacheline alignment of the DMA). It must be a power of 2.
3221  *
3222  * Various parts of the networking layer expect at least 32 bytes of
3223  * headroom, you should not reduce this.
3224  *
3225  * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
3226  * to reduce average number of cache lines per packet.
3227  * get_rps_cpu() for example only access one 64 bytes aligned block :
3228  * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
3229  */
3230 #ifndef NET_SKB_PAD
3231 #define NET_SKB_PAD	max(32, L1_CACHE_BYTES)
3232 #endif
3233 
3234 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
3235 
__skb_set_length(struct sk_buff * skb,unsigned int len)3236 static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
3237 {
3238 	if (WARN_ON(skb_is_nonlinear(skb)))
3239 		return;
3240 	skb->len = len;
3241 	skb_set_tail_pointer(skb, len);
3242 }
3243 
__skb_trim(struct sk_buff * skb,unsigned int len)3244 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
3245 {
3246 	__skb_set_length(skb, len);
3247 }
3248 
3249 void skb_trim(struct sk_buff *skb, unsigned int len);
3250 
__pskb_trim(struct sk_buff * skb,unsigned int len)3251 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
3252 {
3253 	if (skb->data_len)
3254 		return ___pskb_trim(skb, len);
3255 	__skb_trim(skb, len);
3256 	return 0;
3257 }
3258 
pskb_trim(struct sk_buff * skb,unsigned int len)3259 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
3260 {
3261 	skb_might_realloc(skb);
3262 	return (len < skb->len) ? __pskb_trim(skb, len) : 0;
3263 }
3264 
3265 /**
3266  *	pskb_trim_unique - remove end from a paged unique (not cloned) buffer
3267  *	@skb: buffer to alter
3268  *	@len: new length
3269  *
3270  *	This is identical to pskb_trim except that the caller knows that
3271  *	the skb is not cloned so we should never get an error due to out-
3272  *	of-memory.
3273  */
pskb_trim_unique(struct sk_buff * skb,unsigned int len)3274 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
3275 {
3276 	int err = pskb_trim(skb, len);
3277 	BUG_ON(err);
3278 }
3279 
__skb_grow(struct sk_buff * skb,unsigned int len)3280 static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
3281 {
3282 	unsigned int diff = len - skb->len;
3283 
3284 	if (skb_tailroom(skb) < diff) {
3285 		int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
3286 					   GFP_ATOMIC);
3287 		if (ret)
3288 			return ret;
3289 	}
3290 	__skb_set_length(skb, len);
3291 	return 0;
3292 }
3293 
3294 /**
3295  *	skb_orphan - orphan a buffer
3296  *	@skb: buffer to orphan
3297  *
3298  *	If a buffer currently has an owner then we call the owner's
3299  *	destructor function and make the @skb unowned. The buffer continues
3300  *	to exist but is no longer charged to its former owner.
3301  */
skb_orphan(struct sk_buff * skb)3302 static inline void skb_orphan(struct sk_buff *skb)
3303 {
3304 	if (skb->destructor) {
3305 		skb->destructor(skb);
3306 		skb->destructor = NULL;
3307 		skb->sk		= NULL;
3308 	} else {
3309 		BUG_ON(skb->sk);
3310 	}
3311 }
3312 
3313 /**
3314  *	skb_orphan_frags - orphan the frags contained in a buffer
3315  *	@skb: buffer to orphan frags from
3316  *	@gfp_mask: allocation mask for replacement pages
3317  *
3318  *	For each frag in the SKB which needs a destructor (i.e. has an
3319  *	owner) create a copy of that frag and release the original
3320  *	page by calling the destructor.
3321  */
skb_orphan_frags(struct sk_buff * skb,gfp_t gfp_mask)3322 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
3323 {
3324 	if (likely(!skb_zcopy(skb)))
3325 		return 0;
3326 	if (skb_shinfo(skb)->flags & SKBFL_DONT_ORPHAN)
3327 		return 0;
3328 	return skb_copy_ubufs(skb, gfp_mask);
3329 }
3330 
3331 /* 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)3332 static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
3333 {
3334 	if (likely(!skb_zcopy(skb)))
3335 		return 0;
3336 	return skb_copy_ubufs(skb, gfp_mask);
3337 }
3338 
3339 /**
3340  *	__skb_queue_purge_reason - empty a list
3341  *	@list: list to empty
3342  *	@reason: drop reason
3343  *
3344  *	Delete all buffers on an &sk_buff list. Each buffer is removed from
3345  *	the list and one reference dropped. This function does not take the
3346  *	list lock and the caller must hold the relevant locks to use it.
3347  */
__skb_queue_purge_reason(struct sk_buff_head * list,enum skb_drop_reason reason)3348 static inline void __skb_queue_purge_reason(struct sk_buff_head *list,
3349 					    enum skb_drop_reason reason)
3350 {
3351 	struct sk_buff *skb;
3352 
3353 	while ((skb = __skb_dequeue(list)) != NULL)
3354 		kfree_skb_reason(skb, reason);
3355 }
3356 
__skb_queue_purge(struct sk_buff_head * list)3357 static inline void __skb_queue_purge(struct sk_buff_head *list)
3358 {
3359 	__skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE);
3360 }
3361 
3362 void skb_queue_purge_reason(struct sk_buff_head *list,
3363 			    enum skb_drop_reason reason);
3364 
skb_queue_purge(struct sk_buff_head * list)3365 static inline void skb_queue_purge(struct sk_buff_head *list)
3366 {
3367 	skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE);
3368 }
3369 
3370 unsigned int skb_rbtree_purge(struct rb_root *root);
3371 void skb_errqueue_purge(struct sk_buff_head *list);
3372 
3373 void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3374 
3375 /**
3376  * netdev_alloc_frag - allocate a page fragment
3377  * @fragsz: fragment size
3378  *
3379  * Allocates a frag from a page for receive buffer.
3380  * Uses GFP_ATOMIC allocations.
3381  */
netdev_alloc_frag(unsigned int fragsz)3382 static inline void *netdev_alloc_frag(unsigned int fragsz)
3383 {
3384 	return __netdev_alloc_frag_align(fragsz, ~0u);
3385 }
3386 
netdev_alloc_frag_align(unsigned int fragsz,unsigned int align)3387 static inline void *netdev_alloc_frag_align(unsigned int fragsz,
3388 					    unsigned int align)
3389 {
3390 	WARN_ON_ONCE(!is_power_of_2(align));
3391 	return __netdev_alloc_frag_align(fragsz, -align);
3392 }
3393 
3394 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
3395 				   gfp_t gfp_mask);
3396 
3397 /**
3398  *	netdev_alloc_skb - allocate an skbuff for rx on a specific device
3399  *	@dev: network device to receive on
3400  *	@length: length to allocate
3401  *
3402  *	Allocate a new &sk_buff and assign it a usage count of one. The
3403  *	buffer has unspecified headroom built in. Users should allocate
3404  *	the headroom they think they need without accounting for the
3405  *	built in space. The built in space is used for optimisations.
3406  *
3407  *	%NULL is returned if there is no free memory. Although this function
3408  *	allocates memory it can be called from an interrupt.
3409  */
netdev_alloc_skb(struct net_device * dev,unsigned int length)3410 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
3411 					       unsigned int length)
3412 {
3413 	return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
3414 }
3415 
3416 /* legacy helper around __netdev_alloc_skb() */
__dev_alloc_skb(unsigned int length,gfp_t gfp_mask)3417 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
3418 					      gfp_t gfp_mask)
3419 {
3420 	return __netdev_alloc_skb(NULL, length, gfp_mask);
3421 }
3422 
3423 /* legacy helper around netdev_alloc_skb() */
dev_alloc_skb(unsigned int length)3424 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
3425 {
3426 	return netdev_alloc_skb(NULL, length);
3427 }
3428 
3429 
__netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length,gfp_t gfp)3430 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
3431 		unsigned int length, gfp_t gfp)
3432 {
3433 	struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
3434 
3435 	if (NET_IP_ALIGN && skb)
3436 		skb_reserve(skb, NET_IP_ALIGN);
3437 	return skb;
3438 }
3439 
netdev_alloc_skb_ip_align(struct net_device * dev,unsigned int length)3440 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
3441 		unsigned int length)
3442 {
3443 	return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
3444 }
3445 
skb_free_frag(void * addr)3446 static inline void skb_free_frag(void *addr)
3447 {
3448 	page_frag_free(addr);
3449 }
3450 
3451 void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask);
3452 
napi_alloc_frag(unsigned int fragsz)3453 static inline void *napi_alloc_frag(unsigned int fragsz)
3454 {
3455 	return __napi_alloc_frag_align(fragsz, ~0u);
3456 }
3457 
napi_alloc_frag_align(unsigned int fragsz,unsigned int align)3458 static inline void *napi_alloc_frag_align(unsigned int fragsz,
3459 					  unsigned int align)
3460 {
3461 	WARN_ON_ONCE(!is_power_of_2(align));
3462 	return __napi_alloc_frag_align(fragsz, -align);
3463 }
3464 
3465 struct sk_buff *napi_alloc_skb(struct napi_struct *napi, unsigned int length);
3466 void napi_consume_skb(struct sk_buff *skb, int budget);
3467 
3468 void napi_skb_free_stolen_head(struct sk_buff *skb);
3469 void __napi_kfree_skb(struct sk_buff *skb, enum skb_drop_reason reason);
3470 
3471 /**
3472  * __dev_alloc_pages - allocate page for network Rx
3473  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3474  * @order: size of the allocation
3475  *
3476  * Allocate a new page.
3477  *
3478  * %NULL is returned if there is no free memory.
3479 */
__dev_alloc_pages_noprof(gfp_t gfp_mask,unsigned int order)3480 static inline struct page *__dev_alloc_pages_noprof(gfp_t gfp_mask,
3481 					     unsigned int order)
3482 {
3483 	/* This piece of code contains several assumptions.
3484 	 * 1.  This is for device Rx, therefore a cold page is preferred.
3485 	 * 2.  The expectation is the user wants a compound page.
3486 	 * 3.  If requesting a order 0 page it will not be compound
3487 	 *     due to the check to see if order has a value in prep_new_page
3488 	 * 4.  __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
3489 	 *     code in gfp_to_alloc_flags that should be enforcing this.
3490 	 */
3491 	gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
3492 
3493 	return alloc_pages_node_noprof(NUMA_NO_NODE, gfp_mask, order);
3494 }
3495 #define __dev_alloc_pages(...)	alloc_hooks(__dev_alloc_pages_noprof(__VA_ARGS__))
3496 
3497 /*
3498  * This specialized allocator has to be a macro for its allocations to be
3499  * accounted separately (to have a separate alloc_tag).
3500  */
3501 #define dev_alloc_pages(_order) __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, _order)
3502 
3503 /**
3504  * __dev_alloc_page - allocate a page for network Rx
3505  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
3506  *
3507  * Allocate a new page.
3508  *
3509  * %NULL is returned if there is no free memory.
3510  */
__dev_alloc_page_noprof(gfp_t gfp_mask)3511 static inline struct page *__dev_alloc_page_noprof(gfp_t gfp_mask)
3512 {
3513 	return __dev_alloc_pages_noprof(gfp_mask, 0);
3514 }
3515 #define __dev_alloc_page(...)	alloc_hooks(__dev_alloc_page_noprof(__VA_ARGS__))
3516 
3517 /*
3518  * This specialized allocator has to be a macro for its allocations to be
3519  * accounted separately (to have a separate alloc_tag).
3520  */
3521 #define dev_alloc_page()	dev_alloc_pages(0)
3522 
3523 /**
3524  * dev_page_is_reusable - check whether a page can be reused for network Rx
3525  * @page: the page to test
3526  *
3527  * A page shouldn't be considered for reusing/recycling if it was allocated
3528  * under memory pressure or at a distant memory node.
3529  *
3530  * Returns: false if this page should be returned to page allocator, true
3531  * otherwise.
3532  */
dev_page_is_reusable(const struct page * page)3533 static inline bool dev_page_is_reusable(const struct page *page)
3534 {
3535 	return likely(page_to_nid(page) == numa_mem_id() &&
3536 		      !page_is_pfmemalloc(page));
3537 }
3538 
3539 /**
3540  *	skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
3541  *	@page: The page that was allocated from skb_alloc_page
3542  *	@skb: The skb that may need pfmemalloc set
3543  */
skb_propagate_pfmemalloc(const struct page * page,struct sk_buff * skb)3544 static inline void skb_propagate_pfmemalloc(const struct page *page,
3545 					    struct sk_buff *skb)
3546 {
3547 	if (page_is_pfmemalloc(page))
3548 		skb->pfmemalloc = true;
3549 }
3550 
3551 /**
3552  * skb_frag_off() - Returns the offset of a skb fragment
3553  * @frag: the paged fragment
3554  */
skb_frag_off(const skb_frag_t * frag)3555 static inline unsigned int skb_frag_off(const skb_frag_t *frag)
3556 {
3557 	return frag->offset;
3558 }
3559 
3560 /**
3561  * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
3562  * @frag: skb fragment
3563  * @delta: value to add
3564  */
skb_frag_off_add(skb_frag_t * frag,int delta)3565 static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
3566 {
3567 	frag->offset += delta;
3568 }
3569 
3570 /**
3571  * skb_frag_off_set() - Sets the offset of a skb fragment
3572  * @frag: skb fragment
3573  * @offset: offset of fragment
3574  */
skb_frag_off_set(skb_frag_t * frag,unsigned int offset)3575 static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
3576 {
3577 	frag->offset = offset;
3578 }
3579 
3580 /**
3581  * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
3582  * @fragto: skb fragment where offset is set
3583  * @fragfrom: skb fragment offset is copied from
3584  */
skb_frag_off_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3585 static inline void skb_frag_off_copy(skb_frag_t *fragto,
3586 				     const skb_frag_t *fragfrom)
3587 {
3588 	fragto->offset = fragfrom->offset;
3589 }
3590 
3591 /* Return: true if the skb_frag contains a net_iov. */
skb_frag_is_net_iov(const skb_frag_t * frag)3592 static inline bool skb_frag_is_net_iov(const skb_frag_t *frag)
3593 {
3594 	return netmem_is_net_iov(frag->netmem);
3595 }
3596 
3597 /**
3598  * skb_frag_net_iov - retrieve the net_iov referred to by fragment
3599  * @frag: the fragment
3600  *
3601  * Return: the &struct net_iov associated with @frag. Returns NULL if this
3602  * frag has no associated net_iov.
3603  */
skb_frag_net_iov(const skb_frag_t * frag)3604 static inline struct net_iov *skb_frag_net_iov(const skb_frag_t *frag)
3605 {
3606 	if (!skb_frag_is_net_iov(frag))
3607 		return NULL;
3608 
3609 	return netmem_to_net_iov(frag->netmem);
3610 }
3611 
3612 /**
3613  * skb_frag_page - retrieve the page referred to by a paged fragment
3614  * @frag: the paged fragment
3615  *
3616  * Return: the &struct page associated with @frag. Returns NULL if this frag
3617  * has no associated page.
3618  */
skb_frag_page(const skb_frag_t * frag)3619 static inline struct page *skb_frag_page(const skb_frag_t *frag)
3620 {
3621 	if (skb_frag_is_net_iov(frag))
3622 		return NULL;
3623 
3624 	return netmem_to_page(frag->netmem);
3625 }
3626 
3627 /**
3628  * skb_frag_netmem - retrieve the netmem referred to by a fragment
3629  * @frag: the fragment
3630  *
3631  * Return: the &netmem_ref associated with @frag.
3632  */
skb_frag_netmem(const skb_frag_t * frag)3633 static inline netmem_ref skb_frag_netmem(const skb_frag_t *frag)
3634 {
3635 	return frag->netmem;
3636 }
3637 
3638 int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb,
3639 		    unsigned int headroom);
3640 int skb_cow_data_for_xdp(struct page_pool *pool, struct sk_buff **pskb,
3641 			 const struct bpf_prog *prog);
3642 
3643 /**
3644  * skb_frag_address - gets the address of the data contained in a paged fragment
3645  * @frag: the paged fragment buffer
3646  *
3647  * Returns: the address of the data within @frag. The page must already
3648  * be mapped.
3649  */
skb_frag_address(const skb_frag_t * frag)3650 static inline void *skb_frag_address(const skb_frag_t *frag)
3651 {
3652 	if (!skb_frag_page(frag))
3653 		return NULL;
3654 
3655 	return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
3656 }
3657 
3658 /**
3659  * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3660  * @frag: the paged fragment buffer
3661  *
3662  * Returns: the address of the data within @frag. Checks that the page
3663  * is mapped and returns %NULL otherwise.
3664  */
skb_frag_address_safe(const skb_frag_t * frag)3665 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3666 {
3667 	void *ptr = page_address(skb_frag_page(frag));
3668 	if (unlikely(!ptr))
3669 		return NULL;
3670 
3671 	return ptr + skb_frag_off(frag);
3672 }
3673 
3674 /**
3675  * skb_frag_page_copy() - sets the page in a fragment from another fragment
3676  * @fragto: skb fragment where page is set
3677  * @fragfrom: skb fragment page is copied from
3678  */
skb_frag_page_copy(skb_frag_t * fragto,const skb_frag_t * fragfrom)3679 static inline void skb_frag_page_copy(skb_frag_t *fragto,
3680 				      const skb_frag_t *fragfrom)
3681 {
3682 	fragto->netmem = fragfrom->netmem;
3683 }
3684 
3685 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3686 
3687 /**
3688  * __skb_frag_dma_map - maps a paged fragment via the DMA API
3689  * @dev: the device to map the fragment to
3690  * @frag: the paged fragment to map
3691  * @offset: the offset within the fragment (starting at the
3692  *          fragment's own offset)
3693  * @size: the number of bytes to map
3694  * @dir: the direction of the mapping (``PCI_DMA_*``)
3695  *
3696  * Maps the page associated with @frag to @device.
3697  */
__skb_frag_dma_map(struct device * dev,const skb_frag_t * frag,size_t offset,size_t size,enum dma_data_direction dir)3698 static inline dma_addr_t __skb_frag_dma_map(struct device *dev,
3699 					    const skb_frag_t *frag,
3700 					    size_t offset, size_t size,
3701 					    enum dma_data_direction dir)
3702 {
3703 	return dma_map_page(dev, skb_frag_page(frag),
3704 			    skb_frag_off(frag) + offset, size, dir);
3705 }
3706 
3707 #define skb_frag_dma_map(dev, frag, ...)				\
3708 	CONCATENATE(_skb_frag_dma_map,					\
3709 		    COUNT_ARGS(__VA_ARGS__))(dev, frag, ##__VA_ARGS__)
3710 
3711 #define __skb_frag_dma_map1(dev, frag, offset, uf, uo) ({		\
3712 	const skb_frag_t *uf = (frag);					\
3713 	size_t uo = (offset);						\
3714 									\
3715 	__skb_frag_dma_map(dev, uf, uo, skb_frag_size(uf) - uo,		\
3716 			   DMA_TO_DEVICE);				\
3717 })
3718 #define _skb_frag_dma_map1(dev, frag, offset)				\
3719 	__skb_frag_dma_map1(dev, frag, offset, __UNIQUE_ID(frag_),	\
3720 			    __UNIQUE_ID(offset_))
3721 #define _skb_frag_dma_map0(dev, frag)					\
3722 	_skb_frag_dma_map1(dev, frag, 0)
3723 #define _skb_frag_dma_map2(dev, frag, offset, size)			\
3724 	__skb_frag_dma_map(dev, frag, offset, size, DMA_TO_DEVICE)
3725 #define _skb_frag_dma_map3(dev, frag, offset, size, dir)		\
3726 	__skb_frag_dma_map(dev, frag, offset, size, dir)
3727 
pskb_copy(struct sk_buff * skb,gfp_t gfp_mask)3728 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3729 					gfp_t gfp_mask)
3730 {
3731 	return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3732 }
3733 
3734 
pskb_copy_for_clone(struct sk_buff * skb,gfp_t gfp_mask)3735 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3736 						  gfp_t gfp_mask)
3737 {
3738 	return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3739 }
3740 
3741 
3742 /**
3743  *	skb_clone_writable - is the header of a clone writable
3744  *	@skb: buffer to check
3745  *	@len: length up to which to write
3746  *
3747  *	Returns true if modifying the header part of the cloned buffer
3748  *	does not requires the data to be copied.
3749  */
skb_clone_writable(const struct sk_buff * skb,unsigned int len)3750 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3751 {
3752 	return !skb_header_cloned(skb) &&
3753 	       skb_headroom(skb) + len <= skb->hdr_len;
3754 }
3755 
skb_try_make_writable(struct sk_buff * skb,unsigned int write_len)3756 static inline int skb_try_make_writable(struct sk_buff *skb,
3757 					unsigned int write_len)
3758 {
3759 	return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3760 	       pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3761 }
3762 
__skb_cow(struct sk_buff * skb,unsigned int headroom,int cloned)3763 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3764 			    int cloned)
3765 {
3766 	int delta = 0;
3767 
3768 	if (headroom > skb_headroom(skb))
3769 		delta = headroom - skb_headroom(skb);
3770 
3771 	if (delta || cloned)
3772 		return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3773 					GFP_ATOMIC);
3774 	return 0;
3775 }
3776 
3777 /**
3778  *	skb_cow - copy header of skb when it is required
3779  *	@skb: buffer to cow
3780  *	@headroom: needed headroom
3781  *
3782  *	If the skb passed lacks sufficient headroom or its data part
3783  *	is shared, data is reallocated. If reallocation fails, an error
3784  *	is returned and original skb is not changed.
3785  *
3786  *	The result is skb with writable area skb->head...skb->tail
3787  *	and at least @headroom of space at head.
3788  */
skb_cow(struct sk_buff * skb,unsigned int headroom)3789 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3790 {
3791 	return __skb_cow(skb, headroom, skb_cloned(skb));
3792 }
3793 
3794 /**
3795  *	skb_cow_head - skb_cow but only making the head writable
3796  *	@skb: buffer to cow
3797  *	@headroom: needed headroom
3798  *
3799  *	This function is identical to skb_cow except that we replace the
3800  *	skb_cloned check by skb_header_cloned.  It should be used when
3801  *	you only need to push on some header and do not need to modify
3802  *	the data.
3803  */
skb_cow_head(struct sk_buff * skb,unsigned int headroom)3804 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3805 {
3806 	return __skb_cow(skb, headroom, skb_header_cloned(skb));
3807 }
3808 
3809 /**
3810  *	skb_padto	- pad an skbuff up to a minimal size
3811  *	@skb: buffer to pad
3812  *	@len: minimal length
3813  *
3814  *	Pads up a buffer to ensure the trailing bytes exist and are
3815  *	blanked. If the buffer already contains sufficient data it
3816  *	is untouched. Otherwise it is extended. Returns zero on
3817  *	success. The skb is freed on error.
3818  */
skb_padto(struct sk_buff * skb,unsigned int len)3819 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3820 {
3821 	unsigned int size = skb->len;
3822 	if (likely(size >= len))
3823 		return 0;
3824 	return skb_pad(skb, len - size);
3825 }
3826 
3827 /**
3828  *	__skb_put_padto - increase size and pad an skbuff up to a minimal size
3829  *	@skb: buffer to pad
3830  *	@len: minimal length
3831  *	@free_on_error: free buffer on error
3832  *
3833  *	Pads up a buffer to ensure the trailing bytes exist and are
3834  *	blanked. If the buffer already contains sufficient data it
3835  *	is untouched. Otherwise it is extended. Returns zero on
3836  *	success. The skb is freed on error if @free_on_error is true.
3837  */
__skb_put_padto(struct sk_buff * skb,unsigned int len,bool free_on_error)3838 static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3839 					       unsigned int len,
3840 					       bool free_on_error)
3841 {
3842 	unsigned int size = skb->len;
3843 
3844 	if (unlikely(size < len)) {
3845 		len -= size;
3846 		if (__skb_pad(skb, len, free_on_error))
3847 			return -ENOMEM;
3848 		__skb_put(skb, len);
3849 	}
3850 	return 0;
3851 }
3852 
3853 /**
3854  *	skb_put_padto - increase size and pad an skbuff up to a minimal size
3855  *	@skb: buffer to pad
3856  *	@len: minimal length
3857  *
3858  *	Pads up a buffer to ensure the trailing bytes exist and are
3859  *	blanked. If the buffer already contains sufficient data it
3860  *	is untouched. Otherwise it is extended. Returns zero on
3861  *	success. The skb is freed on error.
3862  */
skb_put_padto(struct sk_buff * skb,unsigned int len)3863 static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
3864 {
3865 	return __skb_put_padto(skb, len, true);
3866 }
3867 
3868 bool csum_and_copy_from_iter_full(void *addr, size_t bytes, __wsum *csum, struct iov_iter *i)
3869 	__must_check;
3870 
skb_can_coalesce(struct sk_buff * skb,int i,const struct page * page,int off)3871 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3872 				    const struct page *page, int off)
3873 {
3874 	if (skb_zcopy(skb))
3875 		return false;
3876 	if (i) {
3877 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
3878 
3879 		return page == skb_frag_page(frag) &&
3880 		       off == skb_frag_off(frag) + skb_frag_size(frag);
3881 	}
3882 	return false;
3883 }
3884 
__skb_linearize(struct sk_buff * skb)3885 static inline int __skb_linearize(struct sk_buff *skb)
3886 {
3887 	return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3888 }
3889 
3890 /**
3891  *	skb_linearize - convert paged skb to linear one
3892  *	@skb: buffer to linarize
3893  *
3894  *	If there is no free memory -ENOMEM is returned, otherwise zero
3895  *	is returned and the old skb data released.
3896  */
skb_linearize(struct sk_buff * skb)3897 static inline int skb_linearize(struct sk_buff *skb)
3898 {
3899 	return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3900 }
3901 
3902 /**
3903  * skb_has_shared_frag - can any frag be overwritten
3904  * @skb: buffer to test
3905  *
3906  * Return: true if the skb has at least one frag that might be modified
3907  * by an external entity (as in vmsplice()/sendfile())
3908  */
skb_has_shared_frag(const struct sk_buff * skb)3909 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3910 {
3911 	return skb_is_nonlinear(skb) &&
3912 	       skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG;
3913 }
3914 
3915 /**
3916  *	skb_linearize_cow - make sure skb is linear and writable
3917  *	@skb: buffer to process
3918  *
3919  *	If there is no free memory -ENOMEM is returned, otherwise zero
3920  *	is returned and the old skb data released.
3921  */
skb_linearize_cow(struct sk_buff * skb)3922 static inline int skb_linearize_cow(struct sk_buff *skb)
3923 {
3924 	return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3925 	       __skb_linearize(skb) : 0;
3926 }
3927 
3928 static __always_inline void
__skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3929 __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3930 		     unsigned int off)
3931 {
3932 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3933 		skb->csum = csum_block_sub(skb->csum,
3934 					   csum_partial(start, len, 0), off);
3935 	else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3936 		 skb_checksum_start_offset(skb) < 0)
3937 		skb->ip_summed = CHECKSUM_NONE;
3938 }
3939 
3940 /**
3941  *	skb_postpull_rcsum - update checksum for received skb after pull
3942  *	@skb: buffer to update
3943  *	@start: start of data before pull
3944  *	@len: length of data pulled
3945  *
3946  *	After doing a pull on a received packet, you need to call this to
3947  *	update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3948  *	CHECKSUM_NONE so that it can be recomputed from scratch.
3949  */
skb_postpull_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3950 static inline void skb_postpull_rcsum(struct sk_buff *skb,
3951 				      const void *start, unsigned int len)
3952 {
3953 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3954 		skb->csum = wsum_negate(csum_partial(start, len,
3955 						     wsum_negate(skb->csum)));
3956 	else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3957 		 skb_checksum_start_offset(skb) < 0)
3958 		skb->ip_summed = CHECKSUM_NONE;
3959 }
3960 
3961 static __always_inline void
__skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len,unsigned int off)3962 __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3963 		     unsigned int off)
3964 {
3965 	if (skb->ip_summed == CHECKSUM_COMPLETE)
3966 		skb->csum = csum_block_add(skb->csum,
3967 					   csum_partial(start, len, 0), off);
3968 }
3969 
3970 /**
3971  *	skb_postpush_rcsum - update checksum for received skb after push
3972  *	@skb: buffer to update
3973  *	@start: start of data after push
3974  *	@len: length of data pushed
3975  *
3976  *	After doing a push on a received packet, you need to call this to
3977  *	update the CHECKSUM_COMPLETE checksum.
3978  */
skb_postpush_rcsum(struct sk_buff * skb,const void * start,unsigned int len)3979 static inline void skb_postpush_rcsum(struct sk_buff *skb,
3980 				      const void *start, unsigned int len)
3981 {
3982 	__skb_postpush_rcsum(skb, start, len, 0);
3983 }
3984 
3985 void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3986 
3987 /**
3988  *	skb_push_rcsum - push skb and update receive checksum
3989  *	@skb: buffer to update
3990  *	@len: length of data pulled
3991  *
3992  *	This function performs an skb_push on the packet and updates
3993  *	the CHECKSUM_COMPLETE checksum.  It should be used on
3994  *	receive path processing instead of skb_push unless you know
3995  *	that the checksum difference is zero (e.g., a valid IP header)
3996  *	or you are setting ip_summed to CHECKSUM_NONE.
3997  */
skb_push_rcsum(struct sk_buff * skb,unsigned int len)3998 static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3999 {
4000 	skb_push(skb, len);
4001 	skb_postpush_rcsum(skb, skb->data, len);
4002 	return skb->data;
4003 }
4004 
4005 int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
4006 /**
4007  *	pskb_trim_rcsum - trim received skb and update checksum
4008  *	@skb: buffer to trim
4009  *	@len: new length
4010  *
4011  *	This is exactly the same as pskb_trim except that it ensures the
4012  *	checksum of received packets are still valid after the operation.
4013  *	It can change skb pointers.
4014  */
4015 
pskb_trim_rcsum(struct sk_buff * skb,unsigned int len)4016 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
4017 {
4018 	skb_might_realloc(skb);
4019 	if (likely(len >= skb->len))
4020 		return 0;
4021 	return pskb_trim_rcsum_slow(skb, len);
4022 }
4023 
__skb_trim_rcsum(struct sk_buff * skb,unsigned int len)4024 static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
4025 {
4026 	if (skb->ip_summed == CHECKSUM_COMPLETE)
4027 		skb->ip_summed = CHECKSUM_NONE;
4028 	__skb_trim(skb, len);
4029 	return 0;
4030 }
4031 
__skb_grow_rcsum(struct sk_buff * skb,unsigned int len)4032 static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
4033 {
4034 	if (skb->ip_summed == CHECKSUM_COMPLETE)
4035 		skb->ip_summed = CHECKSUM_NONE;
4036 	return __skb_grow(skb, len);
4037 }
4038 
4039 #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
4040 #define skb_rb_first(root) rb_to_skb(rb_first(root))
4041 #define skb_rb_last(root)  rb_to_skb(rb_last(root))
4042 #define skb_rb_next(skb)   rb_to_skb(rb_next(&(skb)->rbnode))
4043 #define skb_rb_prev(skb)   rb_to_skb(rb_prev(&(skb)->rbnode))
4044 
4045 #define skb_queue_walk(queue, skb) \
4046 		for (skb = (queue)->next;					\
4047 		     skb != (struct sk_buff *)(queue);				\
4048 		     skb = skb->next)
4049 
4050 #define skb_queue_walk_safe(queue, skb, tmp)					\
4051 		for (skb = (queue)->next, tmp = skb->next;			\
4052 		     skb != (struct sk_buff *)(queue);				\
4053 		     skb = tmp, tmp = skb->next)
4054 
4055 #define skb_queue_walk_from(queue, skb)						\
4056 		for (; skb != (struct sk_buff *)(queue);			\
4057 		     skb = skb->next)
4058 
4059 #define skb_rbtree_walk(skb, root)						\
4060 		for (skb = skb_rb_first(root); skb != NULL;			\
4061 		     skb = skb_rb_next(skb))
4062 
4063 #define skb_rbtree_walk_from(skb)						\
4064 		for (; skb != NULL;						\
4065 		     skb = skb_rb_next(skb))
4066 
4067 #define skb_rbtree_walk_from_safe(skb, tmp)					\
4068 		for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL);	\
4069 		     skb = tmp)
4070 
4071 #define skb_queue_walk_from_safe(queue, skb, tmp)				\
4072 		for (tmp = skb->next;						\
4073 		     skb != (struct sk_buff *)(queue);				\
4074 		     skb = tmp, tmp = skb->next)
4075 
4076 #define skb_queue_reverse_walk(queue, skb) \
4077 		for (skb = (queue)->prev;					\
4078 		     skb != (struct sk_buff *)(queue);				\
4079 		     skb = skb->prev)
4080 
4081 #define skb_queue_reverse_walk_safe(queue, skb, tmp)				\
4082 		for (skb = (queue)->prev, tmp = skb->prev;			\
4083 		     skb != (struct sk_buff *)(queue);				\
4084 		     skb = tmp, tmp = skb->prev)
4085 
4086 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp)			\
4087 		for (tmp = skb->prev;						\
4088 		     skb != (struct sk_buff *)(queue);				\
4089 		     skb = tmp, tmp = skb->prev)
4090 
skb_has_frag_list(const struct sk_buff * skb)4091 static inline bool skb_has_frag_list(const struct sk_buff *skb)
4092 {
4093 	return skb_shinfo(skb)->frag_list != NULL;
4094 }
4095 
skb_frag_list_init(struct sk_buff * skb)4096 static inline void skb_frag_list_init(struct sk_buff *skb)
4097 {
4098 	skb_shinfo(skb)->frag_list = NULL;
4099 }
4100 
4101 #define skb_walk_frags(skb, iter)	\
4102 	for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
4103 
4104 
4105 int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
4106 				int *err, long *timeo_p,
4107 				const struct sk_buff *skb);
4108 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
4109 					  struct sk_buff_head *queue,
4110 					  unsigned int flags,
4111 					  int *off, int *err,
4112 					  struct sk_buff **last);
4113 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
4114 					struct sk_buff_head *queue,
4115 					unsigned int flags, int *off, int *err,
4116 					struct sk_buff **last);
4117 struct sk_buff *__skb_recv_datagram(struct sock *sk,
4118 				    struct sk_buff_head *sk_queue,
4119 				    unsigned int flags, int *off, int *err);
4120 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err);
4121 __poll_t datagram_poll(struct file *file, struct socket *sock,
4122 			   struct poll_table_struct *wait);
4123 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
4124 			   struct iov_iter *to, int size);
skb_copy_datagram_msg(const struct sk_buff * from,int offset,struct msghdr * msg,int size)4125 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
4126 					struct msghdr *msg, int size)
4127 {
4128 	return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
4129 }
4130 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
4131 				   struct msghdr *msg);
4132 int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
4133 			   struct iov_iter *to, int len,
4134 			   struct ahash_request *hash);
4135 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
4136 				 struct iov_iter *from, int len);
4137 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
4138 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
4139 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
4140 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
4141 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
4142 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
4143 			      int len);
4144 int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
4145 		    struct pipe_inode_info *pipe, unsigned int len,
4146 		    unsigned int flags);
4147 int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
4148 			 int len);
4149 int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
4150 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
4151 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
4152 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
4153 		 int len, int hlen);
4154 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
4155 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
4156 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
4157 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
4158 struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
4159 				 unsigned int offset);
4160 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
4161 int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len);
4162 int skb_ensure_writable_head_tail(struct sk_buff *skb, struct net_device *dev);
4163 int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
4164 int skb_vlan_pop(struct sk_buff *skb);
4165 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
4166 int skb_eth_pop(struct sk_buff *skb);
4167 int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
4168 		 const unsigned char *src);
4169 int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
4170 		  int mac_len, bool ethernet);
4171 int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
4172 		 bool ethernet);
4173 int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
4174 int skb_mpls_dec_ttl(struct sk_buff *skb);
4175 struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
4176 			     gfp_t gfp);
4177 
memcpy_from_msg(void * data,struct msghdr * msg,int len)4178 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
4179 {
4180 	return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
4181 }
4182 
memcpy_to_msg(struct msghdr * msg,void * data,int len)4183 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
4184 {
4185 	return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
4186 }
4187 
4188 struct skb_checksum_ops {
4189 	__wsum (*update)(const void *mem, int len, __wsum wsum);
4190 	__wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
4191 };
4192 
4193 extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
4194 
4195 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
4196 		      __wsum csum, const struct skb_checksum_ops *ops);
4197 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
4198 		    __wsum csum);
4199 
4200 static inline void * __must_check
__skb_header_pointer(const struct sk_buff * skb,int offset,int len,const void * data,int hlen,void * buffer)4201 __skb_header_pointer(const struct sk_buff *skb, int offset, int len,
4202 		     const void *data, int hlen, void *buffer)
4203 {
4204 	if (likely(hlen - offset >= len))
4205 		return (void *)data + offset;
4206 
4207 	if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0))
4208 		return NULL;
4209 
4210 	return buffer;
4211 }
4212 
4213 static inline void * __must_check
skb_header_pointer(const struct sk_buff * skb,int offset,int len,void * buffer)4214 skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
4215 {
4216 	return __skb_header_pointer(skb, offset, len, skb->data,
4217 				    skb_headlen(skb), buffer);
4218 }
4219 
4220 static inline void * __must_check
skb_pointer_if_linear(const struct sk_buff * skb,int offset,int len)4221 skb_pointer_if_linear(const struct sk_buff *skb, int offset, int len)
4222 {
4223 	if (likely(skb_headlen(skb) - offset >= len))
4224 		return skb->data + offset;
4225 	return NULL;
4226 }
4227 
4228 /**
4229  *	skb_needs_linearize - check if we need to linearize a given skb
4230  *			      depending on the given device features.
4231  *	@skb: socket buffer to check
4232  *	@features: net device features
4233  *
4234  *	Returns true if either:
4235  *	1. skb has frag_list and the device doesn't support FRAGLIST, or
4236  *	2. skb is fragmented and the device does not support SG.
4237  */
skb_needs_linearize(struct sk_buff * skb,netdev_features_t features)4238 static inline bool skb_needs_linearize(struct sk_buff *skb,
4239 				       netdev_features_t features)
4240 {
4241 	return skb_is_nonlinear(skb) &&
4242 	       ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
4243 		(skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
4244 }
4245 
skb_copy_from_linear_data(const struct sk_buff * skb,void * to,const unsigned int len)4246 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
4247 					     void *to,
4248 					     const unsigned int len)
4249 {
4250 	memcpy(to, skb->data, len);
4251 }
4252 
skb_copy_from_linear_data_offset(const struct sk_buff * skb,const int offset,void * to,const unsigned int len)4253 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
4254 						    const int offset, void *to,
4255 						    const unsigned int len)
4256 {
4257 	memcpy(to, skb->data + offset, len);
4258 }
4259 
skb_copy_to_linear_data(struct sk_buff * skb,const void * from,const unsigned int len)4260 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
4261 					   const void *from,
4262 					   const unsigned int len)
4263 {
4264 	memcpy(skb->data, from, len);
4265 }
4266 
skb_copy_to_linear_data_offset(struct sk_buff * skb,const int offset,const void * from,const unsigned int len)4267 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
4268 						  const int offset,
4269 						  const void *from,
4270 						  const unsigned int len)
4271 {
4272 	memcpy(skb->data + offset, from, len);
4273 }
4274 
4275 void skb_init(void);
4276 
skb_get_ktime(const struct sk_buff * skb)4277 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
4278 {
4279 	return skb->tstamp;
4280 }
4281 
4282 /**
4283  *	skb_get_timestamp - get timestamp from a skb
4284  *	@skb: skb to get stamp from
4285  *	@stamp: pointer to struct __kernel_old_timeval to store stamp in
4286  *
4287  *	Timestamps are stored in the skb as offsets to a base timestamp.
4288  *	This function converts the offset back to a struct timeval and stores
4289  *	it in stamp.
4290  */
skb_get_timestamp(const struct sk_buff * skb,struct __kernel_old_timeval * stamp)4291 static inline void skb_get_timestamp(const struct sk_buff *skb,
4292 				     struct __kernel_old_timeval *stamp)
4293 {
4294 	*stamp = ns_to_kernel_old_timeval(skb->tstamp);
4295 }
4296 
skb_get_new_timestamp(const struct sk_buff * skb,struct __kernel_sock_timeval * stamp)4297 static inline void skb_get_new_timestamp(const struct sk_buff *skb,
4298 					 struct __kernel_sock_timeval *stamp)
4299 {
4300 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4301 
4302 	stamp->tv_sec = ts.tv_sec;
4303 	stamp->tv_usec = ts.tv_nsec / 1000;
4304 }
4305 
skb_get_timestampns(const struct sk_buff * skb,struct __kernel_old_timespec * stamp)4306 static inline void skb_get_timestampns(const struct sk_buff *skb,
4307 				       struct __kernel_old_timespec *stamp)
4308 {
4309 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4310 
4311 	stamp->tv_sec = ts.tv_sec;
4312 	stamp->tv_nsec = ts.tv_nsec;
4313 }
4314 
skb_get_new_timestampns(const struct sk_buff * skb,struct __kernel_timespec * stamp)4315 static inline void skb_get_new_timestampns(const struct sk_buff *skb,
4316 					   struct __kernel_timespec *stamp)
4317 {
4318 	struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
4319 
4320 	stamp->tv_sec = ts.tv_sec;
4321 	stamp->tv_nsec = ts.tv_nsec;
4322 }
4323 
__net_timestamp(struct sk_buff * skb)4324 static inline void __net_timestamp(struct sk_buff *skb)
4325 {
4326 	skb->tstamp = ktime_get_real();
4327 	skb->tstamp_type = SKB_CLOCK_REALTIME;
4328 }
4329 
net_timedelta(ktime_t t)4330 static inline ktime_t net_timedelta(ktime_t t)
4331 {
4332 	return ktime_sub(ktime_get_real(), t);
4333 }
4334 
skb_set_delivery_time(struct sk_buff * skb,ktime_t kt,u8 tstamp_type)4335 static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt,
4336 					 u8 tstamp_type)
4337 {
4338 	skb->tstamp = kt;
4339 
4340 	if (kt)
4341 		skb->tstamp_type = tstamp_type;
4342 	else
4343 		skb->tstamp_type = SKB_CLOCK_REALTIME;
4344 }
4345 
skb_set_delivery_type_by_clockid(struct sk_buff * skb,ktime_t kt,clockid_t clockid)4346 static inline void skb_set_delivery_type_by_clockid(struct sk_buff *skb,
4347 						    ktime_t kt, clockid_t clockid)
4348 {
4349 	u8 tstamp_type = SKB_CLOCK_REALTIME;
4350 
4351 	switch (clockid) {
4352 	case CLOCK_REALTIME:
4353 		break;
4354 	case CLOCK_MONOTONIC:
4355 		tstamp_type = SKB_CLOCK_MONOTONIC;
4356 		break;
4357 	case CLOCK_TAI:
4358 		tstamp_type = SKB_CLOCK_TAI;
4359 		break;
4360 	default:
4361 		WARN_ON_ONCE(1);
4362 		kt = 0;
4363 	}
4364 
4365 	skb_set_delivery_time(skb, kt, tstamp_type);
4366 }
4367 
4368 DECLARE_STATIC_KEY_FALSE(netstamp_needed_key);
4369 
4370 /* It is used in the ingress path to clear the delivery_time.
4371  * If needed, set the skb->tstamp to the (rcv) timestamp.
4372  */
skb_clear_delivery_time(struct sk_buff * skb)4373 static inline void skb_clear_delivery_time(struct sk_buff *skb)
4374 {
4375 	if (skb->tstamp_type) {
4376 		skb->tstamp_type = SKB_CLOCK_REALTIME;
4377 		if (static_branch_unlikely(&netstamp_needed_key))
4378 			skb->tstamp = ktime_get_real();
4379 		else
4380 			skb->tstamp = 0;
4381 	}
4382 }
4383 
skb_clear_tstamp(struct sk_buff * skb)4384 static inline void skb_clear_tstamp(struct sk_buff *skb)
4385 {
4386 	if (skb->tstamp_type)
4387 		return;
4388 
4389 	skb->tstamp = 0;
4390 }
4391 
skb_tstamp(const struct sk_buff * skb)4392 static inline ktime_t skb_tstamp(const struct sk_buff *skb)
4393 {
4394 	if (skb->tstamp_type)
4395 		return 0;
4396 
4397 	return skb->tstamp;
4398 }
4399 
skb_tstamp_cond(const struct sk_buff * skb,bool cond)4400 static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond)
4401 {
4402 	if (skb->tstamp_type != SKB_CLOCK_MONOTONIC && skb->tstamp)
4403 		return skb->tstamp;
4404 
4405 	if (static_branch_unlikely(&netstamp_needed_key) || cond)
4406 		return ktime_get_real();
4407 
4408 	return 0;
4409 }
4410 
skb_metadata_len(const struct sk_buff * skb)4411 static inline u8 skb_metadata_len(const struct sk_buff *skb)
4412 {
4413 	return skb_shinfo(skb)->meta_len;
4414 }
4415 
skb_metadata_end(const struct sk_buff * skb)4416 static inline void *skb_metadata_end(const struct sk_buff *skb)
4417 {
4418 	return skb_mac_header(skb);
4419 }
4420 
__skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b,u8 meta_len)4421 static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
4422 					  const struct sk_buff *skb_b,
4423 					  u8 meta_len)
4424 {
4425 	const void *a = skb_metadata_end(skb_a);
4426 	const void *b = skb_metadata_end(skb_b);
4427 	u64 diffs = 0;
4428 
4429 	if (!IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) ||
4430 	    BITS_PER_LONG != 64)
4431 		goto slow;
4432 
4433 	/* Using more efficient variant than plain call to memcmp(). */
4434 	switch (meta_len) {
4435 #define __it(x, op) (x -= sizeof(u##op))
4436 #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
4437 	case 32: diffs |= __it_diff(a, b, 64);
4438 		fallthrough;
4439 	case 24: diffs |= __it_diff(a, b, 64);
4440 		fallthrough;
4441 	case 16: diffs |= __it_diff(a, b, 64);
4442 		fallthrough;
4443 	case  8: diffs |= __it_diff(a, b, 64);
4444 		break;
4445 	case 28: diffs |= __it_diff(a, b, 64);
4446 		fallthrough;
4447 	case 20: diffs |= __it_diff(a, b, 64);
4448 		fallthrough;
4449 	case 12: diffs |= __it_diff(a, b, 64);
4450 		fallthrough;
4451 	case  4: diffs |= __it_diff(a, b, 32);
4452 		break;
4453 	default:
4454 slow:
4455 		return memcmp(a - meta_len, b - meta_len, meta_len);
4456 	}
4457 	return diffs;
4458 }
4459 
skb_metadata_differs(const struct sk_buff * skb_a,const struct sk_buff * skb_b)4460 static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
4461 					const struct sk_buff *skb_b)
4462 {
4463 	u8 len_a = skb_metadata_len(skb_a);
4464 	u8 len_b = skb_metadata_len(skb_b);
4465 
4466 	if (!(len_a | len_b))
4467 		return false;
4468 
4469 	return len_a != len_b ?
4470 	       true : __skb_metadata_differs(skb_a, skb_b, len_a);
4471 }
4472 
skb_metadata_set(struct sk_buff * skb,u8 meta_len)4473 static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
4474 {
4475 	skb_shinfo(skb)->meta_len = meta_len;
4476 }
4477 
skb_metadata_clear(struct sk_buff * skb)4478 static inline void skb_metadata_clear(struct sk_buff *skb)
4479 {
4480 	skb_metadata_set(skb, 0);
4481 }
4482 
4483 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
4484 
4485 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
4486 
4487 void skb_clone_tx_timestamp(struct sk_buff *skb);
4488 bool skb_defer_rx_timestamp(struct sk_buff *skb);
4489 
4490 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
4491 
skb_clone_tx_timestamp(struct sk_buff * skb)4492 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
4493 {
4494 }
4495 
skb_defer_rx_timestamp(struct sk_buff * skb)4496 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
4497 {
4498 	return false;
4499 }
4500 
4501 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
4502 
4503 /**
4504  * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
4505  *
4506  * PHY drivers may accept clones of transmitted packets for
4507  * timestamping via their phy_driver.txtstamp method. These drivers
4508  * must call this function to return the skb back to the stack with a
4509  * timestamp.
4510  *
4511  * @skb: clone of the original outgoing packet
4512  * @hwtstamps: hardware time stamps
4513  *
4514  */
4515 void skb_complete_tx_timestamp(struct sk_buff *skb,
4516 			       struct skb_shared_hwtstamps *hwtstamps);
4517 
4518 void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb,
4519 		     struct skb_shared_hwtstamps *hwtstamps,
4520 		     struct sock *sk, int tstype);
4521 
4522 /**
4523  * skb_tstamp_tx - queue clone of skb with send time stamps
4524  * @orig_skb:	the original outgoing packet
4525  * @hwtstamps:	hardware time stamps, may be NULL if not available
4526  *
4527  * If the skb has a socket associated, then this function clones the
4528  * skb (thus sharing the actual data and optional structures), stores
4529  * the optional hardware time stamping information (if non NULL) or
4530  * generates a software time stamp (otherwise), then queues the clone
4531  * to the error queue of the socket.  Errors are silently ignored.
4532  */
4533 void skb_tstamp_tx(struct sk_buff *orig_skb,
4534 		   struct skb_shared_hwtstamps *hwtstamps);
4535 
4536 /**
4537  * skb_tx_timestamp() - Driver hook for transmit timestamping
4538  *
4539  * Ethernet MAC Drivers should call this function in their hard_xmit()
4540  * function immediately before giving the sk_buff to the MAC hardware.
4541  *
4542  * Specifically, one should make absolutely sure that this function is
4543  * called before TX completion of this packet can trigger.  Otherwise
4544  * the packet could potentially already be freed.
4545  *
4546  * @skb: A socket buffer.
4547  */
skb_tx_timestamp(struct sk_buff * skb)4548 static inline void skb_tx_timestamp(struct sk_buff *skb)
4549 {
4550 	skb_clone_tx_timestamp(skb);
4551 	if (skb_shinfo(skb)->tx_flags & (SKBTX_SW_TSTAMP | SKBTX_BPF))
4552 		skb_tstamp_tx(skb, NULL);
4553 }
4554 
4555 /**
4556  * skb_complete_wifi_ack - deliver skb with wifi status
4557  *
4558  * @skb: the original outgoing packet
4559  * @acked: ack status
4560  *
4561  */
4562 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
4563 
4564 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
4565 __sum16 __skb_checksum_complete(struct sk_buff *skb);
4566 
skb_csum_unnecessary(const struct sk_buff * skb)4567 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
4568 {
4569 	return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
4570 		skb->csum_valid ||
4571 		(skb->ip_summed == CHECKSUM_PARTIAL &&
4572 		 skb_checksum_start_offset(skb) >= 0));
4573 }
4574 
4575 /**
4576  *	skb_checksum_complete - Calculate checksum of an entire packet
4577  *	@skb: packet to process
4578  *
4579  *	This function calculates the checksum over the entire packet plus
4580  *	the value of skb->csum.  The latter can be used to supply the
4581  *	checksum of a pseudo header as used by TCP/UDP.  It returns the
4582  *	checksum.
4583  *
4584  *	For protocols that contain complete checksums such as ICMP/TCP/UDP,
4585  *	this function can be used to verify that checksum on received
4586  *	packets.  In that case the function should return zero if the
4587  *	checksum is correct.  In particular, this function will return zero
4588  *	if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
4589  *	hardware has already verified the correctness of the checksum.
4590  */
skb_checksum_complete(struct sk_buff * skb)4591 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
4592 {
4593 	return skb_csum_unnecessary(skb) ?
4594 	       0 : __skb_checksum_complete(skb);
4595 }
4596 
__skb_decr_checksum_unnecessary(struct sk_buff * skb)4597 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
4598 {
4599 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4600 		if (skb->csum_level == 0)
4601 			skb->ip_summed = CHECKSUM_NONE;
4602 		else
4603 			skb->csum_level--;
4604 	}
4605 }
4606 
__skb_incr_checksum_unnecessary(struct sk_buff * skb)4607 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
4608 {
4609 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4610 		if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
4611 			skb->csum_level++;
4612 	} else if (skb->ip_summed == CHECKSUM_NONE) {
4613 		skb->ip_summed = CHECKSUM_UNNECESSARY;
4614 		skb->csum_level = 0;
4615 	}
4616 }
4617 
__skb_reset_checksum_unnecessary(struct sk_buff * skb)4618 static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
4619 {
4620 	if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
4621 		skb->ip_summed = CHECKSUM_NONE;
4622 		skb->csum_level = 0;
4623 	}
4624 }
4625 
4626 /* Check if we need to perform checksum complete validation.
4627  *
4628  * Returns: true if checksum complete is needed, false otherwise
4629  * (either checksum is unnecessary or zero checksum is allowed).
4630  */
__skb_checksum_validate_needed(struct sk_buff * skb,bool zero_okay,__sum16 check)4631 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
4632 						  bool zero_okay,
4633 						  __sum16 check)
4634 {
4635 	if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
4636 		skb->csum_valid = 1;
4637 		__skb_decr_checksum_unnecessary(skb);
4638 		return false;
4639 	}
4640 
4641 	return true;
4642 }
4643 
4644 /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
4645  * in checksum_init.
4646  */
4647 #define CHECKSUM_BREAK 76
4648 
4649 /* Unset checksum-complete
4650  *
4651  * Unset checksum complete can be done when packet is being modified
4652  * (uncompressed for instance) and checksum-complete value is
4653  * invalidated.
4654  */
skb_checksum_complete_unset(struct sk_buff * skb)4655 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
4656 {
4657 	if (skb->ip_summed == CHECKSUM_COMPLETE)
4658 		skb->ip_summed = CHECKSUM_NONE;
4659 }
4660 
4661 /* Validate (init) checksum based on checksum complete.
4662  *
4663  * Return values:
4664  *   0: checksum is validated or try to in skb_checksum_complete. In the latter
4665  *	case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4666  *	checksum is stored in skb->csum for use in __skb_checksum_complete
4667  *   non-zero: value of invalid checksum
4668  *
4669  */
__skb_checksum_validate_complete(struct sk_buff * skb,bool complete,__wsum psum)4670 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4671 						       bool complete,
4672 						       __wsum psum)
4673 {
4674 	if (skb->ip_summed == CHECKSUM_COMPLETE) {
4675 		if (!csum_fold(csum_add(psum, skb->csum))) {
4676 			skb->csum_valid = 1;
4677 			return 0;
4678 		}
4679 	}
4680 
4681 	skb->csum = psum;
4682 
4683 	if (complete || skb->len <= CHECKSUM_BREAK) {
4684 		__sum16 csum;
4685 
4686 		csum = __skb_checksum_complete(skb);
4687 		skb->csum_valid = !csum;
4688 		return csum;
4689 	}
4690 
4691 	return 0;
4692 }
4693 
null_compute_pseudo(struct sk_buff * skb,int proto)4694 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4695 {
4696 	return 0;
4697 }
4698 
4699 /* Perform checksum validate (init). Note that this is a macro since we only
4700  * want to calculate the pseudo header which is an input function if necessary.
4701  * First we try to validate without any computation (checksum unnecessary) and
4702  * then calculate based on checksum complete calling the function to compute
4703  * pseudo header.
4704  *
4705  * Return values:
4706  *   0: checksum is validated or try to in skb_checksum_complete
4707  *   non-zero: value of invalid checksum
4708  */
4709 #define __skb_checksum_validate(skb, proto, complete,			\
4710 				zero_okay, check, compute_pseudo)	\
4711 ({									\
4712 	__sum16 __ret = 0;						\
4713 	skb->csum_valid = 0;						\
4714 	if (__skb_checksum_validate_needed(skb, zero_okay, check))	\
4715 		__ret = __skb_checksum_validate_complete(skb,		\
4716 				complete, compute_pseudo(skb, proto));	\
4717 	__ret;								\
4718 })
4719 
4720 #define skb_checksum_init(skb, proto, compute_pseudo)			\
4721 	__skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4722 
4723 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo)	\
4724 	__skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4725 
4726 #define skb_checksum_validate(skb, proto, compute_pseudo)		\
4727 	__skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4728 
4729 #define skb_checksum_validate_zero_check(skb, proto, check,		\
4730 					 compute_pseudo)		\
4731 	__skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
4732 
4733 #define skb_checksum_simple_validate(skb)				\
4734 	__skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4735 
__skb_checksum_convert_check(struct sk_buff * skb)4736 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4737 {
4738 	return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4739 }
4740 
__skb_checksum_convert(struct sk_buff * skb,__wsum pseudo)4741 static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
4742 {
4743 	skb->csum = ~pseudo;
4744 	skb->ip_summed = CHECKSUM_COMPLETE;
4745 }
4746 
4747 #define skb_checksum_try_convert(skb, proto, compute_pseudo)	\
4748 do {									\
4749 	if (__skb_checksum_convert_check(skb))				\
4750 		__skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
4751 } while (0)
4752 
skb_remcsum_adjust_partial(struct sk_buff * skb,void * ptr,u16 start,u16 offset)4753 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4754 					      u16 start, u16 offset)
4755 {
4756 	skb->ip_summed = CHECKSUM_PARTIAL;
4757 	skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4758 	skb->csum_offset = offset - start;
4759 }
4760 
4761 /* Update skbuf and packet to reflect the remote checksum offload operation.
4762  * When called, ptr indicates the starting point for skb->csum when
4763  * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4764  * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4765  */
skb_remcsum_process(struct sk_buff * skb,void * ptr,int start,int offset,bool nopartial)4766 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4767 				       int start, int offset, bool nopartial)
4768 {
4769 	__wsum delta;
4770 
4771 	if (!nopartial) {
4772 		skb_remcsum_adjust_partial(skb, ptr, start, offset);
4773 		return;
4774 	}
4775 
4776 	if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4777 		__skb_checksum_complete(skb);
4778 		skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4779 	}
4780 
4781 	delta = remcsum_adjust(ptr, skb->csum, start, offset);
4782 
4783 	/* Adjust skb->csum since we changed the packet */
4784 	skb->csum = csum_add(skb->csum, delta);
4785 }
4786 
skb_nfct(const struct sk_buff * skb)4787 static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4788 {
4789 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4790 	return (void *)(skb->_nfct & NFCT_PTRMASK);
4791 #else
4792 	return NULL;
4793 #endif
4794 }
4795 
skb_get_nfct(const struct sk_buff * skb)4796 static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
4797 {
4798 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4799 	return skb->_nfct;
4800 #else
4801 	return 0UL;
4802 #endif
4803 }
4804 
skb_set_nfct(struct sk_buff * skb,unsigned long nfct)4805 static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4806 {
4807 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
4808 	skb->slow_gro |= !!nfct;
4809 	skb->_nfct = nfct;
4810 #endif
4811 }
4812 
4813 #ifdef CONFIG_SKB_EXTENSIONS
4814 enum skb_ext_id {
4815 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4816 	SKB_EXT_BRIDGE_NF,
4817 #endif
4818 #ifdef CONFIG_XFRM
4819 	SKB_EXT_SEC_PATH,
4820 #endif
4821 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4822 	TC_SKB_EXT,
4823 #endif
4824 #if IS_ENABLED(CONFIG_MPTCP)
4825 	SKB_EXT_MPTCP,
4826 #endif
4827 #if IS_ENABLED(CONFIG_MCTP_FLOWS)
4828 	SKB_EXT_MCTP,
4829 #endif
4830 	SKB_EXT_NUM, /* must be last */
4831 };
4832 
4833 /**
4834  *	struct skb_ext - sk_buff extensions
4835  *	@refcnt: 1 on allocation, deallocated on 0
4836  *	@offset: offset to add to @data to obtain extension address
4837  *	@chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4838  *	@data: start of extension data, variable sized
4839  *
4840  *	Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4841  *	to use 'u8' types while allowing up to 2kb worth of extension data.
4842  */
4843 struct skb_ext {
4844 	refcount_t refcnt;
4845 	u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4846 	u8 chunks;		/* same */
4847 	char data[] __aligned(8);
4848 };
4849 
4850 struct skb_ext *__skb_ext_alloc(gfp_t flags);
4851 void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4852 		    struct skb_ext *ext);
4853 void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4854 void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4855 void __skb_ext_put(struct skb_ext *ext);
4856 
skb_ext_put(struct sk_buff * skb)4857 static inline void skb_ext_put(struct sk_buff *skb)
4858 {
4859 	if (skb->active_extensions)
4860 		__skb_ext_put(skb->extensions);
4861 }
4862 
__skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4863 static inline void __skb_ext_copy(struct sk_buff *dst,
4864 				  const struct sk_buff *src)
4865 {
4866 	dst->active_extensions = src->active_extensions;
4867 
4868 	if (src->active_extensions) {
4869 		struct skb_ext *ext = src->extensions;
4870 
4871 		refcount_inc(&ext->refcnt);
4872 		dst->extensions = ext;
4873 	}
4874 }
4875 
skb_ext_copy(struct sk_buff * dst,const struct sk_buff * src)4876 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4877 {
4878 	skb_ext_put(dst);
4879 	__skb_ext_copy(dst, src);
4880 }
4881 
__skb_ext_exist(const struct skb_ext * ext,enum skb_ext_id i)4882 static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4883 {
4884 	return !!ext->offset[i];
4885 }
4886 
skb_ext_exist(const struct sk_buff * skb,enum skb_ext_id id)4887 static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4888 {
4889 	return skb->active_extensions & (1 << id);
4890 }
4891 
skb_ext_del(struct sk_buff * skb,enum skb_ext_id id)4892 static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4893 {
4894 	if (skb_ext_exist(skb, id))
4895 		__skb_ext_del(skb, id);
4896 }
4897 
skb_ext_find(const struct sk_buff * skb,enum skb_ext_id id)4898 static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4899 {
4900 	if (skb_ext_exist(skb, id)) {
4901 		struct skb_ext *ext = skb->extensions;
4902 
4903 		return (void *)ext + (ext->offset[id] << 3);
4904 	}
4905 
4906 	return NULL;
4907 }
4908 
skb_ext_reset(struct sk_buff * skb)4909 static inline void skb_ext_reset(struct sk_buff *skb)
4910 {
4911 	if (unlikely(skb->active_extensions)) {
4912 		__skb_ext_put(skb->extensions);
4913 		skb->active_extensions = 0;
4914 	}
4915 }
4916 
skb_has_extensions(struct sk_buff * skb)4917 static inline bool skb_has_extensions(struct sk_buff *skb)
4918 {
4919 	return unlikely(skb->active_extensions);
4920 }
4921 #else
skb_ext_put(struct sk_buff * skb)4922 static inline void skb_ext_put(struct sk_buff *skb) {}
skb_ext_reset(struct sk_buff * skb)4923 static inline void skb_ext_reset(struct sk_buff *skb) {}
skb_ext_del(struct sk_buff * skb,int unused)4924 static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
__skb_ext_copy(struct sk_buff * d,const struct sk_buff * s)4925 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)4926 static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
skb_has_extensions(struct sk_buff * skb)4927 static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4928 #endif /* CONFIG_SKB_EXTENSIONS */
4929 
nf_reset_ct(struct sk_buff * skb)4930 static inline void nf_reset_ct(struct sk_buff *skb)
4931 {
4932 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4933 	nf_conntrack_put(skb_nfct(skb));
4934 	skb->_nfct = 0;
4935 #endif
4936 }
4937 
nf_reset_trace(struct sk_buff * skb)4938 static inline void nf_reset_trace(struct sk_buff *skb)
4939 {
4940 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4941 	skb->nf_trace = 0;
4942 #endif
4943 }
4944 
ipvs_reset(struct sk_buff * skb)4945 static inline void ipvs_reset(struct sk_buff *skb)
4946 {
4947 #if IS_ENABLED(CONFIG_IP_VS)
4948 	skb->ipvs_property = 0;
4949 #endif
4950 }
4951 
4952 /* Note: This doesn't put any conntrack info in dst. */
__nf_copy(struct sk_buff * dst,const struct sk_buff * src,bool copy)4953 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4954 			     bool copy)
4955 {
4956 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4957 	dst->_nfct = src->_nfct;
4958 	nf_conntrack_get(skb_nfct(src));
4959 #endif
4960 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES)
4961 	if (copy)
4962 		dst->nf_trace = src->nf_trace;
4963 #endif
4964 }
4965 
nf_copy(struct sk_buff * dst,const struct sk_buff * src)4966 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4967 {
4968 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4969 	nf_conntrack_put(skb_nfct(dst));
4970 #endif
4971 	dst->slow_gro = src->slow_gro;
4972 	__nf_copy(dst, src, true);
4973 }
4974 
4975 #ifdef CONFIG_NETWORK_SECMARK
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4976 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4977 {
4978 	to->secmark = from->secmark;
4979 }
4980 
skb_init_secmark(struct sk_buff * skb)4981 static inline void skb_init_secmark(struct sk_buff *skb)
4982 {
4983 	skb->secmark = 0;
4984 }
4985 #else
skb_copy_secmark(struct sk_buff * to,const struct sk_buff * from)4986 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4987 { }
4988 
skb_init_secmark(struct sk_buff * skb)4989 static inline void skb_init_secmark(struct sk_buff *skb)
4990 { }
4991 #endif
4992 
secpath_exists(const struct sk_buff * skb)4993 static inline int secpath_exists(const struct sk_buff *skb)
4994 {
4995 #ifdef CONFIG_XFRM
4996 	return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4997 #else
4998 	return 0;
4999 #endif
5000 }
5001 
skb_irq_freeable(const struct sk_buff * skb)5002 static inline bool skb_irq_freeable(const struct sk_buff *skb)
5003 {
5004 	return !skb->destructor &&
5005 		!secpath_exists(skb) &&
5006 		!skb_nfct(skb) &&
5007 		!skb->_skb_refdst &&
5008 		!skb_has_frag_list(skb);
5009 }
5010 
skb_set_queue_mapping(struct sk_buff * skb,u16 queue_mapping)5011 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
5012 {
5013 	skb->queue_mapping = queue_mapping;
5014 }
5015 
skb_get_queue_mapping(const struct sk_buff * skb)5016 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
5017 {
5018 	return skb->queue_mapping;
5019 }
5020 
skb_copy_queue_mapping(struct sk_buff * to,const struct sk_buff * from)5021 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
5022 {
5023 	to->queue_mapping = from->queue_mapping;
5024 }
5025 
skb_record_rx_queue(struct sk_buff * skb,u16 rx_queue)5026 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
5027 {
5028 	skb->queue_mapping = rx_queue + 1;
5029 }
5030 
skb_get_rx_queue(const struct sk_buff * skb)5031 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
5032 {
5033 	return skb->queue_mapping - 1;
5034 }
5035 
skb_rx_queue_recorded(const struct sk_buff * skb)5036 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
5037 {
5038 	return skb->queue_mapping != 0;
5039 }
5040 
skb_set_dst_pending_confirm(struct sk_buff * skb,u32 val)5041 static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
5042 {
5043 	skb->dst_pending_confirm = val;
5044 }
5045 
skb_get_dst_pending_confirm(const struct sk_buff * skb)5046 static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
5047 {
5048 	return skb->dst_pending_confirm != 0;
5049 }
5050 
skb_sec_path(const struct sk_buff * skb)5051 static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
5052 {
5053 #ifdef CONFIG_XFRM
5054 	return skb_ext_find(skb, SKB_EXT_SEC_PATH);
5055 #else
5056 	return NULL;
5057 #endif
5058 }
5059 
skb_is_gso(const struct sk_buff * skb)5060 static inline bool skb_is_gso(const struct sk_buff *skb)
5061 {
5062 	return skb_shinfo(skb)->gso_size;
5063 }
5064 
5065 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_v6(const struct sk_buff * skb)5066 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
5067 {
5068 	return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
5069 }
5070 
5071 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_sctp(const struct sk_buff * skb)5072 static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
5073 {
5074 	return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
5075 }
5076 
5077 /* Note: Should be called only if skb_is_gso(skb) is true */
skb_is_gso_tcp(const struct sk_buff * skb)5078 static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
5079 {
5080 	return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
5081 }
5082 
skb_gso_reset(struct sk_buff * skb)5083 static inline void skb_gso_reset(struct sk_buff *skb)
5084 {
5085 	skb_shinfo(skb)->gso_size = 0;
5086 	skb_shinfo(skb)->gso_segs = 0;
5087 	skb_shinfo(skb)->gso_type = 0;
5088 }
5089 
skb_increase_gso_size(struct skb_shared_info * shinfo,u16 increment)5090 static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
5091 					 u16 increment)
5092 {
5093 	if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
5094 		return;
5095 	shinfo->gso_size += increment;
5096 }
5097 
skb_decrease_gso_size(struct skb_shared_info * shinfo,u16 decrement)5098 static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
5099 					 u16 decrement)
5100 {
5101 	if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
5102 		return;
5103 	shinfo->gso_size -= decrement;
5104 }
5105 
5106 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
5107 
skb_warn_if_lro(const struct sk_buff * skb)5108 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
5109 {
5110 	/* LRO sets gso_size but not gso_type, whereas if GSO is really
5111 	 * wanted then gso_type will be set. */
5112 	const struct skb_shared_info *shinfo = skb_shinfo(skb);
5113 
5114 	if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
5115 	    unlikely(shinfo->gso_type == 0)) {
5116 		__skb_warn_lro_forwarding(skb);
5117 		return true;
5118 	}
5119 	return false;
5120 }
5121 
skb_forward_csum(struct sk_buff * skb)5122 static inline void skb_forward_csum(struct sk_buff *skb)
5123 {
5124 	/* Unfortunately we don't support this one.  Any brave souls? */
5125 	if (skb->ip_summed == CHECKSUM_COMPLETE)
5126 		skb->ip_summed = CHECKSUM_NONE;
5127 }
5128 
5129 /**
5130  * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
5131  * @skb: skb to check
5132  *
5133  * fresh skbs have their ip_summed set to CHECKSUM_NONE.
5134  * Instead of forcing ip_summed to CHECKSUM_NONE, we can
5135  * use this helper, to document places where we make this assertion.
5136  */
skb_checksum_none_assert(const struct sk_buff * skb)5137 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
5138 {
5139 	DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE);
5140 }
5141 
5142 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
5143 
5144 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
5145 struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
5146 				     unsigned int transport_len,
5147 				     __sum16(*skb_chkf)(struct sk_buff *skb));
5148 
5149 /**
5150  * skb_head_is_locked - Determine if the skb->head is locked down
5151  * @skb: skb to check
5152  *
5153  * The head on skbs build around a head frag can be removed if they are
5154  * not cloned.  This function returns true if the skb head is locked down
5155  * due to either being allocated via kmalloc, or by being a clone with
5156  * multiple references to the head.
5157  */
skb_head_is_locked(const struct sk_buff * skb)5158 static inline bool skb_head_is_locked(const struct sk_buff *skb)
5159 {
5160 	return !skb->head_frag || skb_cloned(skb);
5161 }
5162 
5163 /* Local Checksum Offload.
5164  * Compute outer checksum based on the assumption that the
5165  * inner checksum will be offloaded later.
5166  * See Documentation/networking/checksum-offloads.rst for
5167  * explanation of how this works.
5168  * Fill in outer checksum adjustment (e.g. with sum of outer
5169  * pseudo-header) before calling.
5170  * Also ensure that inner checksum is in linear data area.
5171  */
lco_csum(struct sk_buff * skb)5172 static inline __wsum lco_csum(struct sk_buff *skb)
5173 {
5174 	unsigned char *csum_start = skb_checksum_start(skb);
5175 	unsigned char *l4_hdr = skb_transport_header(skb);
5176 	__wsum partial;
5177 
5178 	/* Start with complement of inner checksum adjustment */
5179 	partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
5180 						    skb->csum_offset));
5181 
5182 	/* Add in checksum of our headers (incl. outer checksum
5183 	 * adjustment filled in by caller) and return result.
5184 	 */
5185 	return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
5186 }
5187 
skb_is_redirected(const struct sk_buff * skb)5188 static inline bool skb_is_redirected(const struct sk_buff *skb)
5189 {
5190 	return skb->redirected;
5191 }
5192 
skb_set_redirected(struct sk_buff * skb,bool from_ingress)5193 static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
5194 {
5195 	skb->redirected = 1;
5196 #ifdef CONFIG_NET_REDIRECT
5197 	skb->from_ingress = from_ingress;
5198 	if (skb->from_ingress)
5199 		skb_clear_tstamp(skb);
5200 #endif
5201 }
5202 
skb_reset_redirect(struct sk_buff * skb)5203 static inline void skb_reset_redirect(struct sk_buff *skb)
5204 {
5205 	skb->redirected = 0;
5206 }
5207 
skb_set_redirected_noclear(struct sk_buff * skb,bool from_ingress)5208 static inline void skb_set_redirected_noclear(struct sk_buff *skb,
5209 					      bool from_ingress)
5210 {
5211 	skb->redirected = 1;
5212 #ifdef CONFIG_NET_REDIRECT
5213 	skb->from_ingress = from_ingress;
5214 #endif
5215 }
5216 
skb_csum_is_sctp(struct sk_buff * skb)5217 static inline bool skb_csum_is_sctp(struct sk_buff *skb)
5218 {
5219 #if IS_ENABLED(CONFIG_IP_SCTP)
5220 	return skb->csum_not_inet;
5221 #else
5222 	return 0;
5223 #endif
5224 }
5225 
skb_reset_csum_not_inet(struct sk_buff * skb)5226 static inline void skb_reset_csum_not_inet(struct sk_buff *skb)
5227 {
5228 	skb->ip_summed = CHECKSUM_NONE;
5229 #if IS_ENABLED(CONFIG_IP_SCTP)
5230 	skb->csum_not_inet = 0;
5231 #endif
5232 }
5233 
skb_set_kcov_handle(struct sk_buff * skb,const u64 kcov_handle)5234 static inline void skb_set_kcov_handle(struct sk_buff *skb,
5235 				       const u64 kcov_handle)
5236 {
5237 #ifdef CONFIG_KCOV
5238 	skb->kcov_handle = kcov_handle;
5239 #endif
5240 }
5241 
skb_get_kcov_handle(struct sk_buff * skb)5242 static inline u64 skb_get_kcov_handle(struct sk_buff *skb)
5243 {
5244 #ifdef CONFIG_KCOV
5245 	return skb->kcov_handle;
5246 #else
5247 	return 0;
5248 #endif
5249 }
5250 
skb_mark_for_recycle(struct sk_buff * skb)5251 static inline void skb_mark_for_recycle(struct sk_buff *skb)
5252 {
5253 #ifdef CONFIG_PAGE_POOL
5254 	skb->pp_recycle = 1;
5255 #endif
5256 }
5257 
5258 ssize_t skb_splice_from_iter(struct sk_buff *skb, struct iov_iter *iter,
5259 			     ssize_t maxsize, gfp_t gfp);
5260 
5261 #endif	/* __KERNEL__ */
5262 #endif	/* _LINUX_SKBUFF_H */
5263