blob: 8688f7adfda71d955fb93004ade88cb80ab2c5b7 [file] [log] [blame]
David Brazdil0f672f62019-12-10 10:32:29 +00001/* SPDX-License-Identifier: GPL-2.0-or-later */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002/*
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>
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00008 */
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>
David Brazdil0f672f62019-12-10 10:32:29 +000017#include <linux/bvec.h>
Andrew Scullb4b6d4a2019-01-02 15:54:55 +000018#include <linux/cache.h>
19#include <linux/rbtree.h>
20#include <linux/socket.h>
21#include <linux/refcount.h>
22
23#include <linux/atomic.h>
24#include <asm/types.h>
25#include <linux/spinlock.h>
26#include <linux/net.h>
27#include <linux/textsearch.h>
28#include <net/checksum.h>
29#include <linux/rcupdate.h>
30#include <linux/hrtimer.h>
31#include <linux/dma-mapping.h>
32#include <linux/netdev_features.h>
33#include <linux/sched.h>
34#include <linux/sched/clock.h>
35#include <net/flow_dissector.h>
36#include <linux/splice.h>
37#include <linux/in6.h>
38#include <linux/if_packet.h>
39#include <net/flow.h>
David Brazdil0f672f62019-12-10 10:32:29 +000040#if IS_ENABLED(CONFIG_NF_CONNTRACK)
41#include <linux/netfilter/nf_conntrack_common.h>
42#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +000043
44/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
89 *
90 * CHECKSUM_NONE:
91 *
92 * Device did not checksum this packet e.g. due to lack of capabilities.
93 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
112 * FCOE: indicates the CRC in FC frame has been validated.
113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
134 *
135 * CHECKSUM_PARTIAL:
136 *
137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
146 *
147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
149 *
150 * CHECKSUM_PARTIAL:
151 *
152 * The driver is required to checksum the packet as seen by hard_start_xmit()
153 * from skb->csum_start up to the end, and to record/write the checksum at
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
173 *
174 * CHECKSUM_NONE:
175 *
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
178 *
179 * CHECKSUM_UNNECESSARY:
180 *
181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
183 *
184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
187 *
188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
217 */
218
219/* Don't change this without changing skb_csum_unnecessary! */
220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
224
225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
229#define SKB_WITH_OVERHEAD(X) \
230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
241struct net_device;
242struct scatterlist;
243struct pipe_inode_info;
244struct iov_iter;
245struct napi_struct;
David Brazdil0f672f62019-12-10 10:32:29 +0000246struct bpf_prog;
247union bpf_attr;
248struct skb_ext;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000249
250#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
251struct nf_bridge_info {
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000252 enum {
253 BRNF_PROTO_UNCHANGED,
254 BRNF_PROTO_8021Q,
255 BRNF_PROTO_PPPOE
256 } orig_proto:8;
257 u8 pkt_otherhost:1;
258 u8 in_prerouting:1;
259 u8 bridged_dnat:1;
260 __u16 frag_max_size;
261 struct net_device *physindev;
262
263 /* always valid & non-NULL from FORWARD on, for physdev match */
264 struct net_device *physoutdev;
265 union {
266 /* prerouting: detect dnat in orig/reply direction */
267 __be32 ipv4_daddr;
268 struct in6_addr ipv6_daddr;
269
270 /* after prerouting + nat detected: store original source
271 * mac since neigh resolution overwrites it, only used while
272 * skb is out in neigh layer.
273 */
274 char neigh_header[8];
275 };
276};
277#endif
278
David Brazdil0f672f62019-12-10 10:32:29 +0000279#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
280/* Chain in tc_skb_ext will be used to share the tc chain with
281 * ovs recirc_id. It will be set to the current chain by tc
282 * and read by ovs to recirc_id.
283 */
284struct tc_skb_ext {
285 __u32 chain;
286};
287#endif
288
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000289struct sk_buff_head {
290 /* These two members must be first. */
291 struct sk_buff *next;
292 struct sk_buff *prev;
293
294 __u32 qlen;
295 spinlock_t lock;
296};
297
298struct sk_buff;
299
300/* To allow 64K frame to be packed as single skb without frag_list we
301 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
302 * buffers which do not start on a page boundary.
303 *
304 * Since GRO uses frags we allocate at least 16 regardless of page
305 * size.
306 */
307#if (65536/PAGE_SIZE + 1) < 16
308#define MAX_SKB_FRAGS 16UL
309#else
310#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
311#endif
312extern int sysctl_max_skb_frags;
313
314/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
315 * segment using its current segmentation instead.
316 */
317#define GSO_BY_FRAGS 0xFFFF
318
David Brazdil0f672f62019-12-10 10:32:29 +0000319typedef struct bio_vec skb_frag_t;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000320
David Brazdil0f672f62019-12-10 10:32:29 +0000321/**
322 * skb_frag_size() - Returns the size of a skb fragment
323 * @frag: skb fragment
324 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000325static inline unsigned int skb_frag_size(const skb_frag_t *frag)
326{
David Brazdil0f672f62019-12-10 10:32:29 +0000327 return frag->bv_len;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000328}
329
David Brazdil0f672f62019-12-10 10:32:29 +0000330/**
331 * skb_frag_size_set() - Sets the size of a skb fragment
332 * @frag: skb fragment
333 * @size: size of fragment
334 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000335static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
336{
David Brazdil0f672f62019-12-10 10:32:29 +0000337 frag->bv_len = size;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000338}
339
David Brazdil0f672f62019-12-10 10:32:29 +0000340/**
341 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
342 * @frag: skb fragment
343 * @delta: value to add
344 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000345static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
346{
David Brazdil0f672f62019-12-10 10:32:29 +0000347 frag->bv_len += delta;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000348}
349
David Brazdil0f672f62019-12-10 10:32:29 +0000350/**
351 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
352 * @frag: skb fragment
353 * @delta: value to subtract
354 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000355static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
356{
David Brazdil0f672f62019-12-10 10:32:29 +0000357 frag->bv_len -= delta;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000358}
359
David Brazdil0f672f62019-12-10 10:32:29 +0000360/**
361 * skb_frag_must_loop - Test if %p is a high memory page
362 * @p: fragment's page
363 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000364static inline bool skb_frag_must_loop(struct page *p)
365{
366#if defined(CONFIG_HIGHMEM)
367 if (PageHighMem(p))
368 return true;
369#endif
370 return false;
371}
372
373/**
374 * skb_frag_foreach_page - loop over pages in a fragment
375 *
376 * @f: skb frag to operate on
David Brazdil0f672f62019-12-10 10:32:29 +0000377 * @f_off: offset from start of f->bv_page
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000378 * @f_len: length from f_off to loop over
379 * @p: (temp var) current page
380 * @p_off: (temp var) offset from start of current page,
381 * non-zero only on first page.
382 * @p_len: (temp var) length in current page,
383 * < PAGE_SIZE only on first and last page.
384 * @copied: (temp var) length so far, excluding current p_len.
385 *
386 * A fragment can hold a compound page, in which case per-page
387 * operations, notably kmap_atomic, must be called for each
388 * regular page.
389 */
390#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
391 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
392 p_off = (f_off) & (PAGE_SIZE - 1), \
393 p_len = skb_frag_must_loop(p) ? \
394 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
395 copied = 0; \
396 copied < f_len; \
397 copied += p_len, p++, p_off = 0, \
398 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
399
400#define HAVE_HW_TIME_STAMP
401
402/**
403 * struct skb_shared_hwtstamps - hardware time stamps
404 * @hwtstamp: hardware time stamp transformed into duration
405 * since arbitrary point in time
406 *
407 * Software time stamps generated by ktime_get_real() are stored in
408 * skb->tstamp.
409 *
410 * hwtstamps can only be compared against other hwtstamps from
411 * the same device.
412 *
413 * This structure is attached to packets as part of the
414 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
415 */
416struct skb_shared_hwtstamps {
417 ktime_t hwtstamp;
418};
419
420/* Definitions for tx_flags in struct skb_shared_info */
421enum {
422 /* generate hardware time stamp */
423 SKBTX_HW_TSTAMP = 1 << 0,
424
425 /* generate software time stamp when queueing packet to NIC */
426 SKBTX_SW_TSTAMP = 1 << 1,
427
428 /* device driver is going to provide hardware time stamp */
429 SKBTX_IN_PROGRESS = 1 << 2,
430
431 /* device driver supports TX zero-copy buffers */
432 SKBTX_DEV_ZEROCOPY = 1 << 3,
433
434 /* generate wifi status information (where possible) */
435 SKBTX_WIFI_STATUS = 1 << 4,
436
437 /* This indicates at least one fragment might be overwritten
438 * (as in vmsplice(), sendfile() ...)
439 * If we need to compute a TX checksum, we'll need to copy
440 * all frags to avoid possible bad checksum
441 */
442 SKBTX_SHARED_FRAG = 1 << 5,
443
444 /* generate software time stamp when entering packet scheduling */
445 SKBTX_SCHED_TSTAMP = 1 << 6,
446};
447
448#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
449#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
450 SKBTX_SCHED_TSTAMP)
451#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
452
453/*
454 * The callback notifies userspace to release buffers when skb DMA is done in
455 * lower device, the skb last reference should be 0 when calling this.
456 * The zerocopy_success argument is true if zero copy transmit occurred,
457 * false on data copy or out of memory error caused by data copy attempt.
458 * The ctx field is used to track device context.
459 * The desc field is used to track userspace buffer index.
460 */
461struct ubuf_info {
462 void (*callback)(struct ubuf_info *, bool zerocopy_success);
463 union {
464 struct {
465 unsigned long desc;
466 void *ctx;
467 };
468 struct {
469 u32 id;
470 u16 len;
471 u16 zerocopy:1;
472 u32 bytelen;
473 };
474 };
475 refcount_t refcnt;
476
477 struct mmpin {
478 struct user_struct *user;
479 unsigned int num_pg;
480 } mmp;
481};
482
483#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
484
485int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
486void mm_unaccount_pinned_pages(struct mmpin *mmp);
487
488struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
489struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
490 struct ubuf_info *uarg);
491
492static inline void sock_zerocopy_get(struct ubuf_info *uarg)
493{
494 refcount_inc(&uarg->refcnt);
495}
496
497void sock_zerocopy_put(struct ubuf_info *uarg);
David Brazdil0f672f62019-12-10 10:32:29 +0000498void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000499
500void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
501
David Brazdil0f672f62019-12-10 10:32:29 +0000502int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000503int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
504 struct msghdr *msg, int len,
505 struct ubuf_info *uarg);
506
507/* This data is invariant across clones and lives at
508 * the end of the header data, ie. at skb->end.
509 */
510struct skb_shared_info {
511 __u8 __unused;
512 __u8 meta_len;
513 __u8 nr_frags;
514 __u8 tx_flags;
515 unsigned short gso_size;
516 /* Warning: this field is not always filled in (UFO)! */
517 unsigned short gso_segs;
518 struct sk_buff *frag_list;
519 struct skb_shared_hwtstamps hwtstamps;
520 unsigned int gso_type;
521 u32 tskey;
522
523 /*
524 * Warning : all fields before dataref are cleared in __alloc_skb()
525 */
526 atomic_t dataref;
527
528 /* Intermediate layers must ensure that destructor_arg
529 * remains valid until skb destructor */
530 void * destructor_arg;
531
532 /* must be last field, see pskb_expand_head() */
533 skb_frag_t frags[MAX_SKB_FRAGS];
534};
535
536/* We divide dataref into two halves. The higher 16 bits hold references
537 * to the payload part of skb->data. The lower 16 bits hold references to
538 * the entire skb->data. A clone of a headerless skb holds the length of
539 * the header in skb->hdr_len.
540 *
541 * All users must obey the rule that the skb->data reference count must be
542 * greater than or equal to the payload reference count.
543 *
544 * Holding a reference to the payload part means that the user does not
545 * care about modifications to the header part of skb->data.
546 */
547#define SKB_DATAREF_SHIFT 16
548#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
549
550
551enum {
552 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
553 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
554 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
555};
556
557enum {
558 SKB_GSO_TCPV4 = 1 << 0,
559
560 /* This indicates the skb is from an untrusted source. */
561 SKB_GSO_DODGY = 1 << 1,
562
563 /* This indicates the tcp segment has CWR set. */
564 SKB_GSO_TCP_ECN = 1 << 2,
565
566 SKB_GSO_TCP_FIXEDID = 1 << 3,
567
568 SKB_GSO_TCPV6 = 1 << 4,
569
570 SKB_GSO_FCOE = 1 << 5,
571
572 SKB_GSO_GRE = 1 << 6,
573
574 SKB_GSO_GRE_CSUM = 1 << 7,
575
576 SKB_GSO_IPXIP4 = 1 << 8,
577
578 SKB_GSO_IPXIP6 = 1 << 9,
579
580 SKB_GSO_UDP_TUNNEL = 1 << 10,
581
582 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
583
584 SKB_GSO_PARTIAL = 1 << 12,
585
586 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
587
588 SKB_GSO_SCTP = 1 << 14,
589
590 SKB_GSO_ESP = 1 << 15,
591
592 SKB_GSO_UDP = 1 << 16,
593
594 SKB_GSO_UDP_L4 = 1 << 17,
595};
596
597#if BITS_PER_LONG > 32
598#define NET_SKBUFF_DATA_USES_OFFSET 1
599#endif
600
601#ifdef NET_SKBUFF_DATA_USES_OFFSET
602typedef unsigned int sk_buff_data_t;
603#else
604typedef unsigned char *sk_buff_data_t;
605#endif
606
David Brazdil0f672f62019-12-10 10:32:29 +0000607/**
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000608 * struct sk_buff - socket buffer
609 * @next: Next buffer in list
610 * @prev: Previous buffer in list
611 * @tstamp: Time we arrived/left
612 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
613 * @sk: Socket we are owned by
614 * @dev: Device we arrived on/are leaving by
615 * @cb: Control buffer. Free for use by every layer. Put private vars here
616 * @_skb_refdst: destination entry (with norefcount bit)
617 * @sp: the security path, used for xfrm
618 * @len: Length of actual data
619 * @data_len: Data length
620 * @mac_len: Length of link layer header
621 * @hdr_len: writable header length of cloned skb
622 * @csum: Checksum (must include start/offset pair)
623 * @csum_start: Offset from skb->head where checksumming should start
624 * @csum_offset: Offset from csum_start where checksum should be stored
625 * @priority: Packet queueing priority
626 * @ignore_df: allow local fragmentation
627 * @cloned: Head may be cloned (check refcnt to be sure)
628 * @ip_summed: Driver fed us an IP checksum
629 * @nohdr: Payload reference only, must not modify header
630 * @pkt_type: Packet class
631 * @fclone: skbuff clone status
632 * @ipvs_property: skbuff is owned by ipvs
David Brazdil0f672f62019-12-10 10:32:29 +0000633 * @offload_fwd_mark: Packet was L2-forwarded in hardware
634 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000635 * @tc_skip_classify: do not classify packet. set by IFB device
636 * @tc_at_ingress: used within tc_classify to distinguish in/egress
637 * @tc_redirected: packet was redirected by a tc action
638 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
639 * @peeked: this packet has been seen already, so stats have been
640 * done for it, don't do them again
641 * @nf_trace: netfilter packet trace flag
642 * @protocol: Packet protocol from driver
643 * @destructor: Destruct function
644 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
645 * @_nfct: Associated connection, if any (with nfctinfo bits)
646 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
647 * @skb_iif: ifindex of device we arrived on
648 * @tc_index: Traffic control index
649 * @hash: the packet hash
650 * @queue_mapping: Queue mapping for multiqueue devices
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000651 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
David Brazdil0f672f62019-12-10 10:32:29 +0000652 * @active_extensions: active extensions (skb_ext_id types)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000653 * @ndisc_nodetype: router type (from link layer)
654 * @ooo_okay: allow the mapping of a socket to a queue to be changed
655 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
656 * ports.
657 * @sw_hash: indicates hash was computed in software stack
658 * @wifi_acked_valid: wifi_acked was set
659 * @wifi_acked: whether frame was acked on wifi or not
660 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
661 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
662 * @dst_pending_confirm: need to confirm neighbour
663 * @decrypted: Decrypted SKB
David Brazdil0f672f62019-12-10 10:32:29 +0000664 * @napi_id: id of the NAPI struct this skb came from
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000665 * @secmark: security marking
666 * @mark: Generic packet mark
667 * @vlan_proto: vlan encapsulation protocol
668 * @vlan_tci: vlan tag control information
669 * @inner_protocol: Protocol (encapsulation)
670 * @inner_transport_header: Inner transport layer header (encapsulation)
671 * @inner_network_header: Network layer header (encapsulation)
672 * @inner_mac_header: Link layer header (encapsulation)
673 * @transport_header: Transport layer header
674 * @network_header: Network layer header
675 * @mac_header: Link layer header
676 * @tail: Tail pointer
677 * @end: End pointer
678 * @head: Head of buffer
679 * @data: Data head pointer
680 * @truesize: Buffer size
681 * @users: User count - see {datagram,tcp}.c
David Brazdil0f672f62019-12-10 10:32:29 +0000682 * @extensions: allocated extensions, valid if active_extensions is nonzero
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000683 */
684
685struct sk_buff {
686 union {
687 struct {
688 /* These two members must be first. */
689 struct sk_buff *next;
690 struct sk_buff *prev;
691
692 union {
693 struct net_device *dev;
694 /* Some protocols might use this space to store information,
695 * while device pointer would be NULL.
696 * UDP receive path is one user.
697 */
698 unsigned long dev_scratch;
699 };
700 };
701 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
702 struct list_head list;
703 };
704
705 union {
706 struct sock *sk;
707 int ip_defrag_offset;
708 };
709
710 union {
711 ktime_t tstamp;
David Brazdil0f672f62019-12-10 10:32:29 +0000712 u64 skb_mstamp_ns; /* earliest departure time */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000713 };
714 /*
715 * This is the control buffer. It is free to use for every
716 * layer. Please put your private variables there. If you
717 * want to keep them across layers you have to do a skb_clone()
718 * first. This is owned by whoever has the skb queued ATM.
719 */
720 char cb[48] __aligned(8);
721
722 union {
723 struct {
724 unsigned long _skb_refdst;
725 void (*destructor)(struct sk_buff *skb);
726 };
727 struct list_head tcp_tsorted_anchor;
728 };
729
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000730#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
731 unsigned long _nfct;
732#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000733 unsigned int len,
734 data_len;
735 __u16 mac_len,
736 hdr_len;
737
738 /* Following fields are _not_ copied in __copy_skb_header()
739 * Note that queue_mapping is here mostly to fill a hole.
740 */
741 __u16 queue_mapping;
742
743/* if you move cloned around you also must adapt those constants */
744#ifdef __BIG_ENDIAN_BITFIELD
745#define CLONED_MASK (1 << 7)
746#else
747#define CLONED_MASK 1
748#endif
749#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
750
751 __u8 __cloned_offset[0];
752 __u8 cloned:1,
753 nohdr:1,
754 fclone:2,
755 peeked:1,
756 head_frag:1,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000757 pfmemalloc:1;
David Brazdil0f672f62019-12-10 10:32:29 +0000758#ifdef CONFIG_SKB_EXTENSIONS
759 __u8 active_extensions;
760#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000761 /* fields enclosed in headers_start/headers_end are copied
762 * using a single memcpy() in __copy_skb_header()
763 */
764 /* private: */
765 __u32 headers_start[0];
766 /* public: */
767
768/* if you move pkt_type around you also must adapt those constants */
769#ifdef __BIG_ENDIAN_BITFIELD
770#define PKT_TYPE_MAX (7 << 5)
771#else
772#define PKT_TYPE_MAX 7
773#endif
774#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
775
776 __u8 __pkt_type_offset[0];
777 __u8 pkt_type:3;
778 __u8 ignore_df:1;
779 __u8 nf_trace:1;
780 __u8 ip_summed:2;
781 __u8 ooo_okay:1;
782
783 __u8 l4_hash:1;
784 __u8 sw_hash:1;
785 __u8 wifi_acked_valid:1;
786 __u8 wifi_acked:1;
787 __u8 no_fcs:1;
788 /* Indicates the inner headers are valid in the skbuff. */
789 __u8 encapsulation:1;
790 __u8 encap_hdr_csum:1;
791 __u8 csum_valid:1;
792
David Brazdil0f672f62019-12-10 10:32:29 +0000793#ifdef __BIG_ENDIAN_BITFIELD
794#define PKT_VLAN_PRESENT_BIT 7
795#else
796#define PKT_VLAN_PRESENT_BIT 0
797#endif
798#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
799 __u8 __pkt_vlan_present_offset[0];
800 __u8 vlan_present:1;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000801 __u8 csum_complete_sw:1;
802 __u8 csum_level:2;
803 __u8 csum_not_inet:1;
804 __u8 dst_pending_confirm:1;
805#ifdef CONFIG_IPV6_NDISC_NODETYPE
806 __u8 ndisc_nodetype:2;
807#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000808
David Brazdil0f672f62019-12-10 10:32:29 +0000809 __u8 ipvs_property:1;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000810 __u8 inner_protocol_type:1;
811 __u8 remcsum_offload:1;
812#ifdef CONFIG_NET_SWITCHDEV
813 __u8 offload_fwd_mark:1;
David Brazdil0f672f62019-12-10 10:32:29 +0000814 __u8 offload_l3_fwd_mark:1;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000815#endif
816#ifdef CONFIG_NET_CLS_ACT
817 __u8 tc_skip_classify:1;
818 __u8 tc_at_ingress:1;
819 __u8 tc_redirected:1;
820 __u8 tc_from_ingress:1;
821#endif
822#ifdef CONFIG_TLS_DEVICE
823 __u8 decrypted:1;
824#endif
825
826#ifdef CONFIG_NET_SCHED
827 __u16 tc_index; /* traffic control index */
828#endif
829
830 union {
831 __wsum csum;
832 struct {
833 __u16 csum_start;
834 __u16 csum_offset;
835 };
836 };
837 __u32 priority;
838 int skb_iif;
839 __u32 hash;
840 __be16 vlan_proto;
841 __u16 vlan_tci;
842#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
843 union {
844 unsigned int napi_id;
845 unsigned int sender_cpu;
846 };
847#endif
848#ifdef CONFIG_NETWORK_SECMARK
849 __u32 secmark;
850#endif
851
852 union {
853 __u32 mark;
854 __u32 reserved_tailroom;
855 };
856
857 union {
858 __be16 inner_protocol;
859 __u8 inner_ipproto;
860 };
861
862 __u16 inner_transport_header;
863 __u16 inner_network_header;
864 __u16 inner_mac_header;
865
866 __be16 protocol;
867 __u16 transport_header;
868 __u16 network_header;
869 __u16 mac_header;
870
871 /* private: */
872 __u32 headers_end[0];
873 /* public: */
874
875 /* These elements must be at the end, see alloc_skb() for details. */
876 sk_buff_data_t tail;
877 sk_buff_data_t end;
878 unsigned char *head,
879 *data;
880 unsigned int truesize;
881 refcount_t users;
David Brazdil0f672f62019-12-10 10:32:29 +0000882
883#ifdef CONFIG_SKB_EXTENSIONS
884 /* only useable after checking ->active_extensions != 0 */
885 struct skb_ext *extensions;
886#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000887};
888
889#ifdef __KERNEL__
890/*
891 * Handling routines are only of interest to the kernel
892 */
893
894#define SKB_ALLOC_FCLONE 0x01
895#define SKB_ALLOC_RX 0x02
896#define SKB_ALLOC_NAPI 0x04
897
David Brazdil0f672f62019-12-10 10:32:29 +0000898/**
899 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
900 * @skb: buffer
901 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000902static inline bool skb_pfmemalloc(const struct sk_buff *skb)
903{
904 return unlikely(skb->pfmemalloc);
905}
906
907/*
908 * skb might have a dst pointer attached, refcounted or not.
909 * _skb_refdst low order bit is set if refcount was _not_ taken
910 */
911#define SKB_DST_NOREF 1UL
912#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
913
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000914/**
915 * skb_dst - returns skb dst_entry
916 * @skb: buffer
917 *
918 * Returns skb dst_entry, regardless of reference taken or not.
919 */
920static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
921{
David Brazdil0f672f62019-12-10 10:32:29 +0000922 /* If refdst was not refcounted, check we still are in a
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000923 * rcu_read_lock section
924 */
925 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
926 !rcu_read_lock_held() &&
927 !rcu_read_lock_bh_held());
928 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
929}
930
931/**
932 * skb_dst_set - sets skb dst
933 * @skb: buffer
934 * @dst: dst entry
935 *
936 * Sets skb dst, assuming a reference was taken on dst and should
937 * be released by skb_dst_drop()
938 */
939static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
940{
941 skb->_skb_refdst = (unsigned long)dst;
942}
943
944/**
945 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
946 * @skb: buffer
947 * @dst: dst entry
948 *
949 * Sets skb dst, assuming a reference was not taken on dst.
950 * If dst entry is cached, we do not take reference and dst_release
951 * will be avoided by refdst_drop. If dst entry is not cached, we take
952 * reference, so that last dst_release can destroy the dst immediately.
953 */
954static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
955{
956 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
957 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
958}
959
960/**
961 * skb_dst_is_noref - Test if skb dst isn't refcounted
962 * @skb: buffer
963 */
964static inline bool skb_dst_is_noref(const struct sk_buff *skb)
965{
966 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
967}
968
David Brazdil0f672f62019-12-10 10:32:29 +0000969/**
970 * skb_rtable - Returns the skb &rtable
971 * @skb: buffer
972 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000973static inline struct rtable *skb_rtable(const struct sk_buff *skb)
974{
975 return (struct rtable *)skb_dst(skb);
976}
977
978/* For mangling skb->pkt_type from user space side from applications
979 * such as nft, tc, etc, we only allow a conservative subset of
980 * possible pkt_types to be set.
981*/
982static inline bool skb_pkt_type_ok(u32 ptype)
983{
984 return ptype <= PACKET_OTHERHOST;
985}
986
David Brazdil0f672f62019-12-10 10:32:29 +0000987/**
988 * skb_napi_id - Returns the skb's NAPI id
989 * @skb: buffer
990 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +0000991static inline unsigned int skb_napi_id(const struct sk_buff *skb)
992{
993#ifdef CONFIG_NET_RX_BUSY_POLL
994 return skb->napi_id;
995#else
996 return 0;
997#endif
998}
999
David Brazdil0f672f62019-12-10 10:32:29 +00001000/**
1001 * skb_unref - decrement the skb's reference count
1002 * @skb: buffer
1003 *
1004 * Returns true if we can free the skb.
1005 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001006static inline bool skb_unref(struct sk_buff *skb)
1007{
1008 if (unlikely(!skb))
1009 return false;
1010 if (likely(refcount_read(&skb->users) == 1))
1011 smp_rmb();
1012 else if (likely(!refcount_dec_and_test(&skb->users)))
1013 return false;
1014
1015 return true;
1016}
1017
1018void skb_release_head_state(struct sk_buff *skb);
1019void kfree_skb(struct sk_buff *skb);
1020void kfree_skb_list(struct sk_buff *segs);
David Brazdil0f672f62019-12-10 10:32:29 +00001021void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001022void skb_tx_error(struct sk_buff *skb);
1023void consume_skb(struct sk_buff *skb);
1024void __consume_stateless_skb(struct sk_buff *skb);
1025void __kfree_skb(struct sk_buff *skb);
1026extern struct kmem_cache *skbuff_head_cache;
1027
1028void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1029bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1030 bool *fragstolen, int *delta_truesize);
1031
1032struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1033 int node);
1034struct sk_buff *__build_skb(void *data, unsigned int frag_size);
1035struct sk_buff *build_skb(void *data, unsigned int frag_size);
David Brazdil0f672f62019-12-10 10:32:29 +00001036struct sk_buff *build_skb_around(struct sk_buff *skb,
1037 void *data, unsigned int frag_size);
1038
1039/**
1040 * alloc_skb - allocate a network buffer
1041 * @size: size to allocate
1042 * @priority: allocation mask
1043 *
1044 * This function is a convenient wrapper around __alloc_skb().
1045 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001046static inline struct sk_buff *alloc_skb(unsigned int size,
1047 gfp_t priority)
1048{
1049 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
1050}
1051
1052struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1053 unsigned long data_len,
1054 int max_page_order,
1055 int *errcode,
1056 gfp_t gfp_mask);
David Brazdil0f672f62019-12-10 10:32:29 +00001057struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001058
1059/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1060struct sk_buff_fclones {
1061 struct sk_buff skb1;
1062
1063 struct sk_buff skb2;
1064
1065 refcount_t fclone_ref;
1066};
1067
1068/**
1069 * skb_fclone_busy - check if fclone is busy
1070 * @sk: socket
1071 * @skb: buffer
1072 *
1073 * Returns true if skb is a fast clone, and its clone is not freed.
1074 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1075 * so we also check that this didnt happen.
1076 */
1077static inline bool skb_fclone_busy(const struct sock *sk,
1078 const struct sk_buff *skb)
1079{
1080 const struct sk_buff_fclones *fclones;
1081
1082 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1083
1084 return skb->fclone == SKB_FCLONE_ORIG &&
1085 refcount_read(&fclones->fclone_ref) > 1 &&
1086 fclones->skb2.sk == sk;
1087}
1088
David Brazdil0f672f62019-12-10 10:32:29 +00001089/**
1090 * alloc_skb_fclone - allocate a network buffer from fclone cache
1091 * @size: size to allocate
1092 * @priority: allocation mask
1093 *
1094 * This function is a convenient wrapper around __alloc_skb().
1095 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001096static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
1097 gfp_t priority)
1098{
1099 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
1100}
1101
1102struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
1103void skb_headers_offset_update(struct sk_buff *skb, int off);
1104int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1105struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
1106void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
1107struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
1108struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1109 gfp_t gfp_mask, bool fclone);
1110static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1111 gfp_t gfp_mask)
1112{
1113 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1114}
1115
1116int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1117struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1118 unsigned int headroom);
1119struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1120 int newtailroom, gfp_t priority);
1121int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1122 int offset, int len);
1123int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1124 int offset, int len);
1125int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1126int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1127
1128/**
1129 * skb_pad - zero pad the tail of an skb
1130 * @skb: buffer to pad
1131 * @pad: space to pad
1132 *
1133 * Ensure that a buffer is followed by a padding area that is zero
1134 * filled. Used by network drivers which may DMA or transfer data
1135 * beyond the buffer end onto the wire.
1136 *
1137 * May return error in out of memory cases. The skb is freed on error.
1138 */
1139static inline int skb_pad(struct sk_buff *skb, int pad)
1140{
1141 return __skb_pad(skb, pad, true);
1142}
1143#define dev_kfree_skb(a) consume_skb(a)
1144
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001145int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1146 int offset, size_t size);
1147
1148struct skb_seq_state {
1149 __u32 lower_offset;
1150 __u32 upper_offset;
1151 __u32 frag_idx;
1152 __u32 stepped_offset;
1153 struct sk_buff *root_skb;
1154 struct sk_buff *cur_skb;
1155 __u8 *frag_data;
1156};
1157
1158void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1159 unsigned int to, struct skb_seq_state *st);
1160unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1161 struct skb_seq_state *st);
1162void skb_abort_seq_read(struct skb_seq_state *st);
1163
1164unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
1165 unsigned int to, struct ts_config *config);
1166
1167/*
1168 * Packet hash types specify the type of hash in skb_set_hash.
1169 *
1170 * Hash types refer to the protocol layer addresses which are used to
1171 * construct a packet's hash. The hashes are used to differentiate or identify
1172 * flows of the protocol layer for the hash type. Hash types are either
1173 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1174 *
1175 * Properties of hashes:
1176 *
1177 * 1) Two packets in different flows have different hash values
1178 * 2) Two packets in the same flow should have the same hash value
1179 *
1180 * A hash at a higher layer is considered to be more specific. A driver should
1181 * set the most specific hash possible.
1182 *
1183 * A driver cannot indicate a more specific hash than the layer at which a hash
1184 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1185 *
1186 * A driver may indicate a hash level which is less specific than the
1187 * actual layer the hash was computed on. For instance, a hash computed
1188 * at L4 may be considered an L3 hash. This should only be done if the
1189 * driver can't unambiguously determine that the HW computed the hash at
1190 * the higher layer. Note that the "should" in the second property above
1191 * permits this.
1192 */
1193enum pkt_hash_types {
1194 PKT_HASH_TYPE_NONE, /* Undefined type */
1195 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1196 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1197 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1198};
1199
1200static inline void skb_clear_hash(struct sk_buff *skb)
1201{
1202 skb->hash = 0;
1203 skb->sw_hash = 0;
1204 skb->l4_hash = 0;
1205}
1206
1207static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1208{
1209 if (!skb->l4_hash)
1210 skb_clear_hash(skb);
1211}
1212
1213static inline void
1214__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1215{
1216 skb->l4_hash = is_l4;
1217 skb->sw_hash = is_sw;
1218 skb->hash = hash;
1219}
1220
1221static inline void
1222skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1223{
1224 /* Used by drivers to set hash from HW */
1225 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1226}
1227
1228static inline void
1229__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1230{
1231 __skb_set_hash(skb, hash, true, is_l4);
1232}
1233
1234void __skb_get_hash(struct sk_buff *skb);
1235u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
1236u32 skb_get_poff(const struct sk_buff *skb);
1237u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1238 const struct flow_keys_basic *keys, int hlen);
1239__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1240 void *data, int hlen_proto);
1241
1242static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1243 int thoff, u8 ip_proto)
1244{
1245 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1246}
1247
1248void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1249 const struct flow_dissector_key *key,
1250 unsigned int key_count);
1251
David Brazdil0f672f62019-12-10 10:32:29 +00001252#ifdef CONFIG_NET
1253int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1254 union bpf_attr __user *uattr);
1255int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1256 struct bpf_prog *prog);
1257
1258int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
1259#else
1260static inline int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1261 union bpf_attr __user *uattr)
1262{
1263 return -EOPNOTSUPP;
1264}
1265
1266static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1267 struct bpf_prog *prog)
1268{
1269 return -EOPNOTSUPP;
1270}
1271
1272static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
1273{
1274 return -EOPNOTSUPP;
1275}
1276#endif
1277
1278struct bpf_flow_dissector;
1279bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1280 __be16 proto, int nhoff, int hlen, unsigned int flags);
1281
1282bool __skb_flow_dissect(const struct net *net,
1283 const struct sk_buff *skb,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001284 struct flow_dissector *flow_dissector,
1285 void *target_container,
1286 void *data, __be16 proto, int nhoff, int hlen,
1287 unsigned int flags);
1288
1289static inline bool skb_flow_dissect(const struct sk_buff *skb,
1290 struct flow_dissector *flow_dissector,
1291 void *target_container, unsigned int flags)
1292{
David Brazdil0f672f62019-12-10 10:32:29 +00001293 return __skb_flow_dissect(NULL, skb, flow_dissector,
1294 target_container, NULL, 0, 0, 0, flags);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001295}
1296
1297static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
1298 struct flow_keys *flow,
1299 unsigned int flags)
1300{
1301 memset(flow, 0, sizeof(*flow));
David Brazdil0f672f62019-12-10 10:32:29 +00001302 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1303 flow, NULL, 0, 0, 0, flags);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001304}
1305
1306static inline bool
David Brazdil0f672f62019-12-10 10:32:29 +00001307skb_flow_dissect_flow_keys_basic(const struct net *net,
1308 const struct sk_buff *skb,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001309 struct flow_keys_basic *flow, void *data,
1310 __be16 proto, int nhoff, int hlen,
1311 unsigned int flags)
1312{
1313 memset(flow, 0, sizeof(*flow));
David Brazdil0f672f62019-12-10 10:32:29 +00001314 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001315 data, proto, nhoff, hlen, flags);
1316}
1317
David Brazdil0f672f62019-12-10 10:32:29 +00001318void skb_flow_dissect_meta(const struct sk_buff *skb,
1319 struct flow_dissector *flow_dissector,
1320 void *target_container);
1321
1322/* Gets a skb connection tracking info, ctinfo map should be a
1323 * a map of mapsize to translate enum ip_conntrack_info states
1324 * to user states.
1325 */
1326void
1327skb_flow_dissect_ct(const struct sk_buff *skb,
1328 struct flow_dissector *flow_dissector,
1329 void *target_container,
1330 u16 *ctinfo_map,
1331 size_t mapsize);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001332void
1333skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1334 struct flow_dissector *flow_dissector,
1335 void *target_container);
1336
1337static inline __u32 skb_get_hash(struct sk_buff *skb)
1338{
1339 if (!skb->l4_hash && !skb->sw_hash)
1340 __skb_get_hash(skb);
1341
1342 return skb->hash;
1343}
1344
1345static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
1346{
1347 if (!skb->l4_hash && !skb->sw_hash) {
1348 struct flow_keys keys;
1349 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
1350
1351 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1352 }
1353
1354 return skb->hash;
1355}
1356
David Brazdil0f672f62019-12-10 10:32:29 +00001357__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1358 const siphash_key_t *perturb);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001359
1360static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1361{
1362 return skb->hash;
1363}
1364
1365static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1366{
1367 to->hash = from->hash;
1368 to->sw_hash = from->sw_hash;
1369 to->l4_hash = from->l4_hash;
1370};
1371
David Brazdil0f672f62019-12-10 10:32:29 +00001372static inline void skb_copy_decrypted(struct sk_buff *to,
1373 const struct sk_buff *from)
1374{
1375#ifdef CONFIG_TLS_DEVICE
1376 to->decrypted = from->decrypted;
1377#endif
1378}
1379
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001380#ifdef NET_SKBUFF_DATA_USES_OFFSET
1381static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1382{
1383 return skb->head + skb->end;
1384}
1385
1386static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1387{
1388 return skb->end;
1389}
1390#else
1391static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1392{
1393 return skb->end;
1394}
1395
1396static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1397{
1398 return skb->end - skb->head;
1399}
1400#endif
1401
1402/* Internal */
1403#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1404
1405static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1406{
1407 return &skb_shinfo(skb)->hwtstamps;
1408}
1409
1410static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1411{
1412 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1413
1414 return is_zcopy ? skb_uarg(skb) : NULL;
1415}
1416
David Brazdil0f672f62019-12-10 10:32:29 +00001417static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1418 bool *have_ref)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001419{
1420 if (skb && uarg && !skb_zcopy(skb)) {
David Brazdil0f672f62019-12-10 10:32:29 +00001421 if (unlikely(have_ref && *have_ref))
1422 *have_ref = false;
1423 else
1424 sock_zerocopy_get(uarg);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001425 skb_shinfo(skb)->destructor_arg = uarg;
1426 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1427 }
1428}
1429
1430static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1431{
1432 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1433 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1434}
1435
1436static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1437{
1438 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1439}
1440
1441static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1442{
1443 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1444}
1445
1446/* Release a reference on a zerocopy structure */
1447static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1448{
1449 struct ubuf_info *uarg = skb_zcopy(skb);
1450
1451 if (uarg) {
David Brazdil0f672f62019-12-10 10:32:29 +00001452 if (skb_zcopy_is_nouarg(skb)) {
1453 /* no notification callback */
1454 } else if (uarg->callback == sock_zerocopy_callback) {
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001455 uarg->zerocopy = uarg->zerocopy && zerocopy;
1456 sock_zerocopy_put(uarg);
David Brazdil0f672f62019-12-10 10:32:29 +00001457 } else {
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001458 uarg->callback(uarg, zerocopy);
1459 }
1460
1461 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1462 }
1463}
1464
1465/* Abort a zerocopy operation and revert zckey on error in send syscall */
1466static inline void skb_zcopy_abort(struct sk_buff *skb)
1467{
1468 struct ubuf_info *uarg = skb_zcopy(skb);
1469
1470 if (uarg) {
David Brazdil0f672f62019-12-10 10:32:29 +00001471 sock_zerocopy_put_abort(uarg, false);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001472 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1473 }
1474}
1475
1476static inline void skb_mark_not_on_list(struct sk_buff *skb)
1477{
1478 skb->next = NULL;
1479}
1480
1481static inline void skb_list_del_init(struct sk_buff *skb)
1482{
1483 __list_del_entry(&skb->list);
1484 skb_mark_not_on_list(skb);
1485}
1486
1487/**
1488 * skb_queue_empty - check if a queue is empty
1489 * @list: queue head
1490 *
1491 * Returns true if the queue is empty, false otherwise.
1492 */
1493static inline int skb_queue_empty(const struct sk_buff_head *list)
1494{
1495 return list->next == (const struct sk_buff *) list;
1496}
1497
1498/**
David Brazdil0f672f62019-12-10 10:32:29 +00001499 * skb_queue_empty_lockless - check if a queue is empty
1500 * @list: queue head
1501 *
1502 * Returns true if the queue is empty, false otherwise.
1503 * This variant can be used in lockless contexts.
1504 */
1505static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1506{
1507 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1508}
1509
1510
1511/**
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001512 * skb_queue_is_last - check if skb is the last entry in the queue
1513 * @list: queue head
1514 * @skb: buffer
1515 *
1516 * Returns true if @skb is the last buffer on the list.
1517 */
1518static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1519 const struct sk_buff *skb)
1520{
1521 return skb->next == (const struct sk_buff *) list;
1522}
1523
1524/**
1525 * skb_queue_is_first - check if skb is the first entry in the queue
1526 * @list: queue head
1527 * @skb: buffer
1528 *
1529 * Returns true if @skb is the first buffer on the list.
1530 */
1531static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1532 const struct sk_buff *skb)
1533{
1534 return skb->prev == (const struct sk_buff *) list;
1535}
1536
1537/**
1538 * skb_queue_next - return the next packet in the queue
1539 * @list: queue head
1540 * @skb: current buffer
1541 *
1542 * Return the next packet in @list after @skb. It is only valid to
1543 * call this if skb_queue_is_last() evaluates to false.
1544 */
1545static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1546 const struct sk_buff *skb)
1547{
1548 /* This BUG_ON may seem severe, but if we just return then we
1549 * are going to dereference garbage.
1550 */
1551 BUG_ON(skb_queue_is_last(list, skb));
1552 return skb->next;
1553}
1554
1555/**
1556 * skb_queue_prev - return the prev packet in the queue
1557 * @list: queue head
1558 * @skb: current buffer
1559 *
1560 * Return the prev packet in @list before @skb. It is only valid to
1561 * call this if skb_queue_is_first() evaluates to false.
1562 */
1563static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1564 const struct sk_buff *skb)
1565{
1566 /* This BUG_ON may seem severe, but if we just return then we
1567 * are going to dereference garbage.
1568 */
1569 BUG_ON(skb_queue_is_first(list, skb));
1570 return skb->prev;
1571}
1572
1573/**
1574 * skb_get - reference buffer
1575 * @skb: buffer to reference
1576 *
1577 * Makes another reference to a socket buffer and returns a pointer
1578 * to the buffer.
1579 */
1580static inline struct sk_buff *skb_get(struct sk_buff *skb)
1581{
1582 refcount_inc(&skb->users);
1583 return skb;
1584}
1585
1586/*
1587 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1588 */
1589
1590/**
1591 * skb_cloned - is the buffer a clone
1592 * @skb: buffer to check
1593 *
1594 * Returns true if the buffer was generated with skb_clone() and is
1595 * one of multiple shared copies of the buffer. Cloned buffers are
1596 * shared data so must not be written to under normal circumstances.
1597 */
1598static inline int skb_cloned(const struct sk_buff *skb)
1599{
1600 return skb->cloned &&
1601 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1602}
1603
1604static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1605{
1606 might_sleep_if(gfpflags_allow_blocking(pri));
1607
1608 if (skb_cloned(skb))
1609 return pskb_expand_head(skb, 0, 0, pri);
1610
1611 return 0;
1612}
1613
1614/**
1615 * skb_header_cloned - is the header a clone
1616 * @skb: buffer to check
1617 *
1618 * Returns true if modifying the header part of the buffer requires
1619 * the data to be copied.
1620 */
1621static inline int skb_header_cloned(const struct sk_buff *skb)
1622{
1623 int dataref;
1624
1625 if (!skb->cloned)
1626 return 0;
1627
1628 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1629 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1630 return dataref != 1;
1631}
1632
1633static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1634{
1635 might_sleep_if(gfpflags_allow_blocking(pri));
1636
1637 if (skb_header_cloned(skb))
1638 return pskb_expand_head(skb, 0, 0, pri);
1639
1640 return 0;
1641}
1642
1643/**
1644 * __skb_header_release - release reference to header
1645 * @skb: buffer to operate on
1646 */
1647static inline void __skb_header_release(struct sk_buff *skb)
1648{
1649 skb->nohdr = 1;
1650 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1651}
1652
1653
1654/**
1655 * skb_shared - is the buffer shared
1656 * @skb: buffer to check
1657 *
1658 * Returns true if more than one person has a reference to this
1659 * buffer.
1660 */
1661static inline int skb_shared(const struct sk_buff *skb)
1662{
1663 return refcount_read(&skb->users) != 1;
1664}
1665
1666/**
1667 * skb_share_check - check if buffer is shared and if so clone it
1668 * @skb: buffer to check
1669 * @pri: priority for memory allocation
1670 *
1671 * If the buffer is shared the buffer is cloned and the old copy
1672 * drops a reference. A new clone with a single reference is returned.
1673 * If the buffer is not shared the original buffer is returned. When
1674 * being called from interrupt status or with spinlocks held pri must
1675 * be GFP_ATOMIC.
1676 *
1677 * NULL is returned on a memory allocation failure.
1678 */
1679static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1680{
1681 might_sleep_if(gfpflags_allow_blocking(pri));
1682 if (skb_shared(skb)) {
1683 struct sk_buff *nskb = skb_clone(skb, pri);
1684
1685 if (likely(nskb))
1686 consume_skb(skb);
1687 else
1688 kfree_skb(skb);
1689 skb = nskb;
1690 }
1691 return skb;
1692}
1693
1694/*
1695 * Copy shared buffers into a new sk_buff. We effectively do COW on
1696 * packets to handle cases where we have a local reader and forward
1697 * and a couple of other messy ones. The normal one is tcpdumping
1698 * a packet thats being forwarded.
1699 */
1700
1701/**
1702 * skb_unshare - make a copy of a shared buffer
1703 * @skb: buffer to check
1704 * @pri: priority for memory allocation
1705 *
1706 * If the socket buffer is a clone then this function creates a new
1707 * copy of the data, drops a reference count on the old copy and returns
1708 * the new copy with the reference count at 1. If the buffer is not a clone
1709 * the original buffer is returned. When called with a spinlock held or
1710 * from interrupt state @pri must be %GFP_ATOMIC
1711 *
1712 * %NULL is returned on a memory allocation failure.
1713 */
1714static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1715 gfp_t pri)
1716{
1717 might_sleep_if(gfpflags_allow_blocking(pri));
1718 if (skb_cloned(skb)) {
1719 struct sk_buff *nskb = skb_copy(skb, pri);
1720
1721 /* Free our shared copy */
1722 if (likely(nskb))
1723 consume_skb(skb);
1724 else
1725 kfree_skb(skb);
1726 skb = nskb;
1727 }
1728 return skb;
1729}
1730
1731/**
1732 * skb_peek - peek at the head of an &sk_buff_head
1733 * @list_: list to peek at
1734 *
1735 * Peek an &sk_buff. Unlike most other operations you _MUST_
1736 * be careful with this one. A peek leaves the buffer on the
1737 * list and someone else may run off with it. You must hold
1738 * the appropriate locks or have a private queue to do this.
1739 *
1740 * Returns %NULL for an empty list or a pointer to the head element.
1741 * The reference count is not incremented and the reference is therefore
1742 * volatile. Use with caution.
1743 */
1744static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1745{
1746 struct sk_buff *skb = list_->next;
1747
1748 if (skb == (struct sk_buff *)list_)
1749 skb = NULL;
1750 return skb;
1751}
1752
1753/**
David Brazdil0f672f62019-12-10 10:32:29 +00001754 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1755 * @list_: list to peek at
1756 *
1757 * Like skb_peek(), but the caller knows that the list is not empty.
1758 */
1759static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1760{
1761 return list_->next;
1762}
1763
1764/**
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001765 * skb_peek_next - peek skb following the given one from a queue
1766 * @skb: skb to start from
1767 * @list_: list to peek at
1768 *
1769 * Returns %NULL when the end of the list is met or a pointer to the
1770 * next element. The reference count is not incremented and the
1771 * reference is therefore volatile. Use with caution.
1772 */
1773static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1774 const struct sk_buff_head *list_)
1775{
1776 struct sk_buff *next = skb->next;
1777
1778 if (next == (struct sk_buff *)list_)
1779 next = NULL;
1780 return next;
1781}
1782
1783/**
1784 * skb_peek_tail - peek at the tail of an &sk_buff_head
1785 * @list_: list to peek at
1786 *
1787 * Peek an &sk_buff. Unlike most other operations you _MUST_
1788 * be careful with this one. A peek leaves the buffer on the
1789 * list and someone else may run off with it. You must hold
1790 * the appropriate locks or have a private queue to do this.
1791 *
1792 * Returns %NULL for an empty list or a pointer to the tail element.
1793 * The reference count is not incremented and the reference is therefore
1794 * volatile. Use with caution.
1795 */
1796static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1797{
1798 struct sk_buff *skb = list_->prev;
1799
1800 if (skb == (struct sk_buff *)list_)
1801 skb = NULL;
1802 return skb;
1803
1804}
1805
1806/**
1807 * skb_queue_len - get queue length
1808 * @list_: list to measure
1809 *
1810 * Return the length of an &sk_buff queue.
1811 */
1812static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1813{
1814 return list_->qlen;
1815}
1816
1817/**
1818 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1819 * @list: queue to initialize
1820 *
1821 * This initializes only the list and queue length aspects of
1822 * an sk_buff_head object. This allows to initialize the list
1823 * aspects of an sk_buff_head without reinitializing things like
1824 * the spinlock. It can also be used for on-stack sk_buff_head
1825 * objects where the spinlock is known to not be used.
1826 */
1827static inline void __skb_queue_head_init(struct sk_buff_head *list)
1828{
1829 list->prev = list->next = (struct sk_buff *)list;
1830 list->qlen = 0;
1831}
1832
1833/*
1834 * This function creates a split out lock class for each invocation;
1835 * this is needed for now since a whole lot of users of the skb-queue
1836 * infrastructure in drivers have different locking usage (in hardirq)
1837 * than the networking core (in softirq only). In the long run either the
1838 * network layer or drivers should need annotation to consolidate the
1839 * main types of usage into 3 classes.
1840 */
1841static inline void skb_queue_head_init(struct sk_buff_head *list)
1842{
1843 spin_lock_init(&list->lock);
1844 __skb_queue_head_init(list);
1845}
1846
1847static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1848 struct lock_class_key *class)
1849{
1850 skb_queue_head_init(list);
1851 lockdep_set_class(&list->lock, class);
1852}
1853
1854/*
1855 * Insert an sk_buff on a list.
1856 *
1857 * The "__skb_xxxx()" functions are the non-atomic ones that
1858 * can only be called with interrupts disabled.
1859 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001860static inline void __skb_insert(struct sk_buff *newsk,
1861 struct sk_buff *prev, struct sk_buff *next,
1862 struct sk_buff_head *list)
1863{
David Brazdil0f672f62019-12-10 10:32:29 +00001864 /* see skb_queue_empty_lockless() for the opposite READ_ONCE() */
1865 WRITE_ONCE(newsk->next, next);
1866 WRITE_ONCE(newsk->prev, prev);
1867 WRITE_ONCE(next->prev, newsk);
1868 WRITE_ONCE(prev->next, newsk);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001869 list->qlen++;
1870}
1871
1872static inline void __skb_queue_splice(const struct sk_buff_head *list,
1873 struct sk_buff *prev,
1874 struct sk_buff *next)
1875{
1876 struct sk_buff *first = list->next;
1877 struct sk_buff *last = list->prev;
1878
David Brazdil0f672f62019-12-10 10:32:29 +00001879 WRITE_ONCE(first->prev, prev);
1880 WRITE_ONCE(prev->next, first);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001881
David Brazdil0f672f62019-12-10 10:32:29 +00001882 WRITE_ONCE(last->next, next);
1883 WRITE_ONCE(next->prev, last);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001884}
1885
1886/**
1887 * skb_queue_splice - join two skb lists, this is designed for stacks
1888 * @list: the new list to add
1889 * @head: the place to add it in the first list
1890 */
1891static inline void skb_queue_splice(const struct sk_buff_head *list,
1892 struct sk_buff_head *head)
1893{
1894 if (!skb_queue_empty(list)) {
1895 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1896 head->qlen += list->qlen;
1897 }
1898}
1899
1900/**
1901 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1902 * @list: the new list to add
1903 * @head: the place to add it in the first list
1904 *
1905 * The list at @list is reinitialised
1906 */
1907static inline void skb_queue_splice_init(struct sk_buff_head *list,
1908 struct sk_buff_head *head)
1909{
1910 if (!skb_queue_empty(list)) {
1911 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1912 head->qlen += list->qlen;
1913 __skb_queue_head_init(list);
1914 }
1915}
1916
1917/**
1918 * skb_queue_splice_tail - join two skb lists, each list being a queue
1919 * @list: the new list to add
1920 * @head: the place to add it in the first list
1921 */
1922static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1923 struct sk_buff_head *head)
1924{
1925 if (!skb_queue_empty(list)) {
1926 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1927 head->qlen += list->qlen;
1928 }
1929}
1930
1931/**
1932 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1933 * @list: the new list to add
1934 * @head: the place to add it in the first list
1935 *
1936 * Each of the lists is a queue.
1937 * The list at @list is reinitialised
1938 */
1939static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1940 struct sk_buff_head *head)
1941{
1942 if (!skb_queue_empty(list)) {
1943 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1944 head->qlen += list->qlen;
1945 __skb_queue_head_init(list);
1946 }
1947}
1948
1949/**
1950 * __skb_queue_after - queue a buffer at the list head
1951 * @list: list to use
1952 * @prev: place after this buffer
1953 * @newsk: buffer to queue
1954 *
1955 * Queue a buffer int the middle of a list. This function takes no locks
1956 * and you must therefore hold required locks before calling it.
1957 *
1958 * A buffer cannot be placed on two lists at the same time.
1959 */
1960static inline void __skb_queue_after(struct sk_buff_head *list,
1961 struct sk_buff *prev,
1962 struct sk_buff *newsk)
1963{
1964 __skb_insert(newsk, prev, prev->next, list);
1965}
1966
1967void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1968 struct sk_buff_head *list);
1969
1970static inline void __skb_queue_before(struct sk_buff_head *list,
1971 struct sk_buff *next,
1972 struct sk_buff *newsk)
1973{
1974 __skb_insert(newsk, next->prev, next, list);
1975}
1976
1977/**
1978 * __skb_queue_head - queue a buffer at the list head
1979 * @list: list to use
1980 * @newsk: buffer to queue
1981 *
1982 * Queue a buffer at the start of a list. This function takes no locks
1983 * and you must therefore hold required locks before calling it.
1984 *
1985 * A buffer cannot be placed on two lists at the same time.
1986 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001987static inline void __skb_queue_head(struct sk_buff_head *list,
1988 struct sk_buff *newsk)
1989{
1990 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1991}
David Brazdil0f672f62019-12-10 10:32:29 +00001992void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001993
1994/**
1995 * __skb_queue_tail - queue a buffer at the list tail
1996 * @list: list to use
1997 * @newsk: buffer to queue
1998 *
1999 * Queue a buffer at the end of a list. This function takes no locks
2000 * and you must therefore hold required locks before calling it.
2001 *
2002 * A buffer cannot be placed on two lists at the same time.
2003 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002004static inline void __skb_queue_tail(struct sk_buff_head *list,
2005 struct sk_buff *newsk)
2006{
2007 __skb_queue_before(list, (struct sk_buff *)list, newsk);
2008}
David Brazdil0f672f62019-12-10 10:32:29 +00002009void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002010
2011/*
2012 * remove sk_buff from list. _Must_ be called atomically, and with
2013 * the list known..
2014 */
2015void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
2016static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2017{
2018 struct sk_buff *next, *prev;
2019
2020 list->qlen--;
2021 next = skb->next;
2022 prev = skb->prev;
2023 skb->next = skb->prev = NULL;
David Brazdil0f672f62019-12-10 10:32:29 +00002024 WRITE_ONCE(next->prev, prev);
2025 WRITE_ONCE(prev->next, next);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002026}
2027
2028/**
2029 * __skb_dequeue - remove from the head of the queue
2030 * @list: list to dequeue from
2031 *
2032 * Remove the head of the list. This function does not take any locks
2033 * so must be used with appropriate locks held only. The head item is
2034 * returned or %NULL if the list is empty.
2035 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002036static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2037{
2038 struct sk_buff *skb = skb_peek(list);
2039 if (skb)
2040 __skb_unlink(skb, list);
2041 return skb;
2042}
David Brazdil0f672f62019-12-10 10:32:29 +00002043struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002044
2045/**
2046 * __skb_dequeue_tail - remove from the tail of the queue
2047 * @list: list to dequeue from
2048 *
2049 * Remove the tail of the list. This function does not take any locks
2050 * so must be used with appropriate locks held only. The tail item is
2051 * returned or %NULL if the list is empty.
2052 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002053static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2054{
2055 struct sk_buff *skb = skb_peek_tail(list);
2056 if (skb)
2057 __skb_unlink(skb, list);
2058 return skb;
2059}
David Brazdil0f672f62019-12-10 10:32:29 +00002060struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002061
2062
2063static inline bool skb_is_nonlinear(const struct sk_buff *skb)
2064{
2065 return skb->data_len;
2066}
2067
2068static inline unsigned int skb_headlen(const struct sk_buff *skb)
2069{
2070 return skb->len - skb->data_len;
2071}
2072
2073static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
2074{
2075 unsigned int i, len = 0;
2076
2077 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
2078 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
2079 return len;
2080}
2081
2082static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2083{
2084 return skb_headlen(skb) + __skb_pagelen(skb);
2085}
2086
2087/**
2088 * __skb_fill_page_desc - initialise a paged fragment in an skb
2089 * @skb: buffer containing fragment to be initialised
2090 * @i: paged fragment index to initialise
2091 * @page: the page to use for this fragment
2092 * @off: the offset to the data with @page
2093 * @size: the length of the data
2094 *
2095 * Initialises the @i'th fragment of @skb to point to &size bytes at
2096 * offset @off within @page.
2097 *
2098 * Does not take any additional reference on the fragment.
2099 */
2100static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2101 struct page *page, int off, int size)
2102{
2103 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2104
2105 /*
2106 * Propagate page pfmemalloc to the skb if we can. The problem is
2107 * that not all callers have unique ownership of the page but rely
2108 * on page_is_pfmemalloc doing the right thing(tm).
2109 */
David Brazdil0f672f62019-12-10 10:32:29 +00002110 frag->bv_page = page;
2111 frag->bv_offset = off;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002112 skb_frag_size_set(frag, size);
2113
2114 page = compound_head(page);
2115 if (page_is_pfmemalloc(page))
2116 skb->pfmemalloc = true;
2117}
2118
2119/**
2120 * skb_fill_page_desc - initialise a paged fragment in an skb
2121 * @skb: buffer containing fragment to be initialised
2122 * @i: paged fragment index to initialise
2123 * @page: the page to use for this fragment
2124 * @off: the offset to the data with @page
2125 * @size: the length of the data
2126 *
2127 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
2128 * @skb to point to @size bytes at offset @off within @page. In
2129 * addition updates @skb such that @i is the last fragment.
2130 *
2131 * Does not take any additional reference on the fragment.
2132 */
2133static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2134 struct page *page, int off, int size)
2135{
2136 __skb_fill_page_desc(skb, i, page, off, size);
2137 skb_shinfo(skb)->nr_frags = i + 1;
2138}
2139
2140void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2141 int size, unsigned int truesize);
2142
2143void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2144 unsigned int truesize);
2145
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002146#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2147
2148#ifdef NET_SKBUFF_DATA_USES_OFFSET
2149static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2150{
2151 return skb->head + skb->tail;
2152}
2153
2154static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2155{
2156 skb->tail = skb->data - skb->head;
2157}
2158
2159static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2160{
2161 skb_reset_tail_pointer(skb);
2162 skb->tail += offset;
2163}
2164
2165#else /* NET_SKBUFF_DATA_USES_OFFSET */
2166static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2167{
2168 return skb->tail;
2169}
2170
2171static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2172{
2173 skb->tail = skb->data;
2174}
2175
2176static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2177{
2178 skb->tail = skb->data + offset;
2179}
2180
2181#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2182
2183/*
2184 * Add data to an sk_buff
2185 */
2186void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2187void *skb_put(struct sk_buff *skb, unsigned int len);
2188static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
2189{
2190 void *tmp = skb_tail_pointer(skb);
2191 SKB_LINEAR_ASSERT(skb);
2192 skb->tail += len;
2193 skb->len += len;
2194 return tmp;
2195}
2196
2197static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2198{
2199 void *tmp = __skb_put(skb, len);
2200
2201 memset(tmp, 0, len);
2202 return tmp;
2203}
2204
2205static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2206 unsigned int len)
2207{
2208 void *tmp = __skb_put(skb, len);
2209
2210 memcpy(tmp, data, len);
2211 return tmp;
2212}
2213
2214static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2215{
2216 *(u8 *)__skb_put(skb, 1) = val;
2217}
2218
2219static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
2220{
2221 void *tmp = skb_put(skb, len);
2222
2223 memset(tmp, 0, len);
2224
2225 return tmp;
2226}
2227
2228static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2229 unsigned int len)
2230{
2231 void *tmp = skb_put(skb, len);
2232
2233 memcpy(tmp, data, len);
2234
2235 return tmp;
2236}
2237
2238static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2239{
2240 *(u8 *)skb_put(skb, 1) = val;
2241}
2242
2243void *skb_push(struct sk_buff *skb, unsigned int len);
2244static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
2245{
2246 skb->data -= len;
2247 skb->len += len;
2248 return skb->data;
2249}
2250
2251void *skb_pull(struct sk_buff *skb, unsigned int len);
2252static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
2253{
2254 skb->len -= len;
2255 BUG_ON(skb->len < skb->data_len);
2256 return skb->data += len;
2257}
2258
2259static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
2260{
2261 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2262}
2263
2264void *__pskb_pull_tail(struct sk_buff *skb, int delta);
2265
2266static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
2267{
2268 if (len > skb_headlen(skb) &&
2269 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
2270 return NULL;
2271 skb->len -= len;
2272 return skb->data += len;
2273}
2274
2275static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
2276{
2277 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2278}
2279
2280static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2281{
2282 if (likely(len <= skb_headlen(skb)))
2283 return 1;
2284 if (unlikely(len > skb->len))
2285 return 0;
2286 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
2287}
2288
2289void skb_condense(struct sk_buff *skb);
2290
2291/**
2292 * skb_headroom - bytes at buffer head
2293 * @skb: buffer to check
2294 *
2295 * Return the number of bytes of free space at the head of an &sk_buff.
2296 */
2297static inline unsigned int skb_headroom(const struct sk_buff *skb)
2298{
2299 return skb->data - skb->head;
2300}
2301
2302/**
2303 * skb_tailroom - bytes at buffer end
2304 * @skb: buffer to check
2305 *
2306 * Return the number of bytes of free space at the tail of an sk_buff
2307 */
2308static inline int skb_tailroom(const struct sk_buff *skb)
2309{
2310 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
2311}
2312
2313/**
2314 * skb_availroom - bytes at buffer end
2315 * @skb: buffer to check
2316 *
2317 * Return the number of bytes of free space at the tail of an sk_buff
2318 * allocated by sk_stream_alloc()
2319 */
2320static inline int skb_availroom(const struct sk_buff *skb)
2321{
2322 if (skb_is_nonlinear(skb))
2323 return 0;
2324
2325 return skb->end - skb->tail - skb->reserved_tailroom;
2326}
2327
2328/**
2329 * skb_reserve - adjust headroom
2330 * @skb: buffer to alter
2331 * @len: bytes to move
2332 *
2333 * Increase the headroom of an empty &sk_buff by reducing the tail
2334 * room. This is only allowed for an empty buffer.
2335 */
2336static inline void skb_reserve(struct sk_buff *skb, int len)
2337{
2338 skb->data += len;
2339 skb->tail += len;
2340}
2341
2342/**
2343 * skb_tailroom_reserve - adjust reserved_tailroom
2344 * @skb: buffer to alter
2345 * @mtu: maximum amount of headlen permitted
2346 * @needed_tailroom: minimum amount of reserved_tailroom
2347 *
2348 * Set reserved_tailroom so that headlen can be as large as possible but
2349 * not larger than mtu and tailroom cannot be smaller than
2350 * needed_tailroom.
2351 * The required headroom should already have been reserved before using
2352 * this function.
2353 */
2354static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2355 unsigned int needed_tailroom)
2356{
2357 SKB_LINEAR_ASSERT(skb);
2358 if (mtu < skb_tailroom(skb) - needed_tailroom)
2359 /* use at most mtu */
2360 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2361 else
2362 /* use up to all available space */
2363 skb->reserved_tailroom = needed_tailroom;
2364}
2365
2366#define ENCAP_TYPE_ETHER 0
2367#define ENCAP_TYPE_IPPROTO 1
2368
2369static inline void skb_set_inner_protocol(struct sk_buff *skb,
2370 __be16 protocol)
2371{
2372 skb->inner_protocol = protocol;
2373 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2374}
2375
2376static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2377 __u8 ipproto)
2378{
2379 skb->inner_ipproto = ipproto;
2380 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2381}
2382
2383static inline void skb_reset_inner_headers(struct sk_buff *skb)
2384{
2385 skb->inner_mac_header = skb->mac_header;
2386 skb->inner_network_header = skb->network_header;
2387 skb->inner_transport_header = skb->transport_header;
2388}
2389
2390static inline void skb_reset_mac_len(struct sk_buff *skb)
2391{
2392 skb->mac_len = skb->network_header - skb->mac_header;
2393}
2394
2395static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2396 *skb)
2397{
2398 return skb->head + skb->inner_transport_header;
2399}
2400
2401static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2402{
2403 return skb_inner_transport_header(skb) - skb->data;
2404}
2405
2406static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2407{
2408 skb->inner_transport_header = skb->data - skb->head;
2409}
2410
2411static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2412 const int offset)
2413{
2414 skb_reset_inner_transport_header(skb);
2415 skb->inner_transport_header += offset;
2416}
2417
2418static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2419{
2420 return skb->head + skb->inner_network_header;
2421}
2422
2423static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2424{
2425 skb->inner_network_header = skb->data - skb->head;
2426}
2427
2428static inline void skb_set_inner_network_header(struct sk_buff *skb,
2429 const int offset)
2430{
2431 skb_reset_inner_network_header(skb);
2432 skb->inner_network_header += offset;
2433}
2434
2435static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2436{
2437 return skb->head + skb->inner_mac_header;
2438}
2439
2440static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2441{
2442 skb->inner_mac_header = skb->data - skb->head;
2443}
2444
2445static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2446 const int offset)
2447{
2448 skb_reset_inner_mac_header(skb);
2449 skb->inner_mac_header += offset;
2450}
2451static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2452{
2453 return skb->transport_header != (typeof(skb->transport_header))~0U;
2454}
2455
2456static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2457{
2458 return skb->head + skb->transport_header;
2459}
2460
2461static inline void skb_reset_transport_header(struct sk_buff *skb)
2462{
2463 skb->transport_header = skb->data - skb->head;
2464}
2465
2466static inline void skb_set_transport_header(struct sk_buff *skb,
2467 const int offset)
2468{
2469 skb_reset_transport_header(skb);
2470 skb->transport_header += offset;
2471}
2472
2473static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2474{
2475 return skb->head + skb->network_header;
2476}
2477
2478static inline void skb_reset_network_header(struct sk_buff *skb)
2479{
2480 skb->network_header = skb->data - skb->head;
2481}
2482
2483static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2484{
2485 skb_reset_network_header(skb);
2486 skb->network_header += offset;
2487}
2488
2489static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2490{
2491 return skb->head + skb->mac_header;
2492}
2493
2494static inline int skb_mac_offset(const struct sk_buff *skb)
2495{
2496 return skb_mac_header(skb) - skb->data;
2497}
2498
2499static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2500{
2501 return skb->network_header - skb->mac_header;
2502}
2503
2504static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2505{
2506 return skb->mac_header != (typeof(skb->mac_header))~0U;
2507}
2508
2509static inline void skb_reset_mac_header(struct sk_buff *skb)
2510{
2511 skb->mac_header = skb->data - skb->head;
2512}
2513
2514static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2515{
2516 skb_reset_mac_header(skb);
2517 skb->mac_header += offset;
2518}
2519
2520static inline void skb_pop_mac_header(struct sk_buff *skb)
2521{
2522 skb->mac_header = skb->network_header;
2523}
2524
David Brazdil0f672f62019-12-10 10:32:29 +00002525static inline void skb_probe_transport_header(struct sk_buff *skb)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002526{
2527 struct flow_keys_basic keys;
2528
2529 if (skb_transport_header_was_set(skb))
2530 return;
2531
David Brazdil0f672f62019-12-10 10:32:29 +00002532 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2533 NULL, 0, 0, 0, 0))
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002534 skb_set_transport_header(skb, keys.control.thoff);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002535}
2536
2537static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2538{
2539 if (skb_mac_header_was_set(skb)) {
2540 const unsigned char *old_mac = skb_mac_header(skb);
2541
2542 skb_set_mac_header(skb, -skb->mac_len);
2543 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2544 }
2545}
2546
2547static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2548{
2549 return skb->csum_start - skb_headroom(skb);
2550}
2551
2552static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2553{
2554 return skb->head + skb->csum_start;
2555}
2556
2557static inline int skb_transport_offset(const struct sk_buff *skb)
2558{
2559 return skb_transport_header(skb) - skb->data;
2560}
2561
2562static inline u32 skb_network_header_len(const struct sk_buff *skb)
2563{
2564 return skb->transport_header - skb->network_header;
2565}
2566
2567static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2568{
2569 return skb->inner_transport_header - skb->inner_network_header;
2570}
2571
2572static inline int skb_network_offset(const struct sk_buff *skb)
2573{
2574 return skb_network_header(skb) - skb->data;
2575}
2576
2577static inline int skb_inner_network_offset(const struct sk_buff *skb)
2578{
2579 return skb_inner_network_header(skb) - skb->data;
2580}
2581
2582static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2583{
2584 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2585}
2586
2587/*
2588 * CPUs often take a performance hit when accessing unaligned memory
2589 * locations. The actual performance hit varies, it can be small if the
2590 * hardware handles it or large if we have to take an exception and fix it
2591 * in software.
2592 *
2593 * Since an ethernet header is 14 bytes network drivers often end up with
2594 * the IP header at an unaligned offset. The IP header can be aligned by
2595 * shifting the start of the packet by 2 bytes. Drivers should do this
2596 * with:
2597 *
2598 * skb_reserve(skb, NET_IP_ALIGN);
2599 *
2600 * The downside to this alignment of the IP header is that the DMA is now
2601 * unaligned. On some architectures the cost of an unaligned DMA is high
2602 * and this cost outweighs the gains made by aligning the IP header.
2603 *
2604 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2605 * to be overridden.
2606 */
2607#ifndef NET_IP_ALIGN
2608#define NET_IP_ALIGN 2
2609#endif
2610
2611/*
2612 * The networking layer reserves some headroom in skb data (via
2613 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2614 * the header has to grow. In the default case, if the header has to grow
2615 * 32 bytes or less we avoid the reallocation.
2616 *
2617 * Unfortunately this headroom changes the DMA alignment of the resulting
2618 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2619 * on some architectures. An architecture can override this value,
2620 * perhaps setting it to a cacheline in size (since that will maintain
2621 * cacheline alignment of the DMA). It must be a power of 2.
2622 *
2623 * Various parts of the networking layer expect at least 32 bytes of
2624 * headroom, you should not reduce this.
2625 *
2626 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2627 * to reduce average number of cache lines per packet.
2628 * get_rps_cpus() for example only access one 64 bytes aligned block :
2629 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2630 */
2631#ifndef NET_SKB_PAD
2632#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
2633#endif
2634
2635int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2636
2637static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
2638{
David Brazdil0f672f62019-12-10 10:32:29 +00002639 if (WARN_ON(skb_is_nonlinear(skb)))
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002640 return;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002641 skb->len = len;
2642 skb_set_tail_pointer(skb, len);
2643}
2644
2645static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2646{
2647 __skb_set_length(skb, len);
2648}
2649
2650void skb_trim(struct sk_buff *skb, unsigned int len);
2651
2652static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2653{
2654 if (skb->data_len)
2655 return ___pskb_trim(skb, len);
2656 __skb_trim(skb, len);
2657 return 0;
2658}
2659
2660static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2661{
2662 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2663}
2664
2665/**
2666 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2667 * @skb: buffer to alter
2668 * @len: new length
2669 *
2670 * This is identical to pskb_trim except that the caller knows that
2671 * the skb is not cloned so we should never get an error due to out-
2672 * of-memory.
2673 */
2674static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2675{
2676 int err = pskb_trim(skb, len);
2677 BUG_ON(err);
2678}
2679
2680static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2681{
2682 unsigned int diff = len - skb->len;
2683
2684 if (skb_tailroom(skb) < diff) {
2685 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2686 GFP_ATOMIC);
2687 if (ret)
2688 return ret;
2689 }
2690 __skb_set_length(skb, len);
2691 return 0;
2692}
2693
2694/**
2695 * skb_orphan - orphan a buffer
2696 * @skb: buffer to orphan
2697 *
2698 * If a buffer currently has an owner then we call the owner's
2699 * destructor function and make the @skb unowned. The buffer continues
2700 * to exist but is no longer charged to its former owner.
2701 */
2702static inline void skb_orphan(struct sk_buff *skb)
2703{
2704 if (skb->destructor) {
2705 skb->destructor(skb);
2706 skb->destructor = NULL;
2707 skb->sk = NULL;
2708 } else {
2709 BUG_ON(skb->sk);
2710 }
2711}
2712
2713/**
2714 * skb_orphan_frags - orphan the frags contained in a buffer
2715 * @skb: buffer to orphan frags from
2716 * @gfp_mask: allocation mask for replacement pages
2717 *
2718 * For each frag in the SKB which needs a destructor (i.e. has an
2719 * owner) create a copy of that frag and release the original
2720 * page by calling the destructor.
2721 */
2722static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2723{
2724 if (likely(!skb_zcopy(skb)))
2725 return 0;
David Brazdil0f672f62019-12-10 10:32:29 +00002726 if (!skb_zcopy_is_nouarg(skb) &&
2727 skb_uarg(skb)->callback == sock_zerocopy_callback)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002728 return 0;
2729 return skb_copy_ubufs(skb, gfp_mask);
2730}
2731
2732/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2733static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2734{
2735 if (likely(!skb_zcopy(skb)))
2736 return 0;
2737 return skb_copy_ubufs(skb, gfp_mask);
2738}
2739
2740/**
2741 * __skb_queue_purge - empty a list
2742 * @list: list to empty
2743 *
2744 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2745 * the list and one reference dropped. This function does not take the
2746 * list lock and the caller must hold the relevant locks to use it.
2747 */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002748static inline void __skb_queue_purge(struct sk_buff_head *list)
2749{
2750 struct sk_buff *skb;
2751 while ((skb = __skb_dequeue(list)) != NULL)
2752 kfree_skb(skb);
2753}
David Brazdil0f672f62019-12-10 10:32:29 +00002754void skb_queue_purge(struct sk_buff_head *list);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002755
2756unsigned int skb_rbtree_purge(struct rb_root *root);
2757
2758void *netdev_alloc_frag(unsigned int fragsz);
2759
2760struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2761 gfp_t gfp_mask);
2762
2763/**
2764 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2765 * @dev: network device to receive on
2766 * @length: length to allocate
2767 *
2768 * Allocate a new &sk_buff and assign it a usage count of one. The
2769 * buffer has unspecified headroom built in. Users should allocate
2770 * the headroom they think they need without accounting for the
2771 * built in space. The built in space is used for optimisations.
2772 *
2773 * %NULL is returned if there is no free memory. Although this function
2774 * allocates memory it can be called from an interrupt.
2775 */
2776static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2777 unsigned int length)
2778{
2779 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2780}
2781
2782/* legacy helper around __netdev_alloc_skb() */
2783static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2784 gfp_t gfp_mask)
2785{
2786 return __netdev_alloc_skb(NULL, length, gfp_mask);
2787}
2788
2789/* legacy helper around netdev_alloc_skb() */
2790static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2791{
2792 return netdev_alloc_skb(NULL, length);
2793}
2794
2795
2796static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2797 unsigned int length, gfp_t gfp)
2798{
2799 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2800
2801 if (NET_IP_ALIGN && skb)
2802 skb_reserve(skb, NET_IP_ALIGN);
2803 return skb;
2804}
2805
2806static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2807 unsigned int length)
2808{
2809 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2810}
2811
2812static inline void skb_free_frag(void *addr)
2813{
2814 page_frag_free(addr);
2815}
2816
2817void *napi_alloc_frag(unsigned int fragsz);
2818struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2819 unsigned int length, gfp_t gfp_mask);
2820static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2821 unsigned int length)
2822{
2823 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2824}
2825void napi_consume_skb(struct sk_buff *skb, int budget);
2826
2827void __kfree_skb_flush(void);
2828void __kfree_skb_defer(struct sk_buff *skb);
2829
2830/**
2831 * __dev_alloc_pages - allocate page for network Rx
2832 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2833 * @order: size of the allocation
2834 *
2835 * Allocate a new page.
2836 *
2837 * %NULL is returned if there is no free memory.
2838*/
2839static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2840 unsigned int order)
2841{
2842 /* This piece of code contains several assumptions.
2843 * 1. This is for device Rx, therefor a cold page is preferred.
2844 * 2. The expectation is the user wants a compound page.
2845 * 3. If requesting a order 0 page it will not be compound
2846 * due to the check to see if order has a value in prep_new_page
2847 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2848 * code in gfp_to_alloc_flags that should be enforcing this.
2849 */
2850 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
2851
2852 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2853}
2854
2855static inline struct page *dev_alloc_pages(unsigned int order)
2856{
2857 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
2858}
2859
2860/**
2861 * __dev_alloc_page - allocate a page for network Rx
2862 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2863 *
2864 * Allocate a new page.
2865 *
2866 * %NULL is returned if there is no free memory.
2867 */
2868static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2869{
2870 return __dev_alloc_pages(gfp_mask, 0);
2871}
2872
2873static inline struct page *dev_alloc_page(void)
2874{
2875 return dev_alloc_pages(0);
2876}
2877
2878/**
2879 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2880 * @page: The page that was allocated from skb_alloc_page
2881 * @skb: The skb that may need pfmemalloc set
2882 */
2883static inline void skb_propagate_pfmemalloc(struct page *page,
2884 struct sk_buff *skb)
2885{
2886 if (page_is_pfmemalloc(page))
2887 skb->pfmemalloc = true;
2888}
2889
2890/**
David Brazdil0f672f62019-12-10 10:32:29 +00002891 * skb_frag_off() - Returns the offset of a skb fragment
2892 * @frag: the paged fragment
2893 */
2894static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2895{
2896 return frag->bv_offset;
2897}
2898
2899/**
2900 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2901 * @frag: skb fragment
2902 * @delta: value to add
2903 */
2904static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2905{
2906 frag->bv_offset += delta;
2907}
2908
2909/**
2910 * skb_frag_off_set() - Sets the offset of a skb fragment
2911 * @frag: skb fragment
2912 * @offset: offset of fragment
2913 */
2914static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2915{
2916 frag->bv_offset = offset;
2917}
2918
2919/**
2920 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2921 * @fragto: skb fragment where offset is set
2922 * @fragfrom: skb fragment offset is copied from
2923 */
2924static inline void skb_frag_off_copy(skb_frag_t *fragto,
2925 const skb_frag_t *fragfrom)
2926{
2927 fragto->bv_offset = fragfrom->bv_offset;
2928}
2929
2930/**
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002931 * skb_frag_page - retrieve the page referred to by a paged fragment
2932 * @frag: the paged fragment
2933 *
2934 * Returns the &struct page associated with @frag.
2935 */
2936static inline struct page *skb_frag_page(const skb_frag_t *frag)
2937{
David Brazdil0f672f62019-12-10 10:32:29 +00002938 return frag->bv_page;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002939}
2940
2941/**
2942 * __skb_frag_ref - take an addition reference on a paged fragment.
2943 * @frag: the paged fragment
2944 *
2945 * Takes an additional reference on the paged fragment @frag.
2946 */
2947static inline void __skb_frag_ref(skb_frag_t *frag)
2948{
2949 get_page(skb_frag_page(frag));
2950}
2951
2952/**
2953 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2954 * @skb: the buffer
2955 * @f: the fragment offset.
2956 *
2957 * Takes an additional reference on the @f'th paged fragment of @skb.
2958 */
2959static inline void skb_frag_ref(struct sk_buff *skb, int f)
2960{
2961 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2962}
2963
2964/**
2965 * __skb_frag_unref - release a reference on a paged fragment.
2966 * @frag: the paged fragment
2967 *
2968 * Releases a reference on the paged fragment @frag.
2969 */
2970static inline void __skb_frag_unref(skb_frag_t *frag)
2971{
2972 put_page(skb_frag_page(frag));
2973}
2974
2975/**
2976 * skb_frag_unref - release a reference on a paged fragment of an skb.
2977 * @skb: the buffer
2978 * @f: the fragment offset
2979 *
2980 * Releases a reference on the @f'th paged fragment of @skb.
2981 */
2982static inline void skb_frag_unref(struct sk_buff *skb, int f)
2983{
2984 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2985}
2986
2987/**
2988 * skb_frag_address - gets the address of the data contained in a paged fragment
2989 * @frag: the paged fragment buffer
2990 *
2991 * Returns the address of the data within @frag. The page must already
2992 * be mapped.
2993 */
2994static inline void *skb_frag_address(const skb_frag_t *frag)
2995{
David Brazdil0f672f62019-12-10 10:32:29 +00002996 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00002997}
2998
2999/**
3000 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3001 * @frag: the paged fragment buffer
3002 *
3003 * Returns the address of the data within @frag. Checks that the page
3004 * is mapped and returns %NULL otherwise.
3005 */
3006static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3007{
3008 void *ptr = page_address(skb_frag_page(frag));
3009 if (unlikely(!ptr))
3010 return NULL;
3011
David Brazdil0f672f62019-12-10 10:32:29 +00003012 return ptr + skb_frag_off(frag);
3013}
3014
3015/**
3016 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3017 * @fragto: skb fragment where page is set
3018 * @fragfrom: skb fragment page is copied from
3019 */
3020static inline void skb_frag_page_copy(skb_frag_t *fragto,
3021 const skb_frag_t *fragfrom)
3022{
3023 fragto->bv_page = fragfrom->bv_page;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003024}
3025
3026/**
3027 * __skb_frag_set_page - sets the page contained in a paged fragment
3028 * @frag: the paged fragment
3029 * @page: the page to set
3030 *
3031 * Sets the fragment @frag to contain @page.
3032 */
3033static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3034{
David Brazdil0f672f62019-12-10 10:32:29 +00003035 frag->bv_page = page;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003036}
3037
3038/**
3039 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3040 * @skb: the buffer
3041 * @f: the fragment offset
3042 * @page: the page to set
3043 *
3044 * Sets the @f'th fragment of @skb to contain @page.
3045 */
3046static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3047 struct page *page)
3048{
3049 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3050}
3051
3052bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3053
3054/**
3055 * skb_frag_dma_map - maps a paged fragment via the DMA API
3056 * @dev: the device to map the fragment to
3057 * @frag: the paged fragment to map
3058 * @offset: the offset within the fragment (starting at the
3059 * fragment's own offset)
3060 * @size: the number of bytes to map
3061 * @dir: the direction of the mapping (``PCI_DMA_*``)
3062 *
3063 * Maps the page associated with @frag to @device.
3064 */
3065static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3066 const skb_frag_t *frag,
3067 size_t offset, size_t size,
3068 enum dma_data_direction dir)
3069{
3070 return dma_map_page(dev, skb_frag_page(frag),
David Brazdil0f672f62019-12-10 10:32:29 +00003071 skb_frag_off(frag) + offset, size, dir);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003072}
3073
3074static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3075 gfp_t gfp_mask)
3076{
3077 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3078}
3079
3080
3081static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3082 gfp_t gfp_mask)
3083{
3084 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3085}
3086
3087
3088/**
3089 * skb_clone_writable - is the header of a clone writable
3090 * @skb: buffer to check
3091 * @len: length up to which to write
3092 *
3093 * Returns true if modifying the header part of the cloned buffer
3094 * does not requires the data to be copied.
3095 */
3096static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
3097{
3098 return !skb_header_cloned(skb) &&
3099 skb_headroom(skb) + len <= skb->hdr_len;
3100}
3101
3102static inline int skb_try_make_writable(struct sk_buff *skb,
3103 unsigned int write_len)
3104{
3105 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3106 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3107}
3108
3109static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3110 int cloned)
3111{
3112 int delta = 0;
3113
3114 if (headroom > skb_headroom(skb))
3115 delta = headroom - skb_headroom(skb);
3116
3117 if (delta || cloned)
3118 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3119 GFP_ATOMIC);
3120 return 0;
3121}
3122
3123/**
3124 * skb_cow - copy header of skb when it is required
3125 * @skb: buffer to cow
3126 * @headroom: needed headroom
3127 *
3128 * If the skb passed lacks sufficient headroom or its data part
3129 * is shared, data is reallocated. If reallocation fails, an error
3130 * is returned and original skb is not changed.
3131 *
3132 * The result is skb with writable area skb->head...skb->tail
3133 * and at least @headroom of space at head.
3134 */
3135static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3136{
3137 return __skb_cow(skb, headroom, skb_cloned(skb));
3138}
3139
3140/**
3141 * skb_cow_head - skb_cow but only making the head writable
3142 * @skb: buffer to cow
3143 * @headroom: needed headroom
3144 *
3145 * This function is identical to skb_cow except that we replace the
3146 * skb_cloned check by skb_header_cloned. It should be used when
3147 * you only need to push on some header and do not need to modify
3148 * the data.
3149 */
3150static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3151{
3152 return __skb_cow(skb, headroom, skb_header_cloned(skb));
3153}
3154
3155/**
3156 * skb_padto - pad an skbuff up to a minimal size
3157 * @skb: buffer to pad
3158 * @len: minimal length
3159 *
3160 * Pads up a buffer to ensure the trailing bytes exist and are
3161 * blanked. If the buffer already contains sufficient data it
3162 * is untouched. Otherwise it is extended. Returns zero on
3163 * success. The skb is freed on error.
3164 */
3165static inline int skb_padto(struct sk_buff *skb, unsigned int len)
3166{
3167 unsigned int size = skb->len;
3168 if (likely(size >= len))
3169 return 0;
3170 return skb_pad(skb, len - size);
3171}
3172
3173/**
David Brazdil0f672f62019-12-10 10:32:29 +00003174 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003175 * @skb: buffer to pad
3176 * @len: minimal length
3177 * @free_on_error: free buffer on error
3178 *
3179 * Pads up a buffer to ensure the trailing bytes exist and are
3180 * blanked. If the buffer already contains sufficient data it
3181 * is untouched. Otherwise it is extended. Returns zero on
3182 * success. The skb is freed on error if @free_on_error is true.
3183 */
3184static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
3185 bool free_on_error)
3186{
3187 unsigned int size = skb->len;
3188
3189 if (unlikely(size < len)) {
3190 len -= size;
3191 if (__skb_pad(skb, len, free_on_error))
3192 return -ENOMEM;
3193 __skb_put(skb, len);
3194 }
3195 return 0;
3196}
3197
3198/**
3199 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3200 * @skb: buffer to pad
3201 * @len: minimal length
3202 *
3203 * Pads up a buffer to ensure the trailing bytes exist and are
3204 * blanked. If the buffer already contains sufficient data it
3205 * is untouched. Otherwise it is extended. Returns zero on
3206 * success. The skb is freed on error.
3207 */
3208static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3209{
3210 return __skb_put_padto(skb, len, true);
3211}
3212
3213static inline int skb_add_data(struct sk_buff *skb,
3214 struct iov_iter *from, int copy)
3215{
3216 const int off = skb->len;
3217
3218 if (skb->ip_summed == CHECKSUM_NONE) {
3219 __wsum csum = 0;
3220 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3221 &csum, from)) {
3222 skb->csum = csum_block_add(skb->csum, csum, off);
3223 return 0;
3224 }
3225 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
3226 return 0;
3227
3228 __skb_trim(skb, off);
3229 return -EFAULT;
3230}
3231
3232static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3233 const struct page *page, int off)
3234{
3235 if (skb_zcopy(skb))
3236 return false;
3237 if (i) {
David Brazdil0f672f62019-12-10 10:32:29 +00003238 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003239
3240 return page == skb_frag_page(frag) &&
David Brazdil0f672f62019-12-10 10:32:29 +00003241 off == skb_frag_off(frag) + skb_frag_size(frag);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003242 }
3243 return false;
3244}
3245
3246static inline int __skb_linearize(struct sk_buff *skb)
3247{
3248 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3249}
3250
3251/**
3252 * skb_linearize - convert paged skb to linear one
3253 * @skb: buffer to linarize
3254 *
3255 * If there is no free memory -ENOMEM is returned, otherwise zero
3256 * is returned and the old skb data released.
3257 */
3258static inline int skb_linearize(struct sk_buff *skb)
3259{
3260 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3261}
3262
3263/**
3264 * skb_has_shared_frag - can any frag be overwritten
3265 * @skb: buffer to test
3266 *
3267 * Return true if the skb has at least one frag that might be modified
3268 * by an external entity (as in vmsplice()/sendfile())
3269 */
3270static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3271{
3272 return skb_is_nonlinear(skb) &&
3273 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
3274}
3275
3276/**
3277 * skb_linearize_cow - make sure skb is linear and writable
3278 * @skb: buffer to process
3279 *
3280 * If there is no free memory -ENOMEM is returned, otherwise zero
3281 * is returned and the old skb data released.
3282 */
3283static inline int skb_linearize_cow(struct sk_buff *skb)
3284{
3285 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3286 __skb_linearize(skb) : 0;
3287}
3288
3289static __always_inline void
3290__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3291 unsigned int off)
3292{
3293 if (skb->ip_summed == CHECKSUM_COMPLETE)
3294 skb->csum = csum_block_sub(skb->csum,
3295 csum_partial(start, len, 0), off);
3296 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3297 skb_checksum_start_offset(skb) < 0)
3298 skb->ip_summed = CHECKSUM_NONE;
3299}
3300
3301/**
3302 * skb_postpull_rcsum - update checksum for received skb after pull
3303 * @skb: buffer to update
3304 * @start: start of data before pull
3305 * @len: length of data pulled
3306 *
3307 * After doing a pull on a received packet, you need to call this to
3308 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3309 * CHECKSUM_NONE so that it can be recomputed from scratch.
3310 */
3311static inline void skb_postpull_rcsum(struct sk_buff *skb,
3312 const void *start, unsigned int len)
3313{
3314 __skb_postpull_rcsum(skb, start, len, 0);
3315}
3316
3317static __always_inline void
3318__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3319 unsigned int off)
3320{
3321 if (skb->ip_summed == CHECKSUM_COMPLETE)
3322 skb->csum = csum_block_add(skb->csum,
3323 csum_partial(start, len, 0), off);
3324}
3325
3326/**
3327 * skb_postpush_rcsum - update checksum for received skb after push
3328 * @skb: buffer to update
3329 * @start: start of data after push
3330 * @len: length of data pushed
3331 *
3332 * After doing a push on a received packet, you need to call this to
3333 * update the CHECKSUM_COMPLETE checksum.
3334 */
3335static inline void skb_postpush_rcsum(struct sk_buff *skb,
3336 const void *start, unsigned int len)
3337{
3338 __skb_postpush_rcsum(skb, start, len, 0);
3339}
3340
3341void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
3342
3343/**
3344 * skb_push_rcsum - push skb and update receive checksum
3345 * @skb: buffer to update
3346 * @len: length of data pulled
3347 *
3348 * This function performs an skb_push on the packet and updates
3349 * the CHECKSUM_COMPLETE checksum. It should be used on
3350 * receive path processing instead of skb_push unless you know
3351 * that the checksum difference is zero (e.g., a valid IP header)
3352 * or you are setting ip_summed to CHECKSUM_NONE.
3353 */
3354static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
3355{
3356 skb_push(skb, len);
3357 skb_postpush_rcsum(skb, skb->data, len);
3358 return skb->data;
3359}
3360
3361int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
3362/**
3363 * pskb_trim_rcsum - trim received skb and update checksum
3364 * @skb: buffer to trim
3365 * @len: new length
3366 *
3367 * This is exactly the same as pskb_trim except that it ensures the
3368 * checksum of received packets are still valid after the operation.
David Brazdil0f672f62019-12-10 10:32:29 +00003369 * It can change skb pointers.
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003370 */
3371
3372static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3373{
3374 if (likely(len >= skb->len))
3375 return 0;
3376 return pskb_trim_rcsum_slow(skb, len);
3377}
3378
3379static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3380{
3381 if (skb->ip_summed == CHECKSUM_COMPLETE)
3382 skb->ip_summed = CHECKSUM_NONE;
3383 __skb_trim(skb, len);
3384 return 0;
3385}
3386
3387static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3388{
3389 if (skb->ip_summed == CHECKSUM_COMPLETE)
3390 skb->ip_summed = CHECKSUM_NONE;
3391 return __skb_grow(skb, len);
3392}
3393
3394#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3395#define skb_rb_first(root) rb_to_skb(rb_first(root))
3396#define skb_rb_last(root) rb_to_skb(rb_last(root))
3397#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3398#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3399
3400#define skb_queue_walk(queue, skb) \
3401 for (skb = (queue)->next; \
3402 skb != (struct sk_buff *)(queue); \
3403 skb = skb->next)
3404
3405#define skb_queue_walk_safe(queue, skb, tmp) \
3406 for (skb = (queue)->next, tmp = skb->next; \
3407 skb != (struct sk_buff *)(queue); \
3408 skb = tmp, tmp = skb->next)
3409
3410#define skb_queue_walk_from(queue, skb) \
3411 for (; skb != (struct sk_buff *)(queue); \
3412 skb = skb->next)
3413
3414#define skb_rbtree_walk(skb, root) \
3415 for (skb = skb_rb_first(root); skb != NULL; \
3416 skb = skb_rb_next(skb))
3417
3418#define skb_rbtree_walk_from(skb) \
3419 for (; skb != NULL; \
3420 skb = skb_rb_next(skb))
3421
3422#define skb_rbtree_walk_from_safe(skb, tmp) \
3423 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3424 skb = tmp)
3425
3426#define skb_queue_walk_from_safe(queue, skb, tmp) \
3427 for (tmp = skb->next; \
3428 skb != (struct sk_buff *)(queue); \
3429 skb = tmp, tmp = skb->next)
3430
3431#define skb_queue_reverse_walk(queue, skb) \
3432 for (skb = (queue)->prev; \
3433 skb != (struct sk_buff *)(queue); \
3434 skb = skb->prev)
3435
3436#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3437 for (skb = (queue)->prev, tmp = skb->prev; \
3438 skb != (struct sk_buff *)(queue); \
3439 skb = tmp, tmp = skb->prev)
3440
3441#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3442 for (tmp = skb->prev; \
3443 skb != (struct sk_buff *)(queue); \
3444 skb = tmp, tmp = skb->prev)
3445
3446static inline bool skb_has_frag_list(const struct sk_buff *skb)
3447{
3448 return skb_shinfo(skb)->frag_list != NULL;
3449}
3450
3451static inline void skb_frag_list_init(struct sk_buff *skb)
3452{
3453 skb_shinfo(skb)->frag_list = NULL;
3454}
3455
3456#define skb_walk_frags(skb, iter) \
3457 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3458
3459
3460int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3461 const struct sk_buff *skb);
3462struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3463 struct sk_buff_head *queue,
3464 unsigned int flags,
3465 void (*destructor)(struct sock *sk,
3466 struct sk_buff *skb),
David Brazdil0f672f62019-12-10 10:32:29 +00003467 int *off, int *err,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003468 struct sk_buff **last);
3469struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
3470 void (*destructor)(struct sock *sk,
3471 struct sk_buff *skb),
David Brazdil0f672f62019-12-10 10:32:29 +00003472 int *off, int *err,
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003473 struct sk_buff **last);
3474struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
3475 void (*destructor)(struct sock *sk,
3476 struct sk_buff *skb),
David Brazdil0f672f62019-12-10 10:32:29 +00003477 int *off, int *err);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003478struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3479 int *err);
3480__poll_t datagram_poll(struct file *file, struct socket *sock,
3481 struct poll_table_struct *wait);
3482int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3483 struct iov_iter *to, int size);
3484static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3485 struct msghdr *msg, int size)
3486{
3487 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
3488}
3489int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3490 struct msghdr *msg);
David Brazdil0f672f62019-12-10 10:32:29 +00003491int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3492 struct iov_iter *to, int len,
3493 struct ahash_request *hash);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003494int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3495 struct iov_iter *from, int len);
3496int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
3497void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
3498void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3499static inline void skb_free_datagram_locked(struct sock *sk,
3500 struct sk_buff *skb)
3501{
3502 __skb_free_datagram_locked(sk, skb, 0);
3503}
3504int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
3505int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3506int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3507__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3508 int len, __wsum csum);
3509int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
3510 struct pipe_inode_info *pipe, unsigned int len,
3511 unsigned int flags);
3512int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3513 int len);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003514void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
3515unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
3516int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3517 int len, int hlen);
3518void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3519int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3520void skb_scrub_packet(struct sk_buff *skb, bool xnet);
3521bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
3522bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
3523struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
3524struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
3525int skb_ensure_writable(struct sk_buff *skb, int write_len);
3526int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
3527int skb_vlan_pop(struct sk_buff *skb);
3528int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
David Brazdil0f672f62019-12-10 10:32:29 +00003529int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3530 int mac_len);
3531int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len);
3532int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
3533int skb_mpls_dec_ttl(struct sk_buff *skb);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003534struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3535 gfp_t gfp);
3536
3537static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3538{
3539 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
3540}
3541
3542static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3543{
3544 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
3545}
3546
3547struct skb_checksum_ops {
3548 __wsum (*update)(const void *mem, int len, __wsum wsum);
3549 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3550};
3551
3552extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3553
3554__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3555 __wsum csum, const struct skb_checksum_ops *ops);
3556__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3557 __wsum csum);
3558
3559static inline void * __must_check
3560__skb_header_pointer(const struct sk_buff *skb, int offset,
3561 int len, void *data, int hlen, void *buffer)
3562{
3563 if (hlen - offset >= len)
3564 return data + offset;
3565
3566 if (!skb ||
3567 skb_copy_bits(skb, offset, buffer, len) < 0)
3568 return NULL;
3569
3570 return buffer;
3571}
3572
3573static inline void * __must_check
3574skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
3575{
3576 return __skb_header_pointer(skb, offset, len, skb->data,
3577 skb_headlen(skb), buffer);
3578}
3579
3580/**
3581 * skb_needs_linearize - check if we need to linearize a given skb
3582 * depending on the given device features.
3583 * @skb: socket buffer to check
3584 * @features: net device features
3585 *
3586 * Returns true if either:
3587 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3588 * 2. skb is fragmented and the device does not support SG.
3589 */
3590static inline bool skb_needs_linearize(struct sk_buff *skb,
3591 netdev_features_t features)
3592{
3593 return skb_is_nonlinear(skb) &&
3594 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3595 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3596}
3597
3598static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3599 void *to,
3600 const unsigned int len)
3601{
3602 memcpy(to, skb->data, len);
3603}
3604
3605static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3606 const int offset, void *to,
3607 const unsigned int len)
3608{
3609 memcpy(to, skb->data + offset, len);
3610}
3611
3612static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3613 const void *from,
3614 const unsigned int len)
3615{
3616 memcpy(skb->data, from, len);
3617}
3618
3619static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3620 const int offset,
3621 const void *from,
3622 const unsigned int len)
3623{
3624 memcpy(skb->data + offset, from, len);
3625}
3626
3627void skb_init(void);
3628
3629static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3630{
3631 return skb->tstamp;
3632}
3633
3634/**
3635 * skb_get_timestamp - get timestamp from a skb
3636 * @skb: skb to get stamp from
David Brazdil0f672f62019-12-10 10:32:29 +00003637 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003638 *
3639 * Timestamps are stored in the skb as offsets to a base timestamp.
3640 * This function converts the offset back to a struct timeval and stores
3641 * it in stamp.
3642 */
3643static inline void skb_get_timestamp(const struct sk_buff *skb,
David Brazdil0f672f62019-12-10 10:32:29 +00003644 struct __kernel_old_timeval *stamp)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003645{
David Brazdil0f672f62019-12-10 10:32:29 +00003646 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
3647}
3648
3649static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3650 struct __kernel_sock_timeval *stamp)
3651{
3652 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3653
3654 stamp->tv_sec = ts.tv_sec;
3655 stamp->tv_usec = ts.tv_nsec / 1000;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003656}
3657
3658static inline void skb_get_timestampns(const struct sk_buff *skb,
3659 struct timespec *stamp)
3660{
3661 *stamp = ktime_to_timespec(skb->tstamp);
3662}
3663
David Brazdil0f672f62019-12-10 10:32:29 +00003664static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3665 struct __kernel_timespec *stamp)
3666{
3667 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3668
3669 stamp->tv_sec = ts.tv_sec;
3670 stamp->tv_nsec = ts.tv_nsec;
3671}
3672
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003673static inline void __net_timestamp(struct sk_buff *skb)
3674{
3675 skb->tstamp = ktime_get_real();
3676}
3677
3678static inline ktime_t net_timedelta(ktime_t t)
3679{
3680 return ktime_sub(ktime_get_real(), t);
3681}
3682
3683static inline ktime_t net_invalid_timestamp(void)
3684{
3685 return 0;
3686}
3687
3688static inline u8 skb_metadata_len(const struct sk_buff *skb)
3689{
3690 return skb_shinfo(skb)->meta_len;
3691}
3692
3693static inline void *skb_metadata_end(const struct sk_buff *skb)
3694{
3695 return skb_mac_header(skb);
3696}
3697
3698static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3699 const struct sk_buff *skb_b,
3700 u8 meta_len)
3701{
3702 const void *a = skb_metadata_end(skb_a);
3703 const void *b = skb_metadata_end(skb_b);
3704 /* Using more efficient varaiant than plain call to memcmp(). */
3705#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3706 u64 diffs = 0;
3707
3708 switch (meta_len) {
3709#define __it(x, op) (x -= sizeof(u##op))
3710#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3711 case 32: diffs |= __it_diff(a, b, 64);
David Brazdil0f672f62019-12-10 10:32:29 +00003712 /* fall through */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003713 case 24: diffs |= __it_diff(a, b, 64);
David Brazdil0f672f62019-12-10 10:32:29 +00003714 /* fall through */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003715 case 16: diffs |= __it_diff(a, b, 64);
David Brazdil0f672f62019-12-10 10:32:29 +00003716 /* fall through */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003717 case 8: diffs |= __it_diff(a, b, 64);
3718 break;
3719 case 28: diffs |= __it_diff(a, b, 64);
David Brazdil0f672f62019-12-10 10:32:29 +00003720 /* fall through */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003721 case 20: diffs |= __it_diff(a, b, 64);
David Brazdil0f672f62019-12-10 10:32:29 +00003722 /* fall through */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003723 case 12: diffs |= __it_diff(a, b, 64);
David Brazdil0f672f62019-12-10 10:32:29 +00003724 /* fall through */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00003725 case 4: diffs |= __it_diff(a, b, 32);
3726 break;
3727 }
3728 return diffs;
3729#else
3730 return memcmp(a - meta_len, b - meta_len, meta_len);
3731#endif
3732}
3733
3734static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3735 const struct sk_buff *skb_b)
3736{
3737 u8 len_a = skb_metadata_len(skb_a);
3738 u8 len_b = skb_metadata_len(skb_b);
3739
3740 if (!(len_a | len_b))
3741 return false;
3742
3743 return len_a != len_b ?
3744 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3745}
3746
3747static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3748{
3749 skb_shinfo(skb)->meta_len = meta_len;
3750}
3751
3752static inline void skb_metadata_clear(struct sk_buff *skb)
3753{
3754 skb_metadata_set(skb, 0);
3755}
3756
3757struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3758
3759#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3760
3761void skb_clone_tx_timestamp(struct sk_buff *skb);
3762bool skb_defer_rx_timestamp(struct sk_buff *skb);
3763
3764#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3765
3766static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3767{
3768}
3769
3770static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3771{
3772 return false;
3773}
3774
3775#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3776
3777/**
3778 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3779 *
3780 * PHY drivers may accept clones of transmitted packets for
3781 * timestamping via their phy_driver.txtstamp method. These drivers
3782 * must call this function to return the skb back to the stack with a
3783 * timestamp.
3784 *
3785 * @skb: clone of the the original outgoing packet
3786 * @hwtstamps: hardware time stamps
3787 *
3788 */
3789void skb_complete_tx_timestamp(struct sk_buff *skb,
3790 struct skb_shared_hwtstamps *hwtstamps);
3791
3792void __skb_tstamp_tx(struct sk_buff *orig_skb,
3793 struct skb_shared_hwtstamps *hwtstamps,
3794 struct sock *sk, int tstype);
3795
3796/**
3797 * skb_tstamp_tx - queue clone of skb with send time stamps
3798 * @orig_skb: the original outgoing packet
3799 * @hwtstamps: hardware time stamps, may be NULL if not available
3800 *
3801 * If the skb has a socket associated, then this function clones the
3802 * skb (thus sharing the actual data and optional structures), stores
3803 * the optional hardware time stamping information (if non NULL) or
3804 * generates a software time stamp (otherwise), then queues the clone
3805 * to the error queue of the socket. Errors are silently ignored.
3806 */
3807void skb_tstamp_tx(struct sk_buff *orig_skb,
3808 struct skb_shared_hwtstamps *hwtstamps);
3809
3810/**
3811 * skb_tx_timestamp() - Driver hook for transmit timestamping
3812 *
3813 * Ethernet MAC Drivers should call this function in their hard_xmit()
3814 * function immediately before giving the sk_buff to the MAC hardware.
3815 *
3816 * Specifically, one should make absolutely sure that this function is
3817 * called before TX completion of this packet can trigger. Otherwise
3818 * the packet could potentially already be freed.
3819 *
3820 * @skb: A socket buffer.
3821 */
3822static inline void skb_tx_timestamp(struct sk_buff *skb)
3823{
3824 skb_clone_tx_timestamp(skb);
3825 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3826 skb_tstamp_tx(skb, NULL);
3827}
3828
3829/**
3830 * skb_complete_wifi_ack - deliver skb with wifi status
3831 *
3832 * @skb: the original outgoing packet
3833 * @acked: ack status
3834 *
3835 */
3836void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3837
3838__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3839__sum16 __skb_checksum_complete(struct sk_buff *skb);
3840
3841static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3842{
3843 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3844 skb->csum_valid ||
3845 (skb->ip_summed == CHECKSUM_PARTIAL &&
3846 skb_checksum_start_offset(skb) >= 0));
3847}
3848
3849/**
3850 * skb_checksum_complete - Calculate checksum of an entire packet
3851 * @skb: packet to process
3852 *
3853 * This function calculates the checksum over the entire packet plus
3854 * the value of skb->csum. The latter can be used to supply the
3855 * checksum of a pseudo header as used by TCP/UDP. It returns the
3856 * checksum.
3857 *
3858 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3859 * this function can be used to verify that checksum on received
3860 * packets. In that case the function should return zero if the
3861 * checksum is correct. In particular, this function will return zero
3862 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3863 * hardware has already verified the correctness of the checksum.
3864 */
3865static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
3866{
3867 return skb_csum_unnecessary(skb) ?
3868 0 : __skb_checksum_complete(skb);
3869}
3870
3871static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3872{
3873 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3874 if (skb->csum_level == 0)
3875 skb->ip_summed = CHECKSUM_NONE;
3876 else
3877 skb->csum_level--;
3878 }
3879}
3880
3881static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3882{
3883 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3884 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3885 skb->csum_level++;
3886 } else if (skb->ip_summed == CHECKSUM_NONE) {
3887 skb->ip_summed = CHECKSUM_UNNECESSARY;
3888 skb->csum_level = 0;
3889 }
3890}
3891
3892/* Check if we need to perform checksum complete validation.
3893 *
3894 * Returns true if checksum complete is needed, false otherwise
3895 * (either checksum is unnecessary or zero checksum is allowed).
3896 */
3897static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3898 bool zero_okay,
3899 __sum16 check)
3900{
3901 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3902 skb->csum_valid = 1;
3903 __skb_decr_checksum_unnecessary(skb);
3904 return false;
3905 }
3906
3907 return true;
3908}
3909
3910/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
3911 * in checksum_init.
3912 */
3913#define CHECKSUM_BREAK 76
3914
3915/* Unset checksum-complete
3916 *
3917 * Unset checksum complete can be done when packet is being modified
3918 * (uncompressed for instance) and checksum-complete value is
3919 * invalidated.
3920 */
3921static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3922{
3923 if (skb->ip_summed == CHECKSUM_COMPLETE)
3924 skb->ip_summed = CHECKSUM_NONE;
3925}
3926
3927/* Validate (init) checksum based on checksum complete.
3928 *
3929 * Return values:
3930 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3931 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3932 * checksum is stored in skb->csum for use in __skb_checksum_complete
3933 * non-zero: value of invalid checksum
3934 *
3935 */
3936static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3937 bool complete,
3938 __wsum psum)
3939{
3940 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3941 if (!csum_fold(csum_add(psum, skb->csum))) {
3942 skb->csum_valid = 1;
3943 return 0;
3944 }
3945 }
3946
3947 skb->csum = psum;
3948
3949 if (complete || skb->len <= CHECKSUM_BREAK) {
3950 __sum16 csum;
3951
3952 csum = __skb_checksum_complete(skb);
3953 skb->csum_valid = !csum;
3954 return csum;
3955 }
3956
3957 return 0;
3958}
3959
3960static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3961{
3962 return 0;
3963}
3964
3965/* Perform checksum validate (init). Note that this is a macro since we only
3966 * want to calculate the pseudo header which is an input function if necessary.
3967 * First we try to validate without any computation (checksum unnecessary) and
3968 * then calculate based on checksum complete calling the function to compute
3969 * pseudo header.
3970 *
3971 * Return values:
3972 * 0: checksum is validated or try to in skb_checksum_complete
3973 * non-zero: value of invalid checksum
3974 */
3975#define __skb_checksum_validate(skb, proto, complete, \
3976 zero_okay, check, compute_pseudo) \
3977({ \
3978 __sum16 __ret = 0; \
3979 skb->csum_valid = 0; \
3980 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3981 __ret = __skb_checksum_validate_complete(skb, \
3982 complete, compute_pseudo(skb, proto)); \
3983 __ret; \
3984})
3985
3986#define skb_checksum_init(skb, proto, compute_pseudo) \
3987 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3988
3989#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3990 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3991
3992#define skb_checksum_validate(skb, proto, compute_pseudo) \
3993 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3994
3995#define skb_checksum_validate_zero_check(skb, proto, check, \
3996 compute_pseudo) \
3997 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3998
3999#define skb_checksum_simple_validate(skb) \
4000 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4001
4002static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4003{
4004 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
4005}
4006
David Brazdil0f672f62019-12-10 10:32:29 +00004007static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004008{
4009 skb->csum = ~pseudo;
4010 skb->ip_summed = CHECKSUM_COMPLETE;
4011}
4012
David Brazdil0f672f62019-12-10 10:32:29 +00004013#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004014do { \
4015 if (__skb_checksum_convert_check(skb)) \
David Brazdil0f672f62019-12-10 10:32:29 +00004016 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004017} while (0)
4018
4019static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4020 u16 start, u16 offset)
4021{
4022 skb->ip_summed = CHECKSUM_PARTIAL;
4023 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4024 skb->csum_offset = offset - start;
4025}
4026
4027/* Update skbuf and packet to reflect the remote checksum offload operation.
4028 * When called, ptr indicates the starting point for skb->csum when
4029 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4030 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4031 */
4032static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
4033 int start, int offset, bool nopartial)
4034{
4035 __wsum delta;
4036
4037 if (!nopartial) {
4038 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4039 return;
4040 }
4041
4042 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4043 __skb_checksum_complete(skb);
4044 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4045 }
4046
4047 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4048
4049 /* Adjust skb->csum since we changed the packet */
4050 skb->csum = csum_add(skb->csum, delta);
4051}
4052
4053static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4054{
4055#if IS_ENABLED(CONFIG_NF_CONNTRACK)
David Brazdil0f672f62019-12-10 10:32:29 +00004056 return (void *)(skb->_nfct & NFCT_PTRMASK);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004057#else
4058 return NULL;
4059#endif
4060}
4061
David Brazdil0f672f62019-12-10 10:32:29 +00004062static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004063{
David Brazdil0f672f62019-12-10 10:32:29 +00004064#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4065 return skb->_nfct;
4066#else
4067 return 0UL;
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004068#endif
David Brazdil0f672f62019-12-10 10:32:29 +00004069}
4070
4071static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4072{
4073#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4074 skb->_nfct = nfct;
4075#endif
4076}
4077
4078#ifdef CONFIG_SKB_EXTENSIONS
4079enum skb_ext_id {
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004080#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
David Brazdil0f672f62019-12-10 10:32:29 +00004081 SKB_EXT_BRIDGE_NF,
4082#endif
4083#ifdef CONFIG_XFRM
4084 SKB_EXT_SEC_PATH,
4085#endif
4086#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4087 TC_SKB_EXT,
4088#endif
4089 SKB_EXT_NUM, /* must be last */
4090};
4091
4092/**
4093 * struct skb_ext - sk_buff extensions
4094 * @refcnt: 1 on allocation, deallocated on 0
4095 * @offset: offset to add to @data to obtain extension address
4096 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4097 * @data: start of extension data, variable sized
4098 *
4099 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4100 * to use 'u8' types while allowing up to 2kb worth of extension data.
4101 */
4102struct skb_ext {
4103 refcount_t refcnt;
4104 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4105 u8 chunks; /* same */
4106 char data[0] __aligned(8);
4107};
4108
4109void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4110void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4111void __skb_ext_put(struct skb_ext *ext);
4112
4113static inline void skb_ext_put(struct sk_buff *skb)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004114{
David Brazdil0f672f62019-12-10 10:32:29 +00004115 if (skb->active_extensions)
4116 __skb_ext_put(skb->extensions);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004117}
David Brazdil0f672f62019-12-10 10:32:29 +00004118
4119static inline void __skb_ext_copy(struct sk_buff *dst,
4120 const struct sk_buff *src)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004121{
David Brazdil0f672f62019-12-10 10:32:29 +00004122 dst->active_extensions = src->active_extensions;
4123
4124 if (src->active_extensions) {
4125 struct skb_ext *ext = src->extensions;
4126
4127 refcount_inc(&ext->refcnt);
4128 dst->extensions = ext;
4129 }
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004130}
David Brazdil0f672f62019-12-10 10:32:29 +00004131
4132static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4133{
4134 skb_ext_put(dst);
4135 __skb_ext_copy(dst, src);
4136}
4137
4138static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4139{
4140 return !!ext->offset[i];
4141}
4142
4143static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4144{
4145 return skb->active_extensions & (1 << id);
4146}
4147
4148static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4149{
4150 if (skb_ext_exist(skb, id))
4151 __skb_ext_del(skb, id);
4152}
4153
4154static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4155{
4156 if (skb_ext_exist(skb, id)) {
4157 struct skb_ext *ext = skb->extensions;
4158
4159 return (void *)ext + (ext->offset[id] << 3);
4160 }
4161
4162 return NULL;
4163}
4164
4165static inline void skb_ext_reset(struct sk_buff *skb)
4166{
4167 if (unlikely(skb->active_extensions)) {
4168 __skb_ext_put(skb->extensions);
4169 skb->active_extensions = 0;
4170 }
4171}
4172
4173static inline bool skb_has_extensions(struct sk_buff *skb)
4174{
4175 return unlikely(skb->active_extensions);
4176}
4177#else
4178static inline void skb_ext_put(struct sk_buff *skb) {}
4179static inline void skb_ext_reset(struct sk_buff *skb) {}
4180static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4181static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4182static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4183static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
4184#endif /* CONFIG_SKB_EXTENSIONS */
4185
4186static inline void nf_reset_ct(struct sk_buff *skb)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004187{
4188#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4189 nf_conntrack_put(skb_nfct(skb));
4190 skb->_nfct = 0;
4191#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004192}
4193
4194static inline void nf_reset_trace(struct sk_buff *skb)
4195{
4196#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4197 skb->nf_trace = 0;
4198#endif
4199}
4200
4201static inline void ipvs_reset(struct sk_buff *skb)
4202{
4203#if IS_ENABLED(CONFIG_IP_VS)
4204 skb->ipvs_property = 0;
4205#endif
4206}
4207
David Brazdil0f672f62019-12-10 10:32:29 +00004208/* Note: This doesn't put any conntrack info in dst. */
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004209static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4210 bool copy)
4211{
4212#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4213 dst->_nfct = src->_nfct;
4214 nf_conntrack_get(skb_nfct(src));
4215#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004216#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
4217 if (copy)
4218 dst->nf_trace = src->nf_trace;
4219#endif
4220}
4221
4222static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4223{
4224#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
4225 nf_conntrack_put(skb_nfct(dst));
4226#endif
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004227 __nf_copy(dst, src, true);
4228}
4229
4230#ifdef CONFIG_NETWORK_SECMARK
4231static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4232{
4233 to->secmark = from->secmark;
4234}
4235
4236static inline void skb_init_secmark(struct sk_buff *skb)
4237{
4238 skb->secmark = 0;
4239}
4240#else
4241static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4242{ }
4243
4244static inline void skb_init_secmark(struct sk_buff *skb)
4245{ }
4246#endif
4247
David Brazdil0f672f62019-12-10 10:32:29 +00004248static inline int secpath_exists(const struct sk_buff *skb)
4249{
4250#ifdef CONFIG_XFRM
4251 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
4252#else
4253 return 0;
4254#endif
4255}
4256
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004257static inline bool skb_irq_freeable(const struct sk_buff *skb)
4258{
4259 return !skb->destructor &&
David Brazdil0f672f62019-12-10 10:32:29 +00004260 !secpath_exists(skb) &&
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004261 !skb_nfct(skb) &&
4262 !skb->_skb_refdst &&
4263 !skb_has_frag_list(skb);
4264}
4265
4266static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4267{
4268 skb->queue_mapping = queue_mapping;
4269}
4270
4271static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4272{
4273 return skb->queue_mapping;
4274}
4275
4276static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4277{
4278 to->queue_mapping = from->queue_mapping;
4279}
4280
4281static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4282{
4283 skb->queue_mapping = rx_queue + 1;
4284}
4285
4286static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
4287{
4288 return skb->queue_mapping - 1;
4289}
4290
4291static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
4292{
4293 return skb->queue_mapping != 0;
4294}
4295
4296static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4297{
4298 skb->dst_pending_confirm = val;
4299}
4300
4301static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4302{
4303 return skb->dst_pending_confirm != 0;
4304}
4305
David Brazdil0f672f62019-12-10 10:32:29 +00004306static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004307{
4308#ifdef CONFIG_XFRM
David Brazdil0f672f62019-12-10 10:32:29 +00004309 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004310#else
4311 return NULL;
4312#endif
4313}
4314
4315/* Keeps track of mac header offset relative to skb->head.
4316 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4317 * For non-tunnel skb it points to skb_mac_header() and for
4318 * tunnel skb it points to outer mac header.
4319 * Keeps track of level of encapsulation of network headers.
4320 */
4321struct skb_gso_cb {
4322 union {
4323 int mac_offset;
4324 int data_offset;
4325 };
4326 int encap_level;
4327 __wsum csum;
4328 __u16 csum_start;
4329};
4330#define SKB_SGO_CB_OFFSET 32
4331#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
4332
4333static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4334{
4335 return (skb_mac_header(inner_skb) - inner_skb->head) -
4336 SKB_GSO_CB(inner_skb)->mac_offset;
4337}
4338
4339static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4340{
4341 int new_headroom, headroom;
4342 int ret;
4343
4344 headroom = skb_headroom(skb);
4345 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4346 if (ret)
4347 return ret;
4348
4349 new_headroom = skb_headroom(skb);
4350 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4351 return 0;
4352}
4353
4354static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4355{
4356 /* Do not update partial checksums if remote checksum is enabled. */
4357 if (skb->remcsum_offload)
4358 return;
4359
4360 SKB_GSO_CB(skb)->csum = res;
4361 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4362}
4363
4364/* Compute the checksum for a gso segment. First compute the checksum value
4365 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4366 * then add in skb->csum (checksum from csum_start to end of packet).
4367 * skb->csum and csum_start are then updated to reflect the checksum of the
4368 * resultant packet starting from the transport header-- the resultant checksum
4369 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4370 * header.
4371 */
4372static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4373{
4374 unsigned char *csum_start = skb_transport_header(skb);
4375 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4376 __wsum partial = SKB_GSO_CB(skb)->csum;
4377
4378 SKB_GSO_CB(skb)->csum = res;
4379 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
4380
4381 return csum_fold(csum_partial(csum_start, plen, partial));
4382}
4383
4384static inline bool skb_is_gso(const struct sk_buff *skb)
4385{
4386 return skb_shinfo(skb)->gso_size;
4387}
4388
4389/* Note: Should be called only if skb_is_gso(skb) is true */
4390static inline bool skb_is_gso_v6(const struct sk_buff *skb)
4391{
4392 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4393}
4394
4395/* Note: Should be called only if skb_is_gso(skb) is true */
4396static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4397{
4398 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4399}
4400
David Brazdil0f672f62019-12-10 10:32:29 +00004401/* Note: Should be called only if skb_is_gso(skb) is true */
4402static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4403{
4404 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
4405}
4406
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004407static inline void skb_gso_reset(struct sk_buff *skb)
4408{
4409 skb_shinfo(skb)->gso_size = 0;
4410 skb_shinfo(skb)->gso_segs = 0;
4411 skb_shinfo(skb)->gso_type = 0;
4412}
4413
4414static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4415 u16 increment)
4416{
4417 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4418 return;
4419 shinfo->gso_size += increment;
4420}
4421
4422static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4423 u16 decrement)
4424{
4425 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4426 return;
4427 shinfo->gso_size -= decrement;
4428}
4429
4430void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4431
4432static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4433{
4434 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4435 * wanted then gso_type will be set. */
4436 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4437
4438 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4439 unlikely(shinfo->gso_type == 0)) {
4440 __skb_warn_lro_forwarding(skb);
4441 return true;
4442 }
4443 return false;
4444}
4445
4446static inline void skb_forward_csum(struct sk_buff *skb)
4447{
4448 /* Unfortunately we don't support this one. Any brave souls? */
4449 if (skb->ip_summed == CHECKSUM_COMPLETE)
4450 skb->ip_summed = CHECKSUM_NONE;
4451}
4452
4453/**
4454 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4455 * @skb: skb to check
4456 *
4457 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4458 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4459 * use this helper, to document places where we make this assertion.
4460 */
4461static inline void skb_checksum_none_assert(const struct sk_buff *skb)
4462{
4463#ifdef DEBUG
4464 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4465#endif
4466}
4467
4468bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
4469
4470int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
4471struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4472 unsigned int transport_len,
4473 __sum16(*skb_chkf)(struct sk_buff *skb));
4474
4475/**
4476 * skb_head_is_locked - Determine if the skb->head is locked down
4477 * @skb: skb to check
4478 *
4479 * The head on skbs build around a head frag can be removed if they are
4480 * not cloned. This function returns true if the skb head is locked down
4481 * due to either being allocated via kmalloc, or by being a clone with
4482 * multiple references to the head.
4483 */
4484static inline bool skb_head_is_locked(const struct sk_buff *skb)
4485{
4486 return !skb->head_frag || skb_cloned(skb);
4487}
4488
4489/* Local Checksum Offload.
4490 * Compute outer checksum based on the assumption that the
4491 * inner checksum will be offloaded later.
David Brazdil0f672f62019-12-10 10:32:29 +00004492 * See Documentation/networking/checksum-offloads.rst for
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00004493 * explanation of how this works.
4494 * Fill in outer checksum adjustment (e.g. with sum of outer
4495 * pseudo-header) before calling.
4496 * Also ensure that inner checksum is in linear data area.
4497 */
4498static inline __wsum lco_csum(struct sk_buff *skb)
4499{
4500 unsigned char *csum_start = skb_checksum_start(skb);
4501 unsigned char *l4_hdr = skb_transport_header(skb);
4502 __wsum partial;
4503
4504 /* Start with complement of inner checksum adjustment */
4505 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4506 skb->csum_offset));
4507
4508 /* Add in checksum of our headers (incl. outer checksum
4509 * adjustment filled in by caller) and return result.
4510 */
4511 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
4512}
4513
4514#endif /* __KERNEL__ */
4515#endif /* _LINUX_SKBUFF_H */