~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/include/linux/skbuff.h

Version: ~ [ linux-5.3-rc5 ] ~ [ linux-5.2.9 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.67 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.139 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.189 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.189 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.72 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *      Definitions for the 'struct sk_buff' memory handlers.
  3  *
  4  *      Authors:
  5  *              Alan Cox, <gw4pts@gw4pts.ampr.org>
  6  *              Florian La Roche, <rzsfl@rz.uni-sb.de>
  7  *
  8  *      This program is free software; you can redistribute it and/or
  9  *      modify it under the terms of the GNU General Public License
 10  *      as published by the Free Software Foundation; either version
 11  *      2 of the License, or (at your option) any later version.
 12  */
 13 
 14 #ifndef _LINUX_SKBUFF_H
 15 #define _LINUX_SKBUFF_H
 16 
 17 #include <linux/kernel.h>
 18 #include <linux/kmemcheck.h>
 19 #include <linux/compiler.h>
 20 #include <linux/time.h>
 21 #include <linux/bug.h>
 22 #include <linux/cache.h>
 23 #include <linux/rbtree.h>
 24 #include <linux/socket.h>
 25 
 26 #include <linux/atomic.h>
 27 #include <asm/types.h>
 28 #include <linux/spinlock.h>
 29 #include <linux/net.h>
 30 #include <linux/textsearch.h>
 31 #include <net/checksum.h>
 32 #include <linux/rcupdate.h>
 33 #include <linux/hrtimer.h>
 34 #include <linux/dma-mapping.h>
 35 #include <linux/netdev_features.h>
 36 #include <linux/sched.h>
 37 #include <net/flow_keys.h>
 38 
 39 /* A. Checksumming of received packets by device.
 40  *
 41  * CHECKSUM_NONE:
 42  *
 43  *   Device failed to checksum this packet e.g. due to lack of capabilities.
 44  *   The packet contains full (though not verified) checksum in packet but
 45  *   not in skb->csum. Thus, skb->csum is undefined in this case.
 46  *
 47  * CHECKSUM_UNNECESSARY:
 48  *
 49  *   The hardware you're dealing with doesn't calculate the full checksum
 50  *   (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
 51  *   for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
 52  *   if their checksums are okay. skb->csum is still undefined in this case
 53  *   though. It is a bad option, but, unfortunately, nowadays most vendors do
 54  *   this. Apparently with the secret goal to sell you new devices, when you
 55  *   will add new protocol to your host, f.e. IPv6 8)
 56  *
 57  *   CHECKSUM_UNNECESSARY is applicable to following protocols:
 58  *     TCP: IPv6 and IPv4.
 59  *     UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
 60  *       zero UDP checksum for either IPv4 or IPv6, the networking stack
 61  *       may perform further validation in this case.
 62  *     GRE: only if the checksum is present in the header.
 63  *     SCTP: indicates the CRC in SCTP header has been validated.
 64  *
 65  *   skb->csum_level indicates the number of consecutive checksums found in
 66  *   the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
 67  *   For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
 68  *   and a device is able to verify the checksums for UDP (possibly zero),
 69  *   GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
 70  *   two. If the device were only able to verify the UDP checksum and not
 71  *   GRE, either because it doesn't support GRE checksum of because GRE
 72  *   checksum is bad, skb->csum_level would be set to zero (TCP checksum is
 73  *   not considered in this case).
 74  *
 75  * CHECKSUM_COMPLETE:
 76  *
 77  *   This is the most generic way. The device supplied checksum of the _whole_
 78  *   packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
 79  *   hardware doesn't need to parse L3/L4 headers to implement this.
 80  *
 81  *   Note: Even if device supports only some protocols, but is able to produce
 82  *   skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
 83  *
 84  * CHECKSUM_PARTIAL:
 85  *
 86  *   A checksum is set up to be offloaded to a device as described in the
 87  *   output description for CHECKSUM_PARTIAL. This may occur on a packet
 88  *   received directly from another Linux OS, e.g., a virtualized Linux kernel
 89  *   on the same host, or it may be set in the input path in GRO or remote
 90  *   checksum offload. For the purposes of checksum verification, the checksum
 91  *   referred to by skb->csum_start + skb->csum_offset and any preceding
 92  *   checksums in the packet are considered verified. Any checksums in the
 93  *   packet that are after the checksum being offloaded are not considered to
 94  *   be verified.
 95  *
 96  * B. Checksumming on output.
 97  *
 98  * CHECKSUM_NONE:
 99  *
100  *   The skb was already checksummed by the protocol, or a checksum is not
101  *   required.
102  *
103  * CHECKSUM_PARTIAL:
104  *
105  *   The device is required to checksum the packet as seen by hard_start_xmit()
106  *   from skb->csum_start up to the end, and to record/write the checksum at
107  *   offset skb->csum_start + skb->csum_offset.
108  *
109  *   The device must show its capabilities in dev->features, set up at device
110  *   setup time, e.g. netdev_features.h:
111  *
112  *      NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
113  *      NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
114  *                        IPv4. Sigh. Vendors like this way for an unknown reason.
115  *                        Though, see comment above about CHECKSUM_UNNECESSARY. 8)
116  *      NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
117  *      NETIF_F_...     - Well, you get the picture.
118  *
119  * CHECKSUM_UNNECESSARY:
120  *
121  *   Normally, the device will do per protocol specific checksumming. Protocol
122  *   implementations that do not want the NIC to perform the checksum
123  *   calculation should use this flag in their outgoing skbs.
124  *
125  *      NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
126  *                         offload. Correspondingly, the FCoE protocol driver
127  *                         stack should use CHECKSUM_UNNECESSARY.
128  *
129  * Any questions? No questions, good.           --ANK
130  */
131 
132 /* Don't change this without changing skb_csum_unnecessary! */
133 #define CHECKSUM_NONE           0
134 #define CHECKSUM_UNNECESSARY    1
135 #define CHECKSUM_COMPLETE       2
136 #define CHECKSUM_PARTIAL        3
137 
138 /* Maximum value in skb->csum_level */
139 #define SKB_MAX_CSUM_LEVEL      3
140 
141 #define SKB_DATA_ALIGN(X)       ALIGN(X, SMP_CACHE_BYTES)
142 #define SKB_WITH_OVERHEAD(X)    \
143         ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
144 #define SKB_MAX_ORDER(X, ORDER) \
145         SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
146 #define SKB_MAX_HEAD(X)         (SKB_MAX_ORDER((X), 0))
147 #define SKB_MAX_ALLOC           (SKB_MAX_ORDER(0, 2))
148 
149 /* return minimum truesize of one skb containing X bytes of data */
150 #define SKB_TRUESIZE(X) ((X) +                                          \
151                          SKB_DATA_ALIGN(sizeof(struct sk_buff)) +       \
152                          SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
153 
154 struct net_device;
155 struct scatterlist;
156 struct pipe_inode_info;
157 struct iov_iter;
158 struct napi_struct;
159 
160 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
161 struct nf_conntrack {
162         atomic_t use;
163 };
164 #endif
165 
166 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
167 struct nf_bridge_info {
168         atomic_t                use;
169         unsigned int            mask;
170         struct net_device       *physindev;
171         struct net_device       *physoutdev;
172         unsigned long           data[32 / sizeof(unsigned long)];
173 };
174 #endif
175 
176 struct sk_buff_head {
177         /* These two members must be first. */
178         struct sk_buff  *next;
179         struct sk_buff  *prev;
180 
181         __u32           qlen;
182         spinlock_t      lock;
183 };
184 
185 struct sk_buff;
186 
187 /* To allow 64K frame to be packed as single skb without frag_list we
188  * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
189  * buffers which do not start on a page boundary.
190  *
191  * Since GRO uses frags we allocate at least 16 regardless of page
192  * size.
193  */
194 #if (65536/PAGE_SIZE + 1) < 16
195 #define MAX_SKB_FRAGS 16UL
196 #else
197 #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
198 #endif
199 
200 typedef struct skb_frag_struct skb_frag_t;
201 
202 struct skb_frag_struct {
203         struct {
204                 struct page *p;
205         } page;
206 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
207         __u32 page_offset;
208         __u32 size;
209 #else
210         __u16 page_offset;
211         __u16 size;
212 #endif
213 };
214 
215 static inline unsigned int skb_frag_size(const skb_frag_t *frag)
216 {
217         return frag->size;
218 }
219 
220 static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
221 {
222         frag->size = size;
223 }
224 
225 static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
226 {
227         frag->size += delta;
228 }
229 
230 static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
231 {
232         frag->size -= delta;
233 }
234 
235 #define HAVE_HW_TIME_STAMP
236 
237 /**
238  * struct skb_shared_hwtstamps - hardware time stamps
239  * @hwtstamp:   hardware time stamp transformed into duration
240  *              since arbitrary point in time
241  *
242  * Software time stamps generated by ktime_get_real() are stored in
243  * skb->tstamp.
244  *
245  * hwtstamps can only be compared against other hwtstamps from
246  * the same device.
247  *
248  * This structure is attached to packets as part of the
249  * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
250  */
251 struct skb_shared_hwtstamps {
252         ktime_t hwtstamp;
253 };
254 
255 /* Definitions for tx_flags in struct skb_shared_info */
256 enum {
257         /* generate hardware time stamp */
258         SKBTX_HW_TSTAMP = 1 << 0,
259 
260         /* generate software time stamp when queueing packet to NIC */
261         SKBTX_SW_TSTAMP = 1 << 1,
262 
263         /* device driver is going to provide hardware time stamp */
264         SKBTX_IN_PROGRESS = 1 << 2,
265 
266         /* device driver supports TX zero-copy buffers */
267         SKBTX_DEV_ZEROCOPY = 1 << 3,
268 
269         /* generate wifi status information (where possible) */
270         SKBTX_WIFI_STATUS = 1 << 4,
271 
272         /* This indicates at least one fragment might be overwritten
273          * (as in vmsplice(), sendfile() ...)
274          * If we need to compute a TX checksum, we'll need to copy
275          * all frags to avoid possible bad checksum
276          */
277         SKBTX_SHARED_FRAG = 1 << 5,
278 
279         /* generate software time stamp when entering packet scheduling */
280         SKBTX_SCHED_TSTAMP = 1 << 6,
281 
282         /* generate software timestamp on peer data acknowledgment */
283         SKBTX_ACK_TSTAMP = 1 << 7,
284 };
285 
286 #define SKBTX_ANY_SW_TSTAMP     (SKBTX_SW_TSTAMP    | \
287                                  SKBTX_SCHED_TSTAMP | \
288                                  SKBTX_ACK_TSTAMP)
289 #define SKBTX_ANY_TSTAMP        (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
290 
291 /*
292  * The callback notifies userspace to release buffers when skb DMA is done in
293  * lower device, the skb last reference should be 0 when calling this.
294  * The zerocopy_success argument is true if zero copy transmit occurred,
295  * false on data copy or out of memory error caused by data copy attempt.
296  * The ctx field is used to track device context.
297  * The desc field is used to track userspace buffer index.
298  */
299 struct ubuf_info {
300         void (*callback)(struct ubuf_info *, bool zerocopy_success);
301         void *ctx;
302         unsigned long desc;
303 };
304 
305 /* This data is invariant across clones and lives at
306  * the end of the header data, ie. at skb->end.
307  */
308 struct skb_shared_info {
309         unsigned char   nr_frags;
310         __u8            tx_flags;
311         unsigned short  gso_size;
312         /* Warning: this field is not always filled in (UFO)! */
313         unsigned short  gso_segs;
314         unsigned short  gso_type;
315         struct sk_buff  *frag_list;
316         struct skb_shared_hwtstamps hwtstamps;
317         u32             tskey;
318         __be32          ip6_frag_id;
319 
320         /*
321          * Warning : all fields before dataref are cleared in __alloc_skb()
322          */
323         atomic_t        dataref;
324 
325         /* Intermediate layers must ensure that destructor_arg
326          * remains valid until skb destructor */
327         void *          destructor_arg;
328 
329         /* must be last field, see pskb_expand_head() */
330         skb_frag_t      frags[MAX_SKB_FRAGS];
331 };
332 
333 /* We divide dataref into two halves.  The higher 16 bits hold references
334  * to the payload part of skb->data.  The lower 16 bits hold references to
335  * the entire skb->data.  A clone of a headerless skb holds the length of
336  * the header in skb->hdr_len.
337  *
338  * All users must obey the rule that the skb->data reference count must be
339  * greater than or equal to the payload reference count.
340  *
341  * Holding a reference to the payload part means that the user does not
342  * care about modifications to the header part of skb->data.
343  */
344 #define SKB_DATAREF_SHIFT 16
345 #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
346 
347 
348 enum {
349         SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
350         SKB_FCLONE_ORIG,        /* orig skb (from fclone_cache) */
351         SKB_FCLONE_CLONE,       /* companion fclone skb (from fclone_cache) */
352 };
353 
354 enum {
355         SKB_GSO_TCPV4 = 1 << 0,
356         SKB_GSO_UDP = 1 << 1,
357 
358         /* This indicates the skb is from an untrusted source. */
359         SKB_GSO_DODGY = 1 << 2,
360 
361         /* This indicates the tcp segment has CWR set. */
362         SKB_GSO_TCP_ECN = 1 << 3,
363 
364         SKB_GSO_TCPV6 = 1 << 4,
365 
366         SKB_GSO_FCOE = 1 << 5,
367 
368         SKB_GSO_GRE = 1 << 6,
369 
370         SKB_GSO_GRE_CSUM = 1 << 7,
371 
372         SKB_GSO_IPIP = 1 << 8,
373 
374         SKB_GSO_SIT = 1 << 9,
375 
376         SKB_GSO_UDP_TUNNEL = 1 << 10,
377 
378         SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
379 
380         SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
381 };
382 
383 #if BITS_PER_LONG > 32
384 #define NET_SKBUFF_DATA_USES_OFFSET 1
385 #endif
386 
387 #ifdef NET_SKBUFF_DATA_USES_OFFSET
388 typedef unsigned int sk_buff_data_t;
389 #else
390 typedef unsigned char *sk_buff_data_t;
391 #endif
392 
393 /**
394  * struct skb_mstamp - multi resolution time stamps
395  * @stamp_us: timestamp in us resolution
396  * @stamp_jiffies: timestamp in jiffies
397  */
398 struct skb_mstamp {
399         union {
400                 u64             v64;
401                 struct {
402                         u32     stamp_us;
403                         u32     stamp_jiffies;
404                 };
405         };
406 };
407 
408 /**
409  * skb_mstamp_get - get current timestamp
410  * @cl: place to store timestamps
411  */
412 static inline void skb_mstamp_get(struct skb_mstamp *cl)
413 {
414         u64 val = local_clock();
415 
416         do_div(val, NSEC_PER_USEC);
417         cl->stamp_us = (u32)val;
418         cl->stamp_jiffies = (u32)jiffies;
419 }
420 
421 /**
422  * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
423  * @t1: pointer to newest sample
424  * @t0: pointer to oldest sample
425  */
426 static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
427                                       const struct skb_mstamp *t0)
428 {
429         s32 delta_us = t1->stamp_us - t0->stamp_us;
430         u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
431 
432         /* If delta_us is negative, this might be because interval is too big,
433          * or local_clock() drift is too big : fallback using jiffies.
434          */
435         if (delta_us <= 0 ||
436             delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
437 
438                 delta_us = jiffies_to_usecs(delta_jiffies);
439 
440         return delta_us;
441 }
442 
443 
444 /** 
445  *      struct sk_buff - socket buffer
446  *      @next: Next buffer in list
447  *      @prev: Previous buffer in list
448  *      @tstamp: Time we arrived/left
449  *      @rbnode: RB tree node, alternative to next/prev for netem/tcp
450  *      @sk: Socket we are owned by
451  *      @dev: Device we arrived on/are leaving by
452  *      @cb: Control buffer. Free for use by every layer. Put private vars here
453  *      @_skb_refdst: destination entry (with norefcount bit)
454  *      @sp: the security path, used for xfrm
455  *      @len: Length of actual data
456  *      @data_len: Data length
457  *      @mac_len: Length of link layer header
458  *      @hdr_len: writable header length of cloned skb
459  *      @csum: Checksum (must include start/offset pair)
460  *      @csum_start: Offset from skb->head where checksumming should start
461  *      @csum_offset: Offset from csum_start where checksum should be stored
462  *      @priority: Packet queueing priority
463  *      @ignore_df: allow local fragmentation
464  *      @cloned: Head may be cloned (check refcnt to be sure)
465  *      @ip_summed: Driver fed us an IP checksum
466  *      @nohdr: Payload reference only, must not modify header
467  *      @nfctinfo: Relationship of this skb to the connection
468  *      @pkt_type: Packet class
469  *      @fclone: skbuff clone status
470  *      @ipvs_property: skbuff is owned by ipvs
471  *      @peeked: this packet has been seen already, so stats have been
472  *              done for it, don't do them again
473  *      @nf_trace: netfilter packet trace flag
474  *      @protocol: Packet protocol from driver
475  *      @destructor: Destruct function
476  *      @nfct: Associated connection, if any
477  *      @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
478  *      @skb_iif: ifindex of device we arrived on
479  *      @tc_index: Traffic control index
480  *      @tc_verd: traffic control verdict
481  *      @hash: the packet hash
482  *      @queue_mapping: Queue mapping for multiqueue devices
483  *      @xmit_more: More SKBs are pending for this queue
484  *      @ndisc_nodetype: router type (from link layer)
485  *      @ooo_okay: allow the mapping of a socket to a queue to be changed
486  *      @l4_hash: indicate hash is a canonical 4-tuple hash over transport
487  *              ports.
488  *      @sw_hash: indicates hash was computed in software stack
489  *      @wifi_acked_valid: wifi_acked was set
490  *      @wifi_acked: whether frame was acked on wifi or not
491  *      @no_fcs:  Request NIC to treat last 4 bytes as Ethernet FCS
492   *     @napi_id: id of the NAPI struct this skb came from
493  *      @secmark: security marking
494  *      @mark: Generic packet mark
495  *      @dropcount: total number of sk_receive_queue overflows
496  *      @vlan_proto: vlan encapsulation protocol
497  *      @vlan_tci: vlan tag control information
498  *      @inner_protocol: Protocol (encapsulation)
499  *      @inner_transport_header: Inner transport layer header (encapsulation)
500  *      @inner_network_header: Network layer header (encapsulation)
501  *      @inner_mac_header: Link layer header (encapsulation)
502  *      @transport_header: Transport layer header
503  *      @network_header: Network layer header
504  *      @mac_header: Link layer header
505  *      @tail: Tail pointer
506  *      @end: End pointer
507  *      @head: Head of buffer
508  *      @data: Data head pointer
509  *      @truesize: Buffer size
510  *      @users: User count - see {datagram,tcp}.c
511  */
512 
513 struct sk_buff {
514         union {
515                 struct {
516                         /* These two members must be first. */
517                         struct sk_buff          *next;
518                         struct sk_buff          *prev;
519 
520                         union {
521                                 ktime_t         tstamp;
522                                 struct skb_mstamp skb_mstamp;
523                         };
524                 };
525                 struct rb_node  rbnode; /* used in netem & tcp stack */
526         };
527         struct sock             *sk;
528         struct net_device       *dev;
529 
530         /*
531          * This is the control buffer. It is free to use for every
532          * layer. Please put your private variables there. If you
533          * want to keep them across layers you have to do a skb_clone()
534          * first. This is owned by whoever has the skb queued ATM.
535          */
536         char                    cb[48] __aligned(8);
537 
538         unsigned long           _skb_refdst;
539         void                    (*destructor)(struct sk_buff *skb);
540 #ifdef CONFIG_XFRM
541         struct  sec_path        *sp;
542 #endif
543 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
544         struct nf_conntrack     *nfct;
545 #endif
546 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
547         struct nf_bridge_info   *nf_bridge;
548 #endif
549         unsigned int            len,
550                                 data_len;
551         __u16                   mac_len,
552                                 hdr_len;
553 
554         /* Following fields are _not_ copied in __copy_skb_header()
555          * Note that queue_mapping is here mostly to fill a hole.
556          */
557         kmemcheck_bitfield_begin(flags1);
558         __u16                   queue_mapping;
559         __u8                    cloned:1,
560                                 nohdr:1,
561                                 fclone:2,
562                                 peeked:1,
563                                 head_frag:1,
564                                 xmit_more:1;
565         /* one bit hole */
566         kmemcheck_bitfield_end(flags1);
567 
568         /* fields enclosed in headers_start/headers_end are copied
569          * using a single memcpy() in __copy_skb_header()
570          */
571         /* private: */
572         __u32                   headers_start[0];
573         /* public: */
574 
575 /* if you move pkt_type around you also must adapt those constants */
576 #ifdef __BIG_ENDIAN_BITFIELD
577 #define PKT_TYPE_MAX    (7 << 5)
578 #else
579 #define PKT_TYPE_MAX    7
580 #endif
581 #define PKT_TYPE_OFFSET()       offsetof(struct sk_buff, __pkt_type_offset)
582 
583         __u8                    __pkt_type_offset[0];
584         __u8                    pkt_type:3;
585         __u8                    pfmemalloc:1;
586         __u8                    ignore_df:1;
587         __u8                    nfctinfo:3;
588 
589         __u8                    nf_trace:1;
590         __u8                    ip_summed:2;
591         __u8                    ooo_okay:1;
592         __u8                    l4_hash:1;
593         __u8                    sw_hash:1;
594         __u8                    wifi_acked_valid:1;
595         __u8                    wifi_acked:1;
596 
597         __u8                    no_fcs:1;
598         /* Indicates the inner headers are valid in the skbuff. */
599         __u8                    encapsulation:1;
600         __u8                    encap_hdr_csum:1;
601         __u8                    csum_valid:1;
602         __u8                    csum_complete_sw:1;
603         __u8                    csum_level:2;
604         __u8                    csum_bad:1;
605 
606 #ifdef CONFIG_IPV6_NDISC_NODETYPE
607         __u8                    ndisc_nodetype:2;
608 #endif
609         __u8                    ipvs_property:1;
610         __u8                    inner_protocol_type:1;
611         __u8                    remcsum_offload:1;
612         /* 3 or 5 bit hole */
613 
614 #ifdef CONFIG_NET_SCHED
615         __u16                   tc_index;       /* traffic control index */
616 #ifdef CONFIG_NET_CLS_ACT
617         __u16                   tc_verd;        /* traffic control verdict */
618 #endif
619 #endif
620 
621         union {
622                 __wsum          csum;
623                 struct {
624                         __u16   csum_start;
625                         __u16   csum_offset;
626                 };
627         };
628         __u32                   priority;
629         int                     skb_iif;
630         __u32                   hash;
631         __be16                  vlan_proto;
632         __u16                   vlan_tci;
633 #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
634         union {
635                 unsigned int    napi_id;
636                 unsigned int    sender_cpu;
637         };
638 #endif
639 #ifdef CONFIG_NETWORK_SECMARK
640         __u32                   secmark;
641 #endif
642         union {
643                 __u32           mark;
644                 __u32           dropcount;
645                 __u32           reserved_tailroom;
646         };
647 
648         union {
649                 __be16          inner_protocol;
650                 __u8            inner_ipproto;
651         };
652 
653         __u16                   inner_transport_header;
654         __u16                   inner_network_header;
655         __u16                   inner_mac_header;
656 
657         __be16                  protocol;
658         __u16                   transport_header;
659         __u16                   network_header;
660         __u16                   mac_header;
661 
662         /* private: */
663         __u32                   headers_end[0];
664         /* public: */
665 
666         /* These elements must be at the end, see alloc_skb() for details.  */
667         sk_buff_data_t          tail;
668         sk_buff_data_t          end;
669         unsigned char           *head,
670                                 *data;
671         unsigned int            truesize;
672         atomic_t                users;
673 };
674 
675 #ifdef __KERNEL__
676 /*
677  *      Handling routines are only of interest to the kernel
678  */
679 #include <linux/slab.h>
680 
681 
682 #define SKB_ALLOC_FCLONE        0x01
683 #define SKB_ALLOC_RX            0x02
684 #define SKB_ALLOC_NAPI          0x04
685 
686 /* Returns true if the skb was allocated from PFMEMALLOC reserves */
687 static inline bool skb_pfmemalloc(const struct sk_buff *skb)
688 {
689         return unlikely(skb->pfmemalloc);
690 }
691 
692 /*
693  * skb might have a dst pointer attached, refcounted or not.
694  * _skb_refdst low order bit is set if refcount was _not_ taken
695  */
696 #define SKB_DST_NOREF   1UL
697 #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
698 
699 /**
700  * skb_dst - returns skb dst_entry
701  * @skb: buffer
702  *
703  * Returns skb dst_entry, regardless of reference taken or not.
704  */
705 static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
706 {
707         /* If refdst was not refcounted, check we still are in a 
708          * rcu_read_lock section
709          */
710         WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
711                 !rcu_read_lock_held() &&
712                 !rcu_read_lock_bh_held());
713         return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
714 }
715 
716 /**
717  * skb_dst_set - sets skb dst
718  * @skb: buffer
719  * @dst: dst entry
720  *
721  * Sets skb dst, assuming a reference was taken on dst and should
722  * be released by skb_dst_drop()
723  */
724 static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
725 {
726         skb->_skb_refdst = (unsigned long)dst;
727 }
728 
729 /**
730  * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
731  * @skb: buffer
732  * @dst: dst entry
733  *
734  * Sets skb dst, assuming a reference was not taken on dst.
735  * If dst entry is cached, we do not take reference and dst_release
736  * will be avoided by refdst_drop. If dst entry is not cached, we take
737  * reference, so that last dst_release can destroy the dst immediately.
738  */
739 static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
740 {
741         WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
742         skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
743 }
744 
745 /**
746  * skb_dst_is_noref - Test if skb dst isn't refcounted
747  * @skb: buffer
748  */
749 static inline bool skb_dst_is_noref(const struct sk_buff *skb)
750 {
751         return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
752 }
753 
754 static inline struct rtable *skb_rtable(const struct sk_buff *skb)
755 {
756         return (struct rtable *)skb_dst(skb);
757 }
758 
759 void kfree_skb(struct sk_buff *skb);
760 void kfree_skb_list(struct sk_buff *segs);
761 void skb_tx_error(struct sk_buff *skb);
762 void consume_skb(struct sk_buff *skb);
763 void  __kfree_skb(struct sk_buff *skb);
764 extern struct kmem_cache *skbuff_head_cache;
765 
766 void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
767 bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
768                       bool *fragstolen, int *delta_truesize);
769 
770 struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
771                             int node);
772 struct sk_buff *__build_skb(void *data, unsigned int frag_size);
773 struct sk_buff *build_skb(void *data, unsigned int frag_size);
774 static inline struct sk_buff *alloc_skb(unsigned int size,
775                                         gfp_t priority)
776 {
777         return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
778 }
779 
780 struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
781                                      unsigned long data_len,
782                                      int max_page_order,
783                                      int *errcode,
784                                      gfp_t gfp_mask);
785 
786 /* Layout of fast clones : [skb1][skb2][fclone_ref] */
787 struct sk_buff_fclones {
788         struct sk_buff  skb1;
789 
790         struct sk_buff  skb2;
791 
792         atomic_t        fclone_ref;
793 };
794 
795 /**
796  *      skb_fclone_busy - check if fclone is busy
797  *      @skb: buffer
798  *
799  * Returns true is skb is a fast clone, and its clone is not freed.
800  * Some drivers call skb_orphan() in their ndo_start_xmit(),
801  * so we also check that this didnt happen.
802  */
803 static inline bool skb_fclone_busy(const struct sock *sk,
804                                    const struct sk_buff *skb)
805 {
806         const struct sk_buff_fclones *fclones;
807 
808         fclones = container_of(skb, struct sk_buff_fclones, skb1);
809 
810         return skb->fclone == SKB_FCLONE_ORIG &&
811                atomic_read(&fclones->fclone_ref) > 1 &&
812                fclones->skb2.sk == sk;
813 }
814 
815 static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
816                                                gfp_t priority)
817 {
818         return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
819 }
820 
821 struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
822 static inline struct sk_buff *alloc_skb_head(gfp_t priority)
823 {
824         return __alloc_skb_head(priority, -1);
825 }
826 
827 struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
828 int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
829 struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
830 struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
831 struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
832                                    gfp_t gfp_mask, bool fclone);
833 static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
834                                           gfp_t gfp_mask)
835 {
836         return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
837 }
838 
839 int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
840 struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
841                                      unsigned int headroom);
842 struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
843                                 int newtailroom, gfp_t priority);
844 int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
845                         int offset, int len);
846 int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
847                  int len);
848 int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
849 int skb_pad(struct sk_buff *skb, int pad);
850 #define dev_kfree_skb(a)        consume_skb(a)
851 
852 int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
853                             int getfrag(void *from, char *to, int offset,
854                                         int len, int odd, struct sk_buff *skb),
855                             void *from, int length);
856 
857 struct skb_seq_state {
858         __u32           lower_offset;
859         __u32           upper_offset;
860         __u32           frag_idx;
861         __u32           stepped_offset;
862         struct sk_buff  *root_skb;
863         struct sk_buff  *cur_skb;
864         __u8            *frag_data;
865 };
866 
867 void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
868                           unsigned int to, struct skb_seq_state *st);
869 unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
870                           struct skb_seq_state *st);
871 void skb_abort_seq_read(struct skb_seq_state *st);
872 
873 unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
874                            unsigned int to, struct ts_config *config,
875                            struct ts_state *state);
876 
877 /*
878  * Packet hash types specify the type of hash in skb_set_hash.
879  *
880  * Hash types refer to the protocol layer addresses which are used to
881  * construct a packet's hash. The hashes are used to differentiate or identify
882  * flows of the protocol layer for the hash type. Hash types are either
883  * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
884  *
885  * Properties of hashes:
886  *
887  * 1) Two packets in different flows have different hash values
888  * 2) Two packets in the same flow should have the same hash value
889  *
890  * A hash at a higher layer is considered to be more specific. A driver should
891  * set the most specific hash possible.
892  *
893  * A driver cannot indicate a more specific hash than the layer at which a hash
894  * was computed. For instance an L3 hash cannot be set as an L4 hash.
895  *
896  * A driver may indicate a hash level which is less specific than the
897  * actual layer the hash was computed on. For instance, a hash computed
898  * at L4 may be considered an L3 hash. This should only be done if the
899  * driver can't unambiguously determine that the HW computed the hash at
900  * the higher layer. Note that the "should" in the second property above
901  * permits this.
902  */
903 enum pkt_hash_types {
904         PKT_HASH_TYPE_NONE,     /* Undefined type */
905         PKT_HASH_TYPE_L2,       /* Input: src_MAC, dest_MAC */
906         PKT_HASH_TYPE_L3,       /* Input: src_IP, dst_IP */
907         PKT_HASH_TYPE_L4,       /* Input: src_IP, dst_IP, src_port, dst_port */
908 };
909 
910 static inline void
911 skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
912 {
913         skb->l4_hash = (type == PKT_HASH_TYPE_L4);
914         skb->sw_hash = 0;
915         skb->hash = hash;
916 }
917 
918 void __skb_get_hash(struct sk_buff *skb);
919 static inline __u32 skb_get_hash(struct sk_buff *skb)
920 {
921         if (!skb->l4_hash && !skb->sw_hash)
922                 __skb_get_hash(skb);
923 
924         return skb->hash;
925 }
926 
927 static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
928 {
929         return skb->hash;
930 }
931 
932 static inline void skb_clear_hash(struct sk_buff *skb)
933 {
934         skb->hash = 0;
935         skb->sw_hash = 0;
936         skb->l4_hash = 0;
937 }
938 
939 static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
940 {
941         if (!skb->l4_hash)
942                 skb_clear_hash(skb);
943 }
944 
945 static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
946 {
947         to->hash = from->hash;
948         to->sw_hash = from->sw_hash;
949         to->l4_hash = from->l4_hash;
950 };
951 
952 static inline void skb_sender_cpu_clear(struct sk_buff *skb)
953 {
954 #ifdef CONFIG_XPS
955         skb->sender_cpu = 0;
956 #endif
957 }
958 
959 #ifdef NET_SKBUFF_DATA_USES_OFFSET
960 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
961 {
962         return skb->head + skb->end;
963 }
964 
965 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
966 {
967         return skb->end;
968 }
969 #else
970 static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
971 {
972         return skb->end;
973 }
974 
975 static inline unsigned int skb_end_offset(const struct sk_buff *skb)
976 {
977         return skb->end - skb->head;
978 }
979 #endif
980 
981 /* Internal */
982 #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
983 
984 static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
985 {
986         return &skb_shinfo(skb)->hwtstamps;
987 }
988 
989 /**
990  *      skb_queue_empty - check if a queue is empty
991  *      @list: queue head
992  *
993  *      Returns true if the queue is empty, false otherwise.
994  */
995 static inline int skb_queue_empty(const struct sk_buff_head *list)
996 {
997         return list->next == (const struct sk_buff *) list;
998 }
999 
1000 /**
1001  *      skb_queue_is_last - check if skb is the last entry in the queue
1002  *      @list: queue head
1003  *      @skb: buffer
1004  *
1005  *      Returns true if @skb is the last buffer on the list.
1006  */
1007 static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1008                                      const struct sk_buff *skb)
1009 {
1010         return skb->next == (const struct sk_buff *) list;
1011 }
1012 
1013 /**
1014  *      skb_queue_is_first - check if skb is the first entry in the queue
1015  *      @list: queue head
1016  *      @skb: buffer
1017  *
1018  *      Returns true if @skb is the first buffer on the list.
1019  */
1020 static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1021                                       const struct sk_buff *skb)
1022 {
1023         return skb->prev == (const struct sk_buff *) list;
1024 }
1025 
1026 /**
1027  *      skb_queue_next - return the next packet in the queue
1028  *      @list: queue head
1029  *      @skb: current buffer
1030  *
1031  *      Return the next packet in @list after @skb.  It is only valid to
1032  *      call this if skb_queue_is_last() evaluates to false.
1033  */
1034 static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1035                                              const struct sk_buff *skb)
1036 {
1037         /* This BUG_ON may seem severe, but if we just return then we
1038          * are going to dereference garbage.
1039          */
1040         BUG_ON(skb_queue_is_last(list, skb));
1041         return skb->next;
1042 }
1043 
1044 /**
1045  *      skb_queue_prev - return the prev packet in the queue
1046  *      @list: queue head
1047  *      @skb: current buffer
1048  *
1049  *      Return the prev packet in @list before @skb.  It is only valid to
1050  *      call this if skb_queue_is_first() evaluates to false.
1051  */
1052 static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1053                                              const struct sk_buff *skb)
1054 {
1055         /* This BUG_ON may seem severe, but if we just return then we
1056          * are going to dereference garbage.
1057          */
1058         BUG_ON(skb_queue_is_first(list, skb));
1059         return skb->prev;
1060 }
1061 
1062 /**
1063  *      skb_get - reference buffer
1064  *      @skb: buffer to reference
1065  *
1066  *      Makes another reference to a socket buffer and returns a pointer
1067  *      to the buffer.
1068  */
1069 static inline struct sk_buff *skb_get(struct sk_buff *skb)
1070 {
1071         atomic_inc(&skb->users);
1072         return skb;
1073 }
1074 
1075 /*
1076  * If users == 1, we are the only owner and are can avoid redundant
1077  * atomic change.
1078  */
1079 
1080 /**
1081  *      skb_cloned - is the buffer a clone
1082  *      @skb: buffer to check
1083  *
1084  *      Returns true if the buffer was generated with skb_clone() and is
1085  *      one of multiple shared copies of the buffer. Cloned buffers are
1086  *      shared data so must not be written to under normal circumstances.
1087  */
1088 static inline int skb_cloned(const struct sk_buff *skb)
1089 {
1090         return skb->cloned &&
1091                (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1092 }
1093 
1094 static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1095 {
1096         might_sleep_if(pri & __GFP_WAIT);
1097 
1098         if (skb_cloned(skb))
1099                 return pskb_expand_head(skb, 0, 0, pri);
1100 
1101         return 0;
1102 }
1103 
1104 /**
1105  *      skb_header_cloned - is the header a clone
1106  *      @skb: buffer to check
1107  *
1108  *      Returns true if modifying the header part of the buffer requires
1109  *      the data to be copied.
1110  */
1111 static inline int skb_header_cloned(const struct sk_buff *skb)
1112 {
1113         int dataref;
1114 
1115         if (!skb->cloned)
1116                 return 0;
1117 
1118         dataref = atomic_read(&skb_shinfo(skb)->dataref);
1119         dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1120         return dataref != 1;
1121 }
1122 
1123 /**
1124  *      skb_header_release - release reference to header
1125  *      @skb: buffer to operate on
1126  *
1127  *      Drop a reference to the header part of the buffer.  This is done
1128  *      by acquiring a payload reference.  You must not read from the header
1129  *      part of skb->data after this.
1130  *      Note : Check if you can use __skb_header_release() instead.
1131  */
1132 static inline void skb_header_release(struct sk_buff *skb)
1133 {
1134         BUG_ON(skb->nohdr);
1135         skb->nohdr = 1;
1136         atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1137 }
1138 
1139 /**
1140  *      __skb_header_release - release reference to header
1141  *      @skb: buffer to operate on
1142  *
1143  *      Variant of skb_header_release() assuming skb is private to caller.
1144  *      We can avoid one atomic operation.
1145  */
1146 static inline void __skb_header_release(struct sk_buff *skb)
1147 {
1148         skb->nohdr = 1;
1149         atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1150 }
1151 
1152 
1153 /**
1154  *      skb_shared - is the buffer shared
1155  *      @skb: buffer to check
1156  *
1157  *      Returns true if more than one person has a reference to this
1158  *      buffer.
1159  */
1160 static inline int skb_shared(const struct sk_buff *skb)
1161 {
1162         return atomic_read(&skb->users) != 1;
1163 }
1164 
1165 /**
1166  *      skb_share_check - check if buffer is shared and if so clone it
1167  *      @skb: buffer to check
1168  *      @pri: priority for memory allocation
1169  *
1170  *      If the buffer is shared the buffer is cloned and the old copy
1171  *      drops a reference. A new clone with a single reference is returned.
1172  *      If the buffer is not shared the original buffer is returned. When
1173  *      being called from interrupt status or with spinlocks held pri must
1174  *      be GFP_ATOMIC.
1175  *
1176  *      NULL is returned on a memory allocation failure.
1177  */
1178 static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1179 {
1180         might_sleep_if(pri & __GFP_WAIT);
1181         if (skb_shared(skb)) {
1182                 struct sk_buff *nskb = skb_clone(skb, pri);
1183 
1184                 if (likely(nskb))
1185                         consume_skb(skb);
1186                 else
1187                         kfree_skb(skb);
1188                 skb = nskb;
1189         }
1190         return skb;
1191 }
1192 
1193 /*
1194  *      Copy shared buffers into a new sk_buff. We effectively do COW on
1195  *      packets to handle cases where we have a local reader and forward
1196  *      and a couple of other messy ones. The normal one is tcpdumping
1197  *      a packet thats being forwarded.
1198  */
1199 
1200 /**
1201  *      skb_unshare - make a copy of a shared buffer
1202  *      @skb: buffer to check
1203  *      @pri: priority for memory allocation
1204  *
1205  *      If the socket buffer is a clone then this function creates a new
1206  *      copy of the data, drops a reference count on the old copy and returns
1207  *      the new copy with the reference count at 1. If the buffer is not a clone
1208  *      the original buffer is returned. When called with a spinlock held or
1209  *      from interrupt state @pri must be %GFP_ATOMIC
1210  *
1211  *      %NULL is returned on a memory allocation failure.
1212  */
1213 static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
1214                                           gfp_t pri)
1215 {
1216         might_sleep_if(pri & __GFP_WAIT);
1217         if (skb_cloned(skb)) {
1218                 struct sk_buff *nskb = skb_copy(skb, pri);
1219 
1220                 /* Free our shared copy */
1221                 if (likely(nskb))
1222                         consume_skb(skb);
1223                 else
1224                         kfree_skb(skb);
1225                 skb = nskb;
1226         }
1227         return skb;
1228 }
1229 
1230 /**
1231  *      skb_peek - peek at the head of an &sk_buff_head
1232  *      @list_: list to peek at
1233  *
1234  *      Peek an &sk_buff. Unlike most other operations you _MUST_
1235  *      be careful with this one. A peek leaves the buffer on the
1236  *      list and someone else may run off with it. You must hold
1237  *      the appropriate locks or have a private queue to do this.
1238  *
1239  *      Returns %NULL for an empty list or a pointer to the head element.
1240  *      The reference count is not incremented and the reference is therefore
1241  *      volatile. Use with caution.
1242  */
1243 static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1244 {
1245         struct sk_buff *skb = list_->next;
1246 
1247         if (skb == (struct sk_buff *)list_)
1248                 skb = NULL;
1249         return skb;
1250 }
1251 
1252 /**
1253  *      skb_peek_next - peek skb following the given one from a queue
1254  *      @skb: skb to start from
1255  *      @list_: list to peek at
1256  *
1257  *      Returns %NULL when the end of the list is met or a pointer to the
1258  *      next element. The reference count is not incremented and the
1259  *      reference is therefore volatile. Use with caution.
1260  */
1261 static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1262                 const struct sk_buff_head *list_)
1263 {
1264         struct sk_buff *next = skb->next;
1265 
1266         if (next == (struct sk_buff *)list_)
1267                 next = NULL;
1268         return next;
1269 }
1270 
1271 /**
1272  *      skb_peek_tail - peek at the tail of an &sk_buff_head
1273  *      @list_: list to peek at
1274  *
1275  *      Peek an &sk_buff. Unlike most other operations you _MUST_
1276  *      be careful with this one. A peek leaves the buffer on the
1277  *      list and someone else may run off with it. You must hold
1278  *      the appropriate locks or have a private queue to do this.
1279  *
1280  *      Returns %NULL for an empty list or a pointer to the tail element.
1281  *      The reference count is not incremented and the reference is therefore
1282  *      volatile. Use with caution.
1283  */
1284 static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1285 {
1286         struct sk_buff *skb = list_->prev;
1287 
1288         if (skb == (struct sk_buff *)list_)
1289                 skb = NULL;
1290         return skb;
1291 
1292 }
1293 
1294 /**
1295  *      skb_queue_len   - get queue length
1296  *      @list_: list to measure
1297  *
1298  *      Return the length of an &sk_buff queue.
1299  */
1300 static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1301 {
1302         return list_->qlen;
1303 }
1304 
1305 /**
1306  *      __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1307  *      @list: queue to initialize
1308  *
1309  *      This initializes only the list and queue length aspects of
1310  *      an sk_buff_head object.  This allows to initialize the list
1311  *      aspects of an sk_buff_head without reinitializing things like
1312  *      the spinlock.  It can also be used for on-stack sk_buff_head
1313  *      objects where the spinlock is known to not be used.
1314  */
1315 static inline void __skb_queue_head_init(struct sk_buff_head *list)
1316 {
1317         list->prev = list->next = (struct sk_buff *)list;
1318         list->qlen = 0;
1319 }
1320 
1321 /*
1322  * This function creates a split out lock class for each invocation;
1323  * this is needed for now since a whole lot of users of the skb-queue
1324  * infrastructure in drivers have different locking usage (in hardirq)
1325  * than the networking core (in softirq only). In the long run either the
1326  * network layer or drivers should need annotation to consolidate the
1327  * main types of usage into 3 classes.
1328  */
1329 static inline void skb_queue_head_init(struct sk_buff_head *list)
1330 {
1331         spin_lock_init(&list->lock);
1332         __skb_queue_head_init(list);
1333 }
1334 
1335 static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1336                 struct lock_class_key *class)
1337 {
1338         skb_queue_head_init(list);
1339         lockdep_set_class(&list->lock, class);
1340 }
1341 
1342 /*
1343  *      Insert an sk_buff on a list.
1344  *
1345  *      The "__skb_xxxx()" functions are the non-atomic ones that
1346  *      can only be called with interrupts disabled.
1347  */
1348 void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1349                 struct sk_buff_head *list);
1350 static inline void __skb_insert(struct sk_buff *newsk,
1351                                 struct sk_buff *prev, struct sk_buff *next,
1352                                 struct sk_buff_head *list)
1353 {
1354         newsk->next = next;
1355         newsk->prev = prev;
1356         next->prev  = prev->next = newsk;
1357         list->qlen++;
1358 }
1359 
1360 static inline void __skb_queue_splice(const struct sk_buff_head *list,
1361                                       struct sk_buff *prev,
1362                                       struct sk_buff *next)
1363 {
1364         struct sk_buff *first = list->next;
1365         struct sk_buff *last = list->prev;
1366 
1367         first->prev = prev;
1368         prev->next = first;
1369 
1370         last->next = next;
1371         next->prev = last;
1372 }
1373 
1374 /**
1375  *      skb_queue_splice - join two skb lists, this is designed for stacks
1376  *      @list: the new list to add
1377  *      @head: the place to add it in the first list
1378  */
1379 static inline void skb_queue_splice(const struct sk_buff_head *list,
1380                                     struct sk_buff_head *head)
1381 {
1382         if (!skb_queue_empty(list)) {
1383                 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1384                 head->qlen += list->qlen;
1385         }
1386 }
1387 
1388 /**
1389  *      skb_queue_splice_init - join two skb lists and reinitialise the emptied list
1390  *      @list: the new list to add
1391  *      @head: the place to add it in the first list
1392  *
1393  *      The list at @list is reinitialised
1394  */
1395 static inline void skb_queue_splice_init(struct sk_buff_head *list,
1396                                          struct sk_buff_head *head)
1397 {
1398         if (!skb_queue_empty(list)) {
1399                 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1400                 head->qlen += list->qlen;
1401                 __skb_queue_head_init(list);
1402         }
1403 }
1404 
1405 /**
1406  *      skb_queue_splice_tail - join two skb lists, each list being a queue
1407  *      @list: the new list to add
1408  *      @head: the place to add it in the first list
1409  */
1410 static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1411                                          struct sk_buff_head *head)
1412 {
1413         if (!skb_queue_empty(list)) {
1414                 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1415                 head->qlen += list->qlen;
1416         }
1417 }
1418 
1419 /**
1420  *      skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
1421  *      @list: the new list to add
1422  *      @head: the place to add it in the first list
1423  *
1424  *      Each of the lists is a queue.
1425  *      The list at @list is reinitialised
1426  */
1427 static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1428                                               struct sk_buff_head *head)
1429 {
1430         if (!skb_queue_empty(list)) {
1431                 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1432                 head->qlen += list->qlen;
1433                 __skb_queue_head_init(list);
1434         }
1435 }
1436 
1437 /**
1438  *      __skb_queue_after - queue a buffer at the list head
1439  *      @list: list to use
1440  *      @prev: place after this buffer
1441  *      @newsk: buffer to queue
1442  *
1443  *      Queue a buffer int the middle of a list. This function takes no locks
1444  *      and you must therefore hold required locks before calling it.
1445  *
1446  *      A buffer cannot be placed on two lists at the same time.
1447  */
1448 static inline void __skb_queue_after(struct sk_buff_head *list,
1449                                      struct sk_buff *prev,
1450                                      struct sk_buff *newsk)
1451 {
1452         __skb_insert(newsk, prev, prev->next, list);
1453 }
1454 
1455 void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1456                 struct sk_buff_head *list);
1457 
1458 static inline void __skb_queue_before(struct sk_buff_head *list,
1459                                       struct sk_buff *next,
1460                                       struct sk_buff *newsk)
1461 {
1462         __skb_insert(newsk, next->prev, next, list);
1463 }
1464 
1465 /**
1466  *      __skb_queue_head - queue a buffer at the list head
1467  *      @list: list to use
1468  *      @newsk: buffer to queue
1469  *
1470  *      Queue a buffer at the start of a list. This function takes no locks
1471  *      and you must therefore hold required locks before calling it.
1472  *
1473  *      A buffer cannot be placed on two lists at the same time.
1474  */
1475 void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1476 static inline void __skb_queue_head(struct sk_buff_head *list,
1477                                     struct sk_buff *newsk)
1478 {
1479         __skb_queue_after(list, (struct sk_buff *)list, newsk);
1480 }
1481 
1482 /**
1483  *      __skb_queue_tail - queue a buffer at the list tail
1484  *      @list: list to use
1485  *      @newsk: buffer to queue
1486  *
1487  *      Queue a buffer at the end of a list. This function takes no locks
1488  *      and you must therefore hold required locks before calling it.
1489  *
1490  *      A buffer cannot be placed on two lists at the same time.
1491  */
1492 void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1493 static inline void __skb_queue_tail(struct sk_buff_head *list,
1494                                    struct sk_buff *newsk)
1495 {
1496         __skb_queue_before(list, (struct sk_buff *)list, newsk);
1497 }
1498 
1499 /*
1500  * remove sk_buff from list. _Must_ be called atomically, and with
1501  * the list known..
1502  */
1503 void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1504 static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1505 {
1506         struct sk_buff *next, *prev;
1507 
1508         list->qlen--;
1509         next       = skb->next;
1510         prev       = skb->prev;
1511         skb->next  = skb->prev = NULL;
1512         next->prev = prev;
1513         prev->next = next;
1514 }
1515 
1516 /**
1517  *      __skb_dequeue - remove from the head of the queue
1518  *      @list: list to dequeue from
1519  *
1520  *      Remove the head of the list. This function does not take any locks
1521  *      so must be used with appropriate locks held only. The head item is
1522  *      returned or %NULL if the list is empty.
1523  */
1524 struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1525 static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1526 {
1527         struct sk_buff *skb = skb_peek(list);
1528         if (skb)
1529                 __skb_unlink(skb, list);
1530         return skb;
1531 }
1532 
1533 /**
1534  *      __skb_dequeue_tail - remove from the tail of the queue
1535  *      @list: list to dequeue from
1536  *
1537  *      Remove the tail of the list. This function does not take any locks
1538  *      so must be used with appropriate locks held only. The tail item is
1539  *      returned or %NULL if the list is empty.
1540  */
1541 struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1542 static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1543 {
1544         struct sk_buff *skb = skb_peek_tail(list);
1545         if (skb)
1546                 __skb_unlink(skb, list);
1547         return skb;
1548 }
1549 
1550 
1551 static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1552 {
1553         return skb->data_len;
1554 }
1555 
1556 static inline unsigned int skb_headlen(const struct sk_buff *skb)
1557 {
1558         return skb->len - skb->data_len;
1559 }
1560 
1561 static inline int skb_pagelen(const struct sk_buff *skb)
1562 {
1563         int i, len = 0;
1564 
1565         for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1566                 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1567         return len + skb_headlen(skb);
1568 }
1569 
1570 /**
1571  * __skb_fill_page_desc - initialise a paged fragment in an skb
1572  * @skb: buffer containing fragment to be initialised
1573  * @i: paged fragment index to initialise
1574  * @page: the page to use for this fragment
1575  * @off: the offset to the data with @page
1576  * @size: the length of the data
1577  *
1578  * Initialises the @i'th fragment of @skb to point to &size bytes at
1579  * offset @off within @page.
1580  *
1581  * Does not take any additional reference on the fragment.
1582  */
1583 static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1584                                         struct page *page, int off, int size)
1585 {
1586         skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1587 
1588         /*
1589          * Propagate page->pfmemalloc to the skb if we can. The problem is
1590          * that not all callers have unique ownership of the page. If
1591          * pfmemalloc is set, we check the mapping as a mapping implies
1592          * page->index is set (index and pfmemalloc share space).
1593          * If it's a valid mapping, we cannot use page->pfmemalloc but we
1594          * do not lose pfmemalloc information as the pages would not be
1595          * allocated using __GFP_MEMALLOC.
1596          */
1597         frag->page.p              = page;
1598         frag->page_offset         = off;
1599         skb_frag_size_set(frag, size);
1600 
1601         page = compound_head(page);
1602         if (page->pfmemalloc && !page->mapping)
1603                 skb->pfmemalloc = true;
1604 }
1605 
1606 /**
1607  * skb_fill_page_desc - initialise a paged fragment in an skb
1608  * @skb: buffer containing fragment to be initialised
1609  * @i: paged fragment index to initialise
1610  * @page: the page to use for this fragment
1611  * @off: the offset to the data with @page
1612  * @size: the length of the data
1613  *
1614  * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1615  * @skb to point to @size bytes at offset @off within @page. In
1616  * addition updates @skb such that @i is the last fragment.
1617  *
1618  * Does not take any additional reference on the fragment.
1619  */
1620 static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1621                                       struct page *page, int off, int size)
1622 {
1623         __skb_fill_page_desc(skb, i, page, off, size);
1624         skb_shinfo(skb)->nr_frags = i + 1;
1625 }
1626 
1627 void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1628                      int size, unsigned int truesize);
1629 
1630 void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1631                           unsigned int truesize);
1632 
1633 #define SKB_PAGE_ASSERT(skb)    BUG_ON(skb_shinfo(skb)->nr_frags)
1634 #define SKB_FRAG_ASSERT(skb)    BUG_ON(skb_has_frag_list(skb))
1635 #define SKB_LINEAR_ASSERT(skb)  BUG_ON(skb_is_nonlinear(skb))
1636 
1637 #ifdef NET_SKBUFF_DATA_USES_OFFSET
1638 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1639 {
1640         return skb->head + skb->tail;
1641 }
1642 
1643 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1644 {
1645         skb->tail = skb->data - skb->head;
1646 }
1647 
1648 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1649 {
1650         skb_reset_tail_pointer(skb);
1651         skb->tail += offset;
1652 }
1653 
1654 #else /* NET_SKBUFF_DATA_USES_OFFSET */
1655 static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1656 {
1657         return skb->tail;
1658 }
1659 
1660 static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1661 {
1662         skb->tail = skb->data;
1663 }
1664 
1665 static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1666 {
1667         skb->tail = skb->data + offset;
1668 }
1669 
1670 #endif /* NET_SKBUFF_DATA_USES_OFFSET */
1671 
1672 /*
1673  *      Add data to an sk_buff
1674  */
1675 unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1676 unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1677 static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1678 {
1679         unsigned char *tmp = skb_tail_pointer(skb);
1680         SKB_LINEAR_ASSERT(skb);
1681         skb->tail += len;
1682         skb->len  += len;
1683         return tmp;
1684 }
1685 
1686 unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1687 static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1688 {
1689         skb->data -= len;
1690         skb->len  += len;
1691         return skb->data;
1692 }
1693 
1694 unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1695 static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1696 {
1697         skb->len -= len;
1698         BUG_ON(skb->len < skb->data_len);
1699         return skb->data += len;
1700 }
1701 
1702 static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1703 {
1704         return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1705 }
1706 
1707 unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1708 
1709 static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1710 {
1711         if (len > skb_headlen(skb) &&
1712             !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1713                 return NULL;
1714         skb->len -= len;
1715         return skb->data += len;
1716 }
1717 
1718 static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1719 {
1720         return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1721 }
1722 
1723 static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1724 {
1725         if (likely(len <= skb_headlen(skb)))
1726                 return 1;
1727         if (unlikely(len > skb->len))
1728                 return 0;
1729         return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1730 }
1731 
1732 /**
1733  *      skb_headroom - bytes at buffer head
1734  *      @skb: buffer to check
1735  *
1736  *      Return the number of bytes of free space at the head of an &sk_buff.
1737  */
1738 static inline unsigned int skb_headroom(const struct sk_buff *skb)
1739 {
1740         return skb->data - skb->head;
1741 }
1742 
1743 /**
1744  *      skb_tailroom - bytes at buffer end
1745  *      @skb: buffer to check
1746  *
1747  *      Return the number of bytes of free space at the tail of an sk_buff
1748  */
1749 static inline int skb_tailroom(const struct sk_buff *skb)
1750 {
1751         return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1752 }
1753 
1754 /**
1755  *      skb_availroom - bytes at buffer end
1756  *      @skb: buffer to check
1757  *
1758  *      Return the number of bytes of free space at the tail of an sk_buff
1759  *      allocated by sk_stream_alloc()
1760  */
1761 static inline int skb_availroom(const struct sk_buff *skb)
1762 {
1763         if (skb_is_nonlinear(skb))
1764                 return 0;
1765 
1766         return skb->end - skb->tail - skb->reserved_tailroom;
1767 }
1768 
1769 /**
1770  *      skb_reserve - adjust headroom
1771  *      @skb: buffer to alter
1772  *      @len: bytes to move
1773  *
1774  *      Increase the headroom of an empty &sk_buff by reducing the tail
1775  *      room. This is only allowed for an empty buffer.
1776  */
1777 static inline void skb_reserve(struct sk_buff *skb, int len)
1778 {
1779         skb->data += len;
1780         skb->tail += len;
1781 }
1782 
1783 #define ENCAP_TYPE_ETHER        0
1784 #define ENCAP_TYPE_IPPROTO      1
1785 
1786 static inline void skb_set_inner_protocol(struct sk_buff *skb,
1787                                           __be16 protocol)
1788 {
1789         skb->inner_protocol = protocol;
1790         skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1791 }
1792 
1793 static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1794                                          __u8 ipproto)
1795 {
1796         skb->inner_ipproto = ipproto;
1797         skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1798 }
1799 
1800 static inline void skb_reset_inner_headers(struct sk_buff *skb)
1801 {
1802         skb->inner_mac_header = skb->mac_header;
1803         skb->inner_network_header = skb->network_header;
1804         skb->inner_transport_header = skb->transport_header;
1805 }
1806 
1807 static inline void skb_reset_mac_len(struct sk_buff *skb)
1808 {
1809         skb->mac_len = skb->network_header - skb->mac_header;
1810 }
1811 
1812 static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1813                                                         *skb)
1814 {
1815         return skb->head + skb->inner_transport_header;
1816 }
1817 
1818 static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1819 {
1820         skb->inner_transport_header = skb->data - skb->head;
1821 }
1822 
1823 static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1824                                                    const int offset)
1825 {
1826         skb_reset_inner_transport_header(skb);
1827         skb->inner_transport_header += offset;
1828 }
1829 
1830 static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1831 {
1832         return skb->head + skb->inner_network_header;
1833 }
1834 
1835 static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1836 {
1837         skb->inner_network_header = skb->data - skb->head;
1838 }
1839 
1840 static inline void skb_set_inner_network_header(struct sk_buff *skb,
1841                                                 const int offset)
1842 {
1843         skb_reset_inner_network_header(skb);
1844         skb->inner_network_header += offset;
1845 }
1846 
1847 static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1848 {
1849         return skb->head + skb->inner_mac_header;
1850 }
1851 
1852 static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1853 {
1854         skb->inner_mac_header = skb->data - skb->head;
1855 }
1856 
1857 static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1858                                             const int offset)
1859 {
1860         skb_reset_inner_mac_header(skb);
1861         skb->inner_mac_header += offset;
1862 }
1863 static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1864 {
1865         return skb->transport_header != (typeof(skb->transport_header))~0U;
1866 }
1867 
1868 static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1869 {
1870         return skb->head + skb->transport_header;
1871 }
1872 
1873 static inline void skb_reset_transport_header(struct sk_buff *skb)
1874 {
1875         skb->transport_header = skb->data - skb->head;
1876 }
1877 
1878 static inline void skb_set_transport_header(struct sk_buff *skb,
1879                                             const int offset)
1880 {
1881         skb_reset_transport_header(skb);
1882         skb->transport_header += offset;
1883 }
1884 
1885 static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1886 {
1887         return skb->head + skb->network_header;
1888 }
1889 
1890 static inline void skb_reset_network_header(struct sk_buff *skb)
1891 {
1892         skb->network_header = skb->data - skb->head;
1893 }
1894 
1895 static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1896 {
1897         skb_reset_network_header(skb);
1898         skb->network_header += offset;
1899 }
1900 
1901 static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1902 {
1903         return skb->head + skb->mac_header;
1904 }
1905 
1906 static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1907 {
1908         return skb->mac_header != (typeof(skb->mac_header))~0U;
1909 }
1910 
1911 static inline void skb_reset_mac_header(struct sk_buff *skb)
1912 {
1913         skb->mac_header = skb->data - skb->head;
1914 }
1915 
1916 static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1917 {
1918         skb_reset_mac_header(skb);
1919         skb->mac_header += offset;
1920 }
1921 
1922 static inline void skb_pop_mac_header(struct sk_buff *skb)
1923 {
1924         skb->mac_header = skb->network_header;
1925 }
1926 
1927 static inline void skb_probe_transport_header(struct sk_buff *skb,
1928                                               const int offset_hint)
1929 {
1930         struct flow_keys keys;
1931 
1932         if (skb_transport_header_was_set(skb))
1933                 return;
1934         else if (skb_flow_dissect(skb, &keys))
1935                 skb_set_transport_header(skb, keys.thoff);
1936         else
1937                 skb_set_transport_header(skb, offset_hint);
1938 }
1939 
1940 static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1941 {
1942         if (skb_mac_header_was_set(skb)) {
1943                 const unsigned char *old_mac = skb_mac_header(skb);
1944 
1945                 skb_set_mac_header(skb, -skb->mac_len);
1946                 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1947         }
1948 }
1949 
1950 static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1951 {
1952         return skb->csum_start - skb_headroom(skb);
1953 }
1954 
1955 static inline int skb_transport_offset(const struct sk_buff *skb)
1956 {
1957         return skb_transport_header(skb) - skb->data;
1958 }
1959 
1960 static inline u32 skb_network_header_len(const struct sk_buff *skb)
1961 {
1962         return skb->transport_header - skb->network_header;
1963 }
1964 
1965 static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1966 {
1967         return skb->inner_transport_header - skb->inner_network_header;
1968 }
1969 
1970 static inline int skb_network_offset(const struct sk_buff *skb)
1971 {
1972         return skb_network_header(skb) - skb->data;
1973 }
1974 
1975 static inline int skb_inner_network_offset(const struct sk_buff *skb)
1976 {
1977         return skb_inner_network_header(skb) - skb->data;
1978 }
1979 
1980 static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1981 {
1982         return pskb_may_pull(skb, skb_network_offset(skb) + len);
1983 }
1984 
1985 /*
1986  * CPUs often take a performance hit when accessing unaligned memory
1987  * locations. The actual performance hit varies, it can be small if the
1988  * hardware handles it or large if we have to take an exception and fix it
1989  * in software.
1990  *
1991  * Since an ethernet header is 14 bytes network drivers often end up with
1992  * the IP header at an unaligned offset. The IP header can be aligned by
1993  * shifting the start of the packet by 2 bytes. Drivers should do this
1994  * with:
1995  *
1996  * skb_reserve(skb, NET_IP_ALIGN);
1997  *
1998  * The downside to this alignment of the IP header is that the DMA is now
1999  * unaligned. On some architectures the cost of an unaligned DMA is high
2000  * and this cost outweighs the gains made by aligning the IP header.
2001  *
2002  * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2003  * to be overridden.
2004  */
2005 #ifndef NET_IP_ALIGN
2006 #define NET_IP_ALIGN    2
2007 #endif
2008 
2009 /*
2010  * The networking layer reserves some headroom in skb data (via
2011  * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2012  * the header has to grow. In the default case, if the header has to grow
2013  * 32 bytes or less we avoid the reallocation.
2014  *
2015  * Unfortunately this headroom changes the DMA alignment of the resulting
2016  * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2017  * on some architectures. An architecture can override this value,
2018  * perhaps setting it to a cacheline in size (since that will maintain
2019  * cacheline alignment of the DMA). It must be a power of 2.
2020  *
2021  * Various parts of the networking layer expect at least 32 bytes of
2022  * headroom, you should not reduce this.
2023  *
2024  * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2025  * to reduce average number of cache lines per packet.
2026  * get_rps_cpus() for example only access one 64 bytes aligned block :
2027  * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
2028  */
2029 #ifndef NET_SKB_PAD
2030 #define NET_SKB_PAD     max(32, L1_CACHE_BYTES)
2031 #endif
2032 
2033 int ___pskb_trim(struct sk_buff *skb, unsigned int len);
2034 
2035 static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2036 {
2037         if (unlikely(skb_is_nonlinear(skb))) {
2038                 WARN_ON(1);
2039                 return;
2040         }
2041         skb->len = len;
2042         skb_set_tail_pointer(skb, len);
2043 }
2044 
2045 void skb_trim(struct sk_buff *skb, unsigned int len);
2046 
2047 static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2048 {
2049         if (skb->data_len)
2050                 return ___pskb_trim(skb, len);
2051         __skb_trim(skb, len);
2052         return 0;
2053 }
2054 
2055 static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2056 {
2057         return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2058 }
2059 
2060 /**
2061  *      pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2062  *      @skb: buffer to alter
2063  *      @len: new length
2064  *
2065  *      This is identical to pskb_trim except that the caller knows that
2066  *      the skb is not cloned so we should never get an error due to out-
2067  *      of-memory.
2068  */
2069 static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2070 {
2071         int err = pskb_trim(skb, len);
2072         BUG_ON(err);
2073 }
2074 
2075 /**
2076  *      skb_orphan - orphan a buffer
2077  *      @skb: buffer to orphan
2078  *
2079  *      If a buffer currently has an owner then we call the owner's
2080  *      destructor function and make the @skb unowned. The buffer continues
2081  *      to exist but is no longer charged to its former owner.
2082  */
2083 static inline void skb_orphan(struct sk_buff *skb)
2084 {
2085         if (skb->destructor) {
2086                 skb->destructor(skb);
2087                 skb->destructor = NULL;
2088                 skb->sk         = NULL;
2089         } else {
2090                 BUG_ON(skb->sk);
2091         }
2092 }
2093 
2094 /**
2095  *      skb_orphan_frags - orphan the frags contained in a buffer
2096  *      @skb: buffer to orphan frags from
2097  *      @gfp_mask: allocation mask for replacement pages
2098  *
2099  *      For each frag in the SKB which needs a destructor (i.e. has an
2100  *      owner) create a copy of that frag and release the original
2101  *      page by calling the destructor.
2102  */
2103 static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2104 {
2105         if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2106                 return 0;
2107         return skb_copy_ubufs(skb, gfp_mask);
2108 }
2109 
2110 /**
2111  *      __skb_queue_purge - empty a list
2112  *      @list: list to empty
2113  *
2114  *      Delete all buffers on an &sk_buff list. Each buffer is removed from
2115  *      the list and one reference dropped. This function does not take the
2116  *      list lock and the caller must hold the relevant locks to use it.
2117  */
2118 void skb_queue_purge(struct sk_buff_head *list);
2119 static inline void __skb_queue_purge(struct sk_buff_head *list)
2120 {
2121         struct sk_buff *skb;
2122         while ((skb = __skb_dequeue(list)) != NULL)
2123                 kfree_skb(skb);
2124 }
2125 
2126 #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2127 #define NETDEV_FRAG_PAGE_MAX_SIZE  (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2128 #define NETDEV_PAGECNT_MAX_BIAS    NETDEV_FRAG_PAGE_MAX_SIZE
2129 
2130 void *netdev_alloc_frag(unsigned int fragsz);
2131 
2132 struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2133                                    gfp_t gfp_mask);
2134 
2135 /**
2136  *      netdev_alloc_skb - allocate an skbuff for rx on a specific device
2137  *      @dev: network device to receive on
2138  *      @length: length to allocate
2139  *
2140  *      Allocate a new &sk_buff and assign it a usage count of one. The
2141  *      buffer has unspecified headroom built in. Users should allocate
2142  *      the headroom they think they need without accounting for the
2143  *      built in space. The built in space is used for optimisations.
2144  *
2145  *      %NULL is returned if there is no free memory. Although this function
2146  *      allocates memory it can be called from an interrupt.
2147  */
2148 static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
2149                                                unsigned int length)
2150 {
2151         return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2152 }
2153 
2154 /* legacy helper around __netdev_alloc_skb() */
2155 static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2156                                               gfp_t gfp_mask)
2157 {
2158         return __netdev_alloc_skb(NULL, length, gfp_mask);
2159 }
2160 
2161 /* legacy helper around netdev_alloc_skb() */
2162 static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2163 {
2164         return netdev_alloc_skb(NULL, length);
2165 }
2166 
2167 
2168 static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2169                 unsigned int length, gfp_t gfp)
2170 {
2171         struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
2172 
2173         if (NET_IP_ALIGN && skb)
2174                 skb_reserve(skb, NET_IP_ALIGN);
2175         return skb;
2176 }
2177 
2178 static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2179                 unsigned int length)
2180 {
2181         return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2182 }
2183 
2184 void *napi_alloc_frag(unsigned int fragsz);
2185 struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2186                                  unsigned int length, gfp_t gfp_mask);
2187 static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2188                                              unsigned int length)
2189 {
2190         return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2191 }
2192 
2193 /**
2194  * __dev_alloc_pages - allocate page for network Rx
2195  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2196  * @order: size of the allocation
2197  *
2198  * Allocate a new page.
2199  *
2200  * %NULL is returned if there is no free memory.
2201 */
2202 static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2203                                              unsigned int order)
2204 {
2205         /* This piece of code contains several assumptions.
2206          * 1.  This is for device Rx, therefor a cold page is preferred.
2207          * 2.  The expectation is the user wants a compound page.
2208          * 3.  If requesting a order 0 page it will not be compound
2209          *     due to the check to see if order has a value in prep_new_page
2210          * 4.  __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2211          *     code in gfp_to_alloc_flags that should be enforcing this.
2212          */
2213         gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2214 
2215         return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2216 }
2217 
2218 static inline struct page *dev_alloc_pages(unsigned int order)
2219 {
2220         return __dev_alloc_pages(GFP_ATOMIC, order);
2221 }
2222 
2223 /**
2224  * __dev_alloc_page - allocate a page for network Rx
2225  * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2226  *
2227  * Allocate a new page.
2228  *
2229  * %NULL is returned if there is no free memory.
2230  */
2231 static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2232 {
2233         return __dev_alloc_pages(gfp_mask, 0);
2234 }
2235 
2236 static inline struct page *dev_alloc_page(void)
2237 {
2238         return __dev_alloc_page(GFP_ATOMIC);
2239 }
2240 
2241 /**
2242  *      skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2243  *      @page: The page that was allocated from skb_alloc_page
2244  *      @skb: The skb that may need pfmemalloc set
2245  */
2246 static inline void skb_propagate_pfmemalloc(struct page *page,
2247                                              struct sk_buff *skb)
2248 {
2249         if (page && page->pfmemalloc)
2250                 skb->pfmemalloc = true;
2251 }
2252 
2253 /**
2254  * skb_frag_page - retrieve the page referred to by a paged fragment
2255  * @frag: the paged fragment
2256  *
2257  * Returns the &struct page associated with @frag.
2258  */
2259 static inline struct page *skb_frag_page(const skb_frag_t *frag)
2260 {
2261         return frag->page.p;
2262 }
2263 
2264 /**
2265  * __skb_frag_ref - take an addition reference on a paged fragment.
2266  * @frag: the paged fragment
2267  *
2268  * Takes an additional reference on the paged fragment @frag.
2269  */
2270 static inline void __skb_frag_ref(skb_frag_t *frag)
2271 {
2272         get_page(skb_frag_page(frag));
2273 }
2274 
2275 /**
2276  * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2277  * @skb: the buffer
2278  * @f: the fragment offset.
2279  *
2280  * Takes an additional reference on the @f'th paged fragment of @skb.
2281  */
2282 static inline void skb_frag_ref(struct sk_buff *skb, int f)
2283 {
2284         __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2285 }
2286 
2287 /**
2288  * __skb_frag_unref - release a reference on a paged fragment.
2289  * @frag: the paged fragment
2290  *
2291  * Releases a reference on the paged fragment @frag.
2292  */
2293 static inline void __skb_frag_unref(skb_frag_t *frag)
2294 {
2295         put_page(skb_frag_page(frag));
2296 }
2297 
2298 /**
2299  * skb_frag_unref - release a reference on a paged fragment of an skb.
2300  * @skb: the buffer
2301  * @f: the fragment offset
2302  *
2303  * Releases a reference on the @f'th paged fragment of @skb.
2304  */
2305 static inline void skb_frag_unref(struct sk_buff *skb, int f)
2306 {
2307         __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2308 }
2309 
2310 /**
2311  * skb_frag_address - gets the address of the data contained in a paged fragment
2312  * @frag: the paged fragment buffer
2313  *
2314  * Returns the address of the data within @frag. The page must already
2315  * be mapped.
2316  */
2317 static inline void *skb_frag_address(const skb_frag_t *frag)
2318 {
2319         return page_address(skb_frag_page(frag)) + frag->page_offset;
2320 }
2321 
2322 /**
2323  * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2324  * @frag: the paged fragment buffer
2325  *
2326  * Returns the address of the data within @frag. Checks that the page
2327  * is mapped and returns %NULL otherwise.
2328  */
2329 static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2330 {
2331         void *ptr = page_address(skb_frag_page(frag));
2332         if (unlikely(!ptr))
2333                 return NULL;
2334 
2335         return ptr + frag->page_offset;
2336 }
2337 
2338 /**
2339  * __skb_frag_set_page - sets the page contained in a paged fragment
2340  * @frag: the paged fragment
2341  * @page: the page to set
2342  *
2343  * Sets the fragment @frag to contain @page.
2344  */
2345 static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2346 {
2347         frag->page.p = page;
2348 }
2349 
2350 /**
2351  * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2352  * @skb: the buffer
2353  * @f: the fragment offset
2354  * @page: the page to set
2355  *
2356  * Sets the @f'th fragment of @skb to contain @page.
2357  */
2358 static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2359                                      struct page *page)
2360 {
2361         __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2362 }
2363 
2364 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2365 
2366 /**
2367  * skb_frag_dma_map - maps a paged fragment via the DMA API
2368  * @dev: the device to map the fragment to
2369  * @frag: the paged fragment to map
2370  * @offset: the offset within the fragment (starting at the
2371  *          fragment's own offset)
2372  * @size: the number of bytes to map
2373  * @dir: the direction of the mapping (%PCI_DMA_*)
2374  *
2375  * Maps the page associated with @frag to @device.
2376  */
2377 static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2378                                           const skb_frag_t *frag,
2379                                           size_t offset, size_t size,
2380                                           enum dma_data_direction dir)
2381 {
2382         return dma_map_page(dev, skb_frag_page(frag),
2383                             frag->page_offset + offset, size, dir);
2384 }
2385 
2386 static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2387                                         gfp_t gfp_mask)
2388 {
2389         return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2390 }
2391 
2392 
2393 static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2394                                                   gfp_t gfp_mask)
2395 {
2396         return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2397 }
2398 
2399 
2400 /**
2401  *      skb_clone_writable - is the header of a clone writable
2402  *      @skb: buffer to check
2403  *      @len: length up to which to write
2404  *
2405  *      Returns true if modifying the header part of the cloned buffer
2406  *      does not requires the data to be copied.
2407  */
2408 static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
2409 {
2410         return !skb_header_cloned(skb) &&
2411                skb_headroom(skb) + len <= skb->hdr_len;
2412 }
2413 
2414 static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2415                             int cloned)
2416 {
2417         int delta = 0;
2418 
2419         if (headroom > skb_headroom(skb))
2420                 delta = headroom - skb_headroom(skb);
2421 
2422         if (delta || cloned)
2423                 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2424                                         GFP_ATOMIC);
2425         return 0;
2426 }
2427 
2428 /**
2429  *      skb_cow - copy header of skb when it is required
2430  *      @skb: buffer to cow
2431  *      @headroom: needed headroom
2432  *
2433  *      If the skb passed lacks sufficient headroom or its data part
2434  *      is shared, data is reallocated. If reallocation fails, an error
2435  *      is returned and original skb is not changed.
2436  *
2437  *      The result is skb with writable area skb->head...skb->tail
2438  *      and at least @headroom of space at head.
2439  */
2440 static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2441 {
2442         return __skb_cow(skb, headroom, skb_cloned(skb));
2443 }
2444 
2445 /**
2446  *      skb_cow_head - skb_cow but only making the head writable
2447  *      @skb: buffer to cow
2448  *      @headroom: needed headroom
2449  *
2450  *      This function is identical to skb_cow except that we replace the
2451  *      skb_cloned check by skb_header_cloned.  It should be used when
2452  *      you only need to push on some header and do not need to modify
2453  *      the data.
2454  */
2455 static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2456 {
2457         return __skb_cow(skb, headroom, skb_header_cloned(skb));
2458 }
2459 
2460 /**
2461  *      skb_padto       - pad an skbuff up to a minimal size
2462  *      @skb: buffer to pad
2463  *      @len: minimal length
2464  *
2465  *      Pads up a buffer to ensure the trailing bytes exist and are
2466  *      blanked. If the buffer already contains sufficient data it
2467  *      is untouched. Otherwise it is extended. Returns zero on
2468  *      success. The skb is freed on error.
2469  */
2470 static inline int skb_padto(struct sk_buff *skb, unsigned int len)
2471 {
2472         unsigned int size = skb->len;
2473         if (likely(size >= len))
2474                 return 0;
2475         return skb_pad(skb, len - size);
2476 }
2477 
2478 /**
2479  *      skb_put_padto - increase size and pad an skbuff up to a minimal size
2480  *      @skb: buffer to pad
2481  *      @len: minimal length
2482  *
2483  *      Pads up a buffer to ensure the trailing bytes exist and are
2484  *      blanked. If the buffer already contains sufficient data it
2485  *      is untouched. Otherwise it is extended. Returns zero on
2486  *      success. The skb is freed on error.
2487  */
2488 static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2489 {
2490         unsigned int size = skb->len;
2491 
2492         if (unlikely(size < len)) {
2493                 len -= size;
2494                 if (skb_pad(skb, len))
2495                         return -ENOMEM;
2496                 __skb_put(skb, len);
2497         }
2498         return 0;
2499 }
2500 
2501 static inline int skb_add_data(struct sk_buff *skb,
2502                                struct iov_iter *from, int copy)
2503 {
2504         const int off = skb->len;
2505 
2506         if (skb->ip_summed == CHECKSUM_NONE) {
2507                 __wsum csum = 0;
2508                 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2509                                             &csum, from) == copy) {
2510                         skb->csum = csum_block_add(skb->csum, csum, off);
2511                         return 0;
2512                 }
2513         } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
2514                 return 0;
2515 
2516         __skb_trim(skb, off);
2517         return -EFAULT;
2518 }
2519 
2520 static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2521                                     const struct page *page, int off)
2522 {
2523         if (i) {
2524                 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
2525 
2526                 return page == skb_frag_page(frag) &&
2527                        off == frag->page_offset + skb_frag_size(frag);
2528         }
2529         return false;
2530 }
2531 
2532 static inline int __skb_linearize(struct sk_buff *skb)
2533 {
2534         return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2535 }
2536 
2537 /**
2538  *      skb_linearize - convert paged skb to linear one
2539  *      @skb: buffer to linarize
2540  *
2541  *      If there is no free memory -ENOMEM is returned, otherwise zero
2542  *      is returned and the old skb data released.
2543  */
2544 static inline int skb_linearize(struct sk_buff *skb)
2545 {
2546         return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2547 }
2548 
2549 /**
2550  * skb_has_shared_frag - can any frag be overwritten
2551  * @skb: buffer to test
2552  *
2553  * Return true if the skb has at least one frag that might be modified
2554  * by an external entity (as in vmsplice()/sendfile())
2555  */
2556 static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2557 {
2558         return skb_is_nonlinear(skb) &&
2559                skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
2560 }
2561 
2562 /**
2563  *      skb_linearize_cow - make sure skb is linear and writable
2564  *      @skb: buffer to process
2565  *
2566  *      If there is no free memory -ENOMEM is returned, otherwise zero
2567  *      is returned and the old skb data released.
2568  */
2569 static inline int skb_linearize_cow(struct sk_buff *skb)
2570 {
2571         return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2572                __skb_linearize(skb) : 0;
2573 }
2574 
2575 /**
2576  *      skb_postpull_rcsum - update checksum for received skb after pull
2577  *      @skb: buffer to update
2578  *      @start: start of data before pull
2579  *      @len: length of data pulled
2580  *
2581  *      After doing a pull on a received packet, you need to call this to
2582  *      update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2583  *      CHECKSUM_NONE so that it can be recomputed from scratch.
2584  */
2585 
2586 static inline void skb_postpull_rcsum(struct sk_buff *skb,
2587                                       const void *start, unsigned int len)
2588 {
2589         if (skb->ip_summed == CHECKSUM_COMPLETE)
2590                 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2591 }
2592 
2593 unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2594 
2595 /**
2596  *      pskb_trim_rcsum - trim received skb and update checksum
2597  *      @skb: buffer to trim
2598  *      @len: new length
2599  *
2600  *      This is exactly the same as pskb_trim except that it ensures the
2601  *      checksum of received packets are still valid after the operation.
2602  */
2603 
2604 static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2605 {
2606         if (likely(len >= skb->len))
2607                 return 0;
2608         if (skb->ip_summed == CHECKSUM_COMPLETE)
2609                 skb->ip_summed = CHECKSUM_NONE;
2610         return __pskb_trim(skb, len);
2611 }
2612 
2613 #define skb_queue_walk(queue, skb) \
2614                 for (skb = (queue)->next;                                       \
2615                      skb != (struct sk_buff *)(queue);                          \
2616                      skb = skb->next)
2617 
2618 #define skb_queue_walk_safe(queue, skb, tmp)                                    \
2619                 for (skb = (queue)->next, tmp = skb->next;                      \
2620                      skb != (struct sk_buff *)(queue);                          \
2621                      skb = tmp, tmp = skb->next)
2622 
2623 #define skb_queue_walk_from(queue, skb)                                         \
2624                 for (; skb != (struct sk_buff *)(queue);                        \
2625                      skb = skb->next)
2626 
2627 #define skb_queue_walk_from_safe(queue, skb, tmp)                               \
2628                 for (tmp = skb->next;                                           \
2629                      skb != (struct sk_buff *)(queue);                          \
2630                      skb = tmp, tmp = skb->next)
2631 
2632 #define skb_queue_reverse_walk(queue, skb) \
2633                 for (skb = (queue)->prev;                                       \
2634                      skb != (struct sk_buff *)(queue);                          \
2635                      skb = skb->prev)
2636 
2637 #define skb_queue_reverse_walk_safe(queue, skb, tmp)                            \
2638                 for (skb = (queue)->prev, tmp = skb->prev;                      \
2639                      skb != (struct sk_buff *)(queue);                          \
2640                      skb = tmp, tmp = skb->prev)
2641 
2642 #define skb_queue_reverse_walk_from_safe(queue, skb, tmp)                       \
2643                 for (tmp = skb->prev;                                           \
2644                      skb != (struct sk_buff *)(queue);                          \
2645                      skb = tmp, tmp = skb->prev)
2646 
2647 static inline bool skb_has_frag_list(const struct sk_buff *skb)
2648 {
2649         return skb_shinfo(skb)->frag_list != NULL;
2650 }
2651 
2652 static inline void skb_frag_list_init(struct sk_buff *skb)
2653 {
2654         skb_shinfo(skb)->frag_list = NULL;
2655 }
2656 
2657 static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2658 {
2659         frag->next = skb_shinfo(skb)->frag_list;
2660         skb_shinfo(skb)->frag_list = frag;
2661 }
2662 
2663 #define skb_walk_frags(skb, iter)       \
2664         for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2665 
2666 struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2667                                     int *peeked, int *off, int *err);
2668 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2669                                   int *err);
2670 unsigned int datagram_poll(struct file *file, struct socket *sock,
2671                            struct poll_table_struct *wait);
2672 int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2673                            struct iov_iter *to, int size);
2674 static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2675                                         struct msghdr *msg, int size)
2676 {
2677         return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
2678 }
2679 int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2680                                    struct msghdr *msg);
2681 int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2682                                  struct iov_iter *from, int len);
2683 int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
2684 void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2685 void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2686 int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
2687 int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2688 int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2689 __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2690                               int len, __wsum csum);
2691 int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2692                     struct pipe_inode_info *pipe, unsigned int len,
2693                     unsigned int flags);
2694 void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2695 unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2696 int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2697                  int len, int hlen);
2698 void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2699 int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2700 void skb_scrub_packet(struct sk_buff *skb, bool xnet);
2701 unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
2702 struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
2703 struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
2704 int skb_ensure_writable(struct sk_buff *skb, int write_len);
2705 int skb_vlan_pop(struct sk_buff *skb);
2706 int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
2707 
2708 static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2709 {
2710         return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2711 }
2712 
2713 static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2714 {
2715         return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
2716 }
2717 
2718 struct skb_checksum_ops {
2719         __wsum (*update)(const void *mem, int len, __wsum wsum);
2720         __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2721 };
2722 
2723 __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2724                       __wsum csum, const struct skb_checksum_ops *ops);
2725 __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2726                     __wsum csum);
2727 
2728 static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2729                                          int len, void *data, int hlen, void *buffer)
2730 {
2731         if (hlen - offset >= len)
2732                 return data + offset;
2733 
2734         if (!skb ||
2735             skb_copy_bits(skb, offset, buffer, len) < 0)
2736                 return NULL;
2737 
2738         return buffer;
2739 }
2740 
2741 static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2742                                        int len, void *buffer)
2743 {
2744         return __skb_header_pointer(skb, offset, len, skb->data,
2745                                     skb_headlen(skb), buffer);
2746 }
2747 
2748 /**
2749  *      skb_needs_linearize - check if we need to linearize a given skb
2750  *                            depending on the given device features.
2751  *      @skb: socket buffer to check
2752  *      @features: net device features
2753  *
2754  *      Returns true if either:
2755  *      1. skb has frag_list and the device doesn't support FRAGLIST, or
2756  *      2. skb is fragmented and the device does not support SG.
2757  */
2758 static inline bool skb_needs_linearize(struct sk_buff *skb,
2759                                        netdev_features_t features)
2760 {
2761         return skb_is_nonlinear(skb) &&
2762                ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2763                 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2764 }
2765 
2766 static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2767                                              void *to,
2768                                              const unsigned int len)
2769 {
2770         memcpy(to, skb->data, len);
2771 }
2772 
2773 static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2774                                                     const int offset, void *to,
2775                                                     const unsigned int len)
2776 {
2777         memcpy(to, skb->data + offset, len);
2778 }
2779 
2780 static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2781                                            const void *from,
2782                                            const unsigned int len)
2783 {
2784         memcpy(skb->data, from, len);
2785 }
2786 
2787 static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2788                                                   const int offset,
2789                                                   const void *from,
2790                                                   const unsigned int len)
2791 {
2792         memcpy(skb->data + offset, from, len);
2793 }
2794 
2795 void skb_init(void);
2796 
2797 static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2798 {
2799         return skb->tstamp;
2800 }
2801 
2802 /**
2803  *      skb_get_timestamp - get timestamp from a skb
2804  *      @skb: skb to get stamp from
2805  *      @stamp: pointer to struct timeval to store stamp in
2806  *
2807  *      Timestamps are stored in the skb as offsets to a base timestamp.
2808  *      This function converts the offset back to a struct timeval and stores
2809  *      it in stamp.
2810  */
2811 static inline void skb_get_timestamp(const struct sk_buff *skb,
2812                                      struct timeval *stamp)
2813 {
2814         *stamp = ktime_to_timeval(skb->tstamp);
2815 }
2816 
2817 static inline void skb_get_timestampns(const struct sk_buff *skb,
2818                                        struct timespec *stamp)
2819 {
2820         *stamp = ktime_to_timespec(skb->tstamp);
2821 }
2822 
2823 static inline void __net_timestamp(struct sk_buff *skb)
2824 {
2825         skb->tstamp = ktime_get_real();
2826 }
2827 
2828 static inline ktime_t net_timedelta(ktime_t t)
2829 {
2830         return ktime_sub(ktime_get_real(), t);
2831 }
2832 
2833 static inline ktime_t net_invalid_timestamp(void)
2834 {
2835         return ktime_set(0, 0);
2836 }
2837 
2838 struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2839 
2840 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2841 
2842 void skb_clone_tx_timestamp(struct sk_buff *skb);
2843 bool skb_defer_rx_timestamp(struct sk_buff *skb);
2844 
2845 #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2846 
2847 static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2848 {
2849 }
2850 
2851 static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2852 {
2853         return false;
2854 }
2855 
2856 #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2857 
2858 /**
2859  * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2860  *
2861  * PHY drivers may accept clones of transmitted packets for
2862  * timestamping via their phy_driver.txtstamp method. These drivers
2863  * must call this function to return the skb back to the stack, with
2864  * or without a timestamp.
2865  *
2866  * @skb: clone of the the original outgoing packet
2867  * @hwtstamps: hardware time stamps, may be NULL if not available
2868  *
2869  */
2870 void skb_complete_tx_timestamp(struct sk_buff *skb,
2871                                struct skb_shared_hwtstamps *hwtstamps);
2872 
2873 void __skb_tstamp_tx(struct sk_buff *orig_skb,
2874                      struct skb_shared_hwtstamps *hwtstamps,
2875                      struct sock *sk, int tstype);
2876 
2877 /**
2878  * skb_tstamp_tx - queue clone of skb with send time stamps
2879  * @orig_skb:   the original outgoing packet
2880  * @hwtstamps:  hardware time stamps, may be NULL if not available
2881  *
2882  * If the skb has a socket associated, then this function clones the
2883  * skb (thus sharing the actual data and optional structures), stores
2884  * the optional hardware time stamping information (if non NULL) or
2885  * generates a software time stamp (otherwise), then queues the clone
2886  * to the error queue of the socket.  Errors are silently ignored.
2887  */
2888 void skb_tstamp_tx(struct sk_buff *orig_skb,
2889                    struct skb_shared_hwtstamps *hwtstamps);
2890 
2891 static inline void sw_tx_timestamp(struct sk_buff *skb)
2892 {
2893         if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2894             !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
2895                 skb_tstamp_tx(skb, NULL);
2896 }
2897 
2898 /**
2899  * skb_tx_timestamp() - Driver hook for transmit timestamping
2900  *
2901  * Ethernet MAC Drivers should call this function in their hard_xmit()
2902  * function immediately before giving the sk_buff to the MAC hardware.
2903  *
2904  * Specifically, one should make absolutely sure that this function is
2905  * called before TX completion of this packet can trigger.  Otherwise
2906  * the packet could potentially already be freed.
2907  *
2908  * @skb: A socket buffer.
2909  */
2910 static inline void skb_tx_timestamp(struct sk_buff *skb)
2911 {
2912         skb_clone_tx_timestamp(skb);
2913         sw_tx_timestamp(skb);
2914 }
2915 
2916 /**
2917  * skb_complete_wifi_ack - deliver skb with wifi status
2918  *
2919  * @skb: the original outgoing packet
2920  * @acked: ack status
2921  *
2922  */
2923 void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2924 
2925 __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2926 __sum16 __skb_checksum_complete(struct sk_buff *skb);
2927 
2928 static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2929 {
2930         return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2931                 skb->csum_valid ||
2932                 (skb->ip_summed == CHECKSUM_PARTIAL &&
2933                  skb_checksum_start_offset(skb) >= 0));
2934 }
2935 
2936 /**
2937  *      skb_checksum_complete - Calculate checksum of an entire packet
2938  *      @skb: packet to process
2939  *
2940  *      This function calculates the checksum over the entire packet plus
2941  *      the value of skb->csum.  The latter can be used to supply the
2942  *      checksum of a pseudo header as used by TCP/UDP.  It returns the
2943  *      checksum.
2944  *
2945  *      For protocols that contain complete checksums such as ICMP/TCP/UDP,
2946  *      this function can be used to verify that checksum on received
2947  *      packets.  In that case the function should return zero if the
2948  *      checksum is correct.  In particular, this function will return zero
2949  *      if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2950  *      hardware has already verified the correctness of the checksum.
2951  */
2952 static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
2953 {
2954         return skb_csum_unnecessary(skb) ?
2955                0 : __skb_checksum_complete(skb);
2956 }
2957 
2958 static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2959 {
2960         if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2961                 if (skb->csum_level == 0)
2962                         skb->ip_summed = CHECKSUM_NONE;
2963                 else
2964                         skb->csum_level--;
2965         }
2966 }
2967 
2968 static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2969 {
2970         if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2971                 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2972                         skb->csum_level++;
2973         } else if (skb->ip_summed == CHECKSUM_NONE) {
2974                 skb->ip_summed = CHECKSUM_UNNECESSARY;
2975                 skb->csum_level = 0;
2976         }
2977 }
2978 
2979 static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2980 {
2981         /* Mark current checksum as bad (typically called from GRO
2982          * path). In the case that ip_summed is CHECKSUM_NONE
2983          * this must be the first checksum encountered in the packet.
2984          * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2985          * checksum after the last one validated. For UDP, a zero
2986          * checksum can not be marked as bad.
2987          */
2988 
2989         if (skb->ip_summed == CHECKSUM_NONE ||
2990             skb->ip_summed == CHECKSUM_UNNECESSARY)
2991                 skb->csum_bad = 1;
2992 }
2993 
2994 /* Check if we need to perform checksum complete validation.
2995  *
2996  * Returns true if checksum complete is needed, false otherwise
2997  * (either checksum is unnecessary or zero checksum is allowed).
2998  */
2999 static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3000                                                   bool zero_okay,
3001                                                   __sum16 check)
3002 {
3003         if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3004                 skb->csum_valid = 1;
3005                 __skb_decr_checksum_unnecessary(skb);
3006                 return false;
3007         }
3008 
3009         return true;
3010 }
3011 
3012 /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3013  * in checksum_init.
3014  */
3015 #define CHECKSUM_BREAK 76
3016 
3017 /* Unset checksum-complete
3018  *
3019  * Unset checksum complete can be done when packet is being modified
3020  * (uncompressed for instance) and checksum-complete value is
3021  * invalidated.
3022  */
3023 static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3024 {
3025         if (skb->ip_summed == CHECKSUM_COMPLETE)
3026                 skb->ip_summed = CHECKSUM_NONE;
3027 }
3028 
3029 /* Validate (init) checksum based on checksum complete.
3030  *
3031  * Return values:
3032  *   0: checksum is validated or try to in skb_checksum_complete. In the latter
3033  *      case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3034  *      checksum is stored in skb->csum for use in __skb_checksum_complete
3035  *   non-zero: value of invalid checksum
3036  *
3037  */
3038 static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3039                                                        bool complete,
3040                                                        __wsum psum)
3041 {
3042         if (skb->ip_summed == CHECKSUM_COMPLETE) {
3043                 if (!csum_fold(csum_add(psum, skb->csum))) {
3044                         skb->csum_valid = 1;
3045                         return 0;
3046                 }
3047         } else if (skb->csum_bad) {
3048                 /* ip_summed == CHECKSUM_NONE in this case */
3049                 return 1;
3050         }
3051 
3052         skb->csum = psum;
3053 
3054         if (complete || skb->len <= CHECKSUM_BREAK) {
3055                 __sum16 csum;
3056 
3057                 csum = __skb_checksum_complete(skb);
3058                 skb->csum_valid = !csum;
3059                 return csum;
3060         }
3061 
3062         return 0;
3063 }
3064 
3065 static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3066 {
3067         return 0;
3068 }
3069 
3070 /* Perform checksum validate (init). Note that this is a macro since we only
3071  * want to calculate the pseudo header which is an input function if necessary.
3072  * First we try to validate without any computation (checksum unnecessary) and
3073  * then calculate based on checksum complete calling the function to compute
3074  * pseudo header.
3075  *
3076  * Return values:
3077  *   0: checksum is validated or try to in skb_checksum_complete
3078  *   non-zero: value of invalid checksum
3079  */
3080 #define __skb_checksum_validate(skb, proto, complete,                   \
3081                                 zero_okay, check, compute_pseudo)       \
3082 ({                                                                      \
3083         __sum16 __ret = 0;                                              \
3084         skb->csum_valid = 0;                                            \
3085         if (__skb_checksum_validate_needed(skb, zero_okay, check))      \
3086                 __ret = __skb_checksum_validate_complete(skb,           \
3087                                 complete, compute_pseudo(skb, proto));  \
3088         __ret;                                                          \
3089 })
3090 
3091 #define skb_checksum_init(skb, proto, compute_pseudo)                   \
3092         __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3093 
3094 #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3095         __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3096 
3097 #define skb_checksum_validate(skb, proto, compute_pseudo)               \
3098         __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3099 
3100 #define skb_checksum_validate_zero_check(skb, proto, check,             \
3101                                          compute_pseudo)                \
3102         __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
3103 
3104 #define skb_checksum_simple_validate(skb)                               \
3105         __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3106 
3107 static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3108 {
3109         return (skb->ip_summed == CHECKSUM_NONE &&
3110                 skb->csum_valid && !skb->csum_bad);
3111 }
3112 
3113 static inline void __skb_checksum_convert(struct sk_buff *skb,
3114                                           __sum16 check, __wsum pseudo)
3115 {
3116         skb->csum = ~pseudo;
3117         skb->ip_summed = CHECKSUM_COMPLETE;
3118 }
3119 
3120 #define skb_checksum_try_convert(skb, proto, check, compute_pseudo)     \
3121 do {                                                                    \
3122         if (__skb_checksum_convert_check(skb))                          \
3123                 __skb_checksum_convert(skb, check,                      \
3124                                        compute_pseudo(skb, proto));     \
3125 } while (0)
3126 
3127 static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3128                                               u16 start, u16 offset)
3129 {
3130         skb->ip_summed = CHECKSUM_PARTIAL;
3131         skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3132         skb->csum_offset = offset - start;
3133 }
3134 
3135 /* Update skbuf and packet to reflect the remote checksum offload operation.
3136  * When called, ptr indicates the starting point for skb->csum when
3137  * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3138  * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3139  */
3140 static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
3141                                        int start, int offset, bool nopartial)
3142 {
3143         __wsum delta;
3144 
3145         if (!nopartial) {
3146                 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3147                 return;
3148         }
3149 
3150          if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3151                 __skb_checksum_complete(skb);
3152                 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3153         }
3154 
3155         delta = remcsum_adjust(ptr, skb->csum, start, offset);
3156 
3157         /* Adjust skb->csum since we changed the packet */
3158         skb->csum = csum_add(skb->csum, delta);
3159 }
3160 
3161 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3162 void nf_conntrack_destroy(struct nf_conntrack *nfct);
3163 static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3164 {
3165         if (nfct && atomic_dec_and_test(&nfct->use))
3166                 nf_conntrack_destroy(nfct);
3167 }
3168 static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3169 {
3170         if (nfct)
3171                 atomic_inc(&nfct->use);
3172 }
3173 #endif
3174 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3175 static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3176 {
3177         if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3178                 kfree(nf_bridge);
3179 }
3180 static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3181 {
3182         if (nf_bridge)
3183                 atomic_inc(&nf_bridge->use);
3184 }
3185 #endif /* CONFIG_BRIDGE_NETFILTER */
3186 static inline void nf_reset(struct sk_buff *skb)
3187 {
3188 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3189         nf_conntrack_put(skb->nfct);
3190         skb->nfct = NULL;
3191 #endif
3192 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3193         nf_bridge_put(skb->nf_bridge);
3194         skb->nf_bridge = NULL;
3195 #endif
3196 }
3197 
3198 static inline void nf_reset_trace(struct sk_buff *skb)
3199 {
3200 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3201         skb->nf_trace = 0;
3202 #endif
3203 }
3204 
3205 /* Note: This doesn't put any conntrack and bridge info in dst. */
3206 static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3207                              bool copy)
3208 {
3209 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3210         dst->nfct = src->nfct;
3211         nf_conntrack_get(src->nfct);
3212         if (copy)
3213                 dst->nfctinfo = src->nfctinfo;
3214 #endif
3215 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3216         dst->nf_bridge  = src->nf_bridge;
3217         nf_bridge_get(src->nf_bridge);
3218 #endif
3219 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
3220         if (copy)
3221                 dst->nf_trace = src->nf_trace;
3222 #endif
3223 }
3224 
3225 static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3226 {
3227 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
3228         nf_conntrack_put(dst->nfct);
3229 #endif
3230 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3231         nf_bridge_put(dst->nf_bridge);
3232 #endif
3233         __nf_copy(dst, src, true);
3234 }
3235 
3236 #ifdef CONFIG_NETWORK_SECMARK
3237 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3238 {
3239         to->secmark = from->secmark;
3240 }
3241 
3242 static inline void skb_init_secmark(struct sk_buff *skb)
3243 {
3244         skb->secmark = 0;
3245 }
3246 #else
3247 static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3248 { }
3249 
3250 static inline void skb_init_secmark(struct sk_buff *skb)
3251 { }
3252 #endif
3253 
3254 static inline bool skb_irq_freeable(const struct sk_buff *skb)
3255 {
3256         return !skb->destructor &&
3257 #if IS_ENABLED(CONFIG_XFRM)
3258                 !skb->sp &&
3259 #endif
3260 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
3261                 !skb->nfct &&
3262 #endif
3263                 !skb->_skb_refdst &&
3264                 !skb_has_frag_list(skb);
3265 }
3266 
3267 static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3268 {
3269         skb->queue_mapping = queue_mapping;
3270 }
3271 
3272 static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
3273 {
3274         return skb->queue_mapping;
3275 }
3276 
3277 static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3278 {
3279         to->queue_mapping = from->queue_mapping;
3280 }
3281 
3282 static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3283 {
3284         skb->queue_mapping = rx_queue + 1;
3285 }
3286 
3287 static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
3288 {
3289         return skb->queue_mapping - 1;
3290 }
3291 
3292 static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
3293 {
3294         return skb->queue_mapping != 0;
3295 }
3296 
3297 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3298                   unsigned int num_tx_queues);
3299 
3300 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3301 {
3302 #ifdef CONFIG_XFRM
3303         return skb->sp;
3304 #else
3305         return NULL;
3306 #endif
3307 }
3308 
3309 /* Keeps track of mac header offset relative to skb->head.
3310  * It is useful for TSO of Tunneling protocol. e.g. GRE.
3311  * For non-tunnel skb it points to skb_mac_header() and for
3312  * tunnel skb it points to outer mac header.
3313  * Keeps track of level of encapsulation of network headers.
3314  */
3315 struct skb_gso_cb {
3316         int     mac_offset;
3317         int     encap_level;
3318         __u16   csum_start;
3319 };
3320 #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3321 
3322 static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3323 {
3324         return (skb_mac_header(inner_skb) - inner_skb->head) -
3325                 SKB_GSO_CB(inner_skb)->mac_offset;
3326 }
3327 
3328 static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3329 {
3330         int new_headroom, headroom;
3331         int ret;
3332 
3333         headroom = skb_headroom(skb);
3334         ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3335         if (ret)
3336                 return ret;
3337 
3338         new_headroom = skb_headroom(skb);
3339         SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3340         return 0;
3341 }
3342 
3343 /* Compute the checksum for a gso segment. First compute the checksum value
3344  * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3345  * then add in skb->csum (checksum from csum_start to end of packet).
3346  * skb->csum and csum_start are then updated to reflect the checksum of the
3347  * resultant packet starting from the transport header-- the resultant checksum
3348  * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3349  * header.
3350  */
3351 static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3352 {
3353         int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3354             skb_transport_offset(skb);
3355         __u16 csum;
3356 
3357         csum = csum_fold(csum_partial(skb_transport_header(skb),
3358                                       plen, skb->csum));
3359         skb->csum = res;
3360         SKB_GSO_CB(skb)->csum_start -= plen;
3361 
3362         return csum;
3363 }
3364 
3365 static inline bool skb_is_gso(const struct sk_buff *skb)
3366 {
3367         return skb_shinfo(skb)->gso_size;
3368 }
3369 
3370 /* Note: Should be called only if skb_is_gso(skb) is true */
3371 static inline bool skb_is_gso_v6(const struct sk_buff *skb)
3372 {
3373         return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3374 }
3375 
3376 void __skb_warn_lro_forwarding(const struct sk_buff *skb);
3377 
3378 static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3379 {
3380         /* LRO sets gso_size but not gso_type, whereas if GSO is really
3381          * wanted then gso_type will be set. */
3382         const struct skb_shared_info *shinfo = skb_shinfo(skb);
3383 
3384         if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3385             unlikely(shinfo->gso_type == 0)) {
3386                 __skb_warn_lro_forwarding(skb);
3387                 return true;
3388         }
3389         return false;
3390 }
3391 
3392 static inline void skb_forward_csum(struct sk_buff *skb)
3393 {
3394         /* Unfortunately we don't support this one.  Any brave souls? */
3395         if (skb->ip_summed == CHECKSUM_COMPLETE)
3396                 skb->ip_summed = CHECKSUM_NONE;
3397 }
3398 
3399 /**
3400  * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3401  * @skb: skb to check
3402  *
3403  * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3404  * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3405  * use this helper, to document places where we make this assertion.
3406  */
3407 static inline void skb_checksum_none_assert(const struct sk_buff *skb)
3408 {
3409 #ifdef DEBUG
3410         BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3411 #endif
3412 }
3413 
3414 bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
3415 
3416 int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3417 
3418 u32 skb_get_poff(const struct sk_buff *skb);
3419 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3420                    const struct flow_keys *keys, int hlen);
3421 
3422 /**
3423  * skb_head_is_locked - Determine if the skb->head is locked down
3424  * @skb: skb to check
3425  *
3426  * The head on skbs build around a head frag can be removed if they are
3427  * not cloned.  This function returns true if the skb head is locked down
3428  * due to either being allocated via kmalloc, or by being a clone with
3429  * multiple references to the head.
3430  */
3431 static inline bool skb_head_is_locked(const struct sk_buff *skb)
3432 {
3433         return !skb->head_frag || skb_cloned(skb);
3434 }
3435 
3436 /**
3437  * skb_gso_network_seglen - Return length of individual segments of a gso packet
3438  *
3439  * @skb: GSO skb
3440  *
3441  * skb_gso_network_seglen is used to determine the real size of the
3442  * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3443  *
3444  * The MAC/L2 header is not accounted for.
3445  */
3446 static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3447 {
3448         unsigned int hdr_len = skb_transport_header(skb) -
3449                                skb_network_header(skb);
3450         return hdr_len + skb_gso_transport_seglen(skb);
3451 }
3452 #endif  /* __KERNEL__ */
3453 #endif  /* _LINUX_SKBUFF_H */
3454 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp