skbuff.h 80 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881
  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. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/kernel.h>
  16. #include <linux/kmemcheck.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/bug.h>
  20. #include <linux/cache.h>
  21. #include <linux/atomic.h>
  22. #include <asm/types.h>
  23. #include <linux/spinlock.h>
  24. #include <linux/net.h>
  25. #include <linux/textsearch.h>
  26. #include <net/checksum.h>
  27. #include <linux/rcupdate.h>
  28. #include <linux/dmaengine.h>
  29. #include <linux/hrtimer.h>
  30. #include <linux/dma-mapping.h>
  31. #include <linux/netdev_features.h>
  32. #include <net/flow_keys.h>
  33. /* Don't change this without changing skb_csum_unnecessary! */
  34. #define CHECKSUM_NONE 0
  35. #define CHECKSUM_UNNECESSARY 1
  36. #define CHECKSUM_COMPLETE 2
  37. #define CHECKSUM_PARTIAL 3
  38. #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
  39. ~(SMP_CACHE_BYTES - 1))
  40. #define SKB_WITH_OVERHEAD(X) \
  41. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  42. #define SKB_MAX_ORDER(X, ORDER) \
  43. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  44. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  45. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  46. /* return minimum truesize of one skb containing X bytes of data */
  47. #define SKB_TRUESIZE(X) ((X) + \
  48. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  49. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  50. /* A. Checksumming of received packets by device.
  51. *
  52. * NONE: device failed to checksum this packet.
  53. * skb->csum is undefined.
  54. *
  55. * UNNECESSARY: device parsed packet and wouldbe verified checksum.
  56. * skb->csum is undefined.
  57. * It is bad option, but, unfortunately, many of vendors do this.
  58. * Apparently with secret goal to sell you new device, when you
  59. * will add new protocol to your host. F.e. IPv6. 8)
  60. *
  61. * COMPLETE: the most generic way. Device supplied checksum of _all_
  62. * the packet as seen by netif_rx in skb->csum.
  63. * NOTE: Even if device supports only some protocols, but
  64. * is able to produce some skb->csum, it MUST use COMPLETE,
  65. * not UNNECESSARY.
  66. *
  67. * PARTIAL: identical to the case for output below. This may occur
  68. * on a packet received directly from another Linux OS, e.g.,
  69. * a virtualised Linux kernel on the same host. The packet can
  70. * be treated in the same way as UNNECESSARY except that on
  71. * output (i.e., forwarding) the checksum must be filled in
  72. * by the OS or the hardware.
  73. *
  74. * B. Checksumming on output.
  75. *
  76. * NONE: skb is checksummed by protocol or csum is not required.
  77. *
  78. * PARTIAL: device is required to csum packet as seen by hard_start_xmit
  79. * from skb->csum_start to the end and to record the checksum
  80. * at skb->csum_start + skb->csum_offset.
  81. *
  82. * Device must show its capabilities in dev->features, set
  83. * at device setup time.
  84. * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
  85. * everything.
  86. * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
  87. * TCP/UDP over IPv4. Sigh. Vendors like this
  88. * way by an unknown reason. Though, see comment above
  89. * about CHECKSUM_UNNECESSARY. 8)
  90. * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
  91. *
  92. * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
  93. * that do not want net to perform the checksum calculation should use
  94. * this flag in their outgoing skbs.
  95. * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
  96. * offload. Correspondingly, the FCoE protocol driver
  97. * stack should use CHECKSUM_UNNECESSARY.
  98. *
  99. * Any questions? No questions, good. --ANK
  100. */
  101. struct net_device;
  102. struct scatterlist;
  103. struct pipe_inode_info;
  104. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  105. struct nf_conntrack {
  106. atomic_t use;
  107. };
  108. #endif
  109. #ifdef CONFIG_BRIDGE_NETFILTER
  110. struct nf_bridge_info {
  111. atomic_t use;
  112. unsigned int mask;
  113. struct net_device *physindev;
  114. struct net_device *physoutdev;
  115. unsigned long data[32 / sizeof(unsigned long)];
  116. };
  117. #endif
  118. struct sk_buff_head {
  119. /* These two members must be first. */
  120. struct sk_buff *next;
  121. struct sk_buff *prev;
  122. __u32 qlen;
  123. spinlock_t lock;
  124. };
  125. struct sk_buff;
  126. /* To allow 64K frame to be packed as single skb without frag_list we
  127. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  128. * buffers which do not start on a page boundary.
  129. *
  130. * Since GRO uses frags we allocate at least 16 regardless of page
  131. * size.
  132. */
  133. #if (65536/PAGE_SIZE + 1) < 16
  134. #define MAX_SKB_FRAGS 16UL
  135. #else
  136. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  137. #endif
  138. typedef struct skb_frag_struct skb_frag_t;
  139. struct skb_frag_struct {
  140. struct {
  141. struct page *p;
  142. } page;
  143. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  144. __u32 page_offset;
  145. __u32 size;
  146. #else
  147. __u16 page_offset;
  148. __u16 size;
  149. #endif
  150. };
  151. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  152. {
  153. return frag->size;
  154. }
  155. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  156. {
  157. frag->size = size;
  158. }
  159. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  160. {
  161. frag->size += delta;
  162. }
  163. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  164. {
  165. frag->size -= delta;
  166. }
  167. #define HAVE_HW_TIME_STAMP
  168. /**
  169. * struct skb_shared_hwtstamps - hardware time stamps
  170. * @hwtstamp: hardware time stamp transformed into duration
  171. * since arbitrary point in time
  172. * @syststamp: hwtstamp transformed to system time base
  173. *
  174. * Software time stamps generated by ktime_get_real() are stored in
  175. * skb->tstamp. The relation between the different kinds of time
  176. * stamps is as follows:
  177. *
  178. * syststamp and tstamp can be compared against each other in
  179. * arbitrary combinations. The accuracy of a
  180. * syststamp/tstamp/"syststamp from other device" comparison is
  181. * limited by the accuracy of the transformation into system time
  182. * base. This depends on the device driver and its underlying
  183. * hardware.
  184. *
  185. * hwtstamps can only be compared against other hwtstamps from
  186. * the same device.
  187. *
  188. * This structure is attached to packets as part of the
  189. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  190. */
  191. struct skb_shared_hwtstamps {
  192. ktime_t hwtstamp;
  193. ktime_t syststamp;
  194. };
  195. /* Definitions for tx_flags in struct skb_shared_info */
  196. enum {
  197. /* generate hardware time stamp */
  198. SKBTX_HW_TSTAMP = 1 << 0,
  199. /* generate software time stamp */
  200. SKBTX_SW_TSTAMP = 1 << 1,
  201. /* device driver is going to provide hardware time stamp */
  202. SKBTX_IN_PROGRESS = 1 << 2,
  203. /* device driver supports TX zero-copy buffers */
  204. SKBTX_DEV_ZEROCOPY = 1 << 3,
  205. /* generate wifi status information (where possible) */
  206. SKBTX_WIFI_STATUS = 1 << 4,
  207. /* This indicates at least one fragment might be overwritten
  208. * (as in vmsplice(), sendfile() ...)
  209. * If we need to compute a TX checksum, we'll need to copy
  210. * all frags to avoid possible bad checksum
  211. */
  212. SKBTX_SHARED_FRAG = 1 << 5,
  213. };
  214. /*
  215. * The callback notifies userspace to release buffers when skb DMA is done in
  216. * lower device, the skb last reference should be 0 when calling this.
  217. * The zerocopy_success argument is true if zero copy transmit occurred,
  218. * false on data copy or out of memory error caused by data copy attempt.
  219. * The ctx field is used to track device context.
  220. * The desc field is used to track userspace buffer index.
  221. */
  222. struct ubuf_info {
  223. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  224. void *ctx;
  225. unsigned long desc;
  226. };
  227. /* This data is invariant across clones and lives at
  228. * the end of the header data, ie. at skb->end.
  229. */
  230. struct skb_shared_info {
  231. unsigned char nr_frags;
  232. __u8 tx_flags;
  233. unsigned short gso_size;
  234. /* Warning: this field is not always filled in (UFO)! */
  235. unsigned short gso_segs;
  236. unsigned short gso_type;
  237. struct sk_buff *frag_list;
  238. struct skb_shared_hwtstamps hwtstamps;
  239. __be32 ip6_frag_id;
  240. /*
  241. * Warning : all fields before dataref are cleared in __alloc_skb()
  242. */
  243. atomic_t dataref;
  244. /* Intermediate layers must ensure that destructor_arg
  245. * remains valid until skb destructor */
  246. void * destructor_arg;
  247. /* must be last field, see pskb_expand_head() */
  248. skb_frag_t frags[MAX_SKB_FRAGS];
  249. };
  250. /* We divide dataref into two halves. The higher 16 bits hold references
  251. * to the payload part of skb->data. The lower 16 bits hold references to
  252. * the entire skb->data. A clone of a headerless skb holds the length of
  253. * the header in skb->hdr_len.
  254. *
  255. * All users must obey the rule that the skb->data reference count must be
  256. * greater than or equal to the payload reference count.
  257. *
  258. * Holding a reference to the payload part means that the user does not
  259. * care about modifications to the header part of skb->data.
  260. */
  261. #define SKB_DATAREF_SHIFT 16
  262. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  263. enum {
  264. SKB_FCLONE_UNAVAILABLE,
  265. SKB_FCLONE_ORIG,
  266. SKB_FCLONE_CLONE,
  267. };
  268. enum {
  269. SKB_GSO_TCPV4 = 1 << 0,
  270. SKB_GSO_UDP = 1 << 1,
  271. /* This indicates the skb is from an untrusted source. */
  272. SKB_GSO_DODGY = 1 << 2,
  273. /* This indicates the tcp segment has CWR set. */
  274. SKB_GSO_TCP_ECN = 1 << 3,
  275. SKB_GSO_TCPV6 = 1 << 4,
  276. SKB_GSO_FCOE = 1 << 5,
  277. SKB_GSO_GRE = 1 << 6,
  278. SKB_GSO_IPIP = 1 << 7,
  279. SKB_GSO_SIT = 1 << 8,
  280. SKB_GSO_UDP_TUNNEL = 1 << 9,
  281. SKB_GSO_MPLS = 1 << 10,
  282. };
  283. #if BITS_PER_LONG > 32
  284. #define NET_SKBUFF_DATA_USES_OFFSET 1
  285. #endif
  286. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  287. typedef unsigned int sk_buff_data_t;
  288. #else
  289. typedef unsigned char *sk_buff_data_t;
  290. #endif
  291. /**
  292. * struct sk_buff - socket buffer
  293. * @next: Next buffer in list
  294. * @prev: Previous buffer in list
  295. * @tstamp: Time we arrived
  296. * @sk: Socket we are owned by
  297. * @dev: Device we arrived on/are leaving by
  298. * @cb: Control buffer. Free for use by every layer. Put private vars here
  299. * @_skb_refdst: destination entry (with norefcount bit)
  300. * @sp: the security path, used for xfrm
  301. * @len: Length of actual data
  302. * @data_len: Data length
  303. * @mac_len: Length of link layer header
  304. * @hdr_len: writable header length of cloned skb
  305. * @csum: Checksum (must include start/offset pair)
  306. * @csum_start: Offset from skb->head where checksumming should start
  307. * @csum_offset: Offset from csum_start where checksum should be stored
  308. * @priority: Packet queueing priority
  309. * @local_df: allow local fragmentation
  310. * @cloned: Head may be cloned (check refcnt to be sure)
  311. * @ip_summed: Driver fed us an IP checksum
  312. * @nohdr: Payload reference only, must not modify header
  313. * @nfctinfo: Relationship of this skb to the connection
  314. * @pkt_type: Packet class
  315. * @fclone: skbuff clone status
  316. * @ipvs_property: skbuff is owned by ipvs
  317. * @peeked: this packet has been seen already, so stats have been
  318. * done for it, don't do them again
  319. * @nf_trace: netfilter packet trace flag
  320. * @protocol: Packet protocol from driver
  321. * @destructor: Destruct function
  322. * @nfct: Associated connection, if any
  323. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  324. * @skb_iif: ifindex of device we arrived on
  325. * @tc_index: Traffic control index
  326. * @tc_verd: traffic control verdict
  327. * @rxhash: the packet hash computed on receive
  328. * @queue_mapping: Queue mapping for multiqueue devices
  329. * @ndisc_nodetype: router type (from link layer)
  330. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  331. * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
  332. * ports.
  333. * @wifi_acked_valid: wifi_acked was set
  334. * @wifi_acked: whether frame was acked on wifi or not
  335. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  336. * @dma_cookie: a cookie to one of several possible DMA operations
  337. * done by skb DMA functions
  338. * @napi_id: id of the NAPI struct this skb came from
  339. * @secmark: security marking
  340. * @mark: Generic packet mark
  341. * @dropcount: total number of sk_receive_queue overflows
  342. * @vlan_proto: vlan encapsulation protocol
  343. * @vlan_tci: vlan tag control information
  344. * @inner_protocol: Protocol (encapsulation)
  345. * @inner_transport_header: Inner transport layer header (encapsulation)
  346. * @inner_network_header: Network layer header (encapsulation)
  347. * @inner_mac_header: Link layer header (encapsulation)
  348. * @transport_header: Transport layer header
  349. * @network_header: Network layer header
  350. * @mac_header: Link layer header
  351. * @tail: Tail pointer
  352. * @end: End pointer
  353. * @head: Head of buffer
  354. * @data: Data head pointer
  355. * @truesize: Buffer size
  356. * @users: User count - see {datagram,tcp}.c
  357. */
  358. struct sk_buff {
  359. /* These two members must be first. */
  360. struct sk_buff *next;
  361. struct sk_buff *prev;
  362. ktime_t tstamp;
  363. struct sock *sk;
  364. struct net_device *dev;
  365. /*
  366. * This is the control buffer. It is free to use for every
  367. * layer. Please put your private variables there. If you
  368. * want to keep them across layers you have to do a skb_clone()
  369. * first. This is owned by whoever has the skb queued ATM.
  370. */
  371. char cb[48] __aligned(8);
  372. unsigned long _skb_refdst;
  373. #ifdef CONFIG_XFRM
  374. struct sec_path *sp;
  375. #endif
  376. unsigned int len,
  377. data_len;
  378. __u16 mac_len,
  379. hdr_len;
  380. union {
  381. __wsum csum;
  382. struct {
  383. __u16 csum_start;
  384. __u16 csum_offset;
  385. };
  386. };
  387. __u32 priority;
  388. kmemcheck_bitfield_begin(flags1);
  389. __u8 local_df:1,
  390. cloned:1,
  391. ip_summed:2,
  392. nohdr:1,
  393. nfctinfo:3;
  394. __u8 pkt_type:3,
  395. fclone:2,
  396. ipvs_property:1,
  397. peeked:1,
  398. nf_trace:1;
  399. kmemcheck_bitfield_end(flags1);
  400. __be16 protocol;
  401. void (*destructor)(struct sk_buff *skb);
  402. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  403. struct nf_conntrack *nfct;
  404. #endif
  405. #ifdef CONFIG_BRIDGE_NETFILTER
  406. struct nf_bridge_info *nf_bridge;
  407. #endif
  408. int skb_iif;
  409. __u32 rxhash;
  410. __be16 vlan_proto;
  411. __u16 vlan_tci;
  412. #ifdef CONFIG_NET_SCHED
  413. __u16 tc_index; /* traffic control index */
  414. #ifdef CONFIG_NET_CLS_ACT
  415. __u16 tc_verd; /* traffic control verdict */
  416. #endif
  417. #endif
  418. __u16 queue_mapping;
  419. kmemcheck_bitfield_begin(flags2);
  420. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  421. __u8 ndisc_nodetype:2;
  422. #endif
  423. __u8 pfmemalloc:1;
  424. __u8 ooo_okay:1;
  425. __u8 l4_rxhash:1;
  426. __u8 wifi_acked_valid:1;
  427. __u8 wifi_acked:1;
  428. __u8 no_fcs:1;
  429. __u8 head_frag:1;
  430. /* Encapsulation protocol and NIC drivers should use
  431. * this flag to indicate to each other if the skb contains
  432. * encapsulated packet or not and maybe use the inner packet
  433. * headers if needed
  434. */
  435. __u8 encapsulation:1;
  436. /* 6/8 bit hole (depending on ndisc_nodetype presence) */
  437. kmemcheck_bitfield_end(flags2);
  438. #if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
  439. union {
  440. unsigned int napi_id;
  441. dma_cookie_t dma_cookie;
  442. };
  443. #endif
  444. #ifdef CONFIG_NETWORK_SECMARK
  445. __u32 secmark;
  446. #endif
  447. union {
  448. __u32 mark;
  449. __u32 dropcount;
  450. __u32 reserved_tailroom;
  451. };
  452. __be16 inner_protocol;
  453. __u16 inner_transport_header;
  454. __u16 inner_network_header;
  455. __u16 inner_mac_header;
  456. __u16 transport_header;
  457. __u16 network_header;
  458. __u16 mac_header;
  459. /* These elements must be at the end, see alloc_skb() for details. */
  460. sk_buff_data_t tail;
  461. sk_buff_data_t end;
  462. unsigned char *head,
  463. *data;
  464. unsigned int truesize;
  465. atomic_t users;
  466. };
  467. #ifdef __KERNEL__
  468. /*
  469. * Handling routines are only of interest to the kernel
  470. */
  471. #include <linux/slab.h>
  472. #define SKB_ALLOC_FCLONE 0x01
  473. #define SKB_ALLOC_RX 0x02
  474. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  475. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  476. {
  477. return unlikely(skb->pfmemalloc);
  478. }
  479. /*
  480. * skb might have a dst pointer attached, refcounted or not.
  481. * _skb_refdst low order bit is set if refcount was _not_ taken
  482. */
  483. #define SKB_DST_NOREF 1UL
  484. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  485. /**
  486. * skb_dst - returns skb dst_entry
  487. * @skb: buffer
  488. *
  489. * Returns skb dst_entry, regardless of reference taken or not.
  490. */
  491. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  492. {
  493. /* If refdst was not refcounted, check we still are in a
  494. * rcu_read_lock section
  495. */
  496. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  497. !rcu_read_lock_held() &&
  498. !rcu_read_lock_bh_held());
  499. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  500. }
  501. /**
  502. * skb_dst_set - sets skb dst
  503. * @skb: buffer
  504. * @dst: dst entry
  505. *
  506. * Sets skb dst, assuming a reference was taken on dst and should
  507. * be released by skb_dst_drop()
  508. */
  509. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  510. {
  511. skb->_skb_refdst = (unsigned long)dst;
  512. }
  513. void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
  514. bool force);
  515. /**
  516. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  517. * @skb: buffer
  518. * @dst: dst entry
  519. *
  520. * Sets skb dst, assuming a reference was not taken on dst.
  521. * If dst entry is cached, we do not take reference and dst_release
  522. * will be avoided by refdst_drop. If dst entry is not cached, we take
  523. * reference, so that last dst_release can destroy the dst immediately.
  524. */
  525. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  526. {
  527. __skb_dst_set_noref(skb, dst, false);
  528. }
  529. /**
  530. * skb_dst_set_noref_force - sets skb dst, without taking reference
  531. * @skb: buffer
  532. * @dst: dst entry
  533. *
  534. * Sets skb dst, assuming a reference was not taken on dst.
  535. * No reference is taken and no dst_release will be called. While for
  536. * cached dsts deferred reclaim is a basic feature, for entries that are
  537. * not cached it is caller's job to guarantee that last dst_release for
  538. * provided dst happens when nobody uses it, eg. after a RCU grace period.
  539. */
  540. static inline void skb_dst_set_noref_force(struct sk_buff *skb,
  541. struct dst_entry *dst)
  542. {
  543. __skb_dst_set_noref(skb, dst, true);
  544. }
  545. /**
  546. * skb_dst_is_noref - Test if skb dst isn't refcounted
  547. * @skb: buffer
  548. */
  549. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  550. {
  551. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  552. }
  553. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  554. {
  555. return (struct rtable *)skb_dst(skb);
  556. }
  557. void kfree_skb(struct sk_buff *skb);
  558. void kfree_skb_list(struct sk_buff *segs);
  559. void skb_tx_error(struct sk_buff *skb);
  560. void consume_skb(struct sk_buff *skb);
  561. void __kfree_skb(struct sk_buff *skb);
  562. extern struct kmem_cache *skbuff_head_cache;
  563. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  564. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  565. bool *fragstolen, int *delta_truesize);
  566. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  567. int node);
  568. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  569. static inline struct sk_buff *alloc_skb(unsigned int size,
  570. gfp_t priority)
  571. {
  572. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  573. }
  574. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  575. gfp_t priority)
  576. {
  577. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  578. }
  579. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  580. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  581. {
  582. return __alloc_skb_head(priority, -1);
  583. }
  584. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  585. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  586. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  587. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  588. struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
  589. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  590. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  591. unsigned int headroom);
  592. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  593. int newtailroom, gfp_t priority);
  594. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  595. int len);
  596. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  597. int skb_pad(struct sk_buff *skb, int pad);
  598. #define dev_kfree_skb(a) consume_skb(a)
  599. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  600. int getfrag(void *from, char *to, int offset,
  601. int len, int odd, struct sk_buff *skb),
  602. void *from, int length);
  603. struct skb_seq_state {
  604. __u32 lower_offset;
  605. __u32 upper_offset;
  606. __u32 frag_idx;
  607. __u32 stepped_offset;
  608. struct sk_buff *root_skb;
  609. struct sk_buff *cur_skb;
  610. __u8 *frag_data;
  611. };
  612. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  613. unsigned int to, struct skb_seq_state *st);
  614. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  615. struct skb_seq_state *st);
  616. void skb_abort_seq_read(struct skb_seq_state *st);
  617. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  618. unsigned int to, struct ts_config *config,
  619. struct ts_state *state);
  620. /*
  621. * Packet hash types specify the type of hash in skb_set_hash.
  622. *
  623. * Hash types refer to the protocol layer addresses which are used to
  624. * construct a packet's hash. The hashes are used to differentiate or identify
  625. * flows of the protocol layer for the hash type. Hash types are either
  626. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  627. *
  628. * Properties of hashes:
  629. *
  630. * 1) Two packets in different flows have different hash values
  631. * 2) Two packets in the same flow should have the same hash value
  632. *
  633. * A hash at a higher layer is considered to be more specific. A driver should
  634. * set the most specific hash possible.
  635. *
  636. * A driver cannot indicate a more specific hash than the layer at which a hash
  637. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  638. *
  639. * A driver may indicate a hash level which is less specific than the
  640. * actual layer the hash was computed on. For instance, a hash computed
  641. * at L4 may be considered an L3 hash. This should only be done if the
  642. * driver can't unambiguously determine that the HW computed the hash at
  643. * the higher layer. Note that the "should" in the second property above
  644. * permits this.
  645. */
  646. enum pkt_hash_types {
  647. PKT_HASH_TYPE_NONE, /* Undefined type */
  648. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  649. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  650. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  651. };
  652. static inline void
  653. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  654. {
  655. skb->l4_rxhash = (type == PKT_HASH_TYPE_L4);
  656. skb->rxhash = hash;
  657. }
  658. void __skb_get_hash(struct sk_buff *skb);
  659. static inline __u32 skb_get_hash(struct sk_buff *skb)
  660. {
  661. if (!skb->l4_rxhash)
  662. __skb_get_hash(skb);
  663. return skb->rxhash;
  664. }
  665. static inline void skb_clear_hash(struct sk_buff *skb)
  666. {
  667. skb->rxhash = 0;
  668. skb->l4_rxhash = 0;
  669. }
  670. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  671. {
  672. if (!skb->l4_rxhash)
  673. skb_clear_hash(skb);
  674. }
  675. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  676. {
  677. to->rxhash = from->rxhash;
  678. to->l4_rxhash = from->l4_rxhash;
  679. };
  680. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  681. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  682. {
  683. return skb->head + skb->end;
  684. }
  685. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  686. {
  687. return skb->end;
  688. }
  689. #else
  690. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  691. {
  692. return skb->end;
  693. }
  694. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  695. {
  696. return skb->end - skb->head;
  697. }
  698. #endif
  699. /* Internal */
  700. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  701. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  702. {
  703. return &skb_shinfo(skb)->hwtstamps;
  704. }
  705. /**
  706. * skb_queue_empty - check if a queue is empty
  707. * @list: queue head
  708. *
  709. * Returns true if the queue is empty, false otherwise.
  710. */
  711. static inline int skb_queue_empty(const struct sk_buff_head *list)
  712. {
  713. return list->next == (struct sk_buff *)list;
  714. }
  715. /**
  716. * skb_queue_is_last - check if skb is the last entry in the queue
  717. * @list: queue head
  718. * @skb: buffer
  719. *
  720. * Returns true if @skb is the last buffer on the list.
  721. */
  722. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  723. const struct sk_buff *skb)
  724. {
  725. return skb->next == (struct sk_buff *)list;
  726. }
  727. /**
  728. * skb_queue_is_first - check if skb is the first entry in the queue
  729. * @list: queue head
  730. * @skb: buffer
  731. *
  732. * Returns true if @skb is the first buffer on the list.
  733. */
  734. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  735. const struct sk_buff *skb)
  736. {
  737. return skb->prev == (struct sk_buff *)list;
  738. }
  739. /**
  740. * skb_queue_next - return the next packet in the queue
  741. * @list: queue head
  742. * @skb: current buffer
  743. *
  744. * Return the next packet in @list after @skb. It is only valid to
  745. * call this if skb_queue_is_last() evaluates to false.
  746. */
  747. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  748. const struct sk_buff *skb)
  749. {
  750. /* This BUG_ON may seem severe, but if we just return then we
  751. * are going to dereference garbage.
  752. */
  753. BUG_ON(skb_queue_is_last(list, skb));
  754. return skb->next;
  755. }
  756. /**
  757. * skb_queue_prev - return the prev packet in the queue
  758. * @list: queue head
  759. * @skb: current buffer
  760. *
  761. * Return the prev packet in @list before @skb. It is only valid to
  762. * call this if skb_queue_is_first() evaluates to false.
  763. */
  764. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  765. const struct sk_buff *skb)
  766. {
  767. /* This BUG_ON may seem severe, but if we just return then we
  768. * are going to dereference garbage.
  769. */
  770. BUG_ON(skb_queue_is_first(list, skb));
  771. return skb->prev;
  772. }
  773. /**
  774. * skb_get - reference buffer
  775. * @skb: buffer to reference
  776. *
  777. * Makes another reference to a socket buffer and returns a pointer
  778. * to the buffer.
  779. */
  780. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  781. {
  782. atomic_inc(&skb->users);
  783. return skb;
  784. }
  785. /*
  786. * If users == 1, we are the only owner and are can avoid redundant
  787. * atomic change.
  788. */
  789. /**
  790. * skb_cloned - is the buffer a clone
  791. * @skb: buffer to check
  792. *
  793. * Returns true if the buffer was generated with skb_clone() and is
  794. * one of multiple shared copies of the buffer. Cloned buffers are
  795. * shared data so must not be written to under normal circumstances.
  796. */
  797. static inline int skb_cloned(const struct sk_buff *skb)
  798. {
  799. return skb->cloned &&
  800. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  801. }
  802. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  803. {
  804. might_sleep_if(pri & __GFP_WAIT);
  805. if (skb_cloned(skb))
  806. return pskb_expand_head(skb, 0, 0, pri);
  807. return 0;
  808. }
  809. /**
  810. * skb_header_cloned - is the header a clone
  811. * @skb: buffer to check
  812. *
  813. * Returns true if modifying the header part of the buffer requires
  814. * the data to be copied.
  815. */
  816. static inline int skb_header_cloned(const struct sk_buff *skb)
  817. {
  818. int dataref;
  819. if (!skb->cloned)
  820. return 0;
  821. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  822. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  823. return dataref != 1;
  824. }
  825. /**
  826. * skb_header_release - release reference to header
  827. * @skb: buffer to operate on
  828. *
  829. * Drop a reference to the header part of the buffer. This is done
  830. * by acquiring a payload reference. You must not read from the header
  831. * part of skb->data after this.
  832. */
  833. static inline void skb_header_release(struct sk_buff *skb)
  834. {
  835. BUG_ON(skb->nohdr);
  836. skb->nohdr = 1;
  837. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  838. }
  839. /**
  840. * skb_shared - is the buffer shared
  841. * @skb: buffer to check
  842. *
  843. * Returns true if more than one person has a reference to this
  844. * buffer.
  845. */
  846. static inline int skb_shared(const struct sk_buff *skb)
  847. {
  848. return atomic_read(&skb->users) != 1;
  849. }
  850. /**
  851. * skb_share_check - check if buffer is shared and if so clone it
  852. * @skb: buffer to check
  853. * @pri: priority for memory allocation
  854. *
  855. * If the buffer is shared the buffer is cloned and the old copy
  856. * drops a reference. A new clone with a single reference is returned.
  857. * If the buffer is not shared the original buffer is returned. When
  858. * being called from interrupt status or with spinlocks held pri must
  859. * be GFP_ATOMIC.
  860. *
  861. * NULL is returned on a memory allocation failure.
  862. */
  863. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  864. {
  865. might_sleep_if(pri & __GFP_WAIT);
  866. if (skb_shared(skb)) {
  867. struct sk_buff *nskb = skb_clone(skb, pri);
  868. if (likely(nskb))
  869. consume_skb(skb);
  870. else
  871. kfree_skb(skb);
  872. skb = nskb;
  873. }
  874. return skb;
  875. }
  876. /*
  877. * Copy shared buffers into a new sk_buff. We effectively do COW on
  878. * packets to handle cases where we have a local reader and forward
  879. * and a couple of other messy ones. The normal one is tcpdumping
  880. * a packet thats being forwarded.
  881. */
  882. /**
  883. * skb_unshare - make a copy of a shared buffer
  884. * @skb: buffer to check
  885. * @pri: priority for memory allocation
  886. *
  887. * If the socket buffer is a clone then this function creates a new
  888. * copy of the data, drops a reference count on the old copy and returns
  889. * the new copy with the reference count at 1. If the buffer is not a clone
  890. * the original buffer is returned. When called with a spinlock held or
  891. * from interrupt state @pri must be %GFP_ATOMIC
  892. *
  893. * %NULL is returned on a memory allocation failure.
  894. */
  895. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  896. gfp_t pri)
  897. {
  898. might_sleep_if(pri & __GFP_WAIT);
  899. if (skb_cloned(skb)) {
  900. struct sk_buff *nskb = skb_copy(skb, pri);
  901. kfree_skb(skb); /* Free our shared copy */
  902. skb = nskb;
  903. }
  904. return skb;
  905. }
  906. /**
  907. * skb_peek - peek at the head of an &sk_buff_head
  908. * @list_: list to peek at
  909. *
  910. * Peek an &sk_buff. Unlike most other operations you _MUST_
  911. * be careful with this one. A peek leaves the buffer on the
  912. * list and someone else may run off with it. You must hold
  913. * the appropriate locks or have a private queue to do this.
  914. *
  915. * Returns %NULL for an empty list or a pointer to the head element.
  916. * The reference count is not incremented and the reference is therefore
  917. * volatile. Use with caution.
  918. */
  919. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  920. {
  921. struct sk_buff *skb = list_->next;
  922. if (skb == (struct sk_buff *)list_)
  923. skb = NULL;
  924. return skb;
  925. }
  926. /**
  927. * skb_peek_next - peek skb following the given one from a queue
  928. * @skb: skb to start from
  929. * @list_: list to peek at
  930. *
  931. * Returns %NULL when the end of the list is met or a pointer to the
  932. * next element. The reference count is not incremented and the
  933. * reference is therefore volatile. Use with caution.
  934. */
  935. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  936. const struct sk_buff_head *list_)
  937. {
  938. struct sk_buff *next = skb->next;
  939. if (next == (struct sk_buff *)list_)
  940. next = NULL;
  941. return next;
  942. }
  943. /**
  944. * skb_peek_tail - peek at the tail of an &sk_buff_head
  945. * @list_: list to peek at
  946. *
  947. * Peek an &sk_buff. Unlike most other operations you _MUST_
  948. * be careful with this one. A peek leaves the buffer on the
  949. * list and someone else may run off with it. You must hold
  950. * the appropriate locks or have a private queue to do this.
  951. *
  952. * Returns %NULL for an empty list or a pointer to the tail element.
  953. * The reference count is not incremented and the reference is therefore
  954. * volatile. Use with caution.
  955. */
  956. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  957. {
  958. struct sk_buff *skb = list_->prev;
  959. if (skb == (struct sk_buff *)list_)
  960. skb = NULL;
  961. return skb;
  962. }
  963. /**
  964. * skb_queue_len - get queue length
  965. * @list_: list to measure
  966. *
  967. * Return the length of an &sk_buff queue.
  968. */
  969. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  970. {
  971. return list_->qlen;
  972. }
  973. /**
  974. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  975. * @list: queue to initialize
  976. *
  977. * This initializes only the list and queue length aspects of
  978. * an sk_buff_head object. This allows to initialize the list
  979. * aspects of an sk_buff_head without reinitializing things like
  980. * the spinlock. It can also be used for on-stack sk_buff_head
  981. * objects where the spinlock is known to not be used.
  982. */
  983. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  984. {
  985. list->prev = list->next = (struct sk_buff *)list;
  986. list->qlen = 0;
  987. }
  988. /*
  989. * This function creates a split out lock class for each invocation;
  990. * this is needed for now since a whole lot of users of the skb-queue
  991. * infrastructure in drivers have different locking usage (in hardirq)
  992. * than the networking core (in softirq only). In the long run either the
  993. * network layer or drivers should need annotation to consolidate the
  994. * main types of usage into 3 classes.
  995. */
  996. static inline void skb_queue_head_init(struct sk_buff_head *list)
  997. {
  998. spin_lock_init(&list->lock);
  999. __skb_queue_head_init(list);
  1000. }
  1001. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1002. struct lock_class_key *class)
  1003. {
  1004. skb_queue_head_init(list);
  1005. lockdep_set_class(&list->lock, class);
  1006. }
  1007. /*
  1008. * Insert an sk_buff on a list.
  1009. *
  1010. * The "__skb_xxxx()" functions are the non-atomic ones that
  1011. * can only be called with interrupts disabled.
  1012. */
  1013. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1014. struct sk_buff_head *list);
  1015. static inline void __skb_insert(struct sk_buff *newsk,
  1016. struct sk_buff *prev, struct sk_buff *next,
  1017. struct sk_buff_head *list)
  1018. {
  1019. newsk->next = next;
  1020. newsk->prev = prev;
  1021. next->prev = prev->next = newsk;
  1022. list->qlen++;
  1023. }
  1024. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1025. struct sk_buff *prev,
  1026. struct sk_buff *next)
  1027. {
  1028. struct sk_buff *first = list->next;
  1029. struct sk_buff *last = list->prev;
  1030. first->prev = prev;
  1031. prev->next = first;
  1032. last->next = next;
  1033. next->prev = last;
  1034. }
  1035. /**
  1036. * skb_queue_splice - join two skb lists, this is designed for stacks
  1037. * @list: the new list to add
  1038. * @head: the place to add it in the first list
  1039. */
  1040. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1041. struct sk_buff_head *head)
  1042. {
  1043. if (!skb_queue_empty(list)) {
  1044. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1045. head->qlen += list->qlen;
  1046. }
  1047. }
  1048. /**
  1049. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1050. * @list: the new list to add
  1051. * @head: the place to add it in the first list
  1052. *
  1053. * The list at @list is reinitialised
  1054. */
  1055. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1056. struct sk_buff_head *head)
  1057. {
  1058. if (!skb_queue_empty(list)) {
  1059. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1060. head->qlen += list->qlen;
  1061. __skb_queue_head_init(list);
  1062. }
  1063. }
  1064. /**
  1065. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1066. * @list: the new list to add
  1067. * @head: the place to add it in the first list
  1068. */
  1069. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1070. struct sk_buff_head *head)
  1071. {
  1072. if (!skb_queue_empty(list)) {
  1073. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1074. head->qlen += list->qlen;
  1075. }
  1076. }
  1077. /**
  1078. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1079. * @list: the new list to add
  1080. * @head: the place to add it in the first list
  1081. *
  1082. * Each of the lists is a queue.
  1083. * The list at @list is reinitialised
  1084. */
  1085. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1086. struct sk_buff_head *head)
  1087. {
  1088. if (!skb_queue_empty(list)) {
  1089. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1090. head->qlen += list->qlen;
  1091. __skb_queue_head_init(list);
  1092. }
  1093. }
  1094. /**
  1095. * __skb_queue_after - queue a buffer at the list head
  1096. * @list: list to use
  1097. * @prev: place after this buffer
  1098. * @newsk: buffer to queue
  1099. *
  1100. * Queue a buffer int the middle of a list. This function takes no locks
  1101. * and you must therefore hold required locks before calling it.
  1102. *
  1103. * A buffer cannot be placed on two lists at the same time.
  1104. */
  1105. static inline void __skb_queue_after(struct sk_buff_head *list,
  1106. struct sk_buff *prev,
  1107. struct sk_buff *newsk)
  1108. {
  1109. __skb_insert(newsk, prev, prev->next, list);
  1110. }
  1111. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1112. struct sk_buff_head *list);
  1113. static inline void __skb_queue_before(struct sk_buff_head *list,
  1114. struct sk_buff *next,
  1115. struct sk_buff *newsk)
  1116. {
  1117. __skb_insert(newsk, next->prev, next, list);
  1118. }
  1119. /**
  1120. * __skb_queue_head - queue a buffer at the list head
  1121. * @list: list to use
  1122. * @newsk: buffer to queue
  1123. *
  1124. * Queue a buffer at the start of a list. This function takes no locks
  1125. * and you must therefore hold required locks before calling it.
  1126. *
  1127. * A buffer cannot be placed on two lists at the same time.
  1128. */
  1129. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1130. static inline void __skb_queue_head(struct sk_buff_head *list,
  1131. struct sk_buff *newsk)
  1132. {
  1133. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1134. }
  1135. /**
  1136. * __skb_queue_tail - queue a buffer at the list tail
  1137. * @list: list to use
  1138. * @newsk: buffer to queue
  1139. *
  1140. * Queue a buffer at the end of a list. This function takes no locks
  1141. * and you must therefore hold required locks before calling it.
  1142. *
  1143. * A buffer cannot be placed on two lists at the same time.
  1144. */
  1145. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1146. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1147. struct sk_buff *newsk)
  1148. {
  1149. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1150. }
  1151. /*
  1152. * remove sk_buff from list. _Must_ be called atomically, and with
  1153. * the list known..
  1154. */
  1155. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1156. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1157. {
  1158. struct sk_buff *next, *prev;
  1159. list->qlen--;
  1160. next = skb->next;
  1161. prev = skb->prev;
  1162. skb->next = skb->prev = NULL;
  1163. next->prev = prev;
  1164. prev->next = next;
  1165. }
  1166. /**
  1167. * __skb_dequeue - remove from the head of the queue
  1168. * @list: list to dequeue from
  1169. *
  1170. * Remove the head of the list. This function does not take any locks
  1171. * so must be used with appropriate locks held only. The head item is
  1172. * returned or %NULL if the list is empty.
  1173. */
  1174. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1175. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1176. {
  1177. struct sk_buff *skb = skb_peek(list);
  1178. if (skb)
  1179. __skb_unlink(skb, list);
  1180. return skb;
  1181. }
  1182. /**
  1183. * __skb_dequeue_tail - remove from the tail of the queue
  1184. * @list: list to dequeue from
  1185. *
  1186. * Remove the tail of the list. This function does not take any locks
  1187. * so must be used with appropriate locks held only. The tail item is
  1188. * returned or %NULL if the list is empty.
  1189. */
  1190. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1191. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1192. {
  1193. struct sk_buff *skb = skb_peek_tail(list);
  1194. if (skb)
  1195. __skb_unlink(skb, list);
  1196. return skb;
  1197. }
  1198. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1199. {
  1200. return skb->data_len;
  1201. }
  1202. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1203. {
  1204. return skb->len - skb->data_len;
  1205. }
  1206. static inline int skb_pagelen(const struct sk_buff *skb)
  1207. {
  1208. int i, len = 0;
  1209. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1210. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1211. return len + skb_headlen(skb);
  1212. }
  1213. /**
  1214. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1215. * @skb: buffer containing fragment to be initialised
  1216. * @i: paged fragment index to initialise
  1217. * @page: the page to use for this fragment
  1218. * @off: the offset to the data with @page
  1219. * @size: the length of the data
  1220. *
  1221. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1222. * offset @off within @page.
  1223. *
  1224. * Does not take any additional reference on the fragment.
  1225. */
  1226. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1227. struct page *page, int off, int size)
  1228. {
  1229. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1230. /*
  1231. * Propagate page->pfmemalloc to the skb if we can. The problem is
  1232. * that not all callers have unique ownership of the page. If
  1233. * pfmemalloc is set, we check the mapping as a mapping implies
  1234. * page->index is set (index and pfmemalloc share space).
  1235. * If it's a valid mapping, we cannot use page->pfmemalloc but we
  1236. * do not lose pfmemalloc information as the pages would not be
  1237. * allocated using __GFP_MEMALLOC.
  1238. */
  1239. frag->page.p = page;
  1240. frag->page_offset = off;
  1241. skb_frag_size_set(frag, size);
  1242. page = compound_head(page);
  1243. if (page->pfmemalloc && !page->mapping)
  1244. skb->pfmemalloc = true;
  1245. }
  1246. /**
  1247. * skb_fill_page_desc - initialise a paged fragment in an skb
  1248. * @skb: buffer containing fragment to be initialised
  1249. * @i: paged fragment index to initialise
  1250. * @page: the page to use for this fragment
  1251. * @off: the offset to the data with @page
  1252. * @size: the length of the data
  1253. *
  1254. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1255. * @skb to point to @size bytes at offset @off within @page. In
  1256. * addition updates @skb such that @i is the last fragment.
  1257. *
  1258. * Does not take any additional reference on the fragment.
  1259. */
  1260. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1261. struct page *page, int off, int size)
  1262. {
  1263. __skb_fill_page_desc(skb, i, page, off, size);
  1264. skb_shinfo(skb)->nr_frags = i + 1;
  1265. }
  1266. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1267. int size, unsigned int truesize);
  1268. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1269. unsigned int truesize);
  1270. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1271. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1272. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1273. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1274. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1275. {
  1276. return skb->head + skb->tail;
  1277. }
  1278. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1279. {
  1280. skb->tail = skb->data - skb->head;
  1281. }
  1282. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1283. {
  1284. skb_reset_tail_pointer(skb);
  1285. skb->tail += offset;
  1286. }
  1287. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1288. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1289. {
  1290. return skb->tail;
  1291. }
  1292. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1293. {
  1294. skb->tail = skb->data;
  1295. }
  1296. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1297. {
  1298. skb->tail = skb->data + offset;
  1299. }
  1300. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1301. /*
  1302. * Add data to an sk_buff
  1303. */
  1304. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1305. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1306. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1307. {
  1308. unsigned char *tmp = skb_tail_pointer(skb);
  1309. SKB_LINEAR_ASSERT(skb);
  1310. skb->tail += len;
  1311. skb->len += len;
  1312. return tmp;
  1313. }
  1314. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1315. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1316. {
  1317. skb->data -= len;
  1318. skb->len += len;
  1319. return skb->data;
  1320. }
  1321. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1322. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1323. {
  1324. skb->len -= len;
  1325. BUG_ON(skb->len < skb->data_len);
  1326. return skb->data += len;
  1327. }
  1328. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1329. {
  1330. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1331. }
  1332. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1333. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1334. {
  1335. if (len > skb_headlen(skb) &&
  1336. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1337. return NULL;
  1338. skb->len -= len;
  1339. return skb->data += len;
  1340. }
  1341. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1342. {
  1343. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1344. }
  1345. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1346. {
  1347. if (likely(len <= skb_headlen(skb)))
  1348. return 1;
  1349. if (unlikely(len > skb->len))
  1350. return 0;
  1351. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1352. }
  1353. /**
  1354. * skb_headroom - bytes at buffer head
  1355. * @skb: buffer to check
  1356. *
  1357. * Return the number of bytes of free space at the head of an &sk_buff.
  1358. */
  1359. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1360. {
  1361. return skb->data - skb->head;
  1362. }
  1363. /**
  1364. * skb_tailroom - bytes at buffer end
  1365. * @skb: buffer to check
  1366. *
  1367. * Return the number of bytes of free space at the tail of an sk_buff
  1368. */
  1369. static inline int skb_tailroom(const struct sk_buff *skb)
  1370. {
  1371. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1372. }
  1373. /**
  1374. * skb_availroom - bytes at buffer end
  1375. * @skb: buffer to check
  1376. *
  1377. * Return the number of bytes of free space at the tail of an sk_buff
  1378. * allocated by sk_stream_alloc()
  1379. */
  1380. static inline int skb_availroom(const struct sk_buff *skb)
  1381. {
  1382. if (skb_is_nonlinear(skb))
  1383. return 0;
  1384. return skb->end - skb->tail - skb->reserved_tailroom;
  1385. }
  1386. /**
  1387. * skb_reserve - adjust headroom
  1388. * @skb: buffer to alter
  1389. * @len: bytes to move
  1390. *
  1391. * Increase the headroom of an empty &sk_buff by reducing the tail
  1392. * room. This is only allowed for an empty buffer.
  1393. */
  1394. static inline void skb_reserve(struct sk_buff *skb, int len)
  1395. {
  1396. skb->data += len;
  1397. skb->tail += len;
  1398. }
  1399. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1400. {
  1401. skb->inner_mac_header = skb->mac_header;
  1402. skb->inner_network_header = skb->network_header;
  1403. skb->inner_transport_header = skb->transport_header;
  1404. }
  1405. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1406. {
  1407. skb->mac_len = skb->network_header - skb->mac_header;
  1408. }
  1409. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1410. *skb)
  1411. {
  1412. return skb->head + skb->inner_transport_header;
  1413. }
  1414. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1415. {
  1416. skb->inner_transport_header = skb->data - skb->head;
  1417. }
  1418. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1419. const int offset)
  1420. {
  1421. skb_reset_inner_transport_header(skb);
  1422. skb->inner_transport_header += offset;
  1423. }
  1424. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1425. {
  1426. return skb->head + skb->inner_network_header;
  1427. }
  1428. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1429. {
  1430. skb->inner_network_header = skb->data - skb->head;
  1431. }
  1432. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1433. const int offset)
  1434. {
  1435. skb_reset_inner_network_header(skb);
  1436. skb->inner_network_header += offset;
  1437. }
  1438. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1439. {
  1440. return skb->head + skb->inner_mac_header;
  1441. }
  1442. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1443. {
  1444. skb->inner_mac_header = skb->data - skb->head;
  1445. }
  1446. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1447. const int offset)
  1448. {
  1449. skb_reset_inner_mac_header(skb);
  1450. skb->inner_mac_header += offset;
  1451. }
  1452. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1453. {
  1454. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1455. }
  1456. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1457. {
  1458. return skb->head + skb->transport_header;
  1459. }
  1460. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1461. {
  1462. skb->transport_header = skb->data - skb->head;
  1463. }
  1464. static inline void skb_set_transport_header(struct sk_buff *skb,
  1465. const int offset)
  1466. {
  1467. skb_reset_transport_header(skb);
  1468. skb->transport_header += offset;
  1469. }
  1470. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1471. {
  1472. return skb->head + skb->network_header;
  1473. }
  1474. static inline void skb_reset_network_header(struct sk_buff *skb)
  1475. {
  1476. skb->network_header = skb->data - skb->head;
  1477. }
  1478. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1479. {
  1480. skb_reset_network_header(skb);
  1481. skb->network_header += offset;
  1482. }
  1483. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1484. {
  1485. return skb->head + skb->mac_header;
  1486. }
  1487. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1488. {
  1489. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1490. }
  1491. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1492. {
  1493. skb->mac_header = skb->data - skb->head;
  1494. }
  1495. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1496. {
  1497. skb_reset_mac_header(skb);
  1498. skb->mac_header += offset;
  1499. }
  1500. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1501. const int offset_hint)
  1502. {
  1503. struct flow_keys keys;
  1504. if (skb_transport_header_was_set(skb))
  1505. return;
  1506. else if (skb_flow_dissect(skb, &keys))
  1507. skb_set_transport_header(skb, keys.thoff);
  1508. else
  1509. skb_set_transport_header(skb, offset_hint);
  1510. }
  1511. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1512. {
  1513. if (skb_mac_header_was_set(skb)) {
  1514. const unsigned char *old_mac = skb_mac_header(skb);
  1515. skb_set_mac_header(skb, -skb->mac_len);
  1516. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1517. }
  1518. }
  1519. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1520. {
  1521. return skb->csum_start - skb_headroom(skb);
  1522. }
  1523. static inline int skb_transport_offset(const struct sk_buff *skb)
  1524. {
  1525. return skb_transport_header(skb) - skb->data;
  1526. }
  1527. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1528. {
  1529. return skb->transport_header - skb->network_header;
  1530. }
  1531. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1532. {
  1533. return skb->inner_transport_header - skb->inner_network_header;
  1534. }
  1535. static inline int skb_network_offset(const struct sk_buff *skb)
  1536. {
  1537. return skb_network_header(skb) - skb->data;
  1538. }
  1539. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1540. {
  1541. return skb_inner_network_header(skb) - skb->data;
  1542. }
  1543. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1544. {
  1545. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1546. }
  1547. /*
  1548. * CPUs often take a performance hit when accessing unaligned memory
  1549. * locations. The actual performance hit varies, it can be small if the
  1550. * hardware handles it or large if we have to take an exception and fix it
  1551. * in software.
  1552. *
  1553. * Since an ethernet header is 14 bytes network drivers often end up with
  1554. * the IP header at an unaligned offset. The IP header can be aligned by
  1555. * shifting the start of the packet by 2 bytes. Drivers should do this
  1556. * with:
  1557. *
  1558. * skb_reserve(skb, NET_IP_ALIGN);
  1559. *
  1560. * The downside to this alignment of the IP header is that the DMA is now
  1561. * unaligned. On some architectures the cost of an unaligned DMA is high
  1562. * and this cost outweighs the gains made by aligning the IP header.
  1563. *
  1564. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1565. * to be overridden.
  1566. */
  1567. #ifndef NET_IP_ALIGN
  1568. #define NET_IP_ALIGN 2
  1569. #endif
  1570. /*
  1571. * The networking layer reserves some headroom in skb data (via
  1572. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1573. * the header has to grow. In the default case, if the header has to grow
  1574. * 32 bytes or less we avoid the reallocation.
  1575. *
  1576. * Unfortunately this headroom changes the DMA alignment of the resulting
  1577. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1578. * on some architectures. An architecture can override this value,
  1579. * perhaps setting it to a cacheline in size (since that will maintain
  1580. * cacheline alignment of the DMA). It must be a power of 2.
  1581. *
  1582. * Various parts of the networking layer expect at least 32 bytes of
  1583. * headroom, you should not reduce this.
  1584. *
  1585. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1586. * to reduce average number of cache lines per packet.
  1587. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1588. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1589. */
  1590. #ifndef NET_SKB_PAD
  1591. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1592. #endif
  1593. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1594. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1595. {
  1596. if (unlikely(skb_is_nonlinear(skb))) {
  1597. WARN_ON(1);
  1598. return;
  1599. }
  1600. skb->len = len;
  1601. skb_set_tail_pointer(skb, len);
  1602. }
  1603. void skb_trim(struct sk_buff *skb, unsigned int len);
  1604. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1605. {
  1606. if (skb->data_len)
  1607. return ___pskb_trim(skb, len);
  1608. __skb_trim(skb, len);
  1609. return 0;
  1610. }
  1611. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1612. {
  1613. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1614. }
  1615. /**
  1616. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1617. * @skb: buffer to alter
  1618. * @len: new length
  1619. *
  1620. * This is identical to pskb_trim except that the caller knows that
  1621. * the skb is not cloned so we should never get an error due to out-
  1622. * of-memory.
  1623. */
  1624. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  1625. {
  1626. int err = pskb_trim(skb, len);
  1627. BUG_ON(err);
  1628. }
  1629. /**
  1630. * skb_orphan - orphan a buffer
  1631. * @skb: buffer to orphan
  1632. *
  1633. * If a buffer currently has an owner then we call the owner's
  1634. * destructor function and make the @skb unowned. The buffer continues
  1635. * to exist but is no longer charged to its former owner.
  1636. */
  1637. static inline void skb_orphan(struct sk_buff *skb)
  1638. {
  1639. if (skb->destructor) {
  1640. skb->destructor(skb);
  1641. skb->destructor = NULL;
  1642. skb->sk = NULL;
  1643. } else {
  1644. BUG_ON(skb->sk);
  1645. }
  1646. }
  1647. /**
  1648. * skb_orphan_frags - orphan the frags contained in a buffer
  1649. * @skb: buffer to orphan frags from
  1650. * @gfp_mask: allocation mask for replacement pages
  1651. *
  1652. * For each frag in the SKB which needs a destructor (i.e. has an
  1653. * owner) create a copy of that frag and release the original
  1654. * page by calling the destructor.
  1655. */
  1656. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  1657. {
  1658. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  1659. return 0;
  1660. return skb_copy_ubufs(skb, gfp_mask);
  1661. }
  1662. /**
  1663. * __skb_queue_purge - empty a list
  1664. * @list: list to empty
  1665. *
  1666. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  1667. * the list and one reference dropped. This function does not take the
  1668. * list lock and the caller must hold the relevant locks to use it.
  1669. */
  1670. void skb_queue_purge(struct sk_buff_head *list);
  1671. static inline void __skb_queue_purge(struct sk_buff_head *list)
  1672. {
  1673. struct sk_buff *skb;
  1674. while ((skb = __skb_dequeue(list)) != NULL)
  1675. kfree_skb(skb);
  1676. }
  1677. #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
  1678. #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
  1679. #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
  1680. void *netdev_alloc_frag(unsigned int fragsz);
  1681. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  1682. gfp_t gfp_mask);
  1683. /**
  1684. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  1685. * @dev: network device to receive on
  1686. * @length: length to allocate
  1687. *
  1688. * Allocate a new &sk_buff and assign it a usage count of one. The
  1689. * buffer has unspecified headroom built in. Users should allocate
  1690. * the headroom they think they need without accounting for the
  1691. * built in space. The built in space is used for optimisations.
  1692. *
  1693. * %NULL is returned if there is no free memory. Although this function
  1694. * allocates memory it can be called from an interrupt.
  1695. */
  1696. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  1697. unsigned int length)
  1698. {
  1699. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  1700. }
  1701. /* legacy helper around __netdev_alloc_skb() */
  1702. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  1703. gfp_t gfp_mask)
  1704. {
  1705. return __netdev_alloc_skb(NULL, length, gfp_mask);
  1706. }
  1707. /* legacy helper around netdev_alloc_skb() */
  1708. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  1709. {
  1710. return netdev_alloc_skb(NULL, length);
  1711. }
  1712. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  1713. unsigned int length, gfp_t gfp)
  1714. {
  1715. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  1716. if (NET_IP_ALIGN && skb)
  1717. skb_reserve(skb, NET_IP_ALIGN);
  1718. return skb;
  1719. }
  1720. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  1721. unsigned int length)
  1722. {
  1723. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  1724. }
  1725. /**
  1726. * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
  1727. * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
  1728. * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
  1729. * @order: size of the allocation
  1730. *
  1731. * Allocate a new page.
  1732. *
  1733. * %NULL is returned if there is no free memory.
  1734. */
  1735. static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
  1736. struct sk_buff *skb,
  1737. unsigned int order)
  1738. {
  1739. struct page *page;
  1740. gfp_mask |= __GFP_COLD;
  1741. if (!(gfp_mask & __GFP_NOMEMALLOC))
  1742. gfp_mask |= __GFP_MEMALLOC;
  1743. page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  1744. if (skb && page && page->pfmemalloc)
  1745. skb->pfmemalloc = true;
  1746. return page;
  1747. }
  1748. /**
  1749. * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
  1750. * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
  1751. * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
  1752. *
  1753. * Allocate a new page.
  1754. *
  1755. * %NULL is returned if there is no free memory.
  1756. */
  1757. static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
  1758. struct sk_buff *skb)
  1759. {
  1760. return __skb_alloc_pages(gfp_mask, skb, 0);
  1761. }
  1762. /**
  1763. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  1764. * @page: The page that was allocated from skb_alloc_page
  1765. * @skb: The skb that may need pfmemalloc set
  1766. */
  1767. static inline void skb_propagate_pfmemalloc(struct page *page,
  1768. struct sk_buff *skb)
  1769. {
  1770. if (page && page->pfmemalloc)
  1771. skb->pfmemalloc = true;
  1772. }
  1773. /**
  1774. * skb_frag_page - retrieve the page refered to by a paged fragment
  1775. * @frag: the paged fragment
  1776. *
  1777. * Returns the &struct page associated with @frag.
  1778. */
  1779. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  1780. {
  1781. return frag->page.p;
  1782. }
  1783. /**
  1784. * __skb_frag_ref - take an addition reference on a paged fragment.
  1785. * @frag: the paged fragment
  1786. *
  1787. * Takes an additional reference on the paged fragment @frag.
  1788. */
  1789. static inline void __skb_frag_ref(skb_frag_t *frag)
  1790. {
  1791. get_page(skb_frag_page(frag));
  1792. }
  1793. /**
  1794. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  1795. * @skb: the buffer
  1796. * @f: the fragment offset.
  1797. *
  1798. * Takes an additional reference on the @f'th paged fragment of @skb.
  1799. */
  1800. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  1801. {
  1802. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  1803. }
  1804. /**
  1805. * __skb_frag_unref - release a reference on a paged fragment.
  1806. * @frag: the paged fragment
  1807. *
  1808. * Releases a reference on the paged fragment @frag.
  1809. */
  1810. static inline void __skb_frag_unref(skb_frag_t *frag)
  1811. {
  1812. put_page(skb_frag_page(frag));
  1813. }
  1814. /**
  1815. * skb_frag_unref - release a reference on a paged fragment of an skb.
  1816. * @skb: the buffer
  1817. * @f: the fragment offset
  1818. *
  1819. * Releases a reference on the @f'th paged fragment of @skb.
  1820. */
  1821. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  1822. {
  1823. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  1824. }
  1825. /**
  1826. * skb_frag_address - gets the address of the data contained in a paged fragment
  1827. * @frag: the paged fragment buffer
  1828. *
  1829. * Returns the address of the data within @frag. The page must already
  1830. * be mapped.
  1831. */
  1832. static inline void *skb_frag_address(const skb_frag_t *frag)
  1833. {
  1834. return page_address(skb_frag_page(frag)) + frag->page_offset;
  1835. }
  1836. /**
  1837. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  1838. * @frag: the paged fragment buffer
  1839. *
  1840. * Returns the address of the data within @frag. Checks that the page
  1841. * is mapped and returns %NULL otherwise.
  1842. */
  1843. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  1844. {
  1845. void *ptr = page_address(skb_frag_page(frag));
  1846. if (unlikely(!ptr))
  1847. return NULL;
  1848. return ptr + frag->page_offset;
  1849. }
  1850. /**
  1851. * __skb_frag_set_page - sets the page contained in a paged fragment
  1852. * @frag: the paged fragment
  1853. * @page: the page to set
  1854. *
  1855. * Sets the fragment @frag to contain @page.
  1856. */
  1857. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  1858. {
  1859. frag->page.p = page;
  1860. }
  1861. /**
  1862. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  1863. * @skb: the buffer
  1864. * @f: the fragment offset
  1865. * @page: the page to set
  1866. *
  1867. * Sets the @f'th fragment of @skb to contain @page.
  1868. */
  1869. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  1870. struct page *page)
  1871. {
  1872. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  1873. }
  1874. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  1875. /**
  1876. * skb_frag_dma_map - maps a paged fragment via the DMA API
  1877. * @dev: the device to map the fragment to
  1878. * @frag: the paged fragment to map
  1879. * @offset: the offset within the fragment (starting at the
  1880. * fragment's own offset)
  1881. * @size: the number of bytes to map
  1882. * @dir: the direction of the mapping (%PCI_DMA_*)
  1883. *
  1884. * Maps the page associated with @frag to @device.
  1885. */
  1886. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  1887. const skb_frag_t *frag,
  1888. size_t offset, size_t size,
  1889. enum dma_data_direction dir)
  1890. {
  1891. return dma_map_page(dev, skb_frag_page(frag),
  1892. frag->page_offset + offset, size, dir);
  1893. }
  1894. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  1895. gfp_t gfp_mask)
  1896. {
  1897. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  1898. }
  1899. /**
  1900. * skb_clone_writable - is the header of a clone writable
  1901. * @skb: buffer to check
  1902. * @len: length up to which to write
  1903. *
  1904. * Returns true if modifying the header part of the cloned buffer
  1905. * does not requires the data to be copied.
  1906. */
  1907. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  1908. {
  1909. return !skb_header_cloned(skb) &&
  1910. skb_headroom(skb) + len <= skb->hdr_len;
  1911. }
  1912. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  1913. int cloned)
  1914. {
  1915. int delta = 0;
  1916. if (headroom > skb_headroom(skb))
  1917. delta = headroom - skb_headroom(skb);
  1918. if (delta || cloned)
  1919. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  1920. GFP_ATOMIC);
  1921. return 0;
  1922. }
  1923. /**
  1924. * skb_cow - copy header of skb when it is required
  1925. * @skb: buffer to cow
  1926. * @headroom: needed headroom
  1927. *
  1928. * If the skb passed lacks sufficient headroom or its data part
  1929. * is shared, data is reallocated. If reallocation fails, an error
  1930. * is returned and original skb is not changed.
  1931. *
  1932. * The result is skb with writable area skb->head...skb->tail
  1933. * and at least @headroom of space at head.
  1934. */
  1935. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  1936. {
  1937. return __skb_cow(skb, headroom, skb_cloned(skb));
  1938. }
  1939. /**
  1940. * skb_cow_head - skb_cow but only making the head writable
  1941. * @skb: buffer to cow
  1942. * @headroom: needed headroom
  1943. *
  1944. * This function is identical to skb_cow except that we replace the
  1945. * skb_cloned check by skb_header_cloned. It should be used when
  1946. * you only need to push on some header and do not need to modify
  1947. * the data.
  1948. */
  1949. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  1950. {
  1951. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  1952. }
  1953. /**
  1954. * skb_padto - pad an skbuff up to a minimal size
  1955. * @skb: buffer to pad
  1956. * @len: minimal length
  1957. *
  1958. * Pads up a buffer to ensure the trailing bytes exist and are
  1959. * blanked. If the buffer already contains sufficient data it
  1960. * is untouched. Otherwise it is extended. Returns zero on
  1961. * success. The skb is freed on error.
  1962. */
  1963. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  1964. {
  1965. unsigned int size = skb->len;
  1966. if (likely(size >= len))
  1967. return 0;
  1968. return skb_pad(skb, len - size);
  1969. }
  1970. static inline int skb_add_data(struct sk_buff *skb,
  1971. char __user *from, int copy)
  1972. {
  1973. const int off = skb->len;
  1974. if (skb->ip_summed == CHECKSUM_NONE) {
  1975. int err = 0;
  1976. __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
  1977. copy, 0, &err);
  1978. if (!err) {
  1979. skb->csum = csum_block_add(skb->csum, csum, off);
  1980. return 0;
  1981. }
  1982. } else if (!copy_from_user(skb_put(skb, copy), from, copy))
  1983. return 0;
  1984. __skb_trim(skb, off);
  1985. return -EFAULT;
  1986. }
  1987. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  1988. const struct page *page, int off)
  1989. {
  1990. if (i) {
  1991. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  1992. return page == skb_frag_page(frag) &&
  1993. off == frag->page_offset + skb_frag_size(frag);
  1994. }
  1995. return false;
  1996. }
  1997. static inline int __skb_linearize(struct sk_buff *skb)
  1998. {
  1999. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2000. }
  2001. /**
  2002. * skb_linearize - convert paged skb to linear one
  2003. * @skb: buffer to linarize
  2004. *
  2005. * If there is no free memory -ENOMEM is returned, otherwise zero
  2006. * is returned and the old skb data released.
  2007. */
  2008. static inline int skb_linearize(struct sk_buff *skb)
  2009. {
  2010. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2011. }
  2012. /**
  2013. * skb_has_shared_frag - can any frag be overwritten
  2014. * @skb: buffer to test
  2015. *
  2016. * Return true if the skb has at least one frag that might be modified
  2017. * by an external entity (as in vmsplice()/sendfile())
  2018. */
  2019. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2020. {
  2021. return skb_is_nonlinear(skb) &&
  2022. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2023. }
  2024. /**
  2025. * skb_linearize_cow - make sure skb is linear and writable
  2026. * @skb: buffer to process
  2027. *
  2028. * If there is no free memory -ENOMEM is returned, otherwise zero
  2029. * is returned and the old skb data released.
  2030. */
  2031. static inline int skb_linearize_cow(struct sk_buff *skb)
  2032. {
  2033. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2034. __skb_linearize(skb) : 0;
  2035. }
  2036. /**
  2037. * skb_postpull_rcsum - update checksum for received skb after pull
  2038. * @skb: buffer to update
  2039. * @start: start of data before pull
  2040. * @len: length of data pulled
  2041. *
  2042. * After doing a pull on a received packet, you need to call this to
  2043. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2044. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2045. */
  2046. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2047. const void *start, unsigned int len)
  2048. {
  2049. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2050. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2051. }
  2052. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2053. /**
  2054. * pskb_trim_rcsum - trim received skb and update checksum
  2055. * @skb: buffer to trim
  2056. * @len: new length
  2057. *
  2058. * This is exactly the same as pskb_trim except that it ensures the
  2059. * checksum of received packets are still valid after the operation.
  2060. */
  2061. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2062. {
  2063. if (likely(len >= skb->len))
  2064. return 0;
  2065. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2066. skb->ip_summed = CHECKSUM_NONE;
  2067. return __pskb_trim(skb, len);
  2068. }
  2069. #define skb_queue_walk(queue, skb) \
  2070. for (skb = (queue)->next; \
  2071. skb != (struct sk_buff *)(queue); \
  2072. skb = skb->next)
  2073. #define skb_queue_walk_safe(queue, skb, tmp) \
  2074. for (skb = (queue)->next, tmp = skb->next; \
  2075. skb != (struct sk_buff *)(queue); \
  2076. skb = tmp, tmp = skb->next)
  2077. #define skb_queue_walk_from(queue, skb) \
  2078. for (; skb != (struct sk_buff *)(queue); \
  2079. skb = skb->next)
  2080. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2081. for (tmp = skb->next; \
  2082. skb != (struct sk_buff *)(queue); \
  2083. skb = tmp, tmp = skb->next)
  2084. #define skb_queue_reverse_walk(queue, skb) \
  2085. for (skb = (queue)->prev; \
  2086. skb != (struct sk_buff *)(queue); \
  2087. skb = skb->prev)
  2088. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2089. for (skb = (queue)->prev, tmp = skb->prev; \
  2090. skb != (struct sk_buff *)(queue); \
  2091. skb = tmp, tmp = skb->prev)
  2092. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2093. for (tmp = skb->prev; \
  2094. skb != (struct sk_buff *)(queue); \
  2095. skb = tmp, tmp = skb->prev)
  2096. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2097. {
  2098. return skb_shinfo(skb)->frag_list != NULL;
  2099. }
  2100. static inline void skb_frag_list_init(struct sk_buff *skb)
  2101. {
  2102. skb_shinfo(skb)->frag_list = NULL;
  2103. }
  2104. static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
  2105. {
  2106. frag->next = skb_shinfo(skb)->frag_list;
  2107. skb_shinfo(skb)->frag_list = frag;
  2108. }
  2109. #define skb_walk_frags(skb, iter) \
  2110. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2111. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2112. int *peeked, int *off, int *err);
  2113. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2114. int *err);
  2115. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2116. struct poll_table_struct *wait);
  2117. int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
  2118. struct iovec *to, int size);
  2119. int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
  2120. struct iovec *iov);
  2121. int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
  2122. const struct iovec *from, int from_offset,
  2123. int len);
  2124. int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
  2125. int offset, size_t count);
  2126. int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
  2127. const struct iovec *to, int to_offset,
  2128. int size);
  2129. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2130. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2131. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2132. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2133. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2134. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2135. int len, __wsum csum);
  2136. int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
  2137. struct pipe_inode_info *pipe, unsigned int len,
  2138. unsigned int flags);
  2139. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2140. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2141. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2142. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2143. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2144. struct skb_checksum_ops {
  2145. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2146. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2147. };
  2148. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2149. __wsum csum, const struct skb_checksum_ops *ops);
  2150. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2151. __wsum csum);
  2152. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  2153. int len, void *buffer)
  2154. {
  2155. int hlen = skb_headlen(skb);
  2156. if (hlen - offset >= len)
  2157. return skb->data + offset;
  2158. if (skb_copy_bits(skb, offset, buffer, len) < 0)
  2159. return NULL;
  2160. return buffer;
  2161. }
  2162. /**
  2163. * skb_needs_linearize - check if we need to linearize a given skb
  2164. * depending on the given device features.
  2165. * @skb: socket buffer to check
  2166. * @features: net device features
  2167. *
  2168. * Returns true if either:
  2169. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2170. * 2. skb is fragmented and the device does not support SG.
  2171. */
  2172. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2173. netdev_features_t features)
  2174. {
  2175. return skb_is_nonlinear(skb) &&
  2176. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2177. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2178. }
  2179. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2180. void *to,
  2181. const unsigned int len)
  2182. {
  2183. memcpy(to, skb->data, len);
  2184. }
  2185. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2186. const int offset, void *to,
  2187. const unsigned int len)
  2188. {
  2189. memcpy(to, skb->data + offset, len);
  2190. }
  2191. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2192. const void *from,
  2193. const unsigned int len)
  2194. {
  2195. memcpy(skb->data, from, len);
  2196. }
  2197. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2198. const int offset,
  2199. const void *from,
  2200. const unsigned int len)
  2201. {
  2202. memcpy(skb->data + offset, from, len);
  2203. }
  2204. void skb_init(void);
  2205. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2206. {
  2207. return skb->tstamp;
  2208. }
  2209. /**
  2210. * skb_get_timestamp - get timestamp from a skb
  2211. * @skb: skb to get stamp from
  2212. * @stamp: pointer to struct timeval to store stamp in
  2213. *
  2214. * Timestamps are stored in the skb as offsets to a base timestamp.
  2215. * This function converts the offset back to a struct timeval and stores
  2216. * it in stamp.
  2217. */
  2218. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2219. struct timeval *stamp)
  2220. {
  2221. *stamp = ktime_to_timeval(skb->tstamp);
  2222. }
  2223. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2224. struct timespec *stamp)
  2225. {
  2226. *stamp = ktime_to_timespec(skb->tstamp);
  2227. }
  2228. static inline void __net_timestamp(struct sk_buff *skb)
  2229. {
  2230. skb->tstamp = ktime_get_real();
  2231. }
  2232. static inline ktime_t net_timedelta(ktime_t t)
  2233. {
  2234. return ktime_sub(ktime_get_real(), t);
  2235. }
  2236. static inline ktime_t net_invalid_timestamp(void)
  2237. {
  2238. return ktime_set(0, 0);
  2239. }
  2240. void skb_timestamping_init(void);
  2241. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2242. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2243. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2244. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2245. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2246. {
  2247. }
  2248. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2249. {
  2250. return false;
  2251. }
  2252. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2253. /**
  2254. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2255. *
  2256. * PHY drivers may accept clones of transmitted packets for
  2257. * timestamping via their phy_driver.txtstamp method. These drivers
  2258. * must call this function to return the skb back to the stack, with
  2259. * or without a timestamp.
  2260. *
  2261. * @skb: clone of the the original outgoing packet
  2262. * @hwtstamps: hardware time stamps, may be NULL if not available
  2263. *
  2264. */
  2265. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2266. struct skb_shared_hwtstamps *hwtstamps);
  2267. /**
  2268. * skb_tstamp_tx - queue clone of skb with send time stamps
  2269. * @orig_skb: the original outgoing packet
  2270. * @hwtstamps: hardware time stamps, may be NULL if not available
  2271. *
  2272. * If the skb has a socket associated, then this function clones the
  2273. * skb (thus sharing the actual data and optional structures), stores
  2274. * the optional hardware time stamping information (if non NULL) or
  2275. * generates a software time stamp (otherwise), then queues the clone
  2276. * to the error queue of the socket. Errors are silently ignored.
  2277. */
  2278. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2279. struct skb_shared_hwtstamps *hwtstamps);
  2280. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2281. {
  2282. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2283. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2284. skb_tstamp_tx(skb, NULL);
  2285. }
  2286. /**
  2287. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2288. *
  2289. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2290. * function immediately before giving the sk_buff to the MAC hardware.
  2291. *
  2292. * @skb: A socket buffer.
  2293. */
  2294. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2295. {
  2296. skb_clone_tx_timestamp(skb);
  2297. sw_tx_timestamp(skb);
  2298. }
  2299. /**
  2300. * skb_complete_wifi_ack - deliver skb with wifi status
  2301. *
  2302. * @skb: the original outgoing packet
  2303. * @acked: ack status
  2304. *
  2305. */
  2306. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2307. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2308. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2309. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2310. {
  2311. return skb->ip_summed & CHECKSUM_UNNECESSARY;
  2312. }
  2313. /**
  2314. * skb_checksum_complete - Calculate checksum of an entire packet
  2315. * @skb: packet to process
  2316. *
  2317. * This function calculates the checksum over the entire packet plus
  2318. * the value of skb->csum. The latter can be used to supply the
  2319. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2320. * checksum.
  2321. *
  2322. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2323. * this function can be used to verify that checksum on received
  2324. * packets. In that case the function should return zero if the
  2325. * checksum is correct. In particular, this function will return zero
  2326. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2327. * hardware has already verified the correctness of the checksum.
  2328. */
  2329. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2330. {
  2331. return skb_csum_unnecessary(skb) ?
  2332. 0 : __skb_checksum_complete(skb);
  2333. }
  2334. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2335. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2336. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2337. {
  2338. if (nfct && atomic_dec_and_test(&nfct->use))
  2339. nf_conntrack_destroy(nfct);
  2340. }
  2341. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2342. {
  2343. if (nfct)
  2344. atomic_inc(&nfct->use);
  2345. }
  2346. #endif
  2347. #ifdef CONFIG_BRIDGE_NETFILTER
  2348. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  2349. {
  2350. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  2351. kfree(nf_bridge);
  2352. }
  2353. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  2354. {
  2355. if (nf_bridge)
  2356. atomic_inc(&nf_bridge->use);
  2357. }
  2358. #endif /* CONFIG_BRIDGE_NETFILTER */
  2359. static inline void nf_reset(struct sk_buff *skb)
  2360. {
  2361. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2362. nf_conntrack_put(skb->nfct);
  2363. skb->nfct = NULL;
  2364. #endif
  2365. #ifdef CONFIG_BRIDGE_NETFILTER
  2366. nf_bridge_put(skb->nf_bridge);
  2367. skb->nf_bridge = NULL;
  2368. #endif
  2369. }
  2370. static inline void nf_reset_trace(struct sk_buff *skb)
  2371. {
  2372. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
  2373. skb->nf_trace = 0;
  2374. #endif
  2375. }
  2376. /* Note: This doesn't put any conntrack and bridge info in dst. */
  2377. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2378. {
  2379. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2380. dst->nfct = src->nfct;
  2381. nf_conntrack_get(src->nfct);
  2382. dst->nfctinfo = src->nfctinfo;
  2383. #endif
  2384. #ifdef CONFIG_BRIDGE_NETFILTER
  2385. dst->nf_bridge = src->nf_bridge;
  2386. nf_bridge_get(src->nf_bridge);
  2387. #endif
  2388. }
  2389. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2390. {
  2391. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2392. nf_conntrack_put(dst->nfct);
  2393. #endif
  2394. #ifdef CONFIG_BRIDGE_NETFILTER
  2395. nf_bridge_put(dst->nf_bridge);
  2396. #endif
  2397. __nf_copy(dst, src);
  2398. }
  2399. #ifdef CONFIG_NETWORK_SECMARK
  2400. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2401. {
  2402. to->secmark = from->secmark;
  2403. }
  2404. static inline void skb_init_secmark(struct sk_buff *skb)
  2405. {
  2406. skb->secmark = 0;
  2407. }
  2408. #else
  2409. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2410. { }
  2411. static inline void skb_init_secmark(struct sk_buff *skb)
  2412. { }
  2413. #endif
  2414. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  2415. {
  2416. skb->queue_mapping = queue_mapping;
  2417. }
  2418. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  2419. {
  2420. return skb->queue_mapping;
  2421. }
  2422. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  2423. {
  2424. to->queue_mapping = from->queue_mapping;
  2425. }
  2426. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  2427. {
  2428. skb->queue_mapping = rx_queue + 1;
  2429. }
  2430. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  2431. {
  2432. return skb->queue_mapping - 1;
  2433. }
  2434. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  2435. {
  2436. return skb->queue_mapping != 0;
  2437. }
  2438. u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
  2439. unsigned int num_tx_queues);
  2440. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  2441. {
  2442. #ifdef CONFIG_XFRM
  2443. return skb->sp;
  2444. #else
  2445. return NULL;
  2446. #endif
  2447. }
  2448. /* Keeps track of mac header offset relative to skb->head.
  2449. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  2450. * For non-tunnel skb it points to skb_mac_header() and for
  2451. * tunnel skb it points to outer mac header.
  2452. * Keeps track of level of encapsulation of network headers.
  2453. */
  2454. struct skb_gso_cb {
  2455. int mac_offset;
  2456. int encap_level;
  2457. };
  2458. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
  2459. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  2460. {
  2461. return (skb_mac_header(inner_skb) - inner_skb->head) -
  2462. SKB_GSO_CB(inner_skb)->mac_offset;
  2463. }
  2464. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  2465. {
  2466. int new_headroom, headroom;
  2467. int ret;
  2468. headroom = skb_headroom(skb);
  2469. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  2470. if (ret)
  2471. return ret;
  2472. new_headroom = skb_headroom(skb);
  2473. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  2474. return 0;
  2475. }
  2476. static inline bool skb_is_gso(const struct sk_buff *skb)
  2477. {
  2478. return skb_shinfo(skb)->gso_size;
  2479. }
  2480. /* Note: Should be called only if skb_is_gso(skb) is true */
  2481. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  2482. {
  2483. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  2484. }
  2485. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  2486. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  2487. {
  2488. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  2489. * wanted then gso_type will be set. */
  2490. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  2491. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  2492. unlikely(shinfo->gso_type == 0)) {
  2493. __skb_warn_lro_forwarding(skb);
  2494. return true;
  2495. }
  2496. return false;
  2497. }
  2498. static inline void skb_forward_csum(struct sk_buff *skb)
  2499. {
  2500. /* Unfortunately we don't support this one. Any brave souls? */
  2501. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2502. skb->ip_summed = CHECKSUM_NONE;
  2503. }
  2504. /**
  2505. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  2506. * @skb: skb to check
  2507. *
  2508. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  2509. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  2510. * use this helper, to document places where we make this assertion.
  2511. */
  2512. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  2513. {
  2514. #ifdef DEBUG
  2515. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  2516. #endif
  2517. }
  2518. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  2519. u32 __skb_get_poff(const struct sk_buff *skb);
  2520. /**
  2521. * skb_head_is_locked - Determine if the skb->head is locked down
  2522. * @skb: skb to check
  2523. *
  2524. * The head on skbs build around a head frag can be removed if they are
  2525. * not cloned. This function returns true if the skb head is locked down
  2526. * due to either being allocated via kmalloc, or by being a clone with
  2527. * multiple references to the head.
  2528. */
  2529. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  2530. {
  2531. return !skb->head_frag || skb_cloned(skb);
  2532. }
  2533. #endif /* __KERNEL__ */
  2534. #endif /* _LINUX_SKBUFF_H */