skbuff.h 97 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460246124622463246424652466246724682469247024712472247324742475247624772478247924802481248224832484248524862487248824892490249124922493249424952496249724982499250025012502250325042505250625072508250925102511251225132514251525162517251825192520252125222523252425252526252725282529253025312532253325342535253625372538253925402541254225432544254525462547254825492550255125522553255425552556255725582559256025612562256325642565256625672568256925702571257225732574257525762577257825792580258125822583258425852586258725882589259025912592259325942595259625972598259926002601260226032604260526062607260826092610261126122613261426152616261726182619262026212622262326242625262626272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697269826992700270127022703270427052706270727082709271027112712271327142715271627172718271927202721272227232724272527262727272827292730273127322733273427352736273727382739274027412742274327442745274627472748274927502751275227532754275527562757275827592760276127622763276427652766276727682769277027712772277327742775277627772778277927802781278227832784278527862787278827892790279127922793279427952796279727982799280028012802280328042805280628072808280928102811281228132814281528162817281828192820282128222823282428252826282728282829283028312832283328342835283628372838283928402841284228432844284528462847284828492850285128522853285428552856285728582859286028612862286328642865286628672868286928702871287228732874287528762877287828792880288128822883288428852886288728882889289028912892289328942895289628972898289929002901290229032904290529062907290829092910291129122913291429152916291729182919292029212922292329242925292629272928292929302931293229332934293529362937293829392940294129422943294429452946294729482949295029512952295329542955295629572958295929602961296229632964296529662967296829692970297129722973297429752976297729782979298029812982298329842985298629872988298929902991299229932994299529962997299829993000300130023003300430053006300730083009301030113012301330143015301630173018301930203021302230233024302530263027302830293030303130323033303430353036303730383039304030413042304330443045304630473048304930503051305230533054305530563057305830593060306130623063306430653066306730683069307030713072307330743075307630773078307930803081308230833084308530863087308830893090309130923093309430953096309730983099310031013102310331043105310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135313631373138313931403141314231433144314531463147314831493150315131523153315431553156315731583159316031613162316331643165316631673168316931703171317231733174317531763177317831793180318131823183318431853186318731883189319031913192319331943195319631973198319932003201320232033204320532063207320832093210321132123213321432153216321732183219322032213222322332243225322632273228322932303231323232333234323532363237323832393240324132423243324432453246324732483249325032513252325332543255325632573258325932603261326232633264326532663267326832693270327132723273327432753276327732783279328032813282328332843285328632873288328932903291329232933294329532963297329832993300330133023303330433053306330733083309331033113312331333143315331633173318331933203321332233233324332533263327332833293330333133323333333433353336333733383339334033413342334333443345334633473348334933503351335233533354335533563357335833593360336133623363336433653366336733683369337033713372337333743375337633773378337933803381338233833384338533863387338833893390339133923393339433953396339733983399340034013402340334043405340634073408340934103411341234133414341534163417341834193420342134223423342434253426342734283429343034313432343334343435343634373438343934403441344234433444344534463447344834493450345134523453345434553456345734583459346034613462346334643465346634673468
  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/rbtree.h>
  22. #include <linux/socket.h>
  23. #include <linux/atomic.h>
  24. #include <asm/types.h>
  25. #include <linux/spinlock.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #include <linux/rcupdate.h>
  30. #include <linux/hrtimer.h>
  31. #include <linux/dma-mapping.h>
  32. #include <linux/netdev_features.h>
  33. #include <linux/sched.h>
  34. #include <net/flow_dissector.h>
  35. #include <linux/splice.h>
  36. #include <linux/in6.h>
  37. /* A. Checksumming of received packets by device.
  38. *
  39. * CHECKSUM_NONE:
  40. *
  41. * Device failed to checksum this packet e.g. due to lack of capabilities.
  42. * The packet contains full (though not verified) checksum in packet but
  43. * not in skb->csum. Thus, skb->csum is undefined in this case.
  44. *
  45. * CHECKSUM_UNNECESSARY:
  46. *
  47. * The hardware you're dealing with doesn't calculate the full checksum
  48. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  49. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  50. * if their checksums are okay. skb->csum is still undefined in this case
  51. * though. It is a bad option, but, unfortunately, nowadays most vendors do
  52. * this. Apparently with the secret goal to sell you new devices, when you
  53. * will add new protocol to your host, f.e. IPv6 8)
  54. *
  55. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  56. * TCP: IPv6 and IPv4.
  57. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  58. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  59. * may perform further validation in this case.
  60. * GRE: only if the checksum is present in the header.
  61. * SCTP: indicates the CRC in SCTP header has been validated.
  62. *
  63. * skb->csum_level indicates the number of consecutive checksums found in
  64. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  65. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  66. * and a device is able to verify the checksums for UDP (possibly zero),
  67. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  68. * two. If the device were only able to verify the UDP checksum and not
  69. * GRE, either because it doesn't support GRE checksum of because GRE
  70. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  71. * not considered in this case).
  72. *
  73. * CHECKSUM_COMPLETE:
  74. *
  75. * This is the most generic way. The device supplied checksum of the _whole_
  76. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  77. * hardware doesn't need to parse L3/L4 headers to implement this.
  78. *
  79. * Note: Even if device supports only some protocols, but is able to produce
  80. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  81. *
  82. * CHECKSUM_PARTIAL:
  83. *
  84. * A checksum is set up to be offloaded to a device as described in the
  85. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  86. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  87. * on the same host, or it may be set in the input path in GRO or remote
  88. * checksum offload. For the purposes of checksum verification, the checksum
  89. * referred to by skb->csum_start + skb->csum_offset and any preceding
  90. * checksums in the packet are considered verified. Any checksums in the
  91. * packet that are after the checksum being offloaded are not considered to
  92. * be verified.
  93. *
  94. * B. Checksumming on output.
  95. *
  96. * CHECKSUM_NONE:
  97. *
  98. * The skb was already checksummed by the protocol, or a checksum is not
  99. * required.
  100. *
  101. * CHECKSUM_PARTIAL:
  102. *
  103. * The device is required to checksum the packet as seen by hard_start_xmit()
  104. * from skb->csum_start up to the end, and to record/write the checksum at
  105. * offset skb->csum_start + skb->csum_offset.
  106. *
  107. * The device must show its capabilities in dev->features, set up at device
  108. * setup time, e.g. netdev_features.h:
  109. *
  110. * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
  111. * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
  112. * IPv4. Sigh. Vendors like this way for an unknown reason.
  113. * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
  114. * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
  115. * NETIF_F_... - Well, you get the picture.
  116. *
  117. * CHECKSUM_UNNECESSARY:
  118. *
  119. * Normally, the device will do per protocol specific checksumming. Protocol
  120. * implementations that do not want the NIC to perform the checksum
  121. * calculation should use this flag in their outgoing skbs.
  122. *
  123. * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
  124. * offload. Correspondingly, the FCoE protocol driver
  125. * stack should use CHECKSUM_UNNECESSARY.
  126. *
  127. * Any questions? No questions, good. --ANK
  128. */
  129. /* Don't change this without changing skb_csum_unnecessary! */
  130. #define CHECKSUM_NONE 0
  131. #define CHECKSUM_UNNECESSARY 1
  132. #define CHECKSUM_COMPLETE 2
  133. #define CHECKSUM_PARTIAL 3
  134. /* Maximum value in skb->csum_level */
  135. #define SKB_MAX_CSUM_LEVEL 3
  136. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  137. #define SKB_WITH_OVERHEAD(X) \
  138. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  139. #define SKB_MAX_ORDER(X, ORDER) \
  140. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  141. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  142. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  143. /* return minimum truesize of one skb containing X bytes of data */
  144. #define SKB_TRUESIZE(X) ((X) + \
  145. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  146. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  147. struct net_device;
  148. struct scatterlist;
  149. struct pipe_inode_info;
  150. struct iov_iter;
  151. struct napi_struct;
  152. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  153. struct nf_conntrack {
  154. atomic_t use;
  155. };
  156. #endif
  157. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  158. struct nf_bridge_info {
  159. atomic_t use;
  160. enum {
  161. BRNF_PROTO_UNCHANGED,
  162. BRNF_PROTO_8021Q,
  163. BRNF_PROTO_PPPOE
  164. } orig_proto:8;
  165. bool pkt_otherhost;
  166. __u16 frag_max_size;
  167. unsigned int mask;
  168. struct net_device *physindev;
  169. union {
  170. struct net_device *physoutdev;
  171. char neigh_header[8];
  172. };
  173. union {
  174. __be32 ipv4_daddr;
  175. struct in6_addr ipv6_daddr;
  176. };
  177. };
  178. #endif
  179. struct sk_buff_head {
  180. /* These two members must be first. */
  181. struct sk_buff *next;
  182. struct sk_buff *prev;
  183. __u32 qlen;
  184. spinlock_t lock;
  185. };
  186. struct sk_buff;
  187. /* To allow 64K frame to be packed as single skb without frag_list we
  188. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  189. * buffers which do not start on a page boundary.
  190. *
  191. * Since GRO uses frags we allocate at least 16 regardless of page
  192. * size.
  193. */
  194. #if (65536/PAGE_SIZE + 1) < 16
  195. #define MAX_SKB_FRAGS 16UL
  196. #else
  197. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  198. #endif
  199. typedef struct skb_frag_struct skb_frag_t;
  200. struct skb_frag_struct {
  201. struct {
  202. struct page *p;
  203. } page;
  204. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  205. __u32 page_offset;
  206. __u32 size;
  207. #else
  208. __u16 page_offset;
  209. __u16 size;
  210. #endif
  211. };
  212. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  213. {
  214. return frag->size;
  215. }
  216. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  217. {
  218. frag->size = size;
  219. }
  220. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  221. {
  222. frag->size += delta;
  223. }
  224. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  225. {
  226. frag->size -= delta;
  227. }
  228. #define HAVE_HW_TIME_STAMP
  229. /**
  230. * struct skb_shared_hwtstamps - hardware time stamps
  231. * @hwtstamp: hardware time stamp transformed into duration
  232. * since arbitrary point in time
  233. *
  234. * Software time stamps generated by ktime_get_real() are stored in
  235. * skb->tstamp.
  236. *
  237. * hwtstamps can only be compared against other hwtstamps from
  238. * the same device.
  239. *
  240. * This structure is attached to packets as part of the
  241. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  242. */
  243. struct skb_shared_hwtstamps {
  244. ktime_t hwtstamp;
  245. };
  246. /* Definitions for tx_flags in struct skb_shared_info */
  247. enum {
  248. /* generate hardware time stamp */
  249. SKBTX_HW_TSTAMP = 1 << 0,
  250. /* generate software time stamp when queueing packet to NIC */
  251. SKBTX_SW_TSTAMP = 1 << 1,
  252. /* device driver is going to provide hardware time stamp */
  253. SKBTX_IN_PROGRESS = 1 << 2,
  254. /* device driver supports TX zero-copy buffers */
  255. SKBTX_DEV_ZEROCOPY = 1 << 3,
  256. /* generate wifi status information (where possible) */
  257. SKBTX_WIFI_STATUS = 1 << 4,
  258. /* This indicates at least one fragment might be overwritten
  259. * (as in vmsplice(), sendfile() ...)
  260. * If we need to compute a TX checksum, we'll need to copy
  261. * all frags to avoid possible bad checksum
  262. */
  263. SKBTX_SHARED_FRAG = 1 << 5,
  264. /* generate software time stamp when entering packet scheduling */
  265. SKBTX_SCHED_TSTAMP = 1 << 6,
  266. /* generate software timestamp on peer data acknowledgment */
  267. SKBTX_ACK_TSTAMP = 1 << 7,
  268. };
  269. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  270. SKBTX_SCHED_TSTAMP | \
  271. SKBTX_ACK_TSTAMP)
  272. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  273. /*
  274. * The callback notifies userspace to release buffers when skb DMA is done in
  275. * lower device, the skb last reference should be 0 when calling this.
  276. * The zerocopy_success argument is true if zero copy transmit occurred,
  277. * false on data copy or out of memory error caused by data copy attempt.
  278. * The ctx field is used to track device context.
  279. * The desc field is used to track userspace buffer index.
  280. */
  281. struct ubuf_info {
  282. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  283. void *ctx;
  284. unsigned long desc;
  285. };
  286. /* This data is invariant across clones and lives at
  287. * the end of the header data, ie. at skb->end.
  288. */
  289. struct skb_shared_info {
  290. unsigned char nr_frags;
  291. __u8 tx_flags;
  292. unsigned short gso_size;
  293. /* Warning: this field is not always filled in (UFO)! */
  294. unsigned short gso_segs;
  295. unsigned short gso_type;
  296. struct sk_buff *frag_list;
  297. struct skb_shared_hwtstamps hwtstamps;
  298. u32 tskey;
  299. __be32 ip6_frag_id;
  300. /*
  301. * Warning : all fields before dataref are cleared in __alloc_skb()
  302. */
  303. atomic_t dataref;
  304. /* Intermediate layers must ensure that destructor_arg
  305. * remains valid until skb destructor */
  306. void * destructor_arg;
  307. /* must be last field, see pskb_expand_head() */
  308. skb_frag_t frags[MAX_SKB_FRAGS];
  309. };
  310. /* We divide dataref into two halves. The higher 16 bits hold references
  311. * to the payload part of skb->data. The lower 16 bits hold references to
  312. * the entire skb->data. A clone of a headerless skb holds the length of
  313. * the header in skb->hdr_len.
  314. *
  315. * All users must obey the rule that the skb->data reference count must be
  316. * greater than or equal to the payload reference count.
  317. *
  318. * Holding a reference to the payload part means that the user does not
  319. * care about modifications to the header part of skb->data.
  320. */
  321. #define SKB_DATAREF_SHIFT 16
  322. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  323. enum {
  324. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  325. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  326. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  327. };
  328. enum {
  329. SKB_GSO_TCPV4 = 1 << 0,
  330. SKB_GSO_UDP = 1 << 1,
  331. /* This indicates the skb is from an untrusted source. */
  332. SKB_GSO_DODGY = 1 << 2,
  333. /* This indicates the tcp segment has CWR set. */
  334. SKB_GSO_TCP_ECN = 1 << 3,
  335. SKB_GSO_TCPV6 = 1 << 4,
  336. SKB_GSO_FCOE = 1 << 5,
  337. SKB_GSO_GRE = 1 << 6,
  338. SKB_GSO_GRE_CSUM = 1 << 7,
  339. SKB_GSO_IPIP = 1 << 8,
  340. SKB_GSO_SIT = 1 << 9,
  341. SKB_GSO_UDP_TUNNEL = 1 << 10,
  342. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  343. SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
  344. };
  345. #if BITS_PER_LONG > 32
  346. #define NET_SKBUFF_DATA_USES_OFFSET 1
  347. #endif
  348. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  349. typedef unsigned int sk_buff_data_t;
  350. #else
  351. typedef unsigned char *sk_buff_data_t;
  352. #endif
  353. /**
  354. * struct skb_mstamp - multi resolution time stamps
  355. * @stamp_us: timestamp in us resolution
  356. * @stamp_jiffies: timestamp in jiffies
  357. */
  358. struct skb_mstamp {
  359. union {
  360. u64 v64;
  361. struct {
  362. u32 stamp_us;
  363. u32 stamp_jiffies;
  364. };
  365. };
  366. };
  367. /**
  368. * skb_mstamp_get - get current timestamp
  369. * @cl: place to store timestamps
  370. */
  371. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  372. {
  373. u64 val = local_clock();
  374. do_div(val, NSEC_PER_USEC);
  375. cl->stamp_us = (u32)val;
  376. cl->stamp_jiffies = (u32)jiffies;
  377. }
  378. /**
  379. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  380. * @t1: pointer to newest sample
  381. * @t0: pointer to oldest sample
  382. */
  383. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  384. const struct skb_mstamp *t0)
  385. {
  386. s32 delta_us = t1->stamp_us - t0->stamp_us;
  387. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  388. /* If delta_us is negative, this might be because interval is too big,
  389. * or local_clock() drift is too big : fallback using jiffies.
  390. */
  391. if (delta_us <= 0 ||
  392. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  393. delta_us = jiffies_to_usecs(delta_jiffies);
  394. return delta_us;
  395. }
  396. /**
  397. * struct sk_buff - socket buffer
  398. * @next: Next buffer in list
  399. * @prev: Previous buffer in list
  400. * @tstamp: Time we arrived/left
  401. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  402. * @sk: Socket we are owned by
  403. * @dev: Device we arrived on/are leaving by
  404. * @cb: Control buffer. Free for use by every layer. Put private vars here
  405. * @_skb_refdst: destination entry (with norefcount bit)
  406. * @sp: the security path, used for xfrm
  407. * @len: Length of actual data
  408. * @data_len: Data length
  409. * @mac_len: Length of link layer header
  410. * @hdr_len: writable header length of cloned skb
  411. * @csum: Checksum (must include start/offset pair)
  412. * @csum_start: Offset from skb->head where checksumming should start
  413. * @csum_offset: Offset from csum_start where checksum should be stored
  414. * @priority: Packet queueing priority
  415. * @ignore_df: allow local fragmentation
  416. * @cloned: Head may be cloned (check refcnt to be sure)
  417. * @ip_summed: Driver fed us an IP checksum
  418. * @nohdr: Payload reference only, must not modify header
  419. * @nfctinfo: Relationship of this skb to the connection
  420. * @pkt_type: Packet class
  421. * @fclone: skbuff clone status
  422. * @ipvs_property: skbuff is owned by ipvs
  423. * @peeked: this packet has been seen already, so stats have been
  424. * done for it, don't do them again
  425. * @nf_trace: netfilter packet trace flag
  426. * @protocol: Packet protocol from driver
  427. * @destructor: Destruct function
  428. * @nfct: Associated connection, if any
  429. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  430. * @skb_iif: ifindex of device we arrived on
  431. * @tc_index: Traffic control index
  432. * @tc_verd: traffic control verdict
  433. * @hash: the packet hash
  434. * @queue_mapping: Queue mapping for multiqueue devices
  435. * @xmit_more: More SKBs are pending for this queue
  436. * @ndisc_nodetype: router type (from link layer)
  437. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  438. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  439. * ports.
  440. * @sw_hash: indicates hash was computed in software stack
  441. * @wifi_acked_valid: wifi_acked was set
  442. * @wifi_acked: whether frame was acked on wifi or not
  443. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  444. * @napi_id: id of the NAPI struct this skb came from
  445. * @secmark: security marking
  446. * @mark: Generic packet mark
  447. * @vlan_proto: vlan encapsulation protocol
  448. * @vlan_tci: vlan tag control information
  449. * @inner_protocol: Protocol (encapsulation)
  450. * @inner_transport_header: Inner transport layer header (encapsulation)
  451. * @inner_network_header: Network layer header (encapsulation)
  452. * @inner_mac_header: Link layer header (encapsulation)
  453. * @transport_header: Transport layer header
  454. * @network_header: Network layer header
  455. * @mac_header: Link layer header
  456. * @tail: Tail pointer
  457. * @end: End pointer
  458. * @head: Head of buffer
  459. * @data: Data head pointer
  460. * @truesize: Buffer size
  461. * @users: User count - see {datagram,tcp}.c
  462. */
  463. struct sk_buff {
  464. union {
  465. struct {
  466. /* These two members must be first. */
  467. struct sk_buff *next;
  468. struct sk_buff *prev;
  469. union {
  470. ktime_t tstamp;
  471. struct skb_mstamp skb_mstamp;
  472. };
  473. };
  474. struct rb_node rbnode; /* used in netem & tcp stack */
  475. };
  476. struct sock *sk;
  477. struct net_device *dev;
  478. /*
  479. * This is the control buffer. It is free to use for every
  480. * layer. Please put your private variables there. If you
  481. * want to keep them across layers you have to do a skb_clone()
  482. * first. This is owned by whoever has the skb queued ATM.
  483. */
  484. char cb[48] __aligned(8);
  485. unsigned long _skb_refdst;
  486. void (*destructor)(struct sk_buff *skb);
  487. #ifdef CONFIG_XFRM
  488. struct sec_path *sp;
  489. #endif
  490. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  491. struct nf_conntrack *nfct;
  492. #endif
  493. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  494. struct nf_bridge_info *nf_bridge;
  495. #endif
  496. unsigned int len,
  497. data_len;
  498. __u16 mac_len,
  499. hdr_len;
  500. /* Following fields are _not_ copied in __copy_skb_header()
  501. * Note that queue_mapping is here mostly to fill a hole.
  502. */
  503. kmemcheck_bitfield_begin(flags1);
  504. __u16 queue_mapping;
  505. __u8 cloned:1,
  506. nohdr:1,
  507. fclone:2,
  508. peeked:1,
  509. head_frag:1,
  510. xmit_more:1;
  511. /* one bit hole */
  512. kmemcheck_bitfield_end(flags1);
  513. /* fields enclosed in headers_start/headers_end are copied
  514. * using a single memcpy() in __copy_skb_header()
  515. */
  516. /* private: */
  517. __u32 headers_start[0];
  518. /* public: */
  519. /* if you move pkt_type around you also must adapt those constants */
  520. #ifdef __BIG_ENDIAN_BITFIELD
  521. #define PKT_TYPE_MAX (7 << 5)
  522. #else
  523. #define PKT_TYPE_MAX 7
  524. #endif
  525. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  526. __u8 __pkt_type_offset[0];
  527. __u8 pkt_type:3;
  528. __u8 pfmemalloc:1;
  529. __u8 ignore_df:1;
  530. __u8 nfctinfo:3;
  531. __u8 nf_trace:1;
  532. __u8 ip_summed:2;
  533. __u8 ooo_okay:1;
  534. __u8 l4_hash:1;
  535. __u8 sw_hash:1;
  536. __u8 wifi_acked_valid:1;
  537. __u8 wifi_acked:1;
  538. __u8 no_fcs:1;
  539. /* Indicates the inner headers are valid in the skbuff. */
  540. __u8 encapsulation:1;
  541. __u8 encap_hdr_csum:1;
  542. __u8 csum_valid:1;
  543. __u8 csum_complete_sw:1;
  544. __u8 csum_level:2;
  545. __u8 csum_bad:1;
  546. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  547. __u8 ndisc_nodetype:2;
  548. #endif
  549. __u8 ipvs_property:1;
  550. __u8 inner_protocol_type:1;
  551. __u8 remcsum_offload:1;
  552. /* 3 or 5 bit hole */
  553. #ifdef CONFIG_NET_SCHED
  554. __u16 tc_index; /* traffic control index */
  555. #ifdef CONFIG_NET_CLS_ACT
  556. __u16 tc_verd; /* traffic control verdict */
  557. #endif
  558. #endif
  559. union {
  560. __wsum csum;
  561. struct {
  562. __u16 csum_start;
  563. __u16 csum_offset;
  564. };
  565. };
  566. __u32 priority;
  567. int skb_iif;
  568. __u32 hash;
  569. __be16 vlan_proto;
  570. __u16 vlan_tci;
  571. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  572. union {
  573. unsigned int napi_id;
  574. unsigned int sender_cpu;
  575. };
  576. #endif
  577. #ifdef CONFIG_NETWORK_SECMARK
  578. __u32 secmark;
  579. #endif
  580. union {
  581. __u32 mark;
  582. __u32 reserved_tailroom;
  583. };
  584. union {
  585. __be16 inner_protocol;
  586. __u8 inner_ipproto;
  587. };
  588. __u16 inner_transport_header;
  589. __u16 inner_network_header;
  590. __u16 inner_mac_header;
  591. __be16 protocol;
  592. __u16 transport_header;
  593. __u16 network_header;
  594. __u16 mac_header;
  595. /* private: */
  596. __u32 headers_end[0];
  597. /* public: */
  598. /* These elements must be at the end, see alloc_skb() for details. */
  599. sk_buff_data_t tail;
  600. sk_buff_data_t end;
  601. unsigned char *head,
  602. *data;
  603. unsigned int truesize;
  604. atomic_t users;
  605. };
  606. #ifdef __KERNEL__
  607. /*
  608. * Handling routines are only of interest to the kernel
  609. */
  610. #include <linux/slab.h>
  611. #define SKB_ALLOC_FCLONE 0x01
  612. #define SKB_ALLOC_RX 0x02
  613. #define SKB_ALLOC_NAPI 0x04
  614. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  615. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  616. {
  617. return unlikely(skb->pfmemalloc);
  618. }
  619. /*
  620. * skb might have a dst pointer attached, refcounted or not.
  621. * _skb_refdst low order bit is set if refcount was _not_ taken
  622. */
  623. #define SKB_DST_NOREF 1UL
  624. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  625. /**
  626. * skb_dst - returns skb dst_entry
  627. * @skb: buffer
  628. *
  629. * Returns skb dst_entry, regardless of reference taken or not.
  630. */
  631. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  632. {
  633. /* If refdst was not refcounted, check we still are in a
  634. * rcu_read_lock section
  635. */
  636. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  637. !rcu_read_lock_held() &&
  638. !rcu_read_lock_bh_held());
  639. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  640. }
  641. /**
  642. * skb_dst_set - sets skb dst
  643. * @skb: buffer
  644. * @dst: dst entry
  645. *
  646. * Sets skb dst, assuming a reference was taken on dst and should
  647. * be released by skb_dst_drop()
  648. */
  649. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  650. {
  651. skb->_skb_refdst = (unsigned long)dst;
  652. }
  653. /**
  654. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  655. * @skb: buffer
  656. * @dst: dst entry
  657. *
  658. * Sets skb dst, assuming a reference was not taken on dst.
  659. * If dst entry is cached, we do not take reference and dst_release
  660. * will be avoided by refdst_drop. If dst entry is not cached, we take
  661. * reference, so that last dst_release can destroy the dst immediately.
  662. */
  663. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  664. {
  665. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  666. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  667. }
  668. /**
  669. * skb_dst_is_noref - Test if skb dst isn't refcounted
  670. * @skb: buffer
  671. */
  672. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  673. {
  674. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  675. }
  676. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  677. {
  678. return (struct rtable *)skb_dst(skb);
  679. }
  680. void kfree_skb(struct sk_buff *skb);
  681. void kfree_skb_list(struct sk_buff *segs);
  682. void skb_tx_error(struct sk_buff *skb);
  683. void consume_skb(struct sk_buff *skb);
  684. void __kfree_skb(struct sk_buff *skb);
  685. extern struct kmem_cache *skbuff_head_cache;
  686. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  687. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  688. bool *fragstolen, int *delta_truesize);
  689. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  690. int node);
  691. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  692. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  693. static inline struct sk_buff *alloc_skb(unsigned int size,
  694. gfp_t priority)
  695. {
  696. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  697. }
  698. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  699. unsigned long data_len,
  700. int max_page_order,
  701. int *errcode,
  702. gfp_t gfp_mask);
  703. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  704. struct sk_buff_fclones {
  705. struct sk_buff skb1;
  706. struct sk_buff skb2;
  707. atomic_t fclone_ref;
  708. };
  709. /**
  710. * skb_fclone_busy - check if fclone is busy
  711. * @skb: buffer
  712. *
  713. * Returns true is skb is a fast clone, and its clone is not freed.
  714. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  715. * so we also check that this didnt happen.
  716. */
  717. static inline bool skb_fclone_busy(const struct sock *sk,
  718. const struct sk_buff *skb)
  719. {
  720. const struct sk_buff_fclones *fclones;
  721. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  722. return skb->fclone == SKB_FCLONE_ORIG &&
  723. atomic_read(&fclones->fclone_ref) > 1 &&
  724. fclones->skb2.sk == sk;
  725. }
  726. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  727. gfp_t priority)
  728. {
  729. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  730. }
  731. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  732. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  733. {
  734. return __alloc_skb_head(priority, -1);
  735. }
  736. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  737. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  738. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  739. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  740. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  741. gfp_t gfp_mask, bool fclone);
  742. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  743. gfp_t gfp_mask)
  744. {
  745. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  746. }
  747. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  748. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  749. unsigned int headroom);
  750. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  751. int newtailroom, gfp_t priority);
  752. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  753. int offset, int len);
  754. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  755. int len);
  756. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  757. int skb_pad(struct sk_buff *skb, int pad);
  758. #define dev_kfree_skb(a) consume_skb(a)
  759. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  760. int getfrag(void *from, char *to, int offset,
  761. int len, int odd, struct sk_buff *skb),
  762. void *from, int length);
  763. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  764. int offset, size_t size);
  765. struct skb_seq_state {
  766. __u32 lower_offset;
  767. __u32 upper_offset;
  768. __u32 frag_idx;
  769. __u32 stepped_offset;
  770. struct sk_buff *root_skb;
  771. struct sk_buff *cur_skb;
  772. __u8 *frag_data;
  773. };
  774. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  775. unsigned int to, struct skb_seq_state *st);
  776. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  777. struct skb_seq_state *st);
  778. void skb_abort_seq_read(struct skb_seq_state *st);
  779. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  780. unsigned int to, struct ts_config *config);
  781. /*
  782. * Packet hash types specify the type of hash in skb_set_hash.
  783. *
  784. * Hash types refer to the protocol layer addresses which are used to
  785. * construct a packet's hash. The hashes are used to differentiate or identify
  786. * flows of the protocol layer for the hash type. Hash types are either
  787. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  788. *
  789. * Properties of hashes:
  790. *
  791. * 1) Two packets in different flows have different hash values
  792. * 2) Two packets in the same flow should have the same hash value
  793. *
  794. * A hash at a higher layer is considered to be more specific. A driver should
  795. * set the most specific hash possible.
  796. *
  797. * A driver cannot indicate a more specific hash than the layer at which a hash
  798. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  799. *
  800. * A driver may indicate a hash level which is less specific than the
  801. * actual layer the hash was computed on. For instance, a hash computed
  802. * at L4 may be considered an L3 hash. This should only be done if the
  803. * driver can't unambiguously determine that the HW computed the hash at
  804. * the higher layer. Note that the "should" in the second property above
  805. * permits this.
  806. */
  807. enum pkt_hash_types {
  808. PKT_HASH_TYPE_NONE, /* Undefined type */
  809. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  810. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  811. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  812. };
  813. static inline void
  814. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  815. {
  816. skb->l4_hash = (type == PKT_HASH_TYPE_L4);
  817. skb->sw_hash = 0;
  818. skb->hash = hash;
  819. }
  820. static inline __u32 skb_get_hash(struct sk_buff *skb)
  821. {
  822. if (!skb->l4_hash && !skb->sw_hash)
  823. __skb_get_hash(skb);
  824. return skb->hash;
  825. }
  826. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  827. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  828. {
  829. return skb->hash;
  830. }
  831. static inline void skb_clear_hash(struct sk_buff *skb)
  832. {
  833. skb->hash = 0;
  834. skb->sw_hash = 0;
  835. skb->l4_hash = 0;
  836. }
  837. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  838. {
  839. if (!skb->l4_hash)
  840. skb_clear_hash(skb);
  841. }
  842. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  843. {
  844. to->hash = from->hash;
  845. to->sw_hash = from->sw_hash;
  846. to->l4_hash = from->l4_hash;
  847. };
  848. static inline void skb_sender_cpu_clear(struct sk_buff *skb)
  849. {
  850. #ifdef CONFIG_XPS
  851. skb->sender_cpu = 0;
  852. #endif
  853. }
  854. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  855. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  856. {
  857. return skb->head + skb->end;
  858. }
  859. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  860. {
  861. return skb->end;
  862. }
  863. #else
  864. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  865. {
  866. return skb->end;
  867. }
  868. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  869. {
  870. return skb->end - skb->head;
  871. }
  872. #endif
  873. /* Internal */
  874. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  875. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  876. {
  877. return &skb_shinfo(skb)->hwtstamps;
  878. }
  879. /**
  880. * skb_queue_empty - check if a queue is empty
  881. * @list: queue head
  882. *
  883. * Returns true if the queue is empty, false otherwise.
  884. */
  885. static inline int skb_queue_empty(const struct sk_buff_head *list)
  886. {
  887. return list->next == (const struct sk_buff *) list;
  888. }
  889. /**
  890. * skb_queue_is_last - check if skb is the last entry in the queue
  891. * @list: queue head
  892. * @skb: buffer
  893. *
  894. * Returns true if @skb is the last buffer on the list.
  895. */
  896. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  897. const struct sk_buff *skb)
  898. {
  899. return skb->next == (const struct sk_buff *) list;
  900. }
  901. /**
  902. * skb_queue_is_first - check if skb is the first entry in the queue
  903. * @list: queue head
  904. * @skb: buffer
  905. *
  906. * Returns true if @skb is the first buffer on the list.
  907. */
  908. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  909. const struct sk_buff *skb)
  910. {
  911. return skb->prev == (const struct sk_buff *) list;
  912. }
  913. /**
  914. * skb_queue_next - return the next packet in the queue
  915. * @list: queue head
  916. * @skb: current buffer
  917. *
  918. * Return the next packet in @list after @skb. It is only valid to
  919. * call this if skb_queue_is_last() evaluates to false.
  920. */
  921. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  922. const struct sk_buff *skb)
  923. {
  924. /* This BUG_ON may seem severe, but if we just return then we
  925. * are going to dereference garbage.
  926. */
  927. BUG_ON(skb_queue_is_last(list, skb));
  928. return skb->next;
  929. }
  930. /**
  931. * skb_queue_prev - return the prev packet in the queue
  932. * @list: queue head
  933. * @skb: current buffer
  934. *
  935. * Return the prev packet in @list before @skb. It is only valid to
  936. * call this if skb_queue_is_first() evaluates to false.
  937. */
  938. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  939. const struct sk_buff *skb)
  940. {
  941. /* This BUG_ON may seem severe, but if we just return then we
  942. * are going to dereference garbage.
  943. */
  944. BUG_ON(skb_queue_is_first(list, skb));
  945. return skb->prev;
  946. }
  947. /**
  948. * skb_get - reference buffer
  949. * @skb: buffer to reference
  950. *
  951. * Makes another reference to a socket buffer and returns a pointer
  952. * to the buffer.
  953. */
  954. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  955. {
  956. atomic_inc(&skb->users);
  957. return skb;
  958. }
  959. /*
  960. * If users == 1, we are the only owner and are can avoid redundant
  961. * atomic change.
  962. */
  963. /**
  964. * skb_cloned - is the buffer a clone
  965. * @skb: buffer to check
  966. *
  967. * Returns true if the buffer was generated with skb_clone() and is
  968. * one of multiple shared copies of the buffer. Cloned buffers are
  969. * shared data so must not be written to under normal circumstances.
  970. */
  971. static inline int skb_cloned(const struct sk_buff *skb)
  972. {
  973. return skb->cloned &&
  974. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  975. }
  976. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  977. {
  978. might_sleep_if(pri & __GFP_WAIT);
  979. if (skb_cloned(skb))
  980. return pskb_expand_head(skb, 0, 0, pri);
  981. return 0;
  982. }
  983. /**
  984. * skb_header_cloned - is the header a clone
  985. * @skb: buffer to check
  986. *
  987. * Returns true if modifying the header part of the buffer requires
  988. * the data to be copied.
  989. */
  990. static inline int skb_header_cloned(const struct sk_buff *skb)
  991. {
  992. int dataref;
  993. if (!skb->cloned)
  994. return 0;
  995. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  996. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  997. return dataref != 1;
  998. }
  999. /**
  1000. * skb_header_release - release reference to header
  1001. * @skb: buffer to operate on
  1002. *
  1003. * Drop a reference to the header part of the buffer. This is done
  1004. * by acquiring a payload reference. You must not read from the header
  1005. * part of skb->data after this.
  1006. * Note : Check if you can use __skb_header_release() instead.
  1007. */
  1008. static inline void skb_header_release(struct sk_buff *skb)
  1009. {
  1010. BUG_ON(skb->nohdr);
  1011. skb->nohdr = 1;
  1012. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1013. }
  1014. /**
  1015. * __skb_header_release - release reference to header
  1016. * @skb: buffer to operate on
  1017. *
  1018. * Variant of skb_header_release() assuming skb is private to caller.
  1019. * We can avoid one atomic operation.
  1020. */
  1021. static inline void __skb_header_release(struct sk_buff *skb)
  1022. {
  1023. skb->nohdr = 1;
  1024. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1025. }
  1026. /**
  1027. * skb_shared - is the buffer shared
  1028. * @skb: buffer to check
  1029. *
  1030. * Returns true if more than one person has a reference to this
  1031. * buffer.
  1032. */
  1033. static inline int skb_shared(const struct sk_buff *skb)
  1034. {
  1035. return atomic_read(&skb->users) != 1;
  1036. }
  1037. /**
  1038. * skb_share_check - check if buffer is shared and if so clone it
  1039. * @skb: buffer to check
  1040. * @pri: priority for memory allocation
  1041. *
  1042. * If the buffer is shared the buffer is cloned and the old copy
  1043. * drops a reference. A new clone with a single reference is returned.
  1044. * If the buffer is not shared the original buffer is returned. When
  1045. * being called from interrupt status or with spinlocks held pri must
  1046. * be GFP_ATOMIC.
  1047. *
  1048. * NULL is returned on a memory allocation failure.
  1049. */
  1050. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1051. {
  1052. might_sleep_if(pri & __GFP_WAIT);
  1053. if (skb_shared(skb)) {
  1054. struct sk_buff *nskb = skb_clone(skb, pri);
  1055. if (likely(nskb))
  1056. consume_skb(skb);
  1057. else
  1058. kfree_skb(skb);
  1059. skb = nskb;
  1060. }
  1061. return skb;
  1062. }
  1063. /*
  1064. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1065. * packets to handle cases where we have a local reader and forward
  1066. * and a couple of other messy ones. The normal one is tcpdumping
  1067. * a packet thats being forwarded.
  1068. */
  1069. /**
  1070. * skb_unshare - make a copy of a shared buffer
  1071. * @skb: buffer to check
  1072. * @pri: priority for memory allocation
  1073. *
  1074. * If the socket buffer is a clone then this function creates a new
  1075. * copy of the data, drops a reference count on the old copy and returns
  1076. * the new copy with the reference count at 1. If the buffer is not a clone
  1077. * the original buffer is returned. When called with a spinlock held or
  1078. * from interrupt state @pri must be %GFP_ATOMIC
  1079. *
  1080. * %NULL is returned on a memory allocation failure.
  1081. */
  1082. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1083. gfp_t pri)
  1084. {
  1085. might_sleep_if(pri & __GFP_WAIT);
  1086. if (skb_cloned(skb)) {
  1087. struct sk_buff *nskb = skb_copy(skb, pri);
  1088. /* Free our shared copy */
  1089. if (likely(nskb))
  1090. consume_skb(skb);
  1091. else
  1092. kfree_skb(skb);
  1093. skb = nskb;
  1094. }
  1095. return skb;
  1096. }
  1097. /**
  1098. * skb_peek - peek at the head of an &sk_buff_head
  1099. * @list_: list to peek at
  1100. *
  1101. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1102. * be careful with this one. A peek leaves the buffer on the
  1103. * list and someone else may run off with it. You must hold
  1104. * the appropriate locks or have a private queue to do this.
  1105. *
  1106. * Returns %NULL for an empty list or a pointer to the head element.
  1107. * The reference count is not incremented and the reference is therefore
  1108. * volatile. Use with caution.
  1109. */
  1110. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1111. {
  1112. struct sk_buff *skb = list_->next;
  1113. if (skb == (struct sk_buff *)list_)
  1114. skb = NULL;
  1115. return skb;
  1116. }
  1117. /**
  1118. * skb_peek_next - peek skb following the given one from a queue
  1119. * @skb: skb to start from
  1120. * @list_: list to peek at
  1121. *
  1122. * Returns %NULL when the end of the list is met or a pointer to the
  1123. * next element. The reference count is not incremented and the
  1124. * reference is therefore volatile. Use with caution.
  1125. */
  1126. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1127. const struct sk_buff_head *list_)
  1128. {
  1129. struct sk_buff *next = skb->next;
  1130. if (next == (struct sk_buff *)list_)
  1131. next = NULL;
  1132. return next;
  1133. }
  1134. /**
  1135. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1136. * @list_: list to peek at
  1137. *
  1138. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1139. * be careful with this one. A peek leaves the buffer on the
  1140. * list and someone else may run off with it. You must hold
  1141. * the appropriate locks or have a private queue to do this.
  1142. *
  1143. * Returns %NULL for an empty list or a pointer to the tail element.
  1144. * The reference count is not incremented and the reference is therefore
  1145. * volatile. Use with caution.
  1146. */
  1147. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1148. {
  1149. struct sk_buff *skb = list_->prev;
  1150. if (skb == (struct sk_buff *)list_)
  1151. skb = NULL;
  1152. return skb;
  1153. }
  1154. /**
  1155. * skb_queue_len - get queue length
  1156. * @list_: list to measure
  1157. *
  1158. * Return the length of an &sk_buff queue.
  1159. */
  1160. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1161. {
  1162. return list_->qlen;
  1163. }
  1164. /**
  1165. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1166. * @list: queue to initialize
  1167. *
  1168. * This initializes only the list and queue length aspects of
  1169. * an sk_buff_head object. This allows to initialize the list
  1170. * aspects of an sk_buff_head without reinitializing things like
  1171. * the spinlock. It can also be used for on-stack sk_buff_head
  1172. * objects where the spinlock is known to not be used.
  1173. */
  1174. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1175. {
  1176. list->prev = list->next = (struct sk_buff *)list;
  1177. list->qlen = 0;
  1178. }
  1179. /*
  1180. * This function creates a split out lock class for each invocation;
  1181. * this is needed for now since a whole lot of users of the skb-queue
  1182. * infrastructure in drivers have different locking usage (in hardirq)
  1183. * than the networking core (in softirq only). In the long run either the
  1184. * network layer or drivers should need annotation to consolidate the
  1185. * main types of usage into 3 classes.
  1186. */
  1187. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1188. {
  1189. spin_lock_init(&list->lock);
  1190. __skb_queue_head_init(list);
  1191. }
  1192. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1193. struct lock_class_key *class)
  1194. {
  1195. skb_queue_head_init(list);
  1196. lockdep_set_class(&list->lock, class);
  1197. }
  1198. /*
  1199. * Insert an sk_buff on a list.
  1200. *
  1201. * The "__skb_xxxx()" functions are the non-atomic ones that
  1202. * can only be called with interrupts disabled.
  1203. */
  1204. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1205. struct sk_buff_head *list);
  1206. static inline void __skb_insert(struct sk_buff *newsk,
  1207. struct sk_buff *prev, struct sk_buff *next,
  1208. struct sk_buff_head *list)
  1209. {
  1210. newsk->next = next;
  1211. newsk->prev = prev;
  1212. next->prev = prev->next = newsk;
  1213. list->qlen++;
  1214. }
  1215. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1216. struct sk_buff *prev,
  1217. struct sk_buff *next)
  1218. {
  1219. struct sk_buff *first = list->next;
  1220. struct sk_buff *last = list->prev;
  1221. first->prev = prev;
  1222. prev->next = first;
  1223. last->next = next;
  1224. next->prev = last;
  1225. }
  1226. /**
  1227. * skb_queue_splice - join two skb lists, this is designed for stacks
  1228. * @list: the new list to add
  1229. * @head: the place to add it in the first list
  1230. */
  1231. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1232. struct sk_buff_head *head)
  1233. {
  1234. if (!skb_queue_empty(list)) {
  1235. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1236. head->qlen += list->qlen;
  1237. }
  1238. }
  1239. /**
  1240. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1241. * @list: the new list to add
  1242. * @head: the place to add it in the first list
  1243. *
  1244. * The list at @list is reinitialised
  1245. */
  1246. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1247. struct sk_buff_head *head)
  1248. {
  1249. if (!skb_queue_empty(list)) {
  1250. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1251. head->qlen += list->qlen;
  1252. __skb_queue_head_init(list);
  1253. }
  1254. }
  1255. /**
  1256. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1257. * @list: the new list to add
  1258. * @head: the place to add it in the first list
  1259. */
  1260. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1261. struct sk_buff_head *head)
  1262. {
  1263. if (!skb_queue_empty(list)) {
  1264. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1265. head->qlen += list->qlen;
  1266. }
  1267. }
  1268. /**
  1269. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1270. * @list: the new list to add
  1271. * @head: the place to add it in the first list
  1272. *
  1273. * Each of the lists is a queue.
  1274. * The list at @list is reinitialised
  1275. */
  1276. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1277. struct sk_buff_head *head)
  1278. {
  1279. if (!skb_queue_empty(list)) {
  1280. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1281. head->qlen += list->qlen;
  1282. __skb_queue_head_init(list);
  1283. }
  1284. }
  1285. /**
  1286. * __skb_queue_after - queue a buffer at the list head
  1287. * @list: list to use
  1288. * @prev: place after this buffer
  1289. * @newsk: buffer to queue
  1290. *
  1291. * Queue a buffer int the middle of a list. This function takes no locks
  1292. * and you must therefore hold required locks before calling it.
  1293. *
  1294. * A buffer cannot be placed on two lists at the same time.
  1295. */
  1296. static inline void __skb_queue_after(struct sk_buff_head *list,
  1297. struct sk_buff *prev,
  1298. struct sk_buff *newsk)
  1299. {
  1300. __skb_insert(newsk, prev, prev->next, list);
  1301. }
  1302. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1303. struct sk_buff_head *list);
  1304. static inline void __skb_queue_before(struct sk_buff_head *list,
  1305. struct sk_buff *next,
  1306. struct sk_buff *newsk)
  1307. {
  1308. __skb_insert(newsk, next->prev, next, list);
  1309. }
  1310. /**
  1311. * __skb_queue_head - queue a buffer at the list head
  1312. * @list: list to use
  1313. * @newsk: buffer to queue
  1314. *
  1315. * Queue a buffer at the start of a list. This function takes no locks
  1316. * and you must therefore hold required locks before calling it.
  1317. *
  1318. * A buffer cannot be placed on two lists at the same time.
  1319. */
  1320. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1321. static inline void __skb_queue_head(struct sk_buff_head *list,
  1322. struct sk_buff *newsk)
  1323. {
  1324. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1325. }
  1326. /**
  1327. * __skb_queue_tail - queue a buffer at the list tail
  1328. * @list: list to use
  1329. * @newsk: buffer to queue
  1330. *
  1331. * Queue a buffer at the end of a list. This function takes no locks
  1332. * and you must therefore hold required locks before calling it.
  1333. *
  1334. * A buffer cannot be placed on two lists at the same time.
  1335. */
  1336. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1337. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1338. struct sk_buff *newsk)
  1339. {
  1340. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1341. }
  1342. /*
  1343. * remove sk_buff from list. _Must_ be called atomically, and with
  1344. * the list known..
  1345. */
  1346. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1347. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1348. {
  1349. struct sk_buff *next, *prev;
  1350. list->qlen--;
  1351. next = skb->next;
  1352. prev = skb->prev;
  1353. skb->next = skb->prev = NULL;
  1354. next->prev = prev;
  1355. prev->next = next;
  1356. }
  1357. /**
  1358. * __skb_dequeue - remove from the head of the queue
  1359. * @list: list to dequeue from
  1360. *
  1361. * Remove the head of the list. This function does not take any locks
  1362. * so must be used with appropriate locks held only. The head item is
  1363. * returned or %NULL if the list is empty.
  1364. */
  1365. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1366. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1367. {
  1368. struct sk_buff *skb = skb_peek(list);
  1369. if (skb)
  1370. __skb_unlink(skb, list);
  1371. return skb;
  1372. }
  1373. /**
  1374. * __skb_dequeue_tail - remove from the tail of the queue
  1375. * @list: list to dequeue from
  1376. *
  1377. * Remove the tail of the list. This function does not take any locks
  1378. * so must be used with appropriate locks held only. The tail item is
  1379. * returned or %NULL if the list is empty.
  1380. */
  1381. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1382. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1383. {
  1384. struct sk_buff *skb = skb_peek_tail(list);
  1385. if (skb)
  1386. __skb_unlink(skb, list);
  1387. return skb;
  1388. }
  1389. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1390. {
  1391. return skb->data_len;
  1392. }
  1393. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1394. {
  1395. return skb->len - skb->data_len;
  1396. }
  1397. static inline int skb_pagelen(const struct sk_buff *skb)
  1398. {
  1399. int i, len = 0;
  1400. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1401. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1402. return len + skb_headlen(skb);
  1403. }
  1404. /**
  1405. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1406. * @skb: buffer containing fragment to be initialised
  1407. * @i: paged fragment index to initialise
  1408. * @page: the page to use for this fragment
  1409. * @off: the offset to the data with @page
  1410. * @size: the length of the data
  1411. *
  1412. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1413. * offset @off within @page.
  1414. *
  1415. * Does not take any additional reference on the fragment.
  1416. */
  1417. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1418. struct page *page, int off, int size)
  1419. {
  1420. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1421. /*
  1422. * Propagate page pfmemalloc to the skb if we can. The problem is
  1423. * that not all callers have unique ownership of the page but rely
  1424. * on page_is_pfmemalloc doing the right thing(tm).
  1425. */
  1426. frag->page.p = page;
  1427. frag->page_offset = off;
  1428. skb_frag_size_set(frag, size);
  1429. page = compound_head(page);
  1430. if (page_is_pfmemalloc(page))
  1431. skb->pfmemalloc = true;
  1432. }
  1433. /**
  1434. * skb_fill_page_desc - initialise a paged fragment in an skb
  1435. * @skb: buffer containing fragment to be initialised
  1436. * @i: paged fragment index to initialise
  1437. * @page: the page to use for this fragment
  1438. * @off: the offset to the data with @page
  1439. * @size: the length of the data
  1440. *
  1441. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1442. * @skb to point to @size bytes at offset @off within @page. In
  1443. * addition updates @skb such that @i is the last fragment.
  1444. *
  1445. * Does not take any additional reference on the fragment.
  1446. */
  1447. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1448. struct page *page, int off, int size)
  1449. {
  1450. __skb_fill_page_desc(skb, i, page, off, size);
  1451. skb_shinfo(skb)->nr_frags = i + 1;
  1452. }
  1453. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1454. int size, unsigned int truesize);
  1455. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1456. unsigned int truesize);
  1457. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1458. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1459. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1460. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1461. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1462. {
  1463. return skb->head + skb->tail;
  1464. }
  1465. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1466. {
  1467. skb->tail = skb->data - skb->head;
  1468. }
  1469. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1470. {
  1471. skb_reset_tail_pointer(skb);
  1472. skb->tail += offset;
  1473. }
  1474. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1475. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1476. {
  1477. return skb->tail;
  1478. }
  1479. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1480. {
  1481. skb->tail = skb->data;
  1482. }
  1483. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1484. {
  1485. skb->tail = skb->data + offset;
  1486. }
  1487. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1488. /*
  1489. * Add data to an sk_buff
  1490. */
  1491. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1492. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1493. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1494. {
  1495. unsigned char *tmp = skb_tail_pointer(skb);
  1496. SKB_LINEAR_ASSERT(skb);
  1497. skb->tail += len;
  1498. skb->len += len;
  1499. return tmp;
  1500. }
  1501. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1502. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1503. {
  1504. skb->data -= len;
  1505. skb->len += len;
  1506. return skb->data;
  1507. }
  1508. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1509. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1510. {
  1511. skb->len -= len;
  1512. BUG_ON(skb->len < skb->data_len);
  1513. return skb->data += len;
  1514. }
  1515. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1516. {
  1517. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1518. }
  1519. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1520. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1521. {
  1522. if (len > skb_headlen(skb) &&
  1523. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1524. return NULL;
  1525. skb->len -= len;
  1526. return skb->data += len;
  1527. }
  1528. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1529. {
  1530. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1531. }
  1532. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1533. {
  1534. if (likely(len <= skb_headlen(skb)))
  1535. return 1;
  1536. if (unlikely(len > skb->len))
  1537. return 0;
  1538. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1539. }
  1540. /**
  1541. * skb_headroom - bytes at buffer head
  1542. * @skb: buffer to check
  1543. *
  1544. * Return the number of bytes of free space at the head of an &sk_buff.
  1545. */
  1546. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1547. {
  1548. return skb->data - skb->head;
  1549. }
  1550. /**
  1551. * skb_tailroom - bytes at buffer end
  1552. * @skb: buffer to check
  1553. *
  1554. * Return the number of bytes of free space at the tail of an sk_buff
  1555. */
  1556. static inline int skb_tailroom(const struct sk_buff *skb)
  1557. {
  1558. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1559. }
  1560. /**
  1561. * skb_availroom - bytes at buffer end
  1562. * @skb: buffer to check
  1563. *
  1564. * Return the number of bytes of free space at the tail of an sk_buff
  1565. * allocated by sk_stream_alloc()
  1566. */
  1567. static inline int skb_availroom(const struct sk_buff *skb)
  1568. {
  1569. if (skb_is_nonlinear(skb))
  1570. return 0;
  1571. return skb->end - skb->tail - skb->reserved_tailroom;
  1572. }
  1573. /**
  1574. * skb_reserve - adjust headroom
  1575. * @skb: buffer to alter
  1576. * @len: bytes to move
  1577. *
  1578. * Increase the headroom of an empty &sk_buff by reducing the tail
  1579. * room. This is only allowed for an empty buffer.
  1580. */
  1581. static inline void skb_reserve(struct sk_buff *skb, int len)
  1582. {
  1583. skb->data += len;
  1584. skb->tail += len;
  1585. }
  1586. #define ENCAP_TYPE_ETHER 0
  1587. #define ENCAP_TYPE_IPPROTO 1
  1588. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1589. __be16 protocol)
  1590. {
  1591. skb->inner_protocol = protocol;
  1592. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1593. }
  1594. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1595. __u8 ipproto)
  1596. {
  1597. skb->inner_ipproto = ipproto;
  1598. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1599. }
  1600. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1601. {
  1602. skb->inner_mac_header = skb->mac_header;
  1603. skb->inner_network_header = skb->network_header;
  1604. skb->inner_transport_header = skb->transport_header;
  1605. }
  1606. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1607. {
  1608. skb->mac_len = skb->network_header - skb->mac_header;
  1609. }
  1610. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1611. *skb)
  1612. {
  1613. return skb->head + skb->inner_transport_header;
  1614. }
  1615. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1616. {
  1617. skb->inner_transport_header = skb->data - skb->head;
  1618. }
  1619. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1620. const int offset)
  1621. {
  1622. skb_reset_inner_transport_header(skb);
  1623. skb->inner_transport_header += offset;
  1624. }
  1625. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1626. {
  1627. return skb->head + skb->inner_network_header;
  1628. }
  1629. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1630. {
  1631. skb->inner_network_header = skb->data - skb->head;
  1632. }
  1633. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1634. const int offset)
  1635. {
  1636. skb_reset_inner_network_header(skb);
  1637. skb->inner_network_header += offset;
  1638. }
  1639. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1640. {
  1641. return skb->head + skb->inner_mac_header;
  1642. }
  1643. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1644. {
  1645. skb->inner_mac_header = skb->data - skb->head;
  1646. }
  1647. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1648. const int offset)
  1649. {
  1650. skb_reset_inner_mac_header(skb);
  1651. skb->inner_mac_header += offset;
  1652. }
  1653. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1654. {
  1655. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1656. }
  1657. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1658. {
  1659. return skb->head + skb->transport_header;
  1660. }
  1661. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1662. {
  1663. skb->transport_header = skb->data - skb->head;
  1664. }
  1665. static inline void skb_set_transport_header(struct sk_buff *skb,
  1666. const int offset)
  1667. {
  1668. skb_reset_transport_header(skb);
  1669. skb->transport_header += offset;
  1670. }
  1671. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1672. {
  1673. return skb->head + skb->network_header;
  1674. }
  1675. static inline void skb_reset_network_header(struct sk_buff *skb)
  1676. {
  1677. skb->network_header = skb->data - skb->head;
  1678. }
  1679. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1680. {
  1681. skb_reset_network_header(skb);
  1682. skb->network_header += offset;
  1683. }
  1684. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1685. {
  1686. return skb->head + skb->mac_header;
  1687. }
  1688. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1689. {
  1690. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1691. }
  1692. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1693. {
  1694. skb->mac_header = skb->data - skb->head;
  1695. }
  1696. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1697. {
  1698. skb_reset_mac_header(skb);
  1699. skb->mac_header += offset;
  1700. }
  1701. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1702. {
  1703. skb->mac_header = skb->network_header;
  1704. }
  1705. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1706. const int offset_hint)
  1707. {
  1708. struct flow_keys keys;
  1709. if (skb_transport_header_was_set(skb))
  1710. return;
  1711. else if (skb_flow_dissect_flow_keys(skb, &keys))
  1712. skb_set_transport_header(skb, keys.control.thoff);
  1713. else
  1714. skb_set_transport_header(skb, offset_hint);
  1715. }
  1716. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1717. {
  1718. if (skb_mac_header_was_set(skb)) {
  1719. const unsigned char *old_mac = skb_mac_header(skb);
  1720. skb_set_mac_header(skb, -skb->mac_len);
  1721. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1722. }
  1723. }
  1724. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1725. {
  1726. return skb->csum_start - skb_headroom(skb);
  1727. }
  1728. static inline int skb_transport_offset(const struct sk_buff *skb)
  1729. {
  1730. return skb_transport_header(skb) - skb->data;
  1731. }
  1732. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1733. {
  1734. return skb->transport_header - skb->network_header;
  1735. }
  1736. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1737. {
  1738. return skb->inner_transport_header - skb->inner_network_header;
  1739. }
  1740. static inline int skb_network_offset(const struct sk_buff *skb)
  1741. {
  1742. return skb_network_header(skb) - skb->data;
  1743. }
  1744. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1745. {
  1746. return skb_inner_network_header(skb) - skb->data;
  1747. }
  1748. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1749. {
  1750. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1751. }
  1752. /*
  1753. * CPUs often take a performance hit when accessing unaligned memory
  1754. * locations. The actual performance hit varies, it can be small if the
  1755. * hardware handles it or large if we have to take an exception and fix it
  1756. * in software.
  1757. *
  1758. * Since an ethernet header is 14 bytes network drivers often end up with
  1759. * the IP header at an unaligned offset. The IP header can be aligned by
  1760. * shifting the start of the packet by 2 bytes. Drivers should do this
  1761. * with:
  1762. *
  1763. * skb_reserve(skb, NET_IP_ALIGN);
  1764. *
  1765. * The downside to this alignment of the IP header is that the DMA is now
  1766. * unaligned. On some architectures the cost of an unaligned DMA is high
  1767. * and this cost outweighs the gains made by aligning the IP header.
  1768. *
  1769. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1770. * to be overridden.
  1771. */
  1772. #ifndef NET_IP_ALIGN
  1773. #define NET_IP_ALIGN 2
  1774. #endif
  1775. /*
  1776. * The networking layer reserves some headroom in skb data (via
  1777. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1778. * the header has to grow. In the default case, if the header has to grow
  1779. * 32 bytes or less we avoid the reallocation.
  1780. *
  1781. * Unfortunately this headroom changes the DMA alignment of the resulting
  1782. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1783. * on some architectures. An architecture can override this value,
  1784. * perhaps setting it to a cacheline in size (since that will maintain
  1785. * cacheline alignment of the DMA). It must be a power of 2.
  1786. *
  1787. * Various parts of the networking layer expect at least 32 bytes of
  1788. * headroom, you should not reduce this.
  1789. *
  1790. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1791. * to reduce average number of cache lines per packet.
  1792. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1793. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1794. */
  1795. #ifndef NET_SKB_PAD
  1796. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1797. #endif
  1798. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1799. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1800. {
  1801. if (unlikely(skb_is_nonlinear(skb))) {
  1802. WARN_ON(1);
  1803. return;
  1804. }
  1805. skb->len = len;
  1806. skb_set_tail_pointer(skb, len);
  1807. }
  1808. void skb_trim(struct sk_buff *skb, unsigned int len);
  1809. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1810. {
  1811. if (skb->data_len)
  1812. return ___pskb_trim(skb, len);
  1813. __skb_trim(skb, len);
  1814. return 0;
  1815. }
  1816. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1817. {
  1818. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1819. }
  1820. /**
  1821. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1822. * @skb: buffer to alter
  1823. * @len: new length
  1824. *
  1825. * This is identical to pskb_trim except that the caller knows that
  1826. * the skb is not cloned so we should never get an error due to out-
  1827. * of-memory.
  1828. */
  1829. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  1830. {
  1831. int err = pskb_trim(skb, len);
  1832. BUG_ON(err);
  1833. }
  1834. /**
  1835. * skb_orphan - orphan a buffer
  1836. * @skb: buffer to orphan
  1837. *
  1838. * If a buffer currently has an owner then we call the owner's
  1839. * destructor function and make the @skb unowned. The buffer continues
  1840. * to exist but is no longer charged to its former owner.
  1841. */
  1842. static inline void skb_orphan(struct sk_buff *skb)
  1843. {
  1844. if (skb->destructor) {
  1845. skb->destructor(skb);
  1846. skb->destructor = NULL;
  1847. skb->sk = NULL;
  1848. } else {
  1849. BUG_ON(skb->sk);
  1850. }
  1851. }
  1852. /**
  1853. * skb_orphan_frags - orphan the frags contained in a buffer
  1854. * @skb: buffer to orphan frags from
  1855. * @gfp_mask: allocation mask for replacement pages
  1856. *
  1857. * For each frag in the SKB which needs a destructor (i.e. has an
  1858. * owner) create a copy of that frag and release the original
  1859. * page by calling the destructor.
  1860. */
  1861. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  1862. {
  1863. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  1864. return 0;
  1865. return skb_copy_ubufs(skb, gfp_mask);
  1866. }
  1867. /**
  1868. * __skb_queue_purge - empty a list
  1869. * @list: list to empty
  1870. *
  1871. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  1872. * the list and one reference dropped. This function does not take the
  1873. * list lock and the caller must hold the relevant locks to use it.
  1874. */
  1875. void skb_queue_purge(struct sk_buff_head *list);
  1876. static inline void __skb_queue_purge(struct sk_buff_head *list)
  1877. {
  1878. struct sk_buff *skb;
  1879. while ((skb = __skb_dequeue(list)) != NULL)
  1880. kfree_skb(skb);
  1881. }
  1882. void *netdev_alloc_frag(unsigned int fragsz);
  1883. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  1884. gfp_t gfp_mask);
  1885. /**
  1886. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  1887. * @dev: network device to receive on
  1888. * @length: length to allocate
  1889. *
  1890. * Allocate a new &sk_buff and assign it a usage count of one. The
  1891. * buffer has unspecified headroom built in. Users should allocate
  1892. * the headroom they think they need without accounting for the
  1893. * built in space. The built in space is used for optimisations.
  1894. *
  1895. * %NULL is returned if there is no free memory. Although this function
  1896. * allocates memory it can be called from an interrupt.
  1897. */
  1898. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  1899. unsigned int length)
  1900. {
  1901. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  1902. }
  1903. /* legacy helper around __netdev_alloc_skb() */
  1904. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  1905. gfp_t gfp_mask)
  1906. {
  1907. return __netdev_alloc_skb(NULL, length, gfp_mask);
  1908. }
  1909. /* legacy helper around netdev_alloc_skb() */
  1910. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  1911. {
  1912. return netdev_alloc_skb(NULL, length);
  1913. }
  1914. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  1915. unsigned int length, gfp_t gfp)
  1916. {
  1917. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  1918. if (NET_IP_ALIGN && skb)
  1919. skb_reserve(skb, NET_IP_ALIGN);
  1920. return skb;
  1921. }
  1922. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  1923. unsigned int length)
  1924. {
  1925. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  1926. }
  1927. static inline void skb_free_frag(void *addr)
  1928. {
  1929. __free_page_frag(addr);
  1930. }
  1931. void *napi_alloc_frag(unsigned int fragsz);
  1932. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  1933. unsigned int length, gfp_t gfp_mask);
  1934. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  1935. unsigned int length)
  1936. {
  1937. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  1938. }
  1939. /**
  1940. * __dev_alloc_pages - allocate page for network Rx
  1941. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1942. * @order: size of the allocation
  1943. *
  1944. * Allocate a new page.
  1945. *
  1946. * %NULL is returned if there is no free memory.
  1947. */
  1948. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  1949. unsigned int order)
  1950. {
  1951. /* This piece of code contains several assumptions.
  1952. * 1. This is for device Rx, therefor a cold page is preferred.
  1953. * 2. The expectation is the user wants a compound page.
  1954. * 3. If requesting a order 0 page it will not be compound
  1955. * due to the check to see if order has a value in prep_new_page
  1956. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  1957. * code in gfp_to_alloc_flags that should be enforcing this.
  1958. */
  1959. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  1960. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  1961. }
  1962. static inline struct page *dev_alloc_pages(unsigned int order)
  1963. {
  1964. return __dev_alloc_pages(GFP_ATOMIC, order);
  1965. }
  1966. /**
  1967. * __dev_alloc_page - allocate a page for network Rx
  1968. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1969. *
  1970. * Allocate a new page.
  1971. *
  1972. * %NULL is returned if there is no free memory.
  1973. */
  1974. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  1975. {
  1976. return __dev_alloc_pages(gfp_mask, 0);
  1977. }
  1978. static inline struct page *dev_alloc_page(void)
  1979. {
  1980. return __dev_alloc_page(GFP_ATOMIC);
  1981. }
  1982. /**
  1983. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  1984. * @page: The page that was allocated from skb_alloc_page
  1985. * @skb: The skb that may need pfmemalloc set
  1986. */
  1987. static inline void skb_propagate_pfmemalloc(struct page *page,
  1988. struct sk_buff *skb)
  1989. {
  1990. if (page_is_pfmemalloc(page))
  1991. skb->pfmemalloc = true;
  1992. }
  1993. /**
  1994. * skb_frag_page - retrieve the page referred to by a paged fragment
  1995. * @frag: the paged fragment
  1996. *
  1997. * Returns the &struct page associated with @frag.
  1998. */
  1999. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2000. {
  2001. return frag->page.p;
  2002. }
  2003. /**
  2004. * __skb_frag_ref - take an addition reference on a paged fragment.
  2005. * @frag: the paged fragment
  2006. *
  2007. * Takes an additional reference on the paged fragment @frag.
  2008. */
  2009. static inline void __skb_frag_ref(skb_frag_t *frag)
  2010. {
  2011. get_page(skb_frag_page(frag));
  2012. }
  2013. /**
  2014. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2015. * @skb: the buffer
  2016. * @f: the fragment offset.
  2017. *
  2018. * Takes an additional reference on the @f'th paged fragment of @skb.
  2019. */
  2020. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2021. {
  2022. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2023. }
  2024. /**
  2025. * __skb_frag_unref - release a reference on a paged fragment.
  2026. * @frag: the paged fragment
  2027. *
  2028. * Releases a reference on the paged fragment @frag.
  2029. */
  2030. static inline void __skb_frag_unref(skb_frag_t *frag)
  2031. {
  2032. put_page(skb_frag_page(frag));
  2033. }
  2034. /**
  2035. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2036. * @skb: the buffer
  2037. * @f: the fragment offset
  2038. *
  2039. * Releases a reference on the @f'th paged fragment of @skb.
  2040. */
  2041. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2042. {
  2043. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2044. }
  2045. /**
  2046. * skb_frag_address - gets the address of the data contained in a paged fragment
  2047. * @frag: the paged fragment buffer
  2048. *
  2049. * Returns the address of the data within @frag. The page must already
  2050. * be mapped.
  2051. */
  2052. static inline void *skb_frag_address(const skb_frag_t *frag)
  2053. {
  2054. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2055. }
  2056. /**
  2057. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2058. * @frag: the paged fragment buffer
  2059. *
  2060. * Returns the address of the data within @frag. Checks that the page
  2061. * is mapped and returns %NULL otherwise.
  2062. */
  2063. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2064. {
  2065. void *ptr = page_address(skb_frag_page(frag));
  2066. if (unlikely(!ptr))
  2067. return NULL;
  2068. return ptr + frag->page_offset;
  2069. }
  2070. /**
  2071. * __skb_frag_set_page - sets the page contained in a paged fragment
  2072. * @frag: the paged fragment
  2073. * @page: the page to set
  2074. *
  2075. * Sets the fragment @frag to contain @page.
  2076. */
  2077. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2078. {
  2079. frag->page.p = page;
  2080. }
  2081. /**
  2082. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2083. * @skb: the buffer
  2084. * @f: the fragment offset
  2085. * @page: the page to set
  2086. *
  2087. * Sets the @f'th fragment of @skb to contain @page.
  2088. */
  2089. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2090. struct page *page)
  2091. {
  2092. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2093. }
  2094. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2095. /**
  2096. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2097. * @dev: the device to map the fragment to
  2098. * @frag: the paged fragment to map
  2099. * @offset: the offset within the fragment (starting at the
  2100. * fragment's own offset)
  2101. * @size: the number of bytes to map
  2102. * @dir: the direction of the mapping (%PCI_DMA_*)
  2103. *
  2104. * Maps the page associated with @frag to @device.
  2105. */
  2106. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2107. const skb_frag_t *frag,
  2108. size_t offset, size_t size,
  2109. enum dma_data_direction dir)
  2110. {
  2111. return dma_map_page(dev, skb_frag_page(frag),
  2112. frag->page_offset + offset, size, dir);
  2113. }
  2114. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2115. gfp_t gfp_mask)
  2116. {
  2117. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2118. }
  2119. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2120. gfp_t gfp_mask)
  2121. {
  2122. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2123. }
  2124. /**
  2125. * skb_clone_writable - is the header of a clone writable
  2126. * @skb: buffer to check
  2127. * @len: length up to which to write
  2128. *
  2129. * Returns true if modifying the header part of the cloned buffer
  2130. * does not requires the data to be copied.
  2131. */
  2132. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2133. {
  2134. return !skb_header_cloned(skb) &&
  2135. skb_headroom(skb) + len <= skb->hdr_len;
  2136. }
  2137. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2138. int cloned)
  2139. {
  2140. int delta = 0;
  2141. if (headroom > skb_headroom(skb))
  2142. delta = headroom - skb_headroom(skb);
  2143. if (delta || cloned)
  2144. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2145. GFP_ATOMIC);
  2146. return 0;
  2147. }
  2148. /**
  2149. * skb_cow - copy header of skb when it is required
  2150. * @skb: buffer to cow
  2151. * @headroom: needed headroom
  2152. *
  2153. * If the skb passed lacks sufficient headroom or its data part
  2154. * is shared, data is reallocated. If reallocation fails, an error
  2155. * is returned and original skb is not changed.
  2156. *
  2157. * The result is skb with writable area skb->head...skb->tail
  2158. * and at least @headroom of space at head.
  2159. */
  2160. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2161. {
  2162. return __skb_cow(skb, headroom, skb_cloned(skb));
  2163. }
  2164. /**
  2165. * skb_cow_head - skb_cow but only making the head writable
  2166. * @skb: buffer to cow
  2167. * @headroom: needed headroom
  2168. *
  2169. * This function is identical to skb_cow except that we replace the
  2170. * skb_cloned check by skb_header_cloned. It should be used when
  2171. * you only need to push on some header and do not need to modify
  2172. * the data.
  2173. */
  2174. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2175. {
  2176. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2177. }
  2178. /**
  2179. * skb_padto - pad an skbuff up to a minimal size
  2180. * @skb: buffer to pad
  2181. * @len: minimal length
  2182. *
  2183. * Pads up a buffer to ensure the trailing bytes exist and are
  2184. * blanked. If the buffer already contains sufficient data it
  2185. * is untouched. Otherwise it is extended. Returns zero on
  2186. * success. The skb is freed on error.
  2187. */
  2188. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2189. {
  2190. unsigned int size = skb->len;
  2191. if (likely(size >= len))
  2192. return 0;
  2193. return skb_pad(skb, len - size);
  2194. }
  2195. /**
  2196. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2197. * @skb: buffer to pad
  2198. * @len: minimal length
  2199. *
  2200. * Pads up a buffer to ensure the trailing bytes exist and are
  2201. * blanked. If the buffer already contains sufficient data it
  2202. * is untouched. Otherwise it is extended. Returns zero on
  2203. * success. The skb is freed on error.
  2204. */
  2205. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2206. {
  2207. unsigned int size = skb->len;
  2208. if (unlikely(size < len)) {
  2209. len -= size;
  2210. if (skb_pad(skb, len))
  2211. return -ENOMEM;
  2212. __skb_put(skb, len);
  2213. }
  2214. return 0;
  2215. }
  2216. static inline int skb_add_data(struct sk_buff *skb,
  2217. struct iov_iter *from, int copy)
  2218. {
  2219. const int off = skb->len;
  2220. if (skb->ip_summed == CHECKSUM_NONE) {
  2221. __wsum csum = 0;
  2222. if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
  2223. &csum, from) == copy) {
  2224. skb->csum = csum_block_add(skb->csum, csum, off);
  2225. return 0;
  2226. }
  2227. } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
  2228. return 0;
  2229. __skb_trim(skb, off);
  2230. return -EFAULT;
  2231. }
  2232. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2233. const struct page *page, int off)
  2234. {
  2235. if (i) {
  2236. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2237. return page == skb_frag_page(frag) &&
  2238. off == frag->page_offset + skb_frag_size(frag);
  2239. }
  2240. return false;
  2241. }
  2242. static inline int __skb_linearize(struct sk_buff *skb)
  2243. {
  2244. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2245. }
  2246. /**
  2247. * skb_linearize - convert paged skb to linear one
  2248. * @skb: buffer to linarize
  2249. *
  2250. * If there is no free memory -ENOMEM is returned, otherwise zero
  2251. * is returned and the old skb data released.
  2252. */
  2253. static inline int skb_linearize(struct sk_buff *skb)
  2254. {
  2255. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2256. }
  2257. /**
  2258. * skb_has_shared_frag - can any frag be overwritten
  2259. * @skb: buffer to test
  2260. *
  2261. * Return true if the skb has at least one frag that might be modified
  2262. * by an external entity (as in vmsplice()/sendfile())
  2263. */
  2264. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2265. {
  2266. return skb_is_nonlinear(skb) &&
  2267. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2268. }
  2269. /**
  2270. * skb_linearize_cow - make sure skb is linear and writable
  2271. * @skb: buffer to process
  2272. *
  2273. * If there is no free memory -ENOMEM is returned, otherwise zero
  2274. * is returned and the old skb data released.
  2275. */
  2276. static inline int skb_linearize_cow(struct sk_buff *skb)
  2277. {
  2278. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2279. __skb_linearize(skb) : 0;
  2280. }
  2281. /**
  2282. * skb_postpull_rcsum - update checksum for received skb after pull
  2283. * @skb: buffer to update
  2284. * @start: start of data before pull
  2285. * @len: length of data pulled
  2286. *
  2287. * After doing a pull on a received packet, you need to call this to
  2288. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2289. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2290. */
  2291. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2292. const void *start, unsigned int len)
  2293. {
  2294. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2295. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2296. }
  2297. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2298. /**
  2299. * pskb_trim_rcsum - trim received skb and update checksum
  2300. * @skb: buffer to trim
  2301. * @len: new length
  2302. *
  2303. * This is exactly the same as pskb_trim except that it ensures the
  2304. * checksum of received packets are still valid after the operation.
  2305. */
  2306. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2307. {
  2308. if (likely(len >= skb->len))
  2309. return 0;
  2310. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2311. skb->ip_summed = CHECKSUM_NONE;
  2312. return __pskb_trim(skb, len);
  2313. }
  2314. #define skb_queue_walk(queue, skb) \
  2315. for (skb = (queue)->next; \
  2316. skb != (struct sk_buff *)(queue); \
  2317. skb = skb->next)
  2318. #define skb_queue_walk_safe(queue, skb, tmp) \
  2319. for (skb = (queue)->next, tmp = skb->next; \
  2320. skb != (struct sk_buff *)(queue); \
  2321. skb = tmp, tmp = skb->next)
  2322. #define skb_queue_walk_from(queue, skb) \
  2323. for (; skb != (struct sk_buff *)(queue); \
  2324. skb = skb->next)
  2325. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2326. for (tmp = skb->next; \
  2327. skb != (struct sk_buff *)(queue); \
  2328. skb = tmp, tmp = skb->next)
  2329. #define skb_queue_reverse_walk(queue, skb) \
  2330. for (skb = (queue)->prev; \
  2331. skb != (struct sk_buff *)(queue); \
  2332. skb = skb->prev)
  2333. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2334. for (skb = (queue)->prev, tmp = skb->prev; \
  2335. skb != (struct sk_buff *)(queue); \
  2336. skb = tmp, tmp = skb->prev)
  2337. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2338. for (tmp = skb->prev; \
  2339. skb != (struct sk_buff *)(queue); \
  2340. skb = tmp, tmp = skb->prev)
  2341. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2342. {
  2343. return skb_shinfo(skb)->frag_list != NULL;
  2344. }
  2345. static inline void skb_frag_list_init(struct sk_buff *skb)
  2346. {
  2347. skb_shinfo(skb)->frag_list = NULL;
  2348. }
  2349. static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
  2350. {
  2351. frag->next = skb_shinfo(skb)->frag_list;
  2352. skb_shinfo(skb)->frag_list = frag;
  2353. }
  2354. #define skb_walk_frags(skb, iter) \
  2355. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2356. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2357. int *peeked, int *off, int *err);
  2358. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2359. int *err);
  2360. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2361. struct poll_table_struct *wait);
  2362. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2363. struct iov_iter *to, int size);
  2364. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2365. struct msghdr *msg, int size)
  2366. {
  2367. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2368. }
  2369. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2370. struct msghdr *msg);
  2371. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2372. struct iov_iter *from, int len);
  2373. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2374. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2375. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2376. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2377. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2378. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2379. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2380. int len, __wsum csum);
  2381. ssize_t skb_socket_splice(struct sock *sk,
  2382. struct pipe_inode_info *pipe,
  2383. struct splice_pipe_desc *spd);
  2384. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2385. struct pipe_inode_info *pipe, unsigned int len,
  2386. unsigned int flags,
  2387. ssize_t (*splice_cb)(struct sock *,
  2388. struct pipe_inode_info *,
  2389. struct splice_pipe_desc *));
  2390. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2391. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2392. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2393. int len, int hlen);
  2394. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2395. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2396. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2397. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2398. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2399. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2400. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2401. int skb_vlan_pop(struct sk_buff *skb);
  2402. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2403. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2404. {
  2405. return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2406. }
  2407. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2408. {
  2409. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2410. }
  2411. struct skb_checksum_ops {
  2412. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2413. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2414. };
  2415. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2416. __wsum csum, const struct skb_checksum_ops *ops);
  2417. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2418. __wsum csum);
  2419. static inline void * __must_check
  2420. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2421. int len, void *data, int hlen, void *buffer)
  2422. {
  2423. if (hlen - offset >= len)
  2424. return data + offset;
  2425. if (!skb ||
  2426. skb_copy_bits(skb, offset, buffer, len) < 0)
  2427. return NULL;
  2428. return buffer;
  2429. }
  2430. static inline void * __must_check
  2431. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2432. {
  2433. return __skb_header_pointer(skb, offset, len, skb->data,
  2434. skb_headlen(skb), buffer);
  2435. }
  2436. /**
  2437. * skb_needs_linearize - check if we need to linearize a given skb
  2438. * depending on the given device features.
  2439. * @skb: socket buffer to check
  2440. * @features: net device features
  2441. *
  2442. * Returns true if either:
  2443. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2444. * 2. skb is fragmented and the device does not support SG.
  2445. */
  2446. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2447. netdev_features_t features)
  2448. {
  2449. return skb_is_nonlinear(skb) &&
  2450. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2451. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2452. }
  2453. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2454. void *to,
  2455. const unsigned int len)
  2456. {
  2457. memcpy(to, skb->data, len);
  2458. }
  2459. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2460. const int offset, void *to,
  2461. const unsigned int len)
  2462. {
  2463. memcpy(to, skb->data + offset, len);
  2464. }
  2465. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2466. const void *from,
  2467. const unsigned int len)
  2468. {
  2469. memcpy(skb->data, from, len);
  2470. }
  2471. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2472. const int offset,
  2473. const void *from,
  2474. const unsigned int len)
  2475. {
  2476. memcpy(skb->data + offset, from, len);
  2477. }
  2478. void skb_init(void);
  2479. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2480. {
  2481. return skb->tstamp;
  2482. }
  2483. /**
  2484. * skb_get_timestamp - get timestamp from a skb
  2485. * @skb: skb to get stamp from
  2486. * @stamp: pointer to struct timeval to store stamp in
  2487. *
  2488. * Timestamps are stored in the skb as offsets to a base timestamp.
  2489. * This function converts the offset back to a struct timeval and stores
  2490. * it in stamp.
  2491. */
  2492. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2493. struct timeval *stamp)
  2494. {
  2495. *stamp = ktime_to_timeval(skb->tstamp);
  2496. }
  2497. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2498. struct timespec *stamp)
  2499. {
  2500. *stamp = ktime_to_timespec(skb->tstamp);
  2501. }
  2502. static inline void __net_timestamp(struct sk_buff *skb)
  2503. {
  2504. skb->tstamp = ktime_get_real();
  2505. }
  2506. static inline ktime_t net_timedelta(ktime_t t)
  2507. {
  2508. return ktime_sub(ktime_get_real(), t);
  2509. }
  2510. static inline ktime_t net_invalid_timestamp(void)
  2511. {
  2512. return ktime_set(0, 0);
  2513. }
  2514. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2515. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2516. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2517. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2518. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2519. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2520. {
  2521. }
  2522. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2523. {
  2524. return false;
  2525. }
  2526. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2527. /**
  2528. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2529. *
  2530. * PHY drivers may accept clones of transmitted packets for
  2531. * timestamping via their phy_driver.txtstamp method. These drivers
  2532. * must call this function to return the skb back to the stack with a
  2533. * timestamp.
  2534. *
  2535. * @skb: clone of the the original outgoing packet
  2536. * @hwtstamps: hardware time stamps
  2537. *
  2538. */
  2539. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2540. struct skb_shared_hwtstamps *hwtstamps);
  2541. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2542. struct skb_shared_hwtstamps *hwtstamps,
  2543. struct sock *sk, int tstype);
  2544. /**
  2545. * skb_tstamp_tx - queue clone of skb with send time stamps
  2546. * @orig_skb: the original outgoing packet
  2547. * @hwtstamps: hardware time stamps, may be NULL if not available
  2548. *
  2549. * If the skb has a socket associated, then this function clones the
  2550. * skb (thus sharing the actual data and optional structures), stores
  2551. * the optional hardware time stamping information (if non NULL) or
  2552. * generates a software time stamp (otherwise), then queues the clone
  2553. * to the error queue of the socket. Errors are silently ignored.
  2554. */
  2555. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2556. struct skb_shared_hwtstamps *hwtstamps);
  2557. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2558. {
  2559. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2560. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2561. skb_tstamp_tx(skb, NULL);
  2562. }
  2563. /**
  2564. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2565. *
  2566. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2567. * function immediately before giving the sk_buff to the MAC hardware.
  2568. *
  2569. * Specifically, one should make absolutely sure that this function is
  2570. * called before TX completion of this packet can trigger. Otherwise
  2571. * the packet could potentially already be freed.
  2572. *
  2573. * @skb: A socket buffer.
  2574. */
  2575. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2576. {
  2577. skb_clone_tx_timestamp(skb);
  2578. sw_tx_timestamp(skb);
  2579. }
  2580. /**
  2581. * skb_complete_wifi_ack - deliver skb with wifi status
  2582. *
  2583. * @skb: the original outgoing packet
  2584. * @acked: ack status
  2585. *
  2586. */
  2587. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2588. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2589. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2590. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2591. {
  2592. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2593. skb->csum_valid ||
  2594. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2595. skb_checksum_start_offset(skb) >= 0));
  2596. }
  2597. /**
  2598. * skb_checksum_complete - Calculate checksum of an entire packet
  2599. * @skb: packet to process
  2600. *
  2601. * This function calculates the checksum over the entire packet plus
  2602. * the value of skb->csum. The latter can be used to supply the
  2603. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2604. * checksum.
  2605. *
  2606. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2607. * this function can be used to verify that checksum on received
  2608. * packets. In that case the function should return zero if the
  2609. * checksum is correct. In particular, this function will return zero
  2610. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2611. * hardware has already verified the correctness of the checksum.
  2612. */
  2613. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2614. {
  2615. return skb_csum_unnecessary(skb) ?
  2616. 0 : __skb_checksum_complete(skb);
  2617. }
  2618. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2619. {
  2620. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2621. if (skb->csum_level == 0)
  2622. skb->ip_summed = CHECKSUM_NONE;
  2623. else
  2624. skb->csum_level--;
  2625. }
  2626. }
  2627. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2628. {
  2629. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2630. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2631. skb->csum_level++;
  2632. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2633. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2634. skb->csum_level = 0;
  2635. }
  2636. }
  2637. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2638. {
  2639. /* Mark current checksum as bad (typically called from GRO
  2640. * path). In the case that ip_summed is CHECKSUM_NONE
  2641. * this must be the first checksum encountered in the packet.
  2642. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2643. * checksum after the last one validated. For UDP, a zero
  2644. * checksum can not be marked as bad.
  2645. */
  2646. if (skb->ip_summed == CHECKSUM_NONE ||
  2647. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2648. skb->csum_bad = 1;
  2649. }
  2650. /* Check if we need to perform checksum complete validation.
  2651. *
  2652. * Returns true if checksum complete is needed, false otherwise
  2653. * (either checksum is unnecessary or zero checksum is allowed).
  2654. */
  2655. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2656. bool zero_okay,
  2657. __sum16 check)
  2658. {
  2659. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2660. skb->csum_valid = 1;
  2661. __skb_decr_checksum_unnecessary(skb);
  2662. return false;
  2663. }
  2664. return true;
  2665. }
  2666. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2667. * in checksum_init.
  2668. */
  2669. #define CHECKSUM_BREAK 76
  2670. /* Unset checksum-complete
  2671. *
  2672. * Unset checksum complete can be done when packet is being modified
  2673. * (uncompressed for instance) and checksum-complete value is
  2674. * invalidated.
  2675. */
  2676. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2677. {
  2678. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2679. skb->ip_summed = CHECKSUM_NONE;
  2680. }
  2681. /* Validate (init) checksum based on checksum complete.
  2682. *
  2683. * Return values:
  2684. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2685. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2686. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2687. * non-zero: value of invalid checksum
  2688. *
  2689. */
  2690. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2691. bool complete,
  2692. __wsum psum)
  2693. {
  2694. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2695. if (!csum_fold(csum_add(psum, skb->csum))) {
  2696. skb->csum_valid = 1;
  2697. return 0;
  2698. }
  2699. } else if (skb->csum_bad) {
  2700. /* ip_summed == CHECKSUM_NONE in this case */
  2701. return (__force __sum16)1;
  2702. }
  2703. skb->csum = psum;
  2704. if (complete || skb->len <= CHECKSUM_BREAK) {
  2705. __sum16 csum;
  2706. csum = __skb_checksum_complete(skb);
  2707. skb->csum_valid = !csum;
  2708. return csum;
  2709. }
  2710. return 0;
  2711. }
  2712. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2713. {
  2714. return 0;
  2715. }
  2716. /* Perform checksum validate (init). Note that this is a macro since we only
  2717. * want to calculate the pseudo header which is an input function if necessary.
  2718. * First we try to validate without any computation (checksum unnecessary) and
  2719. * then calculate based on checksum complete calling the function to compute
  2720. * pseudo header.
  2721. *
  2722. * Return values:
  2723. * 0: checksum is validated or try to in skb_checksum_complete
  2724. * non-zero: value of invalid checksum
  2725. */
  2726. #define __skb_checksum_validate(skb, proto, complete, \
  2727. zero_okay, check, compute_pseudo) \
  2728. ({ \
  2729. __sum16 __ret = 0; \
  2730. skb->csum_valid = 0; \
  2731. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2732. __ret = __skb_checksum_validate_complete(skb, \
  2733. complete, compute_pseudo(skb, proto)); \
  2734. __ret; \
  2735. })
  2736. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2737. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2738. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2739. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2740. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2741. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2742. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2743. compute_pseudo) \
  2744. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  2745. #define skb_checksum_simple_validate(skb) \
  2746. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2747. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2748. {
  2749. return (skb->ip_summed == CHECKSUM_NONE &&
  2750. skb->csum_valid && !skb->csum_bad);
  2751. }
  2752. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2753. __sum16 check, __wsum pseudo)
  2754. {
  2755. skb->csum = ~pseudo;
  2756. skb->ip_summed = CHECKSUM_COMPLETE;
  2757. }
  2758. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  2759. do { \
  2760. if (__skb_checksum_convert_check(skb)) \
  2761. __skb_checksum_convert(skb, check, \
  2762. compute_pseudo(skb, proto)); \
  2763. } while (0)
  2764. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  2765. u16 start, u16 offset)
  2766. {
  2767. skb->ip_summed = CHECKSUM_PARTIAL;
  2768. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  2769. skb->csum_offset = offset - start;
  2770. }
  2771. /* Update skbuf and packet to reflect the remote checksum offload operation.
  2772. * When called, ptr indicates the starting point for skb->csum when
  2773. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  2774. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  2775. */
  2776. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  2777. int start, int offset, bool nopartial)
  2778. {
  2779. __wsum delta;
  2780. if (!nopartial) {
  2781. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  2782. return;
  2783. }
  2784. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  2785. __skb_checksum_complete(skb);
  2786. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  2787. }
  2788. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  2789. /* Adjust skb->csum since we changed the packet */
  2790. skb->csum = csum_add(skb->csum, delta);
  2791. }
  2792. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2793. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2794. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2795. {
  2796. if (nfct && atomic_dec_and_test(&nfct->use))
  2797. nf_conntrack_destroy(nfct);
  2798. }
  2799. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2800. {
  2801. if (nfct)
  2802. atomic_inc(&nfct->use);
  2803. }
  2804. #endif
  2805. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2806. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  2807. {
  2808. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  2809. kfree(nf_bridge);
  2810. }
  2811. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  2812. {
  2813. if (nf_bridge)
  2814. atomic_inc(&nf_bridge->use);
  2815. }
  2816. #endif /* CONFIG_BRIDGE_NETFILTER */
  2817. static inline void nf_reset(struct sk_buff *skb)
  2818. {
  2819. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2820. nf_conntrack_put(skb->nfct);
  2821. skb->nfct = NULL;
  2822. #endif
  2823. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2824. nf_bridge_put(skb->nf_bridge);
  2825. skb->nf_bridge = NULL;
  2826. #endif
  2827. }
  2828. static inline void nf_reset_trace(struct sk_buff *skb)
  2829. {
  2830. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2831. skb->nf_trace = 0;
  2832. #endif
  2833. }
  2834. /* Note: This doesn't put any conntrack and bridge info in dst. */
  2835. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  2836. bool copy)
  2837. {
  2838. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2839. dst->nfct = src->nfct;
  2840. nf_conntrack_get(src->nfct);
  2841. if (copy)
  2842. dst->nfctinfo = src->nfctinfo;
  2843. #endif
  2844. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2845. dst->nf_bridge = src->nf_bridge;
  2846. nf_bridge_get(src->nf_bridge);
  2847. #endif
  2848. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2849. if (copy)
  2850. dst->nf_trace = src->nf_trace;
  2851. #endif
  2852. }
  2853. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2854. {
  2855. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2856. nf_conntrack_put(dst->nfct);
  2857. #endif
  2858. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2859. nf_bridge_put(dst->nf_bridge);
  2860. #endif
  2861. __nf_copy(dst, src, true);
  2862. }
  2863. #ifdef CONFIG_NETWORK_SECMARK
  2864. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2865. {
  2866. to->secmark = from->secmark;
  2867. }
  2868. static inline void skb_init_secmark(struct sk_buff *skb)
  2869. {
  2870. skb->secmark = 0;
  2871. }
  2872. #else
  2873. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2874. { }
  2875. static inline void skb_init_secmark(struct sk_buff *skb)
  2876. { }
  2877. #endif
  2878. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  2879. {
  2880. return !skb->destructor &&
  2881. #if IS_ENABLED(CONFIG_XFRM)
  2882. !skb->sp &&
  2883. #endif
  2884. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  2885. !skb->nfct &&
  2886. #endif
  2887. !skb->_skb_refdst &&
  2888. !skb_has_frag_list(skb);
  2889. }
  2890. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  2891. {
  2892. skb->queue_mapping = queue_mapping;
  2893. }
  2894. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  2895. {
  2896. return skb->queue_mapping;
  2897. }
  2898. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  2899. {
  2900. to->queue_mapping = from->queue_mapping;
  2901. }
  2902. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  2903. {
  2904. skb->queue_mapping = rx_queue + 1;
  2905. }
  2906. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  2907. {
  2908. return skb->queue_mapping - 1;
  2909. }
  2910. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  2911. {
  2912. return skb->queue_mapping != 0;
  2913. }
  2914. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  2915. {
  2916. #ifdef CONFIG_XFRM
  2917. return skb->sp;
  2918. #else
  2919. return NULL;
  2920. #endif
  2921. }
  2922. /* Keeps track of mac header offset relative to skb->head.
  2923. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  2924. * For non-tunnel skb it points to skb_mac_header() and for
  2925. * tunnel skb it points to outer mac header.
  2926. * Keeps track of level of encapsulation of network headers.
  2927. */
  2928. struct skb_gso_cb {
  2929. int mac_offset;
  2930. int encap_level;
  2931. __u16 csum_start;
  2932. };
  2933. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
  2934. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  2935. {
  2936. return (skb_mac_header(inner_skb) - inner_skb->head) -
  2937. SKB_GSO_CB(inner_skb)->mac_offset;
  2938. }
  2939. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  2940. {
  2941. int new_headroom, headroom;
  2942. int ret;
  2943. headroom = skb_headroom(skb);
  2944. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  2945. if (ret)
  2946. return ret;
  2947. new_headroom = skb_headroom(skb);
  2948. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  2949. return 0;
  2950. }
  2951. /* Compute the checksum for a gso segment. First compute the checksum value
  2952. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  2953. * then add in skb->csum (checksum from csum_start to end of packet).
  2954. * skb->csum and csum_start are then updated to reflect the checksum of the
  2955. * resultant packet starting from the transport header-- the resultant checksum
  2956. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  2957. * header.
  2958. */
  2959. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  2960. {
  2961. int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
  2962. skb_transport_offset(skb);
  2963. __wsum partial;
  2964. partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
  2965. skb->csum = res;
  2966. SKB_GSO_CB(skb)->csum_start -= plen;
  2967. return csum_fold(partial);
  2968. }
  2969. static inline bool skb_is_gso(const struct sk_buff *skb)
  2970. {
  2971. return skb_shinfo(skb)->gso_size;
  2972. }
  2973. /* Note: Should be called only if skb_is_gso(skb) is true */
  2974. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  2975. {
  2976. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  2977. }
  2978. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  2979. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  2980. {
  2981. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  2982. * wanted then gso_type will be set. */
  2983. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  2984. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  2985. unlikely(shinfo->gso_type == 0)) {
  2986. __skb_warn_lro_forwarding(skb);
  2987. return true;
  2988. }
  2989. return false;
  2990. }
  2991. static inline void skb_forward_csum(struct sk_buff *skb)
  2992. {
  2993. /* Unfortunately we don't support this one. Any brave souls? */
  2994. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2995. skb->ip_summed = CHECKSUM_NONE;
  2996. }
  2997. /**
  2998. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  2999. * @skb: skb to check
  3000. *
  3001. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3002. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3003. * use this helper, to document places where we make this assertion.
  3004. */
  3005. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3006. {
  3007. #ifdef DEBUG
  3008. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3009. #endif
  3010. }
  3011. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3012. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3013. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3014. unsigned int transport_len,
  3015. __sum16(*skb_chkf)(struct sk_buff *skb));
  3016. /**
  3017. * skb_head_is_locked - Determine if the skb->head is locked down
  3018. * @skb: skb to check
  3019. *
  3020. * The head on skbs build around a head frag can be removed if they are
  3021. * not cloned. This function returns true if the skb head is locked down
  3022. * due to either being allocated via kmalloc, or by being a clone with
  3023. * multiple references to the head.
  3024. */
  3025. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3026. {
  3027. return !skb->head_frag || skb_cloned(skb);
  3028. }
  3029. /**
  3030. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3031. *
  3032. * @skb: GSO skb
  3033. *
  3034. * skb_gso_network_seglen is used to determine the real size of the
  3035. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3036. *
  3037. * The MAC/L2 header is not accounted for.
  3038. */
  3039. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3040. {
  3041. unsigned int hdr_len = skb_transport_header(skb) -
  3042. skb_network_header(skb);
  3043. return hdr_len + skb_gso_transport_seglen(skb);
  3044. }
  3045. #endif /* __KERNEL__ */
  3046. #endif /* _LINUX_SKBUFF_H */