skbuff.h 113 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460246124622463246424652466246724682469247024712472247324742475247624772478247924802481248224832484248524862487248824892490249124922493249424952496249724982499250025012502250325042505250625072508250925102511251225132514251525162517251825192520252125222523252425252526252725282529253025312532253325342535253625372538253925402541254225432544254525462547254825492550255125522553255425552556255725582559256025612562256325642565256625672568256925702571257225732574257525762577257825792580258125822583258425852586258725882589259025912592259325942595259625972598259926002601260226032604260526062607260826092610261126122613261426152616261726182619262026212622262326242625262626272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697269826992700270127022703270427052706270727082709271027112712271327142715271627172718271927202721272227232724272527262727272827292730273127322733273427352736273727382739274027412742274327442745274627472748274927502751275227532754275527562757275827592760276127622763276427652766276727682769277027712772277327742775277627772778277927802781278227832784278527862787278827892790279127922793279427952796279727982799280028012802280328042805280628072808280928102811281228132814281528162817281828192820282128222823282428252826282728282829283028312832283328342835283628372838283928402841284228432844284528462847284828492850285128522853285428552856285728582859286028612862286328642865286628672868286928702871287228732874287528762877287828792880288128822883288428852886288728882889289028912892289328942895289628972898289929002901290229032904290529062907290829092910291129122913291429152916291729182919292029212922292329242925292629272928292929302931293229332934293529362937293829392940294129422943294429452946294729482949295029512952295329542955295629572958295929602961296229632964296529662967296829692970297129722973297429752976297729782979298029812982298329842985298629872988298929902991299229932994299529962997299829993000300130023003300430053006300730083009301030113012301330143015301630173018301930203021302230233024302530263027302830293030303130323033303430353036303730383039304030413042304330443045304630473048304930503051305230533054305530563057305830593060306130623063306430653066306730683069307030713072307330743075307630773078307930803081308230833084308530863087308830893090309130923093309430953096309730983099310031013102310331043105310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135313631373138313931403141314231433144314531463147314831493150315131523153315431553156315731583159316031613162316331643165316631673168316931703171317231733174317531763177317831793180318131823183318431853186318731883189319031913192319331943195319631973198319932003201320232033204320532063207320832093210321132123213321432153216321732183219322032213222322332243225322632273228322932303231323232333234323532363237323832393240324132423243324432453246324732483249325032513252325332543255325632573258325932603261326232633264326532663267326832693270327132723273327432753276327732783279328032813282328332843285328632873288328932903291329232933294329532963297329832993300330133023303330433053306330733083309331033113312331333143315331633173318331933203321332233233324332533263327332833293330333133323333333433353336333733383339334033413342334333443345334633473348334933503351335233533354335533563357335833593360336133623363336433653366336733683369337033713372337333743375337633773378337933803381338233833384338533863387338833893390339133923393339433953396339733983399340034013402340334043405340634073408340934103411341234133414341534163417341834193420342134223423342434253426342734283429343034313432343334343435343634373438343934403441344234433444344534463447344834493450345134523453345434553456345734583459346034613462346334643465346634673468346934703471347234733474347534763477347834793480348134823483348434853486348734883489349034913492349334943495349634973498349935003501350235033504350535063507350835093510351135123513351435153516351735183519352035213522352335243525352635273528352935303531353235333534353535363537353835393540354135423543354435453546354735483549355035513552355335543555355635573558355935603561356235633564356535663567356835693570357135723573357435753576357735783579358035813582358335843585358635873588358935903591359235933594359535963597359835993600360136023603360436053606360736083609361036113612361336143615361636173618361936203621362236233624362536263627362836293630363136323633363436353636363736383639364036413642364336443645364636473648364936503651365236533654365536563657365836593660366136623663366436653666366736683669367036713672367336743675367636773678367936803681368236833684368536863687368836893690369136923693369436953696369736983699370037013702370337043705370637073708370937103711371237133714371537163717371837193720372137223723372437253726372737283729373037313732373337343735373637373738373937403741374237433744374537463747374837493750375137523753375437553756375737583759376037613762376337643765376637673768376937703771377237733774377537763777377837793780378137823783378437853786378737883789379037913792379337943795379637973798379938003801380238033804380538063807380838093810381138123813381438153816381738183819382038213822382338243825382638273828382938303831383238333834383538363837383838393840384138423843384438453846384738483849385038513852385338543855385638573858385938603861386238633864386538663867386838693870387138723873387438753876387738783879388038813882388338843885388638873888388938903891389238933894389538963897389838993900390139023903390439053906390739083909391039113912391339143915391639173918391939203921392239233924392539263927392839293930393139323933393439353936393739383939394039413942394339443945394639473948394939503951395239533954
  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 <linux/sched/clock.h>
  35. #include <net/flow_dissector.h>
  36. #include <linux/splice.h>
  37. #include <linux/in6.h>
  38. #include <linux/if_packet.h>
  39. #include <net/flow.h>
  40. /* The interface for checksum offload between the stack and networking drivers
  41. * is as follows...
  42. *
  43. * A. IP checksum related features
  44. *
  45. * Drivers advertise checksum offload capabilities in the features of a device.
  46. * From the stack's point of view these are capabilities offered by the driver,
  47. * a driver typically only advertises features that it is capable of offloading
  48. * to its device.
  49. *
  50. * The checksum related features are:
  51. *
  52. * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
  53. * IP (one's complement) checksum for any combination
  54. * of protocols or protocol layering. The checksum is
  55. * computed and set in a packet per the CHECKSUM_PARTIAL
  56. * interface (see below).
  57. *
  58. * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
  59. * TCP or UDP packets over IPv4. These are specifically
  60. * unencapsulated packets of the form IPv4|TCP or
  61. * IPv4|UDP where the Protocol field in the IPv4 header
  62. * is TCP or UDP. The IPv4 header may contain IP options
  63. * This feature cannot be set in features for a device
  64. * with NETIF_F_HW_CSUM also set. This feature is being
  65. * DEPRECATED (see below).
  66. *
  67. * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
  68. * TCP or UDP packets over IPv6. These are specifically
  69. * unencapsulated packets of the form IPv6|TCP or
  70. * IPv4|UDP where the Next Header field in the IPv6
  71. * header is either TCP or UDP. IPv6 extension headers
  72. * are not supported with this feature. This feature
  73. * cannot be set in features for a device with
  74. * NETIF_F_HW_CSUM also set. This feature is being
  75. * DEPRECATED (see below).
  76. *
  77. * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
  78. * This flag is used only used to disable the RX checksum
  79. * feature for a device. The stack will accept receive
  80. * checksum indication in packets received on a device
  81. * regardless of whether NETIF_F_RXCSUM is set.
  82. *
  83. * B. Checksumming of received packets by device. Indication of checksum
  84. * verification is in set skb->ip_summed. Possible values are:
  85. *
  86. * CHECKSUM_NONE:
  87. *
  88. * Device did not checksum this packet e.g. due to lack of capabilities.
  89. * The packet contains full (though not verified) checksum in packet but
  90. * not in skb->csum. Thus, skb->csum is undefined in this case.
  91. *
  92. * CHECKSUM_UNNECESSARY:
  93. *
  94. * The hardware you're dealing with doesn't calculate the full checksum
  95. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  96. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  97. * if their checksums are okay. skb->csum is still undefined in this case
  98. * though. A driver or device must never modify the checksum field in the
  99. * packet even if checksum is verified.
  100. *
  101. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  102. * TCP: IPv6 and IPv4.
  103. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  104. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  105. * may perform further validation in this case.
  106. * GRE: only if the checksum is present in the header.
  107. * SCTP: indicates the CRC in SCTP header has been validated.
  108. * FCOE: indicates the CRC in FC frame has been validated.
  109. *
  110. * skb->csum_level indicates the number of consecutive checksums found in
  111. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  112. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  113. * and a device is able to verify the checksums for UDP (possibly zero),
  114. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  115. * two. If the device were only able to verify the UDP checksum and not
  116. * GRE, either because it doesn't support GRE checksum of because GRE
  117. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  118. * not considered in this case).
  119. *
  120. * CHECKSUM_COMPLETE:
  121. *
  122. * This is the most generic way. The device supplied checksum of the _whole_
  123. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  124. * hardware doesn't need to parse L3/L4 headers to implement this.
  125. *
  126. * Notes:
  127. * - Even if device supports only some protocols, but is able to produce
  128. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  129. * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
  130. *
  131. * CHECKSUM_PARTIAL:
  132. *
  133. * A checksum is set up to be offloaded to a device as described in the
  134. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  135. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  136. * on the same host, or it may be set in the input path in GRO or remote
  137. * checksum offload. For the purposes of checksum verification, the checksum
  138. * referred to by skb->csum_start + skb->csum_offset and any preceding
  139. * checksums in the packet are considered verified. Any checksums in the
  140. * packet that are after the checksum being offloaded are not considered to
  141. * be verified.
  142. *
  143. * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
  144. * in the skb->ip_summed for a packet. Values are:
  145. *
  146. * CHECKSUM_PARTIAL:
  147. *
  148. * The driver is required to checksum the packet as seen by hard_start_xmit()
  149. * from skb->csum_start up to the end, and to record/write the checksum at
  150. * offset skb->csum_start + skb->csum_offset. A driver may verify that the
  151. * csum_start and csum_offset values are valid values given the length and
  152. * offset of the packet, however they should not attempt to validate that the
  153. * checksum refers to a legitimate transport layer checksum-- it is the
  154. * purview of the stack to validate that csum_start and csum_offset are set
  155. * correctly.
  156. *
  157. * When the stack requests checksum offload for a packet, the driver MUST
  158. * ensure that the checksum is set correctly. A driver can either offload the
  159. * checksum calculation to the device, or call skb_checksum_help (in the case
  160. * that the device does not support offload for a particular checksum).
  161. *
  162. * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
  163. * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
  164. * checksum offload capability.
  165. * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
  166. * on network device checksumming capabilities: if a packet does not match
  167. * them, skb_checksum_help or skb_crc32c_help (depending on the value of
  168. * csum_not_inet, see item D.) is called to resolve the checksum.
  169. *
  170. * CHECKSUM_NONE:
  171. *
  172. * The skb was already checksummed by the protocol, or a checksum is not
  173. * required.
  174. *
  175. * CHECKSUM_UNNECESSARY:
  176. *
  177. * This has the same meaning on as CHECKSUM_NONE for checksum offload on
  178. * output.
  179. *
  180. * CHECKSUM_COMPLETE:
  181. * Not used in checksum output. If a driver observes a packet with this value
  182. * set in skbuff, if should treat as CHECKSUM_NONE being set.
  183. *
  184. * D. Non-IP checksum (CRC) offloads
  185. *
  186. * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
  187. * offloading the SCTP CRC in a packet. To perform this offload the stack
  188. * will set set csum_start and csum_offset accordingly, set ip_summed to
  189. * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
  190. * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
  191. * A driver that supports both IP checksum offload and SCTP CRC32c offload
  192. * must verify which offload is configured for a packet by testing the
  193. * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
  194. * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
  195. *
  196. * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
  197. * offloading the FCOE CRC in a packet. To perform this offload the stack
  198. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  199. * accordingly. Note the there is no indication in the skbuff that the
  200. * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
  201. * both IP checksum offload and FCOE CRC offload must verify which offload
  202. * is configured for a packet presumably by inspecting packet headers.
  203. *
  204. * E. Checksumming on output with GSO.
  205. *
  206. * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
  207. * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
  208. * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
  209. * part of the GSO operation is implied. If a checksum is being offloaded
  210. * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
  211. * are set to refer to the outermost checksum being offload (two offloaded
  212. * checksums are possible with UDP encapsulation).
  213. */
  214. /* Don't change this without changing skb_csum_unnecessary! */
  215. #define CHECKSUM_NONE 0
  216. #define CHECKSUM_UNNECESSARY 1
  217. #define CHECKSUM_COMPLETE 2
  218. #define CHECKSUM_PARTIAL 3
  219. /* Maximum value in skb->csum_level */
  220. #define SKB_MAX_CSUM_LEVEL 3
  221. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  222. #define SKB_WITH_OVERHEAD(X) \
  223. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  224. #define SKB_MAX_ORDER(X, ORDER) \
  225. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  226. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  227. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  228. /* return minimum truesize of one skb containing X bytes of data */
  229. #define SKB_TRUESIZE(X) ((X) + \
  230. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  231. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  232. struct net_device;
  233. struct scatterlist;
  234. struct pipe_inode_info;
  235. struct iov_iter;
  236. struct napi_struct;
  237. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  238. struct nf_conntrack {
  239. atomic_t use;
  240. };
  241. #endif
  242. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  243. struct nf_bridge_info {
  244. refcount_t use;
  245. enum {
  246. BRNF_PROTO_UNCHANGED,
  247. BRNF_PROTO_8021Q,
  248. BRNF_PROTO_PPPOE
  249. } orig_proto:8;
  250. u8 pkt_otherhost:1;
  251. u8 in_prerouting:1;
  252. u8 bridged_dnat:1;
  253. __u16 frag_max_size;
  254. struct net_device *physindev;
  255. /* always valid & non-NULL from FORWARD on, for physdev match */
  256. struct net_device *physoutdev;
  257. union {
  258. /* prerouting: detect dnat in orig/reply direction */
  259. __be32 ipv4_daddr;
  260. struct in6_addr ipv6_daddr;
  261. /* after prerouting + nat detected: store original source
  262. * mac since neigh resolution overwrites it, only used while
  263. * skb is out in neigh layer.
  264. */
  265. char neigh_header[8];
  266. };
  267. };
  268. #endif
  269. struct sk_buff_head {
  270. /* These two members must be first. */
  271. struct sk_buff *next;
  272. struct sk_buff *prev;
  273. __u32 qlen;
  274. spinlock_t lock;
  275. };
  276. struct sk_buff;
  277. /* To allow 64K frame to be packed as single skb without frag_list we
  278. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  279. * buffers which do not start on a page boundary.
  280. *
  281. * Since GRO uses frags we allocate at least 16 regardless of page
  282. * size.
  283. */
  284. #if (65536/PAGE_SIZE + 1) < 16
  285. #define MAX_SKB_FRAGS 16UL
  286. #else
  287. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  288. #endif
  289. extern int sysctl_max_skb_frags;
  290. /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
  291. * segment using its current segmentation instead.
  292. */
  293. #define GSO_BY_FRAGS 0xFFFF
  294. typedef struct skb_frag_struct skb_frag_t;
  295. struct skb_frag_struct {
  296. struct {
  297. struct page *p;
  298. } page;
  299. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  300. __u32 page_offset;
  301. __u32 size;
  302. #else
  303. __u16 page_offset;
  304. __u16 size;
  305. #endif
  306. };
  307. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  308. {
  309. return frag->size;
  310. }
  311. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  312. {
  313. frag->size = size;
  314. }
  315. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  316. {
  317. frag->size += delta;
  318. }
  319. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  320. {
  321. frag->size -= delta;
  322. }
  323. #define HAVE_HW_TIME_STAMP
  324. /**
  325. * struct skb_shared_hwtstamps - hardware time stamps
  326. * @hwtstamp: hardware time stamp transformed into duration
  327. * since arbitrary point in time
  328. *
  329. * Software time stamps generated by ktime_get_real() are stored in
  330. * skb->tstamp.
  331. *
  332. * hwtstamps can only be compared against other hwtstamps from
  333. * the same device.
  334. *
  335. * This structure is attached to packets as part of the
  336. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  337. */
  338. struct skb_shared_hwtstamps {
  339. ktime_t hwtstamp;
  340. };
  341. /* Definitions for tx_flags in struct skb_shared_info */
  342. enum {
  343. /* generate hardware time stamp */
  344. SKBTX_HW_TSTAMP = 1 << 0,
  345. /* generate software time stamp when queueing packet to NIC */
  346. SKBTX_SW_TSTAMP = 1 << 1,
  347. /* device driver is going to provide hardware time stamp */
  348. SKBTX_IN_PROGRESS = 1 << 2,
  349. /* device driver supports TX zero-copy buffers */
  350. SKBTX_DEV_ZEROCOPY = 1 << 3,
  351. /* generate wifi status information (where possible) */
  352. SKBTX_WIFI_STATUS = 1 << 4,
  353. /* This indicates at least one fragment might be overwritten
  354. * (as in vmsplice(), sendfile() ...)
  355. * If we need to compute a TX checksum, we'll need to copy
  356. * all frags to avoid possible bad checksum
  357. */
  358. SKBTX_SHARED_FRAG = 1 << 5,
  359. /* generate software time stamp when entering packet scheduling */
  360. SKBTX_SCHED_TSTAMP = 1 << 6,
  361. };
  362. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  363. SKBTX_SCHED_TSTAMP)
  364. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  365. /*
  366. * The callback notifies userspace to release buffers when skb DMA is done in
  367. * lower device, the skb last reference should be 0 when calling this.
  368. * The zerocopy_success argument is true if zero copy transmit occurred,
  369. * false on data copy or out of memory error caused by data copy attempt.
  370. * The ctx field is used to track device context.
  371. * The desc field is used to track userspace buffer index.
  372. */
  373. struct ubuf_info {
  374. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  375. void *ctx;
  376. unsigned long desc;
  377. };
  378. /* This data is invariant across clones and lives at
  379. * the end of the header data, ie. at skb->end.
  380. */
  381. struct skb_shared_info {
  382. unsigned short _unused;
  383. unsigned char nr_frags;
  384. __u8 tx_flags;
  385. unsigned short gso_size;
  386. /* Warning: this field is not always filled in (UFO)! */
  387. unsigned short gso_segs;
  388. struct sk_buff *frag_list;
  389. struct skb_shared_hwtstamps hwtstamps;
  390. unsigned int gso_type;
  391. u32 tskey;
  392. __be32 ip6_frag_id;
  393. /*
  394. * Warning : all fields before dataref are cleared in __alloc_skb()
  395. */
  396. atomic_t dataref;
  397. /* Intermediate layers must ensure that destructor_arg
  398. * remains valid until skb destructor */
  399. void * destructor_arg;
  400. /* must be last field, see pskb_expand_head() */
  401. skb_frag_t frags[MAX_SKB_FRAGS];
  402. };
  403. /* We divide dataref into two halves. The higher 16 bits hold references
  404. * to the payload part of skb->data. The lower 16 bits hold references to
  405. * the entire skb->data. A clone of a headerless skb holds the length of
  406. * the header in skb->hdr_len.
  407. *
  408. * All users must obey the rule that the skb->data reference count must be
  409. * greater than or equal to the payload reference count.
  410. *
  411. * Holding a reference to the payload part means that the user does not
  412. * care about modifications to the header part of skb->data.
  413. */
  414. #define SKB_DATAREF_SHIFT 16
  415. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  416. enum {
  417. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  418. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  419. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  420. };
  421. enum {
  422. SKB_GSO_TCPV4 = 1 << 0,
  423. SKB_GSO_UDP = 1 << 1,
  424. /* This indicates the skb is from an untrusted source. */
  425. SKB_GSO_DODGY = 1 << 2,
  426. /* This indicates the tcp segment has CWR set. */
  427. SKB_GSO_TCP_ECN = 1 << 3,
  428. SKB_GSO_TCP_FIXEDID = 1 << 4,
  429. SKB_GSO_TCPV6 = 1 << 5,
  430. SKB_GSO_FCOE = 1 << 6,
  431. SKB_GSO_GRE = 1 << 7,
  432. SKB_GSO_GRE_CSUM = 1 << 8,
  433. SKB_GSO_IPXIP4 = 1 << 9,
  434. SKB_GSO_IPXIP6 = 1 << 10,
  435. SKB_GSO_UDP_TUNNEL = 1 << 11,
  436. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
  437. SKB_GSO_PARTIAL = 1 << 13,
  438. SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
  439. SKB_GSO_SCTP = 1 << 15,
  440. SKB_GSO_ESP = 1 << 16,
  441. };
  442. #if BITS_PER_LONG > 32
  443. #define NET_SKBUFF_DATA_USES_OFFSET 1
  444. #endif
  445. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  446. typedef unsigned int sk_buff_data_t;
  447. #else
  448. typedef unsigned char *sk_buff_data_t;
  449. #endif
  450. /**
  451. * struct sk_buff - socket buffer
  452. * @next: Next buffer in list
  453. * @prev: Previous buffer in list
  454. * @tstamp: Time we arrived/left
  455. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  456. * @sk: Socket we are owned by
  457. * @dev: Device we arrived on/are leaving by
  458. * @cb: Control buffer. Free for use by every layer. Put private vars here
  459. * @_skb_refdst: destination entry (with norefcount bit)
  460. * @sp: the security path, used for xfrm
  461. * @len: Length of actual data
  462. * @data_len: Data length
  463. * @mac_len: Length of link layer header
  464. * @hdr_len: writable header length of cloned skb
  465. * @csum: Checksum (must include start/offset pair)
  466. * @csum_start: Offset from skb->head where checksumming should start
  467. * @csum_offset: Offset from csum_start where checksum should be stored
  468. * @priority: Packet queueing priority
  469. * @ignore_df: allow local fragmentation
  470. * @cloned: Head may be cloned (check refcnt to be sure)
  471. * @ip_summed: Driver fed us an IP checksum
  472. * @nohdr: Payload reference only, must not modify header
  473. * @pkt_type: Packet class
  474. * @fclone: skbuff clone status
  475. * @ipvs_property: skbuff is owned by ipvs
  476. * @tc_skip_classify: do not classify packet. set by IFB device
  477. * @tc_at_ingress: used within tc_classify to distinguish in/egress
  478. * @tc_redirected: packet was redirected by a tc action
  479. * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
  480. * @peeked: this packet has been seen already, so stats have been
  481. * done for it, don't do them again
  482. * @nf_trace: netfilter packet trace flag
  483. * @protocol: Packet protocol from driver
  484. * @destructor: Destruct function
  485. * @_nfct: Associated connection, if any (with nfctinfo bits)
  486. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  487. * @skb_iif: ifindex of device we arrived on
  488. * @tc_index: Traffic control index
  489. * @hash: the packet hash
  490. * @queue_mapping: Queue mapping for multiqueue devices
  491. * @xmit_more: More SKBs are pending for this queue
  492. * @ndisc_nodetype: router type (from link layer)
  493. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  494. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  495. * ports.
  496. * @sw_hash: indicates hash was computed in software stack
  497. * @wifi_acked_valid: wifi_acked was set
  498. * @wifi_acked: whether frame was acked on wifi or not
  499. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  500. * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
  501. * @dst_pending_confirm: need to confirm neighbour
  502. * @napi_id: id of the NAPI struct this skb came from
  503. * @secmark: security marking
  504. * @mark: Generic packet mark
  505. * @vlan_proto: vlan encapsulation protocol
  506. * @vlan_tci: vlan tag control information
  507. * @inner_protocol: Protocol (encapsulation)
  508. * @inner_transport_header: Inner transport layer header (encapsulation)
  509. * @inner_network_header: Network layer header (encapsulation)
  510. * @inner_mac_header: Link layer header (encapsulation)
  511. * @transport_header: Transport layer header
  512. * @network_header: Network layer header
  513. * @mac_header: Link layer header
  514. * @tail: Tail pointer
  515. * @end: End pointer
  516. * @head: Head of buffer
  517. * @data: Data head pointer
  518. * @truesize: Buffer size
  519. * @users: User count - see {datagram,tcp}.c
  520. */
  521. struct sk_buff {
  522. union {
  523. struct {
  524. /* These two members must be first. */
  525. struct sk_buff *next;
  526. struct sk_buff *prev;
  527. union {
  528. ktime_t tstamp;
  529. u64 skb_mstamp;
  530. };
  531. };
  532. struct rb_node rbnode; /* used in netem & tcp stack */
  533. };
  534. struct sock *sk;
  535. union {
  536. struct net_device *dev;
  537. /* Some protocols might use this space to store information,
  538. * while device pointer would be NULL.
  539. * UDP receive path is one user.
  540. */
  541. unsigned long dev_scratch;
  542. };
  543. /*
  544. * This is the control buffer. It is free to use for every
  545. * layer. Please put your private variables there. If you
  546. * want to keep them across layers you have to do a skb_clone()
  547. * first. This is owned by whoever has the skb queued ATM.
  548. */
  549. char cb[48] __aligned(8);
  550. unsigned long _skb_refdst;
  551. void (*destructor)(struct sk_buff *skb);
  552. #ifdef CONFIG_XFRM
  553. struct sec_path *sp;
  554. #endif
  555. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  556. unsigned long _nfct;
  557. #endif
  558. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  559. struct nf_bridge_info *nf_bridge;
  560. #endif
  561. unsigned int len,
  562. data_len;
  563. __u16 mac_len,
  564. hdr_len;
  565. /* Following fields are _not_ copied in __copy_skb_header()
  566. * Note that queue_mapping is here mostly to fill a hole.
  567. */
  568. kmemcheck_bitfield_begin(flags1);
  569. __u16 queue_mapping;
  570. /* if you move cloned around you also must adapt those constants */
  571. #ifdef __BIG_ENDIAN_BITFIELD
  572. #define CLONED_MASK (1 << 7)
  573. #else
  574. #define CLONED_MASK 1
  575. #endif
  576. #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
  577. __u8 __cloned_offset[0];
  578. __u8 cloned:1,
  579. nohdr:1,
  580. fclone:2,
  581. peeked:1,
  582. head_frag:1,
  583. xmit_more:1,
  584. __unused:1; /* one bit hole */
  585. kmemcheck_bitfield_end(flags1);
  586. /* fields enclosed in headers_start/headers_end are copied
  587. * using a single memcpy() in __copy_skb_header()
  588. */
  589. /* private: */
  590. __u32 headers_start[0];
  591. /* public: */
  592. /* if you move pkt_type around you also must adapt those constants */
  593. #ifdef __BIG_ENDIAN_BITFIELD
  594. #define PKT_TYPE_MAX (7 << 5)
  595. #else
  596. #define PKT_TYPE_MAX 7
  597. #endif
  598. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  599. __u8 __pkt_type_offset[0];
  600. __u8 pkt_type:3;
  601. __u8 pfmemalloc:1;
  602. __u8 ignore_df:1;
  603. __u8 nf_trace:1;
  604. __u8 ip_summed:2;
  605. __u8 ooo_okay:1;
  606. __u8 l4_hash:1;
  607. __u8 sw_hash:1;
  608. __u8 wifi_acked_valid:1;
  609. __u8 wifi_acked:1;
  610. __u8 no_fcs:1;
  611. /* Indicates the inner headers are valid in the skbuff. */
  612. __u8 encapsulation:1;
  613. __u8 encap_hdr_csum:1;
  614. __u8 csum_valid:1;
  615. __u8 csum_complete_sw:1;
  616. __u8 csum_level:2;
  617. __u8 csum_not_inet:1;
  618. __u8 dst_pending_confirm:1;
  619. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  620. __u8 ndisc_nodetype:2;
  621. #endif
  622. __u8 ipvs_property:1;
  623. __u8 inner_protocol_type:1;
  624. __u8 remcsum_offload:1;
  625. #ifdef CONFIG_NET_SWITCHDEV
  626. __u8 offload_fwd_mark:1;
  627. #endif
  628. #ifdef CONFIG_NET_CLS_ACT
  629. __u8 tc_skip_classify:1;
  630. __u8 tc_at_ingress:1;
  631. __u8 tc_redirected:1;
  632. __u8 tc_from_ingress:1;
  633. #endif
  634. #ifdef CONFIG_NET_SCHED
  635. __u16 tc_index; /* traffic control index */
  636. #endif
  637. union {
  638. __wsum csum;
  639. struct {
  640. __u16 csum_start;
  641. __u16 csum_offset;
  642. };
  643. };
  644. __u32 priority;
  645. int skb_iif;
  646. __u32 hash;
  647. __be16 vlan_proto;
  648. __u16 vlan_tci;
  649. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  650. union {
  651. unsigned int napi_id;
  652. unsigned int sender_cpu;
  653. };
  654. #endif
  655. #ifdef CONFIG_NETWORK_SECMARK
  656. __u32 secmark;
  657. #endif
  658. union {
  659. __u32 mark;
  660. __u32 reserved_tailroom;
  661. };
  662. union {
  663. __be16 inner_protocol;
  664. __u8 inner_ipproto;
  665. };
  666. __u16 inner_transport_header;
  667. __u16 inner_network_header;
  668. __u16 inner_mac_header;
  669. __be16 protocol;
  670. __u16 transport_header;
  671. __u16 network_header;
  672. __u16 mac_header;
  673. /* private: */
  674. __u32 headers_end[0];
  675. /* public: */
  676. /* These elements must be at the end, see alloc_skb() for details. */
  677. sk_buff_data_t tail;
  678. sk_buff_data_t end;
  679. unsigned char *head,
  680. *data;
  681. unsigned int truesize;
  682. refcount_t users;
  683. };
  684. #ifdef __KERNEL__
  685. /*
  686. * Handling routines are only of interest to the kernel
  687. */
  688. #include <linux/slab.h>
  689. #define SKB_ALLOC_FCLONE 0x01
  690. #define SKB_ALLOC_RX 0x02
  691. #define SKB_ALLOC_NAPI 0x04
  692. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  693. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  694. {
  695. return unlikely(skb->pfmemalloc);
  696. }
  697. /*
  698. * skb might have a dst pointer attached, refcounted or not.
  699. * _skb_refdst low order bit is set if refcount was _not_ taken
  700. */
  701. #define SKB_DST_NOREF 1UL
  702. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  703. #define SKB_NFCT_PTRMASK ~(7UL)
  704. /**
  705. * skb_dst - returns skb dst_entry
  706. * @skb: buffer
  707. *
  708. * Returns skb dst_entry, regardless of reference taken or not.
  709. */
  710. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  711. {
  712. /* If refdst was not refcounted, check we still are in a
  713. * rcu_read_lock section
  714. */
  715. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  716. !rcu_read_lock_held() &&
  717. !rcu_read_lock_bh_held());
  718. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  719. }
  720. /**
  721. * skb_dst_set - sets skb dst
  722. * @skb: buffer
  723. * @dst: dst entry
  724. *
  725. * Sets skb dst, assuming a reference was taken on dst and should
  726. * be released by skb_dst_drop()
  727. */
  728. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  729. {
  730. skb->_skb_refdst = (unsigned long)dst;
  731. }
  732. /**
  733. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  734. * @skb: buffer
  735. * @dst: dst entry
  736. *
  737. * Sets skb dst, assuming a reference was not taken on dst.
  738. * If dst entry is cached, we do not take reference and dst_release
  739. * will be avoided by refdst_drop. If dst entry is not cached, we take
  740. * reference, so that last dst_release can destroy the dst immediately.
  741. */
  742. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  743. {
  744. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  745. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  746. }
  747. /**
  748. * skb_dst_is_noref - Test if skb dst isn't refcounted
  749. * @skb: buffer
  750. */
  751. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  752. {
  753. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  754. }
  755. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  756. {
  757. return (struct rtable *)skb_dst(skb);
  758. }
  759. /* For mangling skb->pkt_type from user space side from applications
  760. * such as nft, tc, etc, we only allow a conservative subset of
  761. * possible pkt_types to be set.
  762. */
  763. static inline bool skb_pkt_type_ok(u32 ptype)
  764. {
  765. return ptype <= PACKET_OTHERHOST;
  766. }
  767. static inline unsigned int skb_napi_id(const struct sk_buff *skb)
  768. {
  769. #ifdef CONFIG_NET_RX_BUSY_POLL
  770. return skb->napi_id;
  771. #else
  772. return 0;
  773. #endif
  774. }
  775. /* decrement the reference count and return true if we can free the skb */
  776. static inline bool skb_unref(struct sk_buff *skb)
  777. {
  778. if (unlikely(!skb))
  779. return false;
  780. if (likely(refcount_read(&skb->users) == 1))
  781. smp_rmb();
  782. else if (likely(!refcount_dec_and_test(&skb->users)))
  783. return false;
  784. return true;
  785. }
  786. void skb_release_head_state(struct sk_buff *skb);
  787. void kfree_skb(struct sk_buff *skb);
  788. void kfree_skb_list(struct sk_buff *segs);
  789. void skb_tx_error(struct sk_buff *skb);
  790. void consume_skb(struct sk_buff *skb);
  791. void consume_stateless_skb(struct sk_buff *skb);
  792. void __kfree_skb(struct sk_buff *skb);
  793. extern struct kmem_cache *skbuff_head_cache;
  794. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  795. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  796. bool *fragstolen, int *delta_truesize);
  797. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  798. int node);
  799. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  800. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  801. static inline struct sk_buff *alloc_skb(unsigned int size,
  802. gfp_t priority)
  803. {
  804. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  805. }
  806. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  807. unsigned long data_len,
  808. int max_page_order,
  809. int *errcode,
  810. gfp_t gfp_mask);
  811. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  812. struct sk_buff_fclones {
  813. struct sk_buff skb1;
  814. struct sk_buff skb2;
  815. refcount_t fclone_ref;
  816. };
  817. /**
  818. * skb_fclone_busy - check if fclone is busy
  819. * @sk: socket
  820. * @skb: buffer
  821. *
  822. * Returns true if skb is a fast clone, and its clone is not freed.
  823. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  824. * so we also check that this didnt happen.
  825. */
  826. static inline bool skb_fclone_busy(const struct sock *sk,
  827. const struct sk_buff *skb)
  828. {
  829. const struct sk_buff_fclones *fclones;
  830. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  831. return skb->fclone == SKB_FCLONE_ORIG &&
  832. refcount_read(&fclones->fclone_ref) > 1 &&
  833. fclones->skb2.sk == sk;
  834. }
  835. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  836. gfp_t priority)
  837. {
  838. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  839. }
  840. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  841. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  842. {
  843. return __alloc_skb_head(priority, -1);
  844. }
  845. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  846. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  847. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  848. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  849. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  850. gfp_t gfp_mask, bool fclone);
  851. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  852. gfp_t gfp_mask)
  853. {
  854. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  855. }
  856. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  857. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  858. unsigned int headroom);
  859. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  860. int newtailroom, gfp_t priority);
  861. int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  862. int offset, int len);
  863. int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
  864. int offset, int len);
  865. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  866. int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
  867. /**
  868. * skb_pad - zero pad the tail of an skb
  869. * @skb: buffer to pad
  870. * @pad: space to pad
  871. *
  872. * Ensure that a buffer is followed by a padding area that is zero
  873. * filled. Used by network drivers which may DMA or transfer data
  874. * beyond the buffer end onto the wire.
  875. *
  876. * May return error in out of memory cases. The skb is freed on error.
  877. */
  878. static inline int skb_pad(struct sk_buff *skb, int pad)
  879. {
  880. return __skb_pad(skb, pad, true);
  881. }
  882. #define dev_kfree_skb(a) consume_skb(a)
  883. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  884. int getfrag(void *from, char *to, int offset,
  885. int len, int odd, struct sk_buff *skb),
  886. void *from, int length);
  887. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  888. int offset, size_t size);
  889. struct skb_seq_state {
  890. __u32 lower_offset;
  891. __u32 upper_offset;
  892. __u32 frag_idx;
  893. __u32 stepped_offset;
  894. struct sk_buff *root_skb;
  895. struct sk_buff *cur_skb;
  896. __u8 *frag_data;
  897. };
  898. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  899. unsigned int to, struct skb_seq_state *st);
  900. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  901. struct skb_seq_state *st);
  902. void skb_abort_seq_read(struct skb_seq_state *st);
  903. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  904. unsigned int to, struct ts_config *config);
  905. /*
  906. * Packet hash types specify the type of hash in skb_set_hash.
  907. *
  908. * Hash types refer to the protocol layer addresses which are used to
  909. * construct a packet's hash. The hashes are used to differentiate or identify
  910. * flows of the protocol layer for the hash type. Hash types are either
  911. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  912. *
  913. * Properties of hashes:
  914. *
  915. * 1) Two packets in different flows have different hash values
  916. * 2) Two packets in the same flow should have the same hash value
  917. *
  918. * A hash at a higher layer is considered to be more specific. A driver should
  919. * set the most specific hash possible.
  920. *
  921. * A driver cannot indicate a more specific hash than the layer at which a hash
  922. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  923. *
  924. * A driver may indicate a hash level which is less specific than the
  925. * actual layer the hash was computed on. For instance, a hash computed
  926. * at L4 may be considered an L3 hash. This should only be done if the
  927. * driver can't unambiguously determine that the HW computed the hash at
  928. * the higher layer. Note that the "should" in the second property above
  929. * permits this.
  930. */
  931. enum pkt_hash_types {
  932. PKT_HASH_TYPE_NONE, /* Undefined type */
  933. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  934. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  935. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  936. };
  937. static inline void skb_clear_hash(struct sk_buff *skb)
  938. {
  939. skb->hash = 0;
  940. skb->sw_hash = 0;
  941. skb->l4_hash = 0;
  942. }
  943. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  944. {
  945. if (!skb->l4_hash)
  946. skb_clear_hash(skb);
  947. }
  948. static inline void
  949. __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
  950. {
  951. skb->l4_hash = is_l4;
  952. skb->sw_hash = is_sw;
  953. skb->hash = hash;
  954. }
  955. static inline void
  956. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  957. {
  958. /* Used by drivers to set hash from HW */
  959. __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
  960. }
  961. static inline void
  962. __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
  963. {
  964. __skb_set_hash(skb, hash, true, is_l4);
  965. }
  966. void __skb_get_hash(struct sk_buff *skb);
  967. u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
  968. u32 skb_get_poff(const struct sk_buff *skb);
  969. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  970. const struct flow_keys *keys, int hlen);
  971. __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
  972. void *data, int hlen_proto);
  973. static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
  974. int thoff, u8 ip_proto)
  975. {
  976. return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
  977. }
  978. void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  979. const struct flow_dissector_key *key,
  980. unsigned int key_count);
  981. bool __skb_flow_dissect(const struct sk_buff *skb,
  982. struct flow_dissector *flow_dissector,
  983. void *target_container,
  984. void *data, __be16 proto, int nhoff, int hlen,
  985. unsigned int flags);
  986. static inline bool skb_flow_dissect(const struct sk_buff *skb,
  987. struct flow_dissector *flow_dissector,
  988. void *target_container, unsigned int flags)
  989. {
  990. return __skb_flow_dissect(skb, flow_dissector, target_container,
  991. NULL, 0, 0, 0, flags);
  992. }
  993. static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
  994. struct flow_keys *flow,
  995. unsigned int flags)
  996. {
  997. memset(flow, 0, sizeof(*flow));
  998. return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
  999. NULL, 0, 0, 0, flags);
  1000. }
  1001. static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
  1002. void *data, __be16 proto,
  1003. int nhoff, int hlen,
  1004. unsigned int flags)
  1005. {
  1006. memset(flow, 0, sizeof(*flow));
  1007. return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
  1008. data, proto, nhoff, hlen, flags);
  1009. }
  1010. static inline __u32 skb_get_hash(struct sk_buff *skb)
  1011. {
  1012. if (!skb->l4_hash && !skb->sw_hash)
  1013. __skb_get_hash(skb);
  1014. return skb->hash;
  1015. }
  1016. __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
  1017. static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
  1018. {
  1019. if (!skb->l4_hash && !skb->sw_hash) {
  1020. struct flow_keys keys;
  1021. __u32 hash = __get_hash_from_flowi6(fl6, &keys);
  1022. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1023. }
  1024. return skb->hash;
  1025. }
  1026. __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
  1027. static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
  1028. {
  1029. if (!skb->l4_hash && !skb->sw_hash) {
  1030. struct flow_keys keys;
  1031. __u32 hash = __get_hash_from_flowi4(fl4, &keys);
  1032. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1033. }
  1034. return skb->hash;
  1035. }
  1036. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  1037. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  1038. {
  1039. return skb->hash;
  1040. }
  1041. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  1042. {
  1043. to->hash = from->hash;
  1044. to->sw_hash = from->sw_hash;
  1045. to->l4_hash = from->l4_hash;
  1046. };
  1047. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1048. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1049. {
  1050. return skb->head + skb->end;
  1051. }
  1052. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1053. {
  1054. return skb->end;
  1055. }
  1056. #else
  1057. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1058. {
  1059. return skb->end;
  1060. }
  1061. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1062. {
  1063. return skb->end - skb->head;
  1064. }
  1065. #endif
  1066. /* Internal */
  1067. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  1068. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  1069. {
  1070. return &skb_shinfo(skb)->hwtstamps;
  1071. }
  1072. /**
  1073. * skb_queue_empty - check if a queue is empty
  1074. * @list: queue head
  1075. *
  1076. * Returns true if the queue is empty, false otherwise.
  1077. */
  1078. static inline int skb_queue_empty(const struct sk_buff_head *list)
  1079. {
  1080. return list->next == (const struct sk_buff *) list;
  1081. }
  1082. /**
  1083. * skb_queue_is_last - check if skb is the last entry in the queue
  1084. * @list: queue head
  1085. * @skb: buffer
  1086. *
  1087. * Returns true if @skb is the last buffer on the list.
  1088. */
  1089. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  1090. const struct sk_buff *skb)
  1091. {
  1092. return skb->next == (const struct sk_buff *) list;
  1093. }
  1094. /**
  1095. * skb_queue_is_first - check if skb is the first entry in the queue
  1096. * @list: queue head
  1097. * @skb: buffer
  1098. *
  1099. * Returns true if @skb is the first buffer on the list.
  1100. */
  1101. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  1102. const struct sk_buff *skb)
  1103. {
  1104. return skb->prev == (const struct sk_buff *) list;
  1105. }
  1106. /**
  1107. * skb_queue_next - return the next packet in the queue
  1108. * @list: queue head
  1109. * @skb: current buffer
  1110. *
  1111. * Return the next packet in @list after @skb. It is only valid to
  1112. * call this if skb_queue_is_last() evaluates to false.
  1113. */
  1114. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  1115. const struct sk_buff *skb)
  1116. {
  1117. /* This BUG_ON may seem severe, but if we just return then we
  1118. * are going to dereference garbage.
  1119. */
  1120. BUG_ON(skb_queue_is_last(list, skb));
  1121. return skb->next;
  1122. }
  1123. /**
  1124. * skb_queue_prev - return the prev packet in the queue
  1125. * @list: queue head
  1126. * @skb: current buffer
  1127. *
  1128. * Return the prev packet in @list before @skb. It is only valid to
  1129. * call this if skb_queue_is_first() evaluates to false.
  1130. */
  1131. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  1132. const struct sk_buff *skb)
  1133. {
  1134. /* This BUG_ON may seem severe, but if we just return then we
  1135. * are going to dereference garbage.
  1136. */
  1137. BUG_ON(skb_queue_is_first(list, skb));
  1138. return skb->prev;
  1139. }
  1140. /**
  1141. * skb_get - reference buffer
  1142. * @skb: buffer to reference
  1143. *
  1144. * Makes another reference to a socket buffer and returns a pointer
  1145. * to the buffer.
  1146. */
  1147. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  1148. {
  1149. refcount_inc(&skb->users);
  1150. return skb;
  1151. }
  1152. /*
  1153. * If users == 1, we are the only owner and are can avoid redundant
  1154. * atomic change.
  1155. */
  1156. /**
  1157. * skb_cloned - is the buffer a clone
  1158. * @skb: buffer to check
  1159. *
  1160. * Returns true if the buffer was generated with skb_clone() and is
  1161. * one of multiple shared copies of the buffer. Cloned buffers are
  1162. * shared data so must not be written to under normal circumstances.
  1163. */
  1164. static inline int skb_cloned(const struct sk_buff *skb)
  1165. {
  1166. return skb->cloned &&
  1167. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  1168. }
  1169. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  1170. {
  1171. might_sleep_if(gfpflags_allow_blocking(pri));
  1172. if (skb_cloned(skb))
  1173. return pskb_expand_head(skb, 0, 0, pri);
  1174. return 0;
  1175. }
  1176. /**
  1177. * skb_header_cloned - is the header a clone
  1178. * @skb: buffer to check
  1179. *
  1180. * Returns true if modifying the header part of the buffer requires
  1181. * the data to be copied.
  1182. */
  1183. static inline int skb_header_cloned(const struct sk_buff *skb)
  1184. {
  1185. int dataref;
  1186. if (!skb->cloned)
  1187. return 0;
  1188. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  1189. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  1190. return dataref != 1;
  1191. }
  1192. static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
  1193. {
  1194. might_sleep_if(gfpflags_allow_blocking(pri));
  1195. if (skb_header_cloned(skb))
  1196. return pskb_expand_head(skb, 0, 0, pri);
  1197. return 0;
  1198. }
  1199. /**
  1200. * skb_header_release - release reference to header
  1201. * @skb: buffer to operate on
  1202. *
  1203. * Drop a reference to the header part of the buffer. This is done
  1204. * by acquiring a payload reference. You must not read from the header
  1205. * part of skb->data after this.
  1206. * Note : Check if you can use __skb_header_release() instead.
  1207. */
  1208. static inline void skb_header_release(struct sk_buff *skb)
  1209. {
  1210. BUG_ON(skb->nohdr);
  1211. skb->nohdr = 1;
  1212. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1213. }
  1214. /**
  1215. * __skb_header_release - release reference to header
  1216. * @skb: buffer to operate on
  1217. *
  1218. * Variant of skb_header_release() assuming skb is private to caller.
  1219. * We can avoid one atomic operation.
  1220. */
  1221. static inline void __skb_header_release(struct sk_buff *skb)
  1222. {
  1223. skb->nohdr = 1;
  1224. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1225. }
  1226. /**
  1227. * skb_shared - is the buffer shared
  1228. * @skb: buffer to check
  1229. *
  1230. * Returns true if more than one person has a reference to this
  1231. * buffer.
  1232. */
  1233. static inline int skb_shared(const struct sk_buff *skb)
  1234. {
  1235. return refcount_read(&skb->users) != 1;
  1236. }
  1237. /**
  1238. * skb_share_check - check if buffer is shared and if so clone it
  1239. * @skb: buffer to check
  1240. * @pri: priority for memory allocation
  1241. *
  1242. * If the buffer is shared the buffer is cloned and the old copy
  1243. * drops a reference. A new clone with a single reference is returned.
  1244. * If the buffer is not shared the original buffer is returned. When
  1245. * being called from interrupt status or with spinlocks held pri must
  1246. * be GFP_ATOMIC.
  1247. *
  1248. * NULL is returned on a memory allocation failure.
  1249. */
  1250. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1251. {
  1252. might_sleep_if(gfpflags_allow_blocking(pri));
  1253. if (skb_shared(skb)) {
  1254. struct sk_buff *nskb = skb_clone(skb, pri);
  1255. if (likely(nskb))
  1256. consume_skb(skb);
  1257. else
  1258. kfree_skb(skb);
  1259. skb = nskb;
  1260. }
  1261. return skb;
  1262. }
  1263. /*
  1264. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1265. * packets to handle cases where we have a local reader and forward
  1266. * and a couple of other messy ones. The normal one is tcpdumping
  1267. * a packet thats being forwarded.
  1268. */
  1269. /**
  1270. * skb_unshare - make a copy of a shared buffer
  1271. * @skb: buffer to check
  1272. * @pri: priority for memory allocation
  1273. *
  1274. * If the socket buffer is a clone then this function creates a new
  1275. * copy of the data, drops a reference count on the old copy and returns
  1276. * the new copy with the reference count at 1. If the buffer is not a clone
  1277. * the original buffer is returned. When called with a spinlock held or
  1278. * from interrupt state @pri must be %GFP_ATOMIC
  1279. *
  1280. * %NULL is returned on a memory allocation failure.
  1281. */
  1282. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1283. gfp_t pri)
  1284. {
  1285. might_sleep_if(gfpflags_allow_blocking(pri));
  1286. if (skb_cloned(skb)) {
  1287. struct sk_buff *nskb = skb_copy(skb, pri);
  1288. /* Free our shared copy */
  1289. if (likely(nskb))
  1290. consume_skb(skb);
  1291. else
  1292. kfree_skb(skb);
  1293. skb = nskb;
  1294. }
  1295. return skb;
  1296. }
  1297. /**
  1298. * skb_peek - peek at the head of an &sk_buff_head
  1299. * @list_: list to peek at
  1300. *
  1301. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1302. * be careful with this one. A peek leaves the buffer on the
  1303. * list and someone else may run off with it. You must hold
  1304. * the appropriate locks or have a private queue to do this.
  1305. *
  1306. * Returns %NULL for an empty list or a pointer to the head element.
  1307. * The reference count is not incremented and the reference is therefore
  1308. * volatile. Use with caution.
  1309. */
  1310. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1311. {
  1312. struct sk_buff *skb = list_->next;
  1313. if (skb == (struct sk_buff *)list_)
  1314. skb = NULL;
  1315. return skb;
  1316. }
  1317. /**
  1318. * skb_peek_next - peek skb following the given one from a queue
  1319. * @skb: skb to start from
  1320. * @list_: list to peek at
  1321. *
  1322. * Returns %NULL when the end of the list is met or a pointer to the
  1323. * next element. The reference count is not incremented and the
  1324. * reference is therefore volatile. Use with caution.
  1325. */
  1326. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1327. const struct sk_buff_head *list_)
  1328. {
  1329. struct sk_buff *next = skb->next;
  1330. if (next == (struct sk_buff *)list_)
  1331. next = NULL;
  1332. return next;
  1333. }
  1334. /**
  1335. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1336. * @list_: list to peek at
  1337. *
  1338. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1339. * be careful with this one. A peek leaves the buffer on the
  1340. * list and someone else may run off with it. You must hold
  1341. * the appropriate locks or have a private queue to do this.
  1342. *
  1343. * Returns %NULL for an empty list or a pointer to the tail element.
  1344. * The reference count is not incremented and the reference is therefore
  1345. * volatile. Use with caution.
  1346. */
  1347. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1348. {
  1349. struct sk_buff *skb = list_->prev;
  1350. if (skb == (struct sk_buff *)list_)
  1351. skb = NULL;
  1352. return skb;
  1353. }
  1354. /**
  1355. * skb_queue_len - get queue length
  1356. * @list_: list to measure
  1357. *
  1358. * Return the length of an &sk_buff queue.
  1359. */
  1360. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1361. {
  1362. return list_->qlen;
  1363. }
  1364. /**
  1365. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1366. * @list: queue to initialize
  1367. *
  1368. * This initializes only the list and queue length aspects of
  1369. * an sk_buff_head object. This allows to initialize the list
  1370. * aspects of an sk_buff_head without reinitializing things like
  1371. * the spinlock. It can also be used for on-stack sk_buff_head
  1372. * objects where the spinlock is known to not be used.
  1373. */
  1374. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1375. {
  1376. list->prev = list->next = (struct sk_buff *)list;
  1377. list->qlen = 0;
  1378. }
  1379. /*
  1380. * This function creates a split out lock class for each invocation;
  1381. * this is needed for now since a whole lot of users of the skb-queue
  1382. * infrastructure in drivers have different locking usage (in hardirq)
  1383. * than the networking core (in softirq only). In the long run either the
  1384. * network layer or drivers should need annotation to consolidate the
  1385. * main types of usage into 3 classes.
  1386. */
  1387. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1388. {
  1389. spin_lock_init(&list->lock);
  1390. __skb_queue_head_init(list);
  1391. }
  1392. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1393. struct lock_class_key *class)
  1394. {
  1395. skb_queue_head_init(list);
  1396. lockdep_set_class(&list->lock, class);
  1397. }
  1398. /*
  1399. * Insert an sk_buff on a list.
  1400. *
  1401. * The "__skb_xxxx()" functions are the non-atomic ones that
  1402. * can only be called with interrupts disabled.
  1403. */
  1404. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1405. struct sk_buff_head *list);
  1406. static inline void __skb_insert(struct sk_buff *newsk,
  1407. struct sk_buff *prev, struct sk_buff *next,
  1408. struct sk_buff_head *list)
  1409. {
  1410. newsk->next = next;
  1411. newsk->prev = prev;
  1412. next->prev = prev->next = newsk;
  1413. list->qlen++;
  1414. }
  1415. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1416. struct sk_buff *prev,
  1417. struct sk_buff *next)
  1418. {
  1419. struct sk_buff *first = list->next;
  1420. struct sk_buff *last = list->prev;
  1421. first->prev = prev;
  1422. prev->next = first;
  1423. last->next = next;
  1424. next->prev = last;
  1425. }
  1426. /**
  1427. * skb_queue_splice - join two skb lists, this is designed for stacks
  1428. * @list: the new list to add
  1429. * @head: the place to add it in the first list
  1430. */
  1431. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1432. struct sk_buff_head *head)
  1433. {
  1434. if (!skb_queue_empty(list)) {
  1435. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1436. head->qlen += list->qlen;
  1437. }
  1438. }
  1439. /**
  1440. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1441. * @list: the new list to add
  1442. * @head: the place to add it in the first list
  1443. *
  1444. * The list at @list is reinitialised
  1445. */
  1446. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1447. struct sk_buff_head *head)
  1448. {
  1449. if (!skb_queue_empty(list)) {
  1450. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1451. head->qlen += list->qlen;
  1452. __skb_queue_head_init(list);
  1453. }
  1454. }
  1455. /**
  1456. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1457. * @list: the new list to add
  1458. * @head: the place to add it in the first list
  1459. */
  1460. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1461. struct sk_buff_head *head)
  1462. {
  1463. if (!skb_queue_empty(list)) {
  1464. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1465. head->qlen += list->qlen;
  1466. }
  1467. }
  1468. /**
  1469. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1470. * @list: the new list to add
  1471. * @head: the place to add it in the first list
  1472. *
  1473. * Each of the lists is a queue.
  1474. * The list at @list is reinitialised
  1475. */
  1476. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1477. struct sk_buff_head *head)
  1478. {
  1479. if (!skb_queue_empty(list)) {
  1480. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1481. head->qlen += list->qlen;
  1482. __skb_queue_head_init(list);
  1483. }
  1484. }
  1485. /**
  1486. * __skb_queue_after - queue a buffer at the list head
  1487. * @list: list to use
  1488. * @prev: place after this buffer
  1489. * @newsk: buffer to queue
  1490. *
  1491. * Queue a buffer int the middle of a list. This function takes no locks
  1492. * and you must therefore hold required locks before calling it.
  1493. *
  1494. * A buffer cannot be placed on two lists at the same time.
  1495. */
  1496. static inline void __skb_queue_after(struct sk_buff_head *list,
  1497. struct sk_buff *prev,
  1498. struct sk_buff *newsk)
  1499. {
  1500. __skb_insert(newsk, prev, prev->next, list);
  1501. }
  1502. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1503. struct sk_buff_head *list);
  1504. static inline void __skb_queue_before(struct sk_buff_head *list,
  1505. struct sk_buff *next,
  1506. struct sk_buff *newsk)
  1507. {
  1508. __skb_insert(newsk, next->prev, next, list);
  1509. }
  1510. /**
  1511. * __skb_queue_head - queue a buffer at the list head
  1512. * @list: list to use
  1513. * @newsk: buffer to queue
  1514. *
  1515. * Queue a buffer at the start of a list. This function takes no locks
  1516. * and you must therefore hold required locks before calling it.
  1517. *
  1518. * A buffer cannot be placed on two lists at the same time.
  1519. */
  1520. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1521. static inline void __skb_queue_head(struct sk_buff_head *list,
  1522. struct sk_buff *newsk)
  1523. {
  1524. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1525. }
  1526. /**
  1527. * __skb_queue_tail - queue a buffer at the list tail
  1528. * @list: list to use
  1529. * @newsk: buffer to queue
  1530. *
  1531. * Queue a buffer at the end of a list. This function takes no locks
  1532. * and you must therefore hold required locks before calling it.
  1533. *
  1534. * A buffer cannot be placed on two lists at the same time.
  1535. */
  1536. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1537. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1538. struct sk_buff *newsk)
  1539. {
  1540. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1541. }
  1542. /*
  1543. * remove sk_buff from list. _Must_ be called atomically, and with
  1544. * the list known..
  1545. */
  1546. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1547. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1548. {
  1549. struct sk_buff *next, *prev;
  1550. list->qlen--;
  1551. next = skb->next;
  1552. prev = skb->prev;
  1553. skb->next = skb->prev = NULL;
  1554. next->prev = prev;
  1555. prev->next = next;
  1556. }
  1557. /**
  1558. * __skb_dequeue - remove from the head of the queue
  1559. * @list: list to dequeue from
  1560. *
  1561. * Remove the head of the list. This function does not take any locks
  1562. * so must be used with appropriate locks held only. The head item is
  1563. * returned or %NULL if the list is empty.
  1564. */
  1565. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1566. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1567. {
  1568. struct sk_buff *skb = skb_peek(list);
  1569. if (skb)
  1570. __skb_unlink(skb, list);
  1571. return skb;
  1572. }
  1573. /**
  1574. * __skb_dequeue_tail - remove from the tail of the queue
  1575. * @list: list to dequeue from
  1576. *
  1577. * Remove the tail of the list. This function does not take any locks
  1578. * so must be used with appropriate locks held only. The tail item is
  1579. * returned or %NULL if the list is empty.
  1580. */
  1581. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1582. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1583. {
  1584. struct sk_buff *skb = skb_peek_tail(list);
  1585. if (skb)
  1586. __skb_unlink(skb, list);
  1587. return skb;
  1588. }
  1589. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1590. {
  1591. return skb->data_len;
  1592. }
  1593. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1594. {
  1595. return skb->len - skb->data_len;
  1596. }
  1597. static inline unsigned int skb_pagelen(const struct sk_buff *skb)
  1598. {
  1599. unsigned int i, len = 0;
  1600. for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
  1601. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1602. return len + skb_headlen(skb);
  1603. }
  1604. /**
  1605. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1606. * @skb: buffer containing fragment to be initialised
  1607. * @i: paged fragment index to initialise
  1608. * @page: the page to use for this fragment
  1609. * @off: the offset to the data with @page
  1610. * @size: the length of the data
  1611. *
  1612. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1613. * offset @off within @page.
  1614. *
  1615. * Does not take any additional reference on the fragment.
  1616. */
  1617. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1618. struct page *page, int off, int size)
  1619. {
  1620. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1621. /*
  1622. * Propagate page pfmemalloc to the skb if we can. The problem is
  1623. * that not all callers have unique ownership of the page but rely
  1624. * on page_is_pfmemalloc doing the right thing(tm).
  1625. */
  1626. frag->page.p = page;
  1627. frag->page_offset = off;
  1628. skb_frag_size_set(frag, size);
  1629. page = compound_head(page);
  1630. if (page_is_pfmemalloc(page))
  1631. skb->pfmemalloc = true;
  1632. }
  1633. /**
  1634. * skb_fill_page_desc - initialise a paged fragment in an skb
  1635. * @skb: buffer containing fragment to be initialised
  1636. * @i: paged fragment index to initialise
  1637. * @page: the page to use for this fragment
  1638. * @off: the offset to the data with @page
  1639. * @size: the length of the data
  1640. *
  1641. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1642. * @skb to point to @size bytes at offset @off within @page. In
  1643. * addition updates @skb such that @i is the last fragment.
  1644. *
  1645. * Does not take any additional reference on the fragment.
  1646. */
  1647. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1648. struct page *page, int off, int size)
  1649. {
  1650. __skb_fill_page_desc(skb, i, page, off, size);
  1651. skb_shinfo(skb)->nr_frags = i + 1;
  1652. }
  1653. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1654. int size, unsigned int truesize);
  1655. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1656. unsigned int truesize);
  1657. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1658. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1659. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1660. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1661. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1662. {
  1663. return skb->head + skb->tail;
  1664. }
  1665. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1666. {
  1667. skb->tail = skb->data - skb->head;
  1668. }
  1669. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1670. {
  1671. skb_reset_tail_pointer(skb);
  1672. skb->tail += offset;
  1673. }
  1674. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1675. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1676. {
  1677. return skb->tail;
  1678. }
  1679. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1680. {
  1681. skb->tail = skb->data;
  1682. }
  1683. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1684. {
  1685. skb->tail = skb->data + offset;
  1686. }
  1687. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1688. /*
  1689. * Add data to an sk_buff
  1690. */
  1691. void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1692. void *skb_put(struct sk_buff *skb, unsigned int len);
  1693. static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
  1694. {
  1695. void *tmp = skb_tail_pointer(skb);
  1696. SKB_LINEAR_ASSERT(skb);
  1697. skb->tail += len;
  1698. skb->len += len;
  1699. return tmp;
  1700. }
  1701. static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
  1702. {
  1703. void *tmp = __skb_put(skb, len);
  1704. memset(tmp, 0, len);
  1705. return tmp;
  1706. }
  1707. static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
  1708. unsigned int len)
  1709. {
  1710. void *tmp = __skb_put(skb, len);
  1711. memcpy(tmp, data, len);
  1712. return tmp;
  1713. }
  1714. static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
  1715. {
  1716. *(u8 *)__skb_put(skb, 1) = val;
  1717. }
  1718. static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
  1719. {
  1720. void *tmp = skb_put(skb, len);
  1721. memset(tmp, 0, len);
  1722. return tmp;
  1723. }
  1724. static inline void *skb_put_data(struct sk_buff *skb, const void *data,
  1725. unsigned int len)
  1726. {
  1727. void *tmp = skb_put(skb, len);
  1728. memcpy(tmp, data, len);
  1729. return tmp;
  1730. }
  1731. static inline void skb_put_u8(struct sk_buff *skb, u8 val)
  1732. {
  1733. *(u8 *)skb_put(skb, 1) = val;
  1734. }
  1735. void *skb_push(struct sk_buff *skb, unsigned int len);
  1736. static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
  1737. {
  1738. skb->data -= len;
  1739. skb->len += len;
  1740. return skb->data;
  1741. }
  1742. void *skb_pull(struct sk_buff *skb, unsigned int len);
  1743. static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
  1744. {
  1745. skb->len -= len;
  1746. BUG_ON(skb->len < skb->data_len);
  1747. return skb->data += len;
  1748. }
  1749. static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1750. {
  1751. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1752. }
  1753. void *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1754. static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1755. {
  1756. if (len > skb_headlen(skb) &&
  1757. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1758. return NULL;
  1759. skb->len -= len;
  1760. return skb->data += len;
  1761. }
  1762. static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
  1763. {
  1764. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1765. }
  1766. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1767. {
  1768. if (likely(len <= skb_headlen(skb)))
  1769. return 1;
  1770. if (unlikely(len > skb->len))
  1771. return 0;
  1772. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1773. }
  1774. void skb_condense(struct sk_buff *skb);
  1775. /**
  1776. * skb_headroom - bytes at buffer head
  1777. * @skb: buffer to check
  1778. *
  1779. * Return the number of bytes of free space at the head of an &sk_buff.
  1780. */
  1781. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1782. {
  1783. return skb->data - skb->head;
  1784. }
  1785. /**
  1786. * skb_tailroom - bytes at buffer end
  1787. * @skb: buffer to check
  1788. *
  1789. * Return the number of bytes of free space at the tail of an sk_buff
  1790. */
  1791. static inline int skb_tailroom(const struct sk_buff *skb)
  1792. {
  1793. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1794. }
  1795. /**
  1796. * skb_availroom - bytes at buffer end
  1797. * @skb: buffer to check
  1798. *
  1799. * Return the number of bytes of free space at the tail of an sk_buff
  1800. * allocated by sk_stream_alloc()
  1801. */
  1802. static inline int skb_availroom(const struct sk_buff *skb)
  1803. {
  1804. if (skb_is_nonlinear(skb))
  1805. return 0;
  1806. return skb->end - skb->tail - skb->reserved_tailroom;
  1807. }
  1808. /**
  1809. * skb_reserve - adjust headroom
  1810. * @skb: buffer to alter
  1811. * @len: bytes to move
  1812. *
  1813. * Increase the headroom of an empty &sk_buff by reducing the tail
  1814. * room. This is only allowed for an empty buffer.
  1815. */
  1816. static inline void skb_reserve(struct sk_buff *skb, int len)
  1817. {
  1818. skb->data += len;
  1819. skb->tail += len;
  1820. }
  1821. /**
  1822. * skb_tailroom_reserve - adjust reserved_tailroom
  1823. * @skb: buffer to alter
  1824. * @mtu: maximum amount of headlen permitted
  1825. * @needed_tailroom: minimum amount of reserved_tailroom
  1826. *
  1827. * Set reserved_tailroom so that headlen can be as large as possible but
  1828. * not larger than mtu and tailroom cannot be smaller than
  1829. * needed_tailroom.
  1830. * The required headroom should already have been reserved before using
  1831. * this function.
  1832. */
  1833. static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
  1834. unsigned int needed_tailroom)
  1835. {
  1836. SKB_LINEAR_ASSERT(skb);
  1837. if (mtu < skb_tailroom(skb) - needed_tailroom)
  1838. /* use at most mtu */
  1839. skb->reserved_tailroom = skb_tailroom(skb) - mtu;
  1840. else
  1841. /* use up to all available space */
  1842. skb->reserved_tailroom = needed_tailroom;
  1843. }
  1844. #define ENCAP_TYPE_ETHER 0
  1845. #define ENCAP_TYPE_IPPROTO 1
  1846. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1847. __be16 protocol)
  1848. {
  1849. skb->inner_protocol = protocol;
  1850. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1851. }
  1852. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1853. __u8 ipproto)
  1854. {
  1855. skb->inner_ipproto = ipproto;
  1856. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1857. }
  1858. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1859. {
  1860. skb->inner_mac_header = skb->mac_header;
  1861. skb->inner_network_header = skb->network_header;
  1862. skb->inner_transport_header = skb->transport_header;
  1863. }
  1864. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1865. {
  1866. skb->mac_len = skb->network_header - skb->mac_header;
  1867. }
  1868. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1869. *skb)
  1870. {
  1871. return skb->head + skb->inner_transport_header;
  1872. }
  1873. static inline int skb_inner_transport_offset(const struct sk_buff *skb)
  1874. {
  1875. return skb_inner_transport_header(skb) - skb->data;
  1876. }
  1877. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1878. {
  1879. skb->inner_transport_header = skb->data - skb->head;
  1880. }
  1881. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1882. const int offset)
  1883. {
  1884. skb_reset_inner_transport_header(skb);
  1885. skb->inner_transport_header += offset;
  1886. }
  1887. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1888. {
  1889. return skb->head + skb->inner_network_header;
  1890. }
  1891. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1892. {
  1893. skb->inner_network_header = skb->data - skb->head;
  1894. }
  1895. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1896. const int offset)
  1897. {
  1898. skb_reset_inner_network_header(skb);
  1899. skb->inner_network_header += offset;
  1900. }
  1901. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1902. {
  1903. return skb->head + skb->inner_mac_header;
  1904. }
  1905. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1906. {
  1907. skb->inner_mac_header = skb->data - skb->head;
  1908. }
  1909. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1910. const int offset)
  1911. {
  1912. skb_reset_inner_mac_header(skb);
  1913. skb->inner_mac_header += offset;
  1914. }
  1915. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1916. {
  1917. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1918. }
  1919. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1920. {
  1921. return skb->head + skb->transport_header;
  1922. }
  1923. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1924. {
  1925. skb->transport_header = skb->data - skb->head;
  1926. }
  1927. static inline void skb_set_transport_header(struct sk_buff *skb,
  1928. const int offset)
  1929. {
  1930. skb_reset_transport_header(skb);
  1931. skb->transport_header += offset;
  1932. }
  1933. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1934. {
  1935. return skb->head + skb->network_header;
  1936. }
  1937. static inline void skb_reset_network_header(struct sk_buff *skb)
  1938. {
  1939. skb->network_header = skb->data - skb->head;
  1940. }
  1941. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1942. {
  1943. skb_reset_network_header(skb);
  1944. skb->network_header += offset;
  1945. }
  1946. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1947. {
  1948. return skb->head + skb->mac_header;
  1949. }
  1950. static inline int skb_mac_offset(const struct sk_buff *skb)
  1951. {
  1952. return skb_mac_header(skb) - skb->data;
  1953. }
  1954. static inline u32 skb_mac_header_len(const struct sk_buff *skb)
  1955. {
  1956. return skb->network_header - skb->mac_header;
  1957. }
  1958. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1959. {
  1960. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1961. }
  1962. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1963. {
  1964. skb->mac_header = skb->data - skb->head;
  1965. }
  1966. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1967. {
  1968. skb_reset_mac_header(skb);
  1969. skb->mac_header += offset;
  1970. }
  1971. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1972. {
  1973. skb->mac_header = skb->network_header;
  1974. }
  1975. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1976. const int offset_hint)
  1977. {
  1978. struct flow_keys keys;
  1979. if (skb_transport_header_was_set(skb))
  1980. return;
  1981. else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
  1982. skb_set_transport_header(skb, keys.control.thoff);
  1983. else
  1984. skb_set_transport_header(skb, offset_hint);
  1985. }
  1986. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1987. {
  1988. if (skb_mac_header_was_set(skb)) {
  1989. const unsigned char *old_mac = skb_mac_header(skb);
  1990. skb_set_mac_header(skb, -skb->mac_len);
  1991. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1992. }
  1993. }
  1994. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1995. {
  1996. return skb->csum_start - skb_headroom(skb);
  1997. }
  1998. static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
  1999. {
  2000. return skb->head + skb->csum_start;
  2001. }
  2002. static inline int skb_transport_offset(const struct sk_buff *skb)
  2003. {
  2004. return skb_transport_header(skb) - skb->data;
  2005. }
  2006. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  2007. {
  2008. return skb->transport_header - skb->network_header;
  2009. }
  2010. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  2011. {
  2012. return skb->inner_transport_header - skb->inner_network_header;
  2013. }
  2014. static inline int skb_network_offset(const struct sk_buff *skb)
  2015. {
  2016. return skb_network_header(skb) - skb->data;
  2017. }
  2018. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  2019. {
  2020. return skb_inner_network_header(skb) - skb->data;
  2021. }
  2022. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  2023. {
  2024. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  2025. }
  2026. /*
  2027. * CPUs often take a performance hit when accessing unaligned memory
  2028. * locations. The actual performance hit varies, it can be small if the
  2029. * hardware handles it or large if we have to take an exception and fix it
  2030. * in software.
  2031. *
  2032. * Since an ethernet header is 14 bytes network drivers often end up with
  2033. * the IP header at an unaligned offset. The IP header can be aligned by
  2034. * shifting the start of the packet by 2 bytes. Drivers should do this
  2035. * with:
  2036. *
  2037. * skb_reserve(skb, NET_IP_ALIGN);
  2038. *
  2039. * The downside to this alignment of the IP header is that the DMA is now
  2040. * unaligned. On some architectures the cost of an unaligned DMA is high
  2041. * and this cost outweighs the gains made by aligning the IP header.
  2042. *
  2043. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  2044. * to be overridden.
  2045. */
  2046. #ifndef NET_IP_ALIGN
  2047. #define NET_IP_ALIGN 2
  2048. #endif
  2049. /*
  2050. * The networking layer reserves some headroom in skb data (via
  2051. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  2052. * the header has to grow. In the default case, if the header has to grow
  2053. * 32 bytes or less we avoid the reallocation.
  2054. *
  2055. * Unfortunately this headroom changes the DMA alignment of the resulting
  2056. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  2057. * on some architectures. An architecture can override this value,
  2058. * perhaps setting it to a cacheline in size (since that will maintain
  2059. * cacheline alignment of the DMA). It must be a power of 2.
  2060. *
  2061. * Various parts of the networking layer expect at least 32 bytes of
  2062. * headroom, you should not reduce this.
  2063. *
  2064. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  2065. * to reduce average number of cache lines per packet.
  2066. * get_rps_cpus() for example only access one 64 bytes aligned block :
  2067. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  2068. */
  2069. #ifndef NET_SKB_PAD
  2070. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  2071. #endif
  2072. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  2073. static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
  2074. {
  2075. if (unlikely(skb_is_nonlinear(skb))) {
  2076. WARN_ON(1);
  2077. return;
  2078. }
  2079. skb->len = len;
  2080. skb_set_tail_pointer(skb, len);
  2081. }
  2082. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  2083. {
  2084. __skb_set_length(skb, len);
  2085. }
  2086. void skb_trim(struct sk_buff *skb, unsigned int len);
  2087. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  2088. {
  2089. if (skb->data_len)
  2090. return ___pskb_trim(skb, len);
  2091. __skb_trim(skb, len);
  2092. return 0;
  2093. }
  2094. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  2095. {
  2096. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  2097. }
  2098. /**
  2099. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  2100. * @skb: buffer to alter
  2101. * @len: new length
  2102. *
  2103. * This is identical to pskb_trim except that the caller knows that
  2104. * the skb is not cloned so we should never get an error due to out-
  2105. * of-memory.
  2106. */
  2107. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  2108. {
  2109. int err = pskb_trim(skb, len);
  2110. BUG_ON(err);
  2111. }
  2112. static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
  2113. {
  2114. unsigned int diff = len - skb->len;
  2115. if (skb_tailroom(skb) < diff) {
  2116. int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
  2117. GFP_ATOMIC);
  2118. if (ret)
  2119. return ret;
  2120. }
  2121. __skb_set_length(skb, len);
  2122. return 0;
  2123. }
  2124. /**
  2125. * skb_orphan - orphan a buffer
  2126. * @skb: buffer to orphan
  2127. *
  2128. * If a buffer currently has an owner then we call the owner's
  2129. * destructor function and make the @skb unowned. The buffer continues
  2130. * to exist but is no longer charged to its former owner.
  2131. */
  2132. static inline void skb_orphan(struct sk_buff *skb)
  2133. {
  2134. if (skb->destructor) {
  2135. skb->destructor(skb);
  2136. skb->destructor = NULL;
  2137. skb->sk = NULL;
  2138. } else {
  2139. BUG_ON(skb->sk);
  2140. }
  2141. }
  2142. /**
  2143. * skb_orphan_frags - orphan the frags contained in a buffer
  2144. * @skb: buffer to orphan frags from
  2145. * @gfp_mask: allocation mask for replacement pages
  2146. *
  2147. * For each frag in the SKB which needs a destructor (i.e. has an
  2148. * owner) create a copy of that frag and release the original
  2149. * page by calling the destructor.
  2150. */
  2151. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  2152. {
  2153. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  2154. return 0;
  2155. return skb_copy_ubufs(skb, gfp_mask);
  2156. }
  2157. /**
  2158. * __skb_queue_purge - empty a list
  2159. * @list: list to empty
  2160. *
  2161. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  2162. * the list and one reference dropped. This function does not take the
  2163. * list lock and the caller must hold the relevant locks to use it.
  2164. */
  2165. void skb_queue_purge(struct sk_buff_head *list);
  2166. static inline void __skb_queue_purge(struct sk_buff_head *list)
  2167. {
  2168. struct sk_buff *skb;
  2169. while ((skb = __skb_dequeue(list)) != NULL)
  2170. kfree_skb(skb);
  2171. }
  2172. void skb_rbtree_purge(struct rb_root *root);
  2173. void *netdev_alloc_frag(unsigned int fragsz);
  2174. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  2175. gfp_t gfp_mask);
  2176. /**
  2177. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  2178. * @dev: network device to receive on
  2179. * @length: length to allocate
  2180. *
  2181. * Allocate a new &sk_buff and assign it a usage count of one. The
  2182. * buffer has unspecified headroom built in. Users should allocate
  2183. * the headroom they think they need without accounting for the
  2184. * built in space. The built in space is used for optimisations.
  2185. *
  2186. * %NULL is returned if there is no free memory. Although this function
  2187. * allocates memory it can be called from an interrupt.
  2188. */
  2189. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  2190. unsigned int length)
  2191. {
  2192. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  2193. }
  2194. /* legacy helper around __netdev_alloc_skb() */
  2195. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  2196. gfp_t gfp_mask)
  2197. {
  2198. return __netdev_alloc_skb(NULL, length, gfp_mask);
  2199. }
  2200. /* legacy helper around netdev_alloc_skb() */
  2201. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  2202. {
  2203. return netdev_alloc_skb(NULL, length);
  2204. }
  2205. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  2206. unsigned int length, gfp_t gfp)
  2207. {
  2208. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  2209. if (NET_IP_ALIGN && skb)
  2210. skb_reserve(skb, NET_IP_ALIGN);
  2211. return skb;
  2212. }
  2213. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  2214. unsigned int length)
  2215. {
  2216. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  2217. }
  2218. static inline void skb_free_frag(void *addr)
  2219. {
  2220. page_frag_free(addr);
  2221. }
  2222. void *napi_alloc_frag(unsigned int fragsz);
  2223. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  2224. unsigned int length, gfp_t gfp_mask);
  2225. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  2226. unsigned int length)
  2227. {
  2228. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  2229. }
  2230. void napi_consume_skb(struct sk_buff *skb, int budget);
  2231. void __kfree_skb_flush(void);
  2232. void __kfree_skb_defer(struct sk_buff *skb);
  2233. /**
  2234. * __dev_alloc_pages - allocate page for network Rx
  2235. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2236. * @order: size of the allocation
  2237. *
  2238. * Allocate a new page.
  2239. *
  2240. * %NULL is returned if there is no free memory.
  2241. */
  2242. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  2243. unsigned int order)
  2244. {
  2245. /* This piece of code contains several assumptions.
  2246. * 1. This is for device Rx, therefor a cold page is preferred.
  2247. * 2. The expectation is the user wants a compound page.
  2248. * 3. If requesting a order 0 page it will not be compound
  2249. * due to the check to see if order has a value in prep_new_page
  2250. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  2251. * code in gfp_to_alloc_flags that should be enforcing this.
  2252. */
  2253. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  2254. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  2255. }
  2256. static inline struct page *dev_alloc_pages(unsigned int order)
  2257. {
  2258. return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
  2259. }
  2260. /**
  2261. * __dev_alloc_page - allocate a page for network Rx
  2262. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2263. *
  2264. * Allocate a new page.
  2265. *
  2266. * %NULL is returned if there is no free memory.
  2267. */
  2268. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  2269. {
  2270. return __dev_alloc_pages(gfp_mask, 0);
  2271. }
  2272. static inline struct page *dev_alloc_page(void)
  2273. {
  2274. return dev_alloc_pages(0);
  2275. }
  2276. /**
  2277. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  2278. * @page: The page that was allocated from skb_alloc_page
  2279. * @skb: The skb that may need pfmemalloc set
  2280. */
  2281. static inline void skb_propagate_pfmemalloc(struct page *page,
  2282. struct sk_buff *skb)
  2283. {
  2284. if (page_is_pfmemalloc(page))
  2285. skb->pfmemalloc = true;
  2286. }
  2287. /**
  2288. * skb_frag_page - retrieve the page referred to by a paged fragment
  2289. * @frag: the paged fragment
  2290. *
  2291. * Returns the &struct page associated with @frag.
  2292. */
  2293. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2294. {
  2295. return frag->page.p;
  2296. }
  2297. /**
  2298. * __skb_frag_ref - take an addition reference on a paged fragment.
  2299. * @frag: the paged fragment
  2300. *
  2301. * Takes an additional reference on the paged fragment @frag.
  2302. */
  2303. static inline void __skb_frag_ref(skb_frag_t *frag)
  2304. {
  2305. get_page(skb_frag_page(frag));
  2306. }
  2307. /**
  2308. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2309. * @skb: the buffer
  2310. * @f: the fragment offset.
  2311. *
  2312. * Takes an additional reference on the @f'th paged fragment of @skb.
  2313. */
  2314. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2315. {
  2316. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2317. }
  2318. /**
  2319. * __skb_frag_unref - release a reference on a paged fragment.
  2320. * @frag: the paged fragment
  2321. *
  2322. * Releases a reference on the paged fragment @frag.
  2323. */
  2324. static inline void __skb_frag_unref(skb_frag_t *frag)
  2325. {
  2326. put_page(skb_frag_page(frag));
  2327. }
  2328. /**
  2329. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2330. * @skb: the buffer
  2331. * @f: the fragment offset
  2332. *
  2333. * Releases a reference on the @f'th paged fragment of @skb.
  2334. */
  2335. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2336. {
  2337. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2338. }
  2339. /**
  2340. * skb_frag_address - gets the address of the data contained in a paged fragment
  2341. * @frag: the paged fragment buffer
  2342. *
  2343. * Returns the address of the data within @frag. The page must already
  2344. * be mapped.
  2345. */
  2346. static inline void *skb_frag_address(const skb_frag_t *frag)
  2347. {
  2348. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2349. }
  2350. /**
  2351. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2352. * @frag: the paged fragment buffer
  2353. *
  2354. * Returns the address of the data within @frag. Checks that the page
  2355. * is mapped and returns %NULL otherwise.
  2356. */
  2357. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2358. {
  2359. void *ptr = page_address(skb_frag_page(frag));
  2360. if (unlikely(!ptr))
  2361. return NULL;
  2362. return ptr + frag->page_offset;
  2363. }
  2364. /**
  2365. * __skb_frag_set_page - sets the page contained in a paged fragment
  2366. * @frag: the paged fragment
  2367. * @page: the page to set
  2368. *
  2369. * Sets the fragment @frag to contain @page.
  2370. */
  2371. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2372. {
  2373. frag->page.p = page;
  2374. }
  2375. /**
  2376. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2377. * @skb: the buffer
  2378. * @f: the fragment offset
  2379. * @page: the page to set
  2380. *
  2381. * Sets the @f'th fragment of @skb to contain @page.
  2382. */
  2383. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2384. struct page *page)
  2385. {
  2386. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2387. }
  2388. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2389. /**
  2390. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2391. * @dev: the device to map the fragment to
  2392. * @frag: the paged fragment to map
  2393. * @offset: the offset within the fragment (starting at the
  2394. * fragment's own offset)
  2395. * @size: the number of bytes to map
  2396. * @dir: the direction of the mapping (``PCI_DMA_*``)
  2397. *
  2398. * Maps the page associated with @frag to @device.
  2399. */
  2400. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2401. const skb_frag_t *frag,
  2402. size_t offset, size_t size,
  2403. enum dma_data_direction dir)
  2404. {
  2405. return dma_map_page(dev, skb_frag_page(frag),
  2406. frag->page_offset + offset, size, dir);
  2407. }
  2408. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2409. gfp_t gfp_mask)
  2410. {
  2411. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2412. }
  2413. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2414. gfp_t gfp_mask)
  2415. {
  2416. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2417. }
  2418. /**
  2419. * skb_clone_writable - is the header of a clone writable
  2420. * @skb: buffer to check
  2421. * @len: length up to which to write
  2422. *
  2423. * Returns true if modifying the header part of the cloned buffer
  2424. * does not requires the data to be copied.
  2425. */
  2426. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2427. {
  2428. return !skb_header_cloned(skb) &&
  2429. skb_headroom(skb) + len <= skb->hdr_len;
  2430. }
  2431. static inline int skb_try_make_writable(struct sk_buff *skb,
  2432. unsigned int write_len)
  2433. {
  2434. return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
  2435. pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
  2436. }
  2437. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2438. int cloned)
  2439. {
  2440. int delta = 0;
  2441. if (headroom > skb_headroom(skb))
  2442. delta = headroom - skb_headroom(skb);
  2443. if (delta || cloned)
  2444. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2445. GFP_ATOMIC);
  2446. return 0;
  2447. }
  2448. /**
  2449. * skb_cow - copy header of skb when it is required
  2450. * @skb: buffer to cow
  2451. * @headroom: needed headroom
  2452. *
  2453. * If the skb passed lacks sufficient headroom or its data part
  2454. * is shared, data is reallocated. If reallocation fails, an error
  2455. * is returned and original skb is not changed.
  2456. *
  2457. * The result is skb with writable area skb->head...skb->tail
  2458. * and at least @headroom of space at head.
  2459. */
  2460. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2461. {
  2462. return __skb_cow(skb, headroom, skb_cloned(skb));
  2463. }
  2464. /**
  2465. * skb_cow_head - skb_cow but only making the head writable
  2466. * @skb: buffer to cow
  2467. * @headroom: needed headroom
  2468. *
  2469. * This function is identical to skb_cow except that we replace the
  2470. * skb_cloned check by skb_header_cloned. It should be used when
  2471. * you only need to push on some header and do not need to modify
  2472. * the data.
  2473. */
  2474. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2475. {
  2476. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2477. }
  2478. /**
  2479. * skb_padto - pad an skbuff up to a minimal size
  2480. * @skb: buffer to pad
  2481. * @len: minimal length
  2482. *
  2483. * Pads up a buffer to ensure the trailing bytes exist and are
  2484. * blanked. If the buffer already contains sufficient data it
  2485. * is untouched. Otherwise it is extended. Returns zero on
  2486. * success. The skb is freed on error.
  2487. */
  2488. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2489. {
  2490. unsigned int size = skb->len;
  2491. if (likely(size >= len))
  2492. return 0;
  2493. return skb_pad(skb, len - size);
  2494. }
  2495. /**
  2496. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2497. * @skb: buffer to pad
  2498. * @len: minimal length
  2499. * @free_on_error: free buffer on error
  2500. *
  2501. * Pads up a buffer to ensure the trailing bytes exist and are
  2502. * blanked. If the buffer already contains sufficient data it
  2503. * is untouched. Otherwise it is extended. Returns zero on
  2504. * success. The skb is freed on error if @free_on_error is true.
  2505. */
  2506. static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
  2507. bool free_on_error)
  2508. {
  2509. unsigned int size = skb->len;
  2510. if (unlikely(size < len)) {
  2511. len -= size;
  2512. if (__skb_pad(skb, len, free_on_error))
  2513. return -ENOMEM;
  2514. __skb_put(skb, len);
  2515. }
  2516. return 0;
  2517. }
  2518. /**
  2519. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2520. * @skb: buffer to pad
  2521. * @len: minimal length
  2522. *
  2523. * Pads up a buffer to ensure the trailing bytes exist and are
  2524. * blanked. If the buffer already contains sufficient data it
  2525. * is untouched. Otherwise it is extended. Returns zero on
  2526. * success. The skb is freed on error.
  2527. */
  2528. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2529. {
  2530. return __skb_put_padto(skb, len, true);
  2531. }
  2532. static inline int skb_add_data(struct sk_buff *skb,
  2533. struct iov_iter *from, int copy)
  2534. {
  2535. const int off = skb->len;
  2536. if (skb->ip_summed == CHECKSUM_NONE) {
  2537. __wsum csum = 0;
  2538. if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
  2539. &csum, from)) {
  2540. skb->csum = csum_block_add(skb->csum, csum, off);
  2541. return 0;
  2542. }
  2543. } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
  2544. return 0;
  2545. __skb_trim(skb, off);
  2546. return -EFAULT;
  2547. }
  2548. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2549. const struct page *page, int off)
  2550. {
  2551. if (i) {
  2552. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2553. return page == skb_frag_page(frag) &&
  2554. off == frag->page_offset + skb_frag_size(frag);
  2555. }
  2556. return false;
  2557. }
  2558. static inline int __skb_linearize(struct sk_buff *skb)
  2559. {
  2560. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2561. }
  2562. /**
  2563. * skb_linearize - convert paged skb to linear one
  2564. * @skb: buffer to linarize
  2565. *
  2566. * If there is no free memory -ENOMEM is returned, otherwise zero
  2567. * is returned and the old skb data released.
  2568. */
  2569. static inline int skb_linearize(struct sk_buff *skb)
  2570. {
  2571. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2572. }
  2573. /**
  2574. * skb_has_shared_frag - can any frag be overwritten
  2575. * @skb: buffer to test
  2576. *
  2577. * Return true if the skb has at least one frag that might be modified
  2578. * by an external entity (as in vmsplice()/sendfile())
  2579. */
  2580. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2581. {
  2582. return skb_is_nonlinear(skb) &&
  2583. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2584. }
  2585. /**
  2586. * skb_linearize_cow - make sure skb is linear and writable
  2587. * @skb: buffer to process
  2588. *
  2589. * If there is no free memory -ENOMEM is returned, otherwise zero
  2590. * is returned and the old skb data released.
  2591. */
  2592. static inline int skb_linearize_cow(struct sk_buff *skb)
  2593. {
  2594. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2595. __skb_linearize(skb) : 0;
  2596. }
  2597. static __always_inline void
  2598. __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2599. unsigned int off)
  2600. {
  2601. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2602. skb->csum = csum_block_sub(skb->csum,
  2603. csum_partial(start, len, 0), off);
  2604. else if (skb->ip_summed == CHECKSUM_PARTIAL &&
  2605. skb_checksum_start_offset(skb) < 0)
  2606. skb->ip_summed = CHECKSUM_NONE;
  2607. }
  2608. /**
  2609. * skb_postpull_rcsum - update checksum for received skb after pull
  2610. * @skb: buffer to update
  2611. * @start: start of data before pull
  2612. * @len: length of data pulled
  2613. *
  2614. * After doing a pull on a received packet, you need to call this to
  2615. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2616. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2617. */
  2618. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2619. const void *start, unsigned int len)
  2620. {
  2621. __skb_postpull_rcsum(skb, start, len, 0);
  2622. }
  2623. static __always_inline void
  2624. __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2625. unsigned int off)
  2626. {
  2627. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2628. skb->csum = csum_block_add(skb->csum,
  2629. csum_partial(start, len, 0), off);
  2630. }
  2631. /**
  2632. * skb_postpush_rcsum - update checksum for received skb after push
  2633. * @skb: buffer to update
  2634. * @start: start of data after push
  2635. * @len: length of data pushed
  2636. *
  2637. * After doing a push on a received packet, you need to call this to
  2638. * update the CHECKSUM_COMPLETE checksum.
  2639. */
  2640. static inline void skb_postpush_rcsum(struct sk_buff *skb,
  2641. const void *start, unsigned int len)
  2642. {
  2643. __skb_postpush_rcsum(skb, start, len, 0);
  2644. }
  2645. void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2646. /**
  2647. * skb_push_rcsum - push skb and update receive checksum
  2648. * @skb: buffer to update
  2649. * @len: length of data pulled
  2650. *
  2651. * This function performs an skb_push on the packet and updates
  2652. * the CHECKSUM_COMPLETE checksum. It should be used on
  2653. * receive path processing instead of skb_push unless you know
  2654. * that the checksum difference is zero (e.g., a valid IP header)
  2655. * or you are setting ip_summed to CHECKSUM_NONE.
  2656. */
  2657. static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
  2658. {
  2659. skb_push(skb, len);
  2660. skb_postpush_rcsum(skb, skb->data, len);
  2661. return skb->data;
  2662. }
  2663. /**
  2664. * pskb_trim_rcsum - trim received skb and update checksum
  2665. * @skb: buffer to trim
  2666. * @len: new length
  2667. *
  2668. * This is exactly the same as pskb_trim except that it ensures the
  2669. * checksum of received packets are still valid after the operation.
  2670. */
  2671. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2672. {
  2673. if (likely(len >= skb->len))
  2674. return 0;
  2675. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2676. skb->ip_summed = CHECKSUM_NONE;
  2677. return __pskb_trim(skb, len);
  2678. }
  2679. static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2680. {
  2681. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2682. skb->ip_summed = CHECKSUM_NONE;
  2683. __skb_trim(skb, len);
  2684. return 0;
  2685. }
  2686. static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
  2687. {
  2688. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2689. skb->ip_summed = CHECKSUM_NONE;
  2690. return __skb_grow(skb, len);
  2691. }
  2692. #define skb_queue_walk(queue, skb) \
  2693. for (skb = (queue)->next; \
  2694. skb != (struct sk_buff *)(queue); \
  2695. skb = skb->next)
  2696. #define skb_queue_walk_safe(queue, skb, tmp) \
  2697. for (skb = (queue)->next, tmp = skb->next; \
  2698. skb != (struct sk_buff *)(queue); \
  2699. skb = tmp, tmp = skb->next)
  2700. #define skb_queue_walk_from(queue, skb) \
  2701. for (; skb != (struct sk_buff *)(queue); \
  2702. skb = skb->next)
  2703. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2704. for (tmp = skb->next; \
  2705. skb != (struct sk_buff *)(queue); \
  2706. skb = tmp, tmp = skb->next)
  2707. #define skb_queue_reverse_walk(queue, skb) \
  2708. for (skb = (queue)->prev; \
  2709. skb != (struct sk_buff *)(queue); \
  2710. skb = skb->prev)
  2711. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2712. for (skb = (queue)->prev, tmp = skb->prev; \
  2713. skb != (struct sk_buff *)(queue); \
  2714. skb = tmp, tmp = skb->prev)
  2715. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2716. for (tmp = skb->prev; \
  2717. skb != (struct sk_buff *)(queue); \
  2718. skb = tmp, tmp = skb->prev)
  2719. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2720. {
  2721. return skb_shinfo(skb)->frag_list != NULL;
  2722. }
  2723. static inline void skb_frag_list_init(struct sk_buff *skb)
  2724. {
  2725. skb_shinfo(skb)->frag_list = NULL;
  2726. }
  2727. #define skb_walk_frags(skb, iter) \
  2728. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2729. int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
  2730. const struct sk_buff *skb);
  2731. struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
  2732. struct sk_buff_head *queue,
  2733. unsigned int flags,
  2734. void (*destructor)(struct sock *sk,
  2735. struct sk_buff *skb),
  2736. int *peeked, int *off, int *err,
  2737. struct sk_buff **last);
  2738. struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
  2739. void (*destructor)(struct sock *sk,
  2740. struct sk_buff *skb),
  2741. int *peeked, int *off, int *err,
  2742. struct sk_buff **last);
  2743. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2744. void (*destructor)(struct sock *sk,
  2745. struct sk_buff *skb),
  2746. int *peeked, int *off, int *err);
  2747. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2748. int *err);
  2749. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2750. struct poll_table_struct *wait);
  2751. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2752. struct iov_iter *to, int size);
  2753. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2754. struct msghdr *msg, int size)
  2755. {
  2756. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2757. }
  2758. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2759. struct msghdr *msg);
  2760. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2761. struct iov_iter *from, int len);
  2762. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2763. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2764. void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
  2765. static inline void skb_free_datagram_locked(struct sock *sk,
  2766. struct sk_buff *skb)
  2767. {
  2768. __skb_free_datagram_locked(sk, skb, 0);
  2769. }
  2770. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2771. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2772. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2773. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2774. int len, __wsum csum);
  2775. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2776. struct pipe_inode_info *pipe, unsigned int len,
  2777. unsigned int flags);
  2778. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2779. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2780. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2781. int len, int hlen);
  2782. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2783. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2784. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2785. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2786. bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
  2787. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2788. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2789. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2790. int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
  2791. int skb_vlan_pop(struct sk_buff *skb);
  2792. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2793. struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
  2794. gfp_t gfp);
  2795. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2796. {
  2797. return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
  2798. }
  2799. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2800. {
  2801. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2802. }
  2803. struct skb_checksum_ops {
  2804. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2805. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2806. };
  2807. extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
  2808. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2809. __wsum csum, const struct skb_checksum_ops *ops);
  2810. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2811. __wsum csum);
  2812. static inline void * __must_check
  2813. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2814. int len, void *data, int hlen, void *buffer)
  2815. {
  2816. if (hlen - offset >= len)
  2817. return data + offset;
  2818. if (!skb ||
  2819. skb_copy_bits(skb, offset, buffer, len) < 0)
  2820. return NULL;
  2821. return buffer;
  2822. }
  2823. static inline void * __must_check
  2824. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2825. {
  2826. return __skb_header_pointer(skb, offset, len, skb->data,
  2827. skb_headlen(skb), buffer);
  2828. }
  2829. /**
  2830. * skb_needs_linearize - check if we need to linearize a given skb
  2831. * depending on the given device features.
  2832. * @skb: socket buffer to check
  2833. * @features: net device features
  2834. *
  2835. * Returns true if either:
  2836. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2837. * 2. skb is fragmented and the device does not support SG.
  2838. */
  2839. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2840. netdev_features_t features)
  2841. {
  2842. return skb_is_nonlinear(skb) &&
  2843. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2844. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2845. }
  2846. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2847. void *to,
  2848. const unsigned int len)
  2849. {
  2850. memcpy(to, skb->data, len);
  2851. }
  2852. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2853. const int offset, void *to,
  2854. const unsigned int len)
  2855. {
  2856. memcpy(to, skb->data + offset, len);
  2857. }
  2858. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2859. const void *from,
  2860. const unsigned int len)
  2861. {
  2862. memcpy(skb->data, from, len);
  2863. }
  2864. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2865. const int offset,
  2866. const void *from,
  2867. const unsigned int len)
  2868. {
  2869. memcpy(skb->data + offset, from, len);
  2870. }
  2871. void skb_init(void);
  2872. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2873. {
  2874. return skb->tstamp;
  2875. }
  2876. /**
  2877. * skb_get_timestamp - get timestamp from a skb
  2878. * @skb: skb to get stamp from
  2879. * @stamp: pointer to struct timeval to store stamp in
  2880. *
  2881. * Timestamps are stored in the skb as offsets to a base timestamp.
  2882. * This function converts the offset back to a struct timeval and stores
  2883. * it in stamp.
  2884. */
  2885. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2886. struct timeval *stamp)
  2887. {
  2888. *stamp = ktime_to_timeval(skb->tstamp);
  2889. }
  2890. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2891. struct timespec *stamp)
  2892. {
  2893. *stamp = ktime_to_timespec(skb->tstamp);
  2894. }
  2895. static inline void __net_timestamp(struct sk_buff *skb)
  2896. {
  2897. skb->tstamp = ktime_get_real();
  2898. }
  2899. static inline ktime_t net_timedelta(ktime_t t)
  2900. {
  2901. return ktime_sub(ktime_get_real(), t);
  2902. }
  2903. static inline ktime_t net_invalid_timestamp(void)
  2904. {
  2905. return 0;
  2906. }
  2907. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2908. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2909. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2910. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2911. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2912. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2913. {
  2914. }
  2915. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2916. {
  2917. return false;
  2918. }
  2919. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2920. /**
  2921. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2922. *
  2923. * PHY drivers may accept clones of transmitted packets for
  2924. * timestamping via their phy_driver.txtstamp method. These drivers
  2925. * must call this function to return the skb back to the stack with a
  2926. * timestamp.
  2927. *
  2928. * @skb: clone of the the original outgoing packet
  2929. * @hwtstamps: hardware time stamps
  2930. *
  2931. */
  2932. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2933. struct skb_shared_hwtstamps *hwtstamps);
  2934. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2935. struct skb_shared_hwtstamps *hwtstamps,
  2936. struct sock *sk, int tstype);
  2937. /**
  2938. * skb_tstamp_tx - queue clone of skb with send time stamps
  2939. * @orig_skb: the original outgoing packet
  2940. * @hwtstamps: hardware time stamps, may be NULL if not available
  2941. *
  2942. * If the skb has a socket associated, then this function clones the
  2943. * skb (thus sharing the actual data and optional structures), stores
  2944. * the optional hardware time stamping information (if non NULL) or
  2945. * generates a software time stamp (otherwise), then queues the clone
  2946. * to the error queue of the socket. Errors are silently ignored.
  2947. */
  2948. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2949. struct skb_shared_hwtstamps *hwtstamps);
  2950. /**
  2951. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2952. *
  2953. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2954. * function immediately before giving the sk_buff to the MAC hardware.
  2955. *
  2956. * Specifically, one should make absolutely sure that this function is
  2957. * called before TX completion of this packet can trigger. Otherwise
  2958. * the packet could potentially already be freed.
  2959. *
  2960. * @skb: A socket buffer.
  2961. */
  2962. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2963. {
  2964. skb_clone_tx_timestamp(skb);
  2965. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
  2966. skb_tstamp_tx(skb, NULL);
  2967. }
  2968. /**
  2969. * skb_complete_wifi_ack - deliver skb with wifi status
  2970. *
  2971. * @skb: the original outgoing packet
  2972. * @acked: ack status
  2973. *
  2974. */
  2975. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2976. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2977. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2978. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2979. {
  2980. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2981. skb->csum_valid ||
  2982. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2983. skb_checksum_start_offset(skb) >= 0));
  2984. }
  2985. /**
  2986. * skb_checksum_complete - Calculate checksum of an entire packet
  2987. * @skb: packet to process
  2988. *
  2989. * This function calculates the checksum over the entire packet plus
  2990. * the value of skb->csum. The latter can be used to supply the
  2991. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2992. * checksum.
  2993. *
  2994. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2995. * this function can be used to verify that checksum on received
  2996. * packets. In that case the function should return zero if the
  2997. * checksum is correct. In particular, this function will return zero
  2998. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2999. * hardware has already verified the correctness of the checksum.
  3000. */
  3001. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  3002. {
  3003. return skb_csum_unnecessary(skb) ?
  3004. 0 : __skb_checksum_complete(skb);
  3005. }
  3006. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  3007. {
  3008. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  3009. if (skb->csum_level == 0)
  3010. skb->ip_summed = CHECKSUM_NONE;
  3011. else
  3012. skb->csum_level--;
  3013. }
  3014. }
  3015. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  3016. {
  3017. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  3018. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  3019. skb->csum_level++;
  3020. } else if (skb->ip_summed == CHECKSUM_NONE) {
  3021. skb->ip_summed = CHECKSUM_UNNECESSARY;
  3022. skb->csum_level = 0;
  3023. }
  3024. }
  3025. /* Check if we need to perform checksum complete validation.
  3026. *
  3027. * Returns true if checksum complete is needed, false otherwise
  3028. * (either checksum is unnecessary or zero checksum is allowed).
  3029. */
  3030. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  3031. bool zero_okay,
  3032. __sum16 check)
  3033. {
  3034. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  3035. skb->csum_valid = 1;
  3036. __skb_decr_checksum_unnecessary(skb);
  3037. return false;
  3038. }
  3039. return true;
  3040. }
  3041. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  3042. * in checksum_init.
  3043. */
  3044. #define CHECKSUM_BREAK 76
  3045. /* Unset checksum-complete
  3046. *
  3047. * Unset checksum complete can be done when packet is being modified
  3048. * (uncompressed for instance) and checksum-complete value is
  3049. * invalidated.
  3050. */
  3051. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  3052. {
  3053. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3054. skb->ip_summed = CHECKSUM_NONE;
  3055. }
  3056. /* Validate (init) checksum based on checksum complete.
  3057. *
  3058. * Return values:
  3059. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  3060. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  3061. * checksum is stored in skb->csum for use in __skb_checksum_complete
  3062. * non-zero: value of invalid checksum
  3063. *
  3064. */
  3065. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  3066. bool complete,
  3067. __wsum psum)
  3068. {
  3069. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  3070. if (!csum_fold(csum_add(psum, skb->csum))) {
  3071. skb->csum_valid = 1;
  3072. return 0;
  3073. }
  3074. }
  3075. skb->csum = psum;
  3076. if (complete || skb->len <= CHECKSUM_BREAK) {
  3077. __sum16 csum;
  3078. csum = __skb_checksum_complete(skb);
  3079. skb->csum_valid = !csum;
  3080. return csum;
  3081. }
  3082. return 0;
  3083. }
  3084. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  3085. {
  3086. return 0;
  3087. }
  3088. /* Perform checksum validate (init). Note that this is a macro since we only
  3089. * want to calculate the pseudo header which is an input function if necessary.
  3090. * First we try to validate without any computation (checksum unnecessary) and
  3091. * then calculate based on checksum complete calling the function to compute
  3092. * pseudo header.
  3093. *
  3094. * Return values:
  3095. * 0: checksum is validated or try to in skb_checksum_complete
  3096. * non-zero: value of invalid checksum
  3097. */
  3098. #define __skb_checksum_validate(skb, proto, complete, \
  3099. zero_okay, check, compute_pseudo) \
  3100. ({ \
  3101. __sum16 __ret = 0; \
  3102. skb->csum_valid = 0; \
  3103. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  3104. __ret = __skb_checksum_validate_complete(skb, \
  3105. complete, compute_pseudo(skb, proto)); \
  3106. __ret; \
  3107. })
  3108. #define skb_checksum_init(skb, proto, compute_pseudo) \
  3109. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  3110. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  3111. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  3112. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  3113. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  3114. #define skb_checksum_validate_zero_check(skb, proto, check, \
  3115. compute_pseudo) \
  3116. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  3117. #define skb_checksum_simple_validate(skb) \
  3118. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  3119. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  3120. {
  3121. return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
  3122. }
  3123. static inline void __skb_checksum_convert(struct sk_buff *skb,
  3124. __sum16 check, __wsum pseudo)
  3125. {
  3126. skb->csum = ~pseudo;
  3127. skb->ip_summed = CHECKSUM_COMPLETE;
  3128. }
  3129. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  3130. do { \
  3131. if (__skb_checksum_convert_check(skb)) \
  3132. __skb_checksum_convert(skb, check, \
  3133. compute_pseudo(skb, proto)); \
  3134. } while (0)
  3135. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  3136. u16 start, u16 offset)
  3137. {
  3138. skb->ip_summed = CHECKSUM_PARTIAL;
  3139. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  3140. skb->csum_offset = offset - start;
  3141. }
  3142. /* Update skbuf and packet to reflect the remote checksum offload operation.
  3143. * When called, ptr indicates the starting point for skb->csum when
  3144. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  3145. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  3146. */
  3147. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  3148. int start, int offset, bool nopartial)
  3149. {
  3150. __wsum delta;
  3151. if (!nopartial) {
  3152. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  3153. return;
  3154. }
  3155. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  3156. __skb_checksum_complete(skb);
  3157. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  3158. }
  3159. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  3160. /* Adjust skb->csum since we changed the packet */
  3161. skb->csum = csum_add(skb->csum, delta);
  3162. }
  3163. static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
  3164. {
  3165. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  3166. return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
  3167. #else
  3168. return NULL;
  3169. #endif
  3170. }
  3171. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3172. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  3173. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  3174. {
  3175. if (nfct && atomic_dec_and_test(&nfct->use))
  3176. nf_conntrack_destroy(nfct);
  3177. }
  3178. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  3179. {
  3180. if (nfct)
  3181. atomic_inc(&nfct->use);
  3182. }
  3183. #endif
  3184. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3185. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  3186. {
  3187. if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
  3188. kfree(nf_bridge);
  3189. }
  3190. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  3191. {
  3192. if (nf_bridge)
  3193. refcount_inc(&nf_bridge->use);
  3194. }
  3195. #endif /* CONFIG_BRIDGE_NETFILTER */
  3196. static inline void nf_reset(struct sk_buff *skb)
  3197. {
  3198. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3199. nf_conntrack_put(skb_nfct(skb));
  3200. skb->_nfct = 0;
  3201. #endif
  3202. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3203. nf_bridge_put(skb->nf_bridge);
  3204. skb->nf_bridge = NULL;
  3205. #endif
  3206. }
  3207. static inline void nf_reset_trace(struct sk_buff *skb)
  3208. {
  3209. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3210. skb->nf_trace = 0;
  3211. #endif
  3212. }
  3213. /* Note: This doesn't put any conntrack and bridge info in dst. */
  3214. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  3215. bool copy)
  3216. {
  3217. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3218. dst->_nfct = src->_nfct;
  3219. nf_conntrack_get(skb_nfct(src));
  3220. #endif
  3221. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3222. dst->nf_bridge = src->nf_bridge;
  3223. nf_bridge_get(src->nf_bridge);
  3224. #endif
  3225. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3226. if (copy)
  3227. dst->nf_trace = src->nf_trace;
  3228. #endif
  3229. }
  3230. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  3231. {
  3232. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3233. nf_conntrack_put(skb_nfct(dst));
  3234. #endif
  3235. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3236. nf_bridge_put(dst->nf_bridge);
  3237. #endif
  3238. __nf_copy(dst, src, true);
  3239. }
  3240. #ifdef CONFIG_NETWORK_SECMARK
  3241. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3242. {
  3243. to->secmark = from->secmark;
  3244. }
  3245. static inline void skb_init_secmark(struct sk_buff *skb)
  3246. {
  3247. skb->secmark = 0;
  3248. }
  3249. #else
  3250. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3251. { }
  3252. static inline void skb_init_secmark(struct sk_buff *skb)
  3253. { }
  3254. #endif
  3255. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  3256. {
  3257. return !skb->destructor &&
  3258. #if IS_ENABLED(CONFIG_XFRM)
  3259. !skb->sp &&
  3260. #endif
  3261. !skb_nfct(skb) &&
  3262. !skb->_skb_refdst &&
  3263. !skb_has_frag_list(skb);
  3264. }
  3265. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  3266. {
  3267. skb->queue_mapping = queue_mapping;
  3268. }
  3269. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  3270. {
  3271. return skb->queue_mapping;
  3272. }
  3273. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  3274. {
  3275. to->queue_mapping = from->queue_mapping;
  3276. }
  3277. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  3278. {
  3279. skb->queue_mapping = rx_queue + 1;
  3280. }
  3281. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  3282. {
  3283. return skb->queue_mapping - 1;
  3284. }
  3285. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  3286. {
  3287. return skb->queue_mapping != 0;
  3288. }
  3289. static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
  3290. {
  3291. skb->dst_pending_confirm = val;
  3292. }
  3293. static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
  3294. {
  3295. return skb->dst_pending_confirm != 0;
  3296. }
  3297. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  3298. {
  3299. #ifdef CONFIG_XFRM
  3300. return skb->sp;
  3301. #else
  3302. return NULL;
  3303. #endif
  3304. }
  3305. /* Keeps track of mac header offset relative to skb->head.
  3306. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  3307. * For non-tunnel skb it points to skb_mac_header() and for
  3308. * tunnel skb it points to outer mac header.
  3309. * Keeps track of level of encapsulation of network headers.
  3310. */
  3311. struct skb_gso_cb {
  3312. union {
  3313. int mac_offset;
  3314. int data_offset;
  3315. };
  3316. int encap_level;
  3317. __wsum csum;
  3318. __u16 csum_start;
  3319. };
  3320. #define SKB_SGO_CB_OFFSET 32
  3321. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
  3322. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  3323. {
  3324. return (skb_mac_header(inner_skb) - inner_skb->head) -
  3325. SKB_GSO_CB(inner_skb)->mac_offset;
  3326. }
  3327. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  3328. {
  3329. int new_headroom, headroom;
  3330. int ret;
  3331. headroom = skb_headroom(skb);
  3332. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  3333. if (ret)
  3334. return ret;
  3335. new_headroom = skb_headroom(skb);
  3336. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  3337. return 0;
  3338. }
  3339. static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
  3340. {
  3341. /* Do not update partial checksums if remote checksum is enabled. */
  3342. if (skb->remcsum_offload)
  3343. return;
  3344. SKB_GSO_CB(skb)->csum = res;
  3345. SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
  3346. }
  3347. /* Compute the checksum for a gso segment. First compute the checksum value
  3348. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  3349. * then add in skb->csum (checksum from csum_start to end of packet).
  3350. * skb->csum and csum_start are then updated to reflect the checksum of the
  3351. * resultant packet starting from the transport header-- the resultant checksum
  3352. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  3353. * header.
  3354. */
  3355. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  3356. {
  3357. unsigned char *csum_start = skb_transport_header(skb);
  3358. int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
  3359. __wsum partial = SKB_GSO_CB(skb)->csum;
  3360. SKB_GSO_CB(skb)->csum = res;
  3361. SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
  3362. return csum_fold(csum_partial(csum_start, plen, partial));
  3363. }
  3364. static inline bool skb_is_gso(const struct sk_buff *skb)
  3365. {
  3366. return skb_shinfo(skb)->gso_size;
  3367. }
  3368. /* Note: Should be called only if skb_is_gso(skb) is true */
  3369. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  3370. {
  3371. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  3372. }
  3373. static inline void skb_gso_reset(struct sk_buff *skb)
  3374. {
  3375. skb_shinfo(skb)->gso_size = 0;
  3376. skb_shinfo(skb)->gso_segs = 0;
  3377. skb_shinfo(skb)->gso_type = 0;
  3378. }
  3379. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  3380. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  3381. {
  3382. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  3383. * wanted then gso_type will be set. */
  3384. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  3385. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  3386. unlikely(shinfo->gso_type == 0)) {
  3387. __skb_warn_lro_forwarding(skb);
  3388. return true;
  3389. }
  3390. return false;
  3391. }
  3392. static inline void skb_forward_csum(struct sk_buff *skb)
  3393. {
  3394. /* Unfortunately we don't support this one. Any brave souls? */
  3395. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3396. skb->ip_summed = CHECKSUM_NONE;
  3397. }
  3398. /**
  3399. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3400. * @skb: skb to check
  3401. *
  3402. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3403. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3404. * use this helper, to document places where we make this assertion.
  3405. */
  3406. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3407. {
  3408. #ifdef DEBUG
  3409. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3410. #endif
  3411. }
  3412. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3413. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3414. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3415. unsigned int transport_len,
  3416. __sum16(*skb_chkf)(struct sk_buff *skb));
  3417. /**
  3418. * skb_head_is_locked - Determine if the skb->head is locked down
  3419. * @skb: skb to check
  3420. *
  3421. * The head on skbs build around a head frag can be removed if they are
  3422. * not cloned. This function returns true if the skb head is locked down
  3423. * due to either being allocated via kmalloc, or by being a clone with
  3424. * multiple references to the head.
  3425. */
  3426. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3427. {
  3428. return !skb->head_frag || skb_cloned(skb);
  3429. }
  3430. /**
  3431. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3432. *
  3433. * @skb: GSO skb
  3434. *
  3435. * skb_gso_network_seglen is used to determine the real size of the
  3436. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3437. *
  3438. * The MAC/L2 header is not accounted for.
  3439. */
  3440. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3441. {
  3442. unsigned int hdr_len = skb_transport_header(skb) -
  3443. skb_network_header(skb);
  3444. return hdr_len + skb_gso_transport_seglen(skb);
  3445. }
  3446. /* Local Checksum Offload.
  3447. * Compute outer checksum based on the assumption that the
  3448. * inner checksum will be offloaded later.
  3449. * See Documentation/networking/checksum-offloads.txt for
  3450. * explanation of how this works.
  3451. * Fill in outer checksum adjustment (e.g. with sum of outer
  3452. * pseudo-header) before calling.
  3453. * Also ensure that inner checksum is in linear data area.
  3454. */
  3455. static inline __wsum lco_csum(struct sk_buff *skb)
  3456. {
  3457. unsigned char *csum_start = skb_checksum_start(skb);
  3458. unsigned char *l4_hdr = skb_transport_header(skb);
  3459. __wsum partial;
  3460. /* Start with complement of inner checksum adjustment */
  3461. partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
  3462. skb->csum_offset));
  3463. /* Add in checksum of our headers (incl. outer checksum
  3464. * adjustment filled in by caller) and return result.
  3465. */
  3466. return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
  3467. }
  3468. #endif /* __KERNEL__ */
  3469. #endif /* _LINUX_SKBUFF_H */