skbuff.h 119 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963296429652966296729682969297029712972297329742975297629772978297929802981298229832984298529862987298829892990299129922993299429952996299729982999300030013002300330043005300630073008300930103011301230133014301530163017301830193020302130223023302430253026302730283029303030313032303330343035303630373038303930403041304230433044304530463047304830493050305130523053305430553056305730583059306030613062306330643065306630673068306930703071307230733074307530763077307830793080308130823083308430853086308730883089309030913092309330943095309630973098309931003101310231033104310531063107310831093110311131123113311431153116311731183119312031213122312331243125312631273128312931303131313231333134313531363137313831393140314131423143314431453146314731483149315031513152315331543155315631573158315931603161316231633164316531663167316831693170317131723173317431753176317731783179318031813182318331843185318631873188318931903191319231933194319531963197319831993200320132023203320432053206320732083209321032113212321332143215321632173218321932203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278327932803281328232833284328532863287328832893290329132923293329432953296329732983299330033013302330333043305330633073308330933103311331233133314331533163317331833193320332133223323332433253326332733283329333033313332333333343335333633373338333933403341334233433344334533463347334833493350335133523353335433553356335733583359336033613362336333643365336633673368336933703371337233733374337533763377337833793380338133823383338433853386338733883389339033913392339333943395339633973398339934003401340234033404340534063407340834093410341134123413341434153416341734183419342034213422342334243425342634273428342934303431343234333434343534363437343834393440344134423443344434453446344734483449345034513452345334543455345634573458345934603461346234633464346534663467346834693470347134723473347434753476347734783479348034813482348334843485348634873488348934903491349234933494349534963497349834993500350135023503350435053506350735083509351035113512351335143515351635173518351935203521352235233524352535263527352835293530353135323533353435353536353735383539354035413542354335443545354635473548354935503551355235533554355535563557355835593560356135623563356435653566356735683569357035713572357335743575357635773578357935803581358235833584358535863587358835893590359135923593359435953596359735983599360036013602360336043605360636073608360936103611361236133614361536163617361836193620362136223623362436253626362736283629363036313632363336343635363636373638363936403641364236433644364536463647364836493650365136523653365436553656365736583659366036613662366336643665366636673668366936703671367236733674367536763677367836793680368136823683368436853686368736883689369036913692369336943695369636973698369937003701370237033704370537063707370837093710371137123713371437153716371737183719372037213722372337243725372637273728372937303731373237333734373537363737373837393740374137423743374437453746374737483749375037513752375337543755375637573758375937603761376237633764376537663767376837693770377137723773377437753776377737783779378037813782378337843785378637873788378937903791379237933794379537963797379837993800380138023803380438053806380738083809381038113812381338143815381638173818381938203821382238233824382538263827382838293830383138323833383438353836383738383839384038413842384338443845384638473848384938503851385238533854385538563857385838593860386138623863386438653866386738683869387038713872387338743875387638773878387938803881388238833884388538863887388838893890389138923893389438953896389738983899390039013902390339043905390639073908390939103911391239133914391539163917391839193920392139223923392439253926392739283929393039313932393339343935393639373938393939403941394239433944394539463947394839493950395139523953395439553956395739583959396039613962396339643965396639673968396939703971397239733974397539763977397839793980398139823983398439853986398739883989399039913992399339943995399639973998399940004001400240034004400540064007400840094010401140124013401440154016401740184019402040214022402340244025402640274028402940304031403240334034403540364037403840394040404140424043404440454046404740484049405040514052405340544055405640574058405940604061406240634064406540664067406840694070407140724073407440754076407740784079408040814082408340844085408640874088408940904091409240934094409540964097409840994100410141024103410441054106410741084109411041114112411341144115411641174118411941204121412241234124412541264127412841294130413141324133413441354136413741384139414041414142414341444145414641474148414941504151415241534154415541564157415841594160416141624163416441654166416741684169417041714172417341744175417641774178417941804181418241834184418541864187418841894190419141924193419441954196
  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/compiler.h>
  17. #include <linux/time.h>
  18. #include <linux/bug.h>
  19. #include <linux/cache.h>
  20. #include <linux/rbtree.h>
  21. #include <linux/socket.h>
  22. #include <linux/refcount.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. static inline bool skb_frag_must_loop(struct page *p)
  324. {
  325. #if defined(CONFIG_HIGHMEM)
  326. if (PageHighMem(p))
  327. return true;
  328. #endif
  329. return false;
  330. }
  331. /**
  332. * skb_frag_foreach_page - loop over pages in a fragment
  333. *
  334. * @f: skb frag to operate on
  335. * @f_off: offset from start of f->page.p
  336. * @f_len: length from f_off to loop over
  337. * @p: (temp var) current page
  338. * @p_off: (temp var) offset from start of current page,
  339. * non-zero only on first page.
  340. * @p_len: (temp var) length in current page,
  341. * < PAGE_SIZE only on first and last page.
  342. * @copied: (temp var) length so far, excluding current p_len.
  343. *
  344. * A fragment can hold a compound page, in which case per-page
  345. * operations, notably kmap_atomic, must be called for each
  346. * regular page.
  347. */
  348. #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
  349. for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
  350. p_off = (f_off) & (PAGE_SIZE - 1), \
  351. p_len = skb_frag_must_loop(p) ? \
  352. min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
  353. copied = 0; \
  354. copied < f_len; \
  355. copied += p_len, p++, p_off = 0, \
  356. p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
  357. #define HAVE_HW_TIME_STAMP
  358. /**
  359. * struct skb_shared_hwtstamps - hardware time stamps
  360. * @hwtstamp: hardware time stamp transformed into duration
  361. * since arbitrary point in time
  362. *
  363. * Software time stamps generated by ktime_get_real() are stored in
  364. * skb->tstamp.
  365. *
  366. * hwtstamps can only be compared against other hwtstamps from
  367. * the same device.
  368. *
  369. * This structure is attached to packets as part of the
  370. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  371. */
  372. struct skb_shared_hwtstamps {
  373. ktime_t hwtstamp;
  374. };
  375. /* Definitions for tx_flags in struct skb_shared_info */
  376. enum {
  377. /* generate hardware time stamp */
  378. SKBTX_HW_TSTAMP = 1 << 0,
  379. /* generate software time stamp when queueing packet to NIC */
  380. SKBTX_SW_TSTAMP = 1 << 1,
  381. /* device driver is going to provide hardware time stamp */
  382. SKBTX_IN_PROGRESS = 1 << 2,
  383. /* device driver supports TX zero-copy buffers */
  384. SKBTX_DEV_ZEROCOPY = 1 << 3,
  385. /* generate wifi status information (where possible) */
  386. SKBTX_WIFI_STATUS = 1 << 4,
  387. /* This indicates at least one fragment might be overwritten
  388. * (as in vmsplice(), sendfile() ...)
  389. * If we need to compute a TX checksum, we'll need to copy
  390. * all frags to avoid possible bad checksum
  391. */
  392. SKBTX_SHARED_FRAG = 1 << 5,
  393. /* generate software time stamp when entering packet scheduling */
  394. SKBTX_SCHED_TSTAMP = 1 << 6,
  395. };
  396. #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
  397. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  398. SKBTX_SCHED_TSTAMP)
  399. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  400. /*
  401. * The callback notifies userspace to release buffers when skb DMA is done in
  402. * lower device, the skb last reference should be 0 when calling this.
  403. * The zerocopy_success argument is true if zero copy transmit occurred,
  404. * false on data copy or out of memory error caused by data copy attempt.
  405. * The ctx field is used to track device context.
  406. * The desc field is used to track userspace buffer index.
  407. */
  408. struct ubuf_info {
  409. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  410. union {
  411. struct {
  412. unsigned long desc;
  413. void *ctx;
  414. };
  415. struct {
  416. u32 id;
  417. u16 len;
  418. u16 zerocopy:1;
  419. u32 bytelen;
  420. };
  421. };
  422. refcount_t refcnt;
  423. struct mmpin {
  424. struct user_struct *user;
  425. unsigned int num_pg;
  426. } mmp;
  427. };
  428. #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
  429. int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
  430. void mm_unaccount_pinned_pages(struct mmpin *mmp);
  431. struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
  432. struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
  433. struct ubuf_info *uarg);
  434. static inline void sock_zerocopy_get(struct ubuf_info *uarg)
  435. {
  436. refcount_inc(&uarg->refcnt);
  437. }
  438. void sock_zerocopy_put(struct ubuf_info *uarg);
  439. void sock_zerocopy_put_abort(struct ubuf_info *uarg);
  440. void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
  441. int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
  442. struct msghdr *msg, int len,
  443. struct ubuf_info *uarg);
  444. /* This data is invariant across clones and lives at
  445. * the end of the header data, ie. at skb->end.
  446. */
  447. struct skb_shared_info {
  448. __u8 __unused;
  449. __u8 meta_len;
  450. __u8 nr_frags;
  451. __u8 tx_flags;
  452. unsigned short gso_size;
  453. /* Warning: this field is not always filled in (UFO)! */
  454. unsigned short gso_segs;
  455. struct sk_buff *frag_list;
  456. struct skb_shared_hwtstamps hwtstamps;
  457. unsigned int gso_type;
  458. u32 tskey;
  459. /*
  460. * Warning : all fields before dataref are cleared in __alloc_skb()
  461. */
  462. atomic_t dataref;
  463. /* Intermediate layers must ensure that destructor_arg
  464. * remains valid until skb destructor */
  465. void * destructor_arg;
  466. /* must be last field, see pskb_expand_head() */
  467. skb_frag_t frags[MAX_SKB_FRAGS];
  468. };
  469. /* We divide dataref into two halves. The higher 16 bits hold references
  470. * to the payload part of skb->data. The lower 16 bits hold references to
  471. * the entire skb->data. A clone of a headerless skb holds the length of
  472. * the header in skb->hdr_len.
  473. *
  474. * All users must obey the rule that the skb->data reference count must be
  475. * greater than or equal to the payload reference count.
  476. *
  477. * Holding a reference to the payload part means that the user does not
  478. * care about modifications to the header part of skb->data.
  479. */
  480. #define SKB_DATAREF_SHIFT 16
  481. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  482. enum {
  483. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  484. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  485. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  486. };
  487. enum {
  488. SKB_GSO_TCPV4 = 1 << 0,
  489. /* This indicates the skb is from an untrusted source. */
  490. SKB_GSO_DODGY = 1 << 1,
  491. /* This indicates the tcp segment has CWR set. */
  492. SKB_GSO_TCP_ECN = 1 << 2,
  493. SKB_GSO_TCP_FIXEDID = 1 << 3,
  494. SKB_GSO_TCPV6 = 1 << 4,
  495. SKB_GSO_FCOE = 1 << 5,
  496. SKB_GSO_GRE = 1 << 6,
  497. SKB_GSO_GRE_CSUM = 1 << 7,
  498. SKB_GSO_IPXIP4 = 1 << 8,
  499. SKB_GSO_IPXIP6 = 1 << 9,
  500. SKB_GSO_UDP_TUNNEL = 1 << 10,
  501. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  502. SKB_GSO_PARTIAL = 1 << 12,
  503. SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
  504. SKB_GSO_SCTP = 1 << 14,
  505. SKB_GSO_ESP = 1 << 15,
  506. SKB_GSO_UDP = 1 << 16,
  507. SKB_GSO_UDP_L4 = 1 << 17,
  508. };
  509. #if BITS_PER_LONG > 32
  510. #define NET_SKBUFF_DATA_USES_OFFSET 1
  511. #endif
  512. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  513. typedef unsigned int sk_buff_data_t;
  514. #else
  515. typedef unsigned char *sk_buff_data_t;
  516. #endif
  517. /**
  518. * struct sk_buff - socket buffer
  519. * @next: Next buffer in list
  520. * @prev: Previous buffer in list
  521. * @tstamp: Time we arrived/left
  522. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  523. * @sk: Socket we are owned by
  524. * @dev: Device we arrived on/are leaving by
  525. * @cb: Control buffer. Free for use by every layer. Put private vars here
  526. * @_skb_refdst: destination entry (with norefcount bit)
  527. * @sp: the security path, used for xfrm
  528. * @len: Length of actual data
  529. * @data_len: Data length
  530. * @mac_len: Length of link layer header
  531. * @hdr_len: writable header length of cloned skb
  532. * @csum: Checksum (must include start/offset pair)
  533. * @csum_start: Offset from skb->head where checksumming should start
  534. * @csum_offset: Offset from csum_start where checksum should be stored
  535. * @priority: Packet queueing priority
  536. * @ignore_df: allow local fragmentation
  537. * @cloned: Head may be cloned (check refcnt to be sure)
  538. * @ip_summed: Driver fed us an IP checksum
  539. * @nohdr: Payload reference only, must not modify header
  540. * @pkt_type: Packet class
  541. * @fclone: skbuff clone status
  542. * @ipvs_property: skbuff is owned by ipvs
  543. * @tc_skip_classify: do not classify packet. set by IFB device
  544. * @tc_at_ingress: used within tc_classify to distinguish in/egress
  545. * @tc_redirected: packet was redirected by a tc action
  546. * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
  547. * @peeked: this packet has been seen already, so stats have been
  548. * done for it, don't do them again
  549. * @nf_trace: netfilter packet trace flag
  550. * @protocol: Packet protocol from driver
  551. * @destructor: Destruct function
  552. * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
  553. * @_nfct: Associated connection, if any (with nfctinfo bits)
  554. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  555. * @skb_iif: ifindex of device we arrived on
  556. * @tc_index: Traffic control index
  557. * @hash: the packet hash
  558. * @queue_mapping: Queue mapping for multiqueue devices
  559. * @xmit_more: More SKBs are pending for this queue
  560. * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
  561. * @ndisc_nodetype: router type (from link layer)
  562. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  563. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  564. * ports.
  565. * @sw_hash: indicates hash was computed in software stack
  566. * @wifi_acked_valid: wifi_acked was set
  567. * @wifi_acked: whether frame was acked on wifi or not
  568. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  569. * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
  570. * @dst_pending_confirm: need to confirm neighbour
  571. * @decrypted: Decrypted SKB
  572. * @napi_id: id of the NAPI struct this skb came from
  573. * @secmark: security marking
  574. * @mark: Generic packet mark
  575. * @vlan_proto: vlan encapsulation protocol
  576. * @vlan_tci: vlan tag control information
  577. * @inner_protocol: Protocol (encapsulation)
  578. * @inner_transport_header: Inner transport layer header (encapsulation)
  579. * @inner_network_header: Network layer header (encapsulation)
  580. * @inner_mac_header: Link layer header (encapsulation)
  581. * @transport_header: Transport layer header
  582. * @network_header: Network layer header
  583. * @mac_header: Link layer header
  584. * @tail: Tail pointer
  585. * @end: End pointer
  586. * @head: Head of buffer
  587. * @data: Data head pointer
  588. * @truesize: Buffer size
  589. * @users: User count - see {datagram,tcp}.c
  590. */
  591. struct sk_buff {
  592. union {
  593. struct {
  594. /* These two members must be first. */
  595. struct sk_buff *next;
  596. struct sk_buff *prev;
  597. union {
  598. struct net_device *dev;
  599. /* Some protocols might use this space to store information,
  600. * while device pointer would be NULL.
  601. * UDP receive path is one user.
  602. */
  603. unsigned long dev_scratch;
  604. };
  605. };
  606. struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
  607. struct list_head list;
  608. };
  609. union {
  610. struct sock *sk;
  611. int ip_defrag_offset;
  612. };
  613. union {
  614. ktime_t tstamp;
  615. u64 skb_mstamp;
  616. };
  617. /*
  618. * This is the control buffer. It is free to use for every
  619. * layer. Please put your private variables there. If you
  620. * want to keep them across layers you have to do a skb_clone()
  621. * first. This is owned by whoever has the skb queued ATM.
  622. */
  623. char cb[48] __aligned(8);
  624. union {
  625. struct {
  626. unsigned long _skb_refdst;
  627. void (*destructor)(struct sk_buff *skb);
  628. };
  629. struct list_head tcp_tsorted_anchor;
  630. };
  631. #ifdef CONFIG_XFRM
  632. struct sec_path *sp;
  633. #endif
  634. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  635. unsigned long _nfct;
  636. #endif
  637. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  638. struct nf_bridge_info *nf_bridge;
  639. #endif
  640. unsigned int len,
  641. data_len;
  642. __u16 mac_len,
  643. hdr_len;
  644. /* Following fields are _not_ copied in __copy_skb_header()
  645. * Note that queue_mapping is here mostly to fill a hole.
  646. */
  647. __u16 queue_mapping;
  648. /* if you move cloned around you also must adapt those constants */
  649. #ifdef __BIG_ENDIAN_BITFIELD
  650. #define CLONED_MASK (1 << 7)
  651. #else
  652. #define CLONED_MASK 1
  653. #endif
  654. #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
  655. __u8 __cloned_offset[0];
  656. __u8 cloned:1,
  657. nohdr:1,
  658. fclone:2,
  659. peeked:1,
  660. head_frag:1,
  661. xmit_more:1,
  662. pfmemalloc:1;
  663. /* fields enclosed in headers_start/headers_end are copied
  664. * using a single memcpy() in __copy_skb_header()
  665. */
  666. /* private: */
  667. __u32 headers_start[0];
  668. /* public: */
  669. /* if you move pkt_type around you also must adapt those constants */
  670. #ifdef __BIG_ENDIAN_BITFIELD
  671. #define PKT_TYPE_MAX (7 << 5)
  672. #else
  673. #define PKT_TYPE_MAX 7
  674. #endif
  675. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  676. __u8 __pkt_type_offset[0];
  677. __u8 pkt_type:3;
  678. __u8 ignore_df:1;
  679. __u8 nf_trace:1;
  680. __u8 ip_summed:2;
  681. __u8 ooo_okay:1;
  682. __u8 l4_hash:1;
  683. __u8 sw_hash:1;
  684. __u8 wifi_acked_valid:1;
  685. __u8 wifi_acked:1;
  686. __u8 no_fcs:1;
  687. /* Indicates the inner headers are valid in the skbuff. */
  688. __u8 encapsulation:1;
  689. __u8 encap_hdr_csum:1;
  690. __u8 csum_valid:1;
  691. __u8 csum_complete_sw:1;
  692. __u8 csum_level:2;
  693. __u8 csum_not_inet:1;
  694. __u8 dst_pending_confirm:1;
  695. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  696. __u8 ndisc_nodetype:2;
  697. #endif
  698. __u8 ipvs_property:1;
  699. __u8 inner_protocol_type:1;
  700. __u8 remcsum_offload:1;
  701. #ifdef CONFIG_NET_SWITCHDEV
  702. __u8 offload_fwd_mark:1;
  703. __u8 offload_mr_fwd_mark:1;
  704. #endif
  705. #ifdef CONFIG_NET_CLS_ACT
  706. __u8 tc_skip_classify:1;
  707. __u8 tc_at_ingress:1;
  708. __u8 tc_redirected:1;
  709. __u8 tc_from_ingress:1;
  710. #endif
  711. #ifdef CONFIG_TLS_DEVICE
  712. __u8 decrypted:1;
  713. #endif
  714. #ifdef CONFIG_NET_SCHED
  715. __u16 tc_index; /* traffic control index */
  716. #endif
  717. union {
  718. __wsum csum;
  719. struct {
  720. __u16 csum_start;
  721. __u16 csum_offset;
  722. };
  723. };
  724. __u32 priority;
  725. int skb_iif;
  726. __u32 hash;
  727. __be16 vlan_proto;
  728. __u16 vlan_tci;
  729. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  730. union {
  731. unsigned int napi_id;
  732. unsigned int sender_cpu;
  733. };
  734. #endif
  735. #ifdef CONFIG_NETWORK_SECMARK
  736. __u32 secmark;
  737. #endif
  738. union {
  739. __u32 mark;
  740. __u32 reserved_tailroom;
  741. };
  742. union {
  743. __be16 inner_protocol;
  744. __u8 inner_ipproto;
  745. };
  746. __u16 inner_transport_header;
  747. __u16 inner_network_header;
  748. __u16 inner_mac_header;
  749. __be16 protocol;
  750. __u16 transport_header;
  751. __u16 network_header;
  752. __u16 mac_header;
  753. /* private: */
  754. __u32 headers_end[0];
  755. /* public: */
  756. /* These elements must be at the end, see alloc_skb() for details. */
  757. sk_buff_data_t tail;
  758. sk_buff_data_t end;
  759. unsigned char *head,
  760. *data;
  761. unsigned int truesize;
  762. refcount_t users;
  763. };
  764. #ifdef __KERNEL__
  765. /*
  766. * Handling routines are only of interest to the kernel
  767. */
  768. #define SKB_ALLOC_FCLONE 0x01
  769. #define SKB_ALLOC_RX 0x02
  770. #define SKB_ALLOC_NAPI 0x04
  771. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  772. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  773. {
  774. return unlikely(skb->pfmemalloc);
  775. }
  776. /*
  777. * skb might have a dst pointer attached, refcounted or not.
  778. * _skb_refdst low order bit is set if refcount was _not_ taken
  779. */
  780. #define SKB_DST_NOREF 1UL
  781. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  782. #define SKB_NFCT_PTRMASK ~(7UL)
  783. /**
  784. * skb_dst - returns skb dst_entry
  785. * @skb: buffer
  786. *
  787. * Returns skb dst_entry, regardless of reference taken or not.
  788. */
  789. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  790. {
  791. /* If refdst was not refcounted, check we still are in a
  792. * rcu_read_lock section
  793. */
  794. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  795. !rcu_read_lock_held() &&
  796. !rcu_read_lock_bh_held());
  797. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  798. }
  799. /**
  800. * skb_dst_set - sets skb dst
  801. * @skb: buffer
  802. * @dst: dst entry
  803. *
  804. * Sets skb dst, assuming a reference was taken on dst and should
  805. * be released by skb_dst_drop()
  806. */
  807. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  808. {
  809. skb->_skb_refdst = (unsigned long)dst;
  810. }
  811. /**
  812. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  813. * @skb: buffer
  814. * @dst: dst entry
  815. *
  816. * Sets skb dst, assuming a reference was not taken on dst.
  817. * If dst entry is cached, we do not take reference and dst_release
  818. * will be avoided by refdst_drop. If dst entry is not cached, we take
  819. * reference, so that last dst_release can destroy the dst immediately.
  820. */
  821. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  822. {
  823. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  824. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  825. }
  826. /**
  827. * skb_dst_is_noref - Test if skb dst isn't refcounted
  828. * @skb: buffer
  829. */
  830. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  831. {
  832. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  833. }
  834. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  835. {
  836. return (struct rtable *)skb_dst(skb);
  837. }
  838. /* For mangling skb->pkt_type from user space side from applications
  839. * such as nft, tc, etc, we only allow a conservative subset of
  840. * possible pkt_types to be set.
  841. */
  842. static inline bool skb_pkt_type_ok(u32 ptype)
  843. {
  844. return ptype <= PACKET_OTHERHOST;
  845. }
  846. static inline unsigned int skb_napi_id(const struct sk_buff *skb)
  847. {
  848. #ifdef CONFIG_NET_RX_BUSY_POLL
  849. return skb->napi_id;
  850. #else
  851. return 0;
  852. #endif
  853. }
  854. /* decrement the reference count and return true if we can free the skb */
  855. static inline bool skb_unref(struct sk_buff *skb)
  856. {
  857. if (unlikely(!skb))
  858. return false;
  859. if (likely(refcount_read(&skb->users) == 1))
  860. smp_rmb();
  861. else if (likely(!refcount_dec_and_test(&skb->users)))
  862. return false;
  863. return true;
  864. }
  865. void skb_release_head_state(struct sk_buff *skb);
  866. void kfree_skb(struct sk_buff *skb);
  867. void kfree_skb_list(struct sk_buff *segs);
  868. void skb_tx_error(struct sk_buff *skb);
  869. void consume_skb(struct sk_buff *skb);
  870. void __consume_stateless_skb(struct sk_buff *skb);
  871. void __kfree_skb(struct sk_buff *skb);
  872. extern struct kmem_cache *skbuff_head_cache;
  873. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  874. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  875. bool *fragstolen, int *delta_truesize);
  876. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  877. int node);
  878. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  879. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  880. static inline struct sk_buff *alloc_skb(unsigned int size,
  881. gfp_t priority)
  882. {
  883. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  884. }
  885. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  886. unsigned long data_len,
  887. int max_page_order,
  888. int *errcode,
  889. gfp_t gfp_mask);
  890. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  891. struct sk_buff_fclones {
  892. struct sk_buff skb1;
  893. struct sk_buff skb2;
  894. refcount_t fclone_ref;
  895. };
  896. /**
  897. * skb_fclone_busy - check if fclone is busy
  898. * @sk: socket
  899. * @skb: buffer
  900. *
  901. * Returns true if skb is a fast clone, and its clone is not freed.
  902. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  903. * so we also check that this didnt happen.
  904. */
  905. static inline bool skb_fclone_busy(const struct sock *sk,
  906. const struct sk_buff *skb)
  907. {
  908. const struct sk_buff_fclones *fclones;
  909. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  910. return skb->fclone == SKB_FCLONE_ORIG &&
  911. refcount_read(&fclones->fclone_ref) > 1 &&
  912. fclones->skb2.sk == sk;
  913. }
  914. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  915. gfp_t priority)
  916. {
  917. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  918. }
  919. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  920. void skb_headers_offset_update(struct sk_buff *skb, int off);
  921. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  922. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  923. void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
  924. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  925. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  926. gfp_t gfp_mask, bool fclone);
  927. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  928. gfp_t gfp_mask)
  929. {
  930. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  931. }
  932. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  933. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  934. unsigned int headroom);
  935. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  936. int newtailroom, gfp_t priority);
  937. int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  938. int offset, int len);
  939. int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
  940. int offset, int len);
  941. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  942. int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
  943. /**
  944. * skb_pad - zero pad the tail of an skb
  945. * @skb: buffer to pad
  946. * @pad: space to pad
  947. *
  948. * Ensure that a buffer is followed by a padding area that is zero
  949. * filled. Used by network drivers which may DMA or transfer data
  950. * beyond the buffer end onto the wire.
  951. *
  952. * May return error in out of memory cases. The skb is freed on error.
  953. */
  954. static inline int skb_pad(struct sk_buff *skb, int pad)
  955. {
  956. return __skb_pad(skb, pad, true);
  957. }
  958. #define dev_kfree_skb(a) consume_skb(a)
  959. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  960. int getfrag(void *from, char *to, int offset,
  961. int len, int odd, struct sk_buff *skb),
  962. void *from, int length);
  963. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  964. int offset, size_t size);
  965. struct skb_seq_state {
  966. __u32 lower_offset;
  967. __u32 upper_offset;
  968. __u32 frag_idx;
  969. __u32 stepped_offset;
  970. struct sk_buff *root_skb;
  971. struct sk_buff *cur_skb;
  972. __u8 *frag_data;
  973. };
  974. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  975. unsigned int to, struct skb_seq_state *st);
  976. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  977. struct skb_seq_state *st);
  978. void skb_abort_seq_read(struct skb_seq_state *st);
  979. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  980. unsigned int to, struct ts_config *config);
  981. /*
  982. * Packet hash types specify the type of hash in skb_set_hash.
  983. *
  984. * Hash types refer to the protocol layer addresses which are used to
  985. * construct a packet's hash. The hashes are used to differentiate or identify
  986. * flows of the protocol layer for the hash type. Hash types are either
  987. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  988. *
  989. * Properties of hashes:
  990. *
  991. * 1) Two packets in different flows have different hash values
  992. * 2) Two packets in the same flow should have the same hash value
  993. *
  994. * A hash at a higher layer is considered to be more specific. A driver should
  995. * set the most specific hash possible.
  996. *
  997. * A driver cannot indicate a more specific hash than the layer at which a hash
  998. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  999. *
  1000. * A driver may indicate a hash level which is less specific than the
  1001. * actual layer the hash was computed on. For instance, a hash computed
  1002. * at L4 may be considered an L3 hash. This should only be done if the
  1003. * driver can't unambiguously determine that the HW computed the hash at
  1004. * the higher layer. Note that the "should" in the second property above
  1005. * permits this.
  1006. */
  1007. enum pkt_hash_types {
  1008. PKT_HASH_TYPE_NONE, /* Undefined type */
  1009. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  1010. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  1011. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  1012. };
  1013. static inline void skb_clear_hash(struct sk_buff *skb)
  1014. {
  1015. skb->hash = 0;
  1016. skb->sw_hash = 0;
  1017. skb->l4_hash = 0;
  1018. }
  1019. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  1020. {
  1021. if (!skb->l4_hash)
  1022. skb_clear_hash(skb);
  1023. }
  1024. static inline void
  1025. __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
  1026. {
  1027. skb->l4_hash = is_l4;
  1028. skb->sw_hash = is_sw;
  1029. skb->hash = hash;
  1030. }
  1031. static inline void
  1032. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  1033. {
  1034. /* Used by drivers to set hash from HW */
  1035. __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
  1036. }
  1037. static inline void
  1038. __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
  1039. {
  1040. __skb_set_hash(skb, hash, true, is_l4);
  1041. }
  1042. void __skb_get_hash(struct sk_buff *skb);
  1043. u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
  1044. u32 skb_get_poff(const struct sk_buff *skb);
  1045. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  1046. const struct flow_keys_basic *keys, int hlen);
  1047. __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
  1048. void *data, int hlen_proto);
  1049. static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
  1050. int thoff, u8 ip_proto)
  1051. {
  1052. return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
  1053. }
  1054. void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  1055. const struct flow_dissector_key *key,
  1056. unsigned int key_count);
  1057. bool __skb_flow_dissect(const struct sk_buff *skb,
  1058. struct flow_dissector *flow_dissector,
  1059. void *target_container,
  1060. void *data, __be16 proto, int nhoff, int hlen,
  1061. unsigned int flags);
  1062. static inline bool skb_flow_dissect(const struct sk_buff *skb,
  1063. struct flow_dissector *flow_dissector,
  1064. void *target_container, unsigned int flags)
  1065. {
  1066. return __skb_flow_dissect(skb, flow_dissector, target_container,
  1067. NULL, 0, 0, 0, flags);
  1068. }
  1069. static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
  1070. struct flow_keys *flow,
  1071. unsigned int flags)
  1072. {
  1073. memset(flow, 0, sizeof(*flow));
  1074. return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
  1075. NULL, 0, 0, 0, flags);
  1076. }
  1077. static inline bool
  1078. skb_flow_dissect_flow_keys_basic(const struct sk_buff *skb,
  1079. struct flow_keys_basic *flow, void *data,
  1080. __be16 proto, int nhoff, int hlen,
  1081. unsigned int flags)
  1082. {
  1083. memset(flow, 0, sizeof(*flow));
  1084. return __skb_flow_dissect(skb, &flow_keys_basic_dissector, flow,
  1085. data, proto, nhoff, hlen, flags);
  1086. }
  1087. void
  1088. skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
  1089. struct flow_dissector *flow_dissector,
  1090. void *target_container);
  1091. static inline __u32 skb_get_hash(struct sk_buff *skb)
  1092. {
  1093. if (!skb->l4_hash && !skb->sw_hash)
  1094. __skb_get_hash(skb);
  1095. return skb->hash;
  1096. }
  1097. static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
  1098. {
  1099. if (!skb->l4_hash && !skb->sw_hash) {
  1100. struct flow_keys keys;
  1101. __u32 hash = __get_hash_from_flowi6(fl6, &keys);
  1102. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1103. }
  1104. return skb->hash;
  1105. }
  1106. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  1107. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  1108. {
  1109. return skb->hash;
  1110. }
  1111. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  1112. {
  1113. to->hash = from->hash;
  1114. to->sw_hash = from->sw_hash;
  1115. to->l4_hash = from->l4_hash;
  1116. };
  1117. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1118. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1119. {
  1120. return skb->head + skb->end;
  1121. }
  1122. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1123. {
  1124. return skb->end;
  1125. }
  1126. #else
  1127. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1128. {
  1129. return skb->end;
  1130. }
  1131. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1132. {
  1133. return skb->end - skb->head;
  1134. }
  1135. #endif
  1136. /* Internal */
  1137. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  1138. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  1139. {
  1140. return &skb_shinfo(skb)->hwtstamps;
  1141. }
  1142. static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
  1143. {
  1144. bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
  1145. return is_zcopy ? skb_uarg(skb) : NULL;
  1146. }
  1147. static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
  1148. {
  1149. if (skb && uarg && !skb_zcopy(skb)) {
  1150. sock_zerocopy_get(uarg);
  1151. skb_shinfo(skb)->destructor_arg = uarg;
  1152. skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
  1153. }
  1154. }
  1155. static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
  1156. {
  1157. skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
  1158. skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
  1159. }
  1160. static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
  1161. {
  1162. return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
  1163. }
  1164. static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
  1165. {
  1166. return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
  1167. }
  1168. /* Release a reference on a zerocopy structure */
  1169. static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
  1170. {
  1171. struct ubuf_info *uarg = skb_zcopy(skb);
  1172. if (uarg) {
  1173. if (uarg->callback == sock_zerocopy_callback) {
  1174. uarg->zerocopy = uarg->zerocopy && zerocopy;
  1175. sock_zerocopy_put(uarg);
  1176. } else if (!skb_zcopy_is_nouarg(skb)) {
  1177. uarg->callback(uarg, zerocopy);
  1178. }
  1179. skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
  1180. }
  1181. }
  1182. /* Abort a zerocopy operation and revert zckey on error in send syscall */
  1183. static inline void skb_zcopy_abort(struct sk_buff *skb)
  1184. {
  1185. struct ubuf_info *uarg = skb_zcopy(skb);
  1186. if (uarg) {
  1187. sock_zerocopy_put_abort(uarg);
  1188. skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
  1189. }
  1190. }
  1191. static inline void skb_mark_not_on_list(struct sk_buff *skb)
  1192. {
  1193. skb->next = NULL;
  1194. }
  1195. static inline void skb_list_del_init(struct sk_buff *skb)
  1196. {
  1197. __list_del_entry(&skb->list);
  1198. skb_mark_not_on_list(skb);
  1199. }
  1200. /**
  1201. * skb_queue_empty - check if a queue is empty
  1202. * @list: queue head
  1203. *
  1204. * Returns true if the queue is empty, false otherwise.
  1205. */
  1206. static inline int skb_queue_empty(const struct sk_buff_head *list)
  1207. {
  1208. return list->next == (const struct sk_buff *) list;
  1209. }
  1210. /**
  1211. * skb_queue_is_last - check if skb is the last entry in the queue
  1212. * @list: queue head
  1213. * @skb: buffer
  1214. *
  1215. * Returns true if @skb is the last buffer on the list.
  1216. */
  1217. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  1218. const struct sk_buff *skb)
  1219. {
  1220. return skb->next == (const struct sk_buff *) list;
  1221. }
  1222. /**
  1223. * skb_queue_is_first - check if skb is the first entry in the queue
  1224. * @list: queue head
  1225. * @skb: buffer
  1226. *
  1227. * Returns true if @skb is the first buffer on the list.
  1228. */
  1229. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  1230. const struct sk_buff *skb)
  1231. {
  1232. return skb->prev == (const struct sk_buff *) list;
  1233. }
  1234. /**
  1235. * skb_queue_next - return the next packet in the queue
  1236. * @list: queue head
  1237. * @skb: current buffer
  1238. *
  1239. * Return the next packet in @list after @skb. It is only valid to
  1240. * call this if skb_queue_is_last() evaluates to false.
  1241. */
  1242. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  1243. const struct sk_buff *skb)
  1244. {
  1245. /* This BUG_ON may seem severe, but if we just return then we
  1246. * are going to dereference garbage.
  1247. */
  1248. BUG_ON(skb_queue_is_last(list, skb));
  1249. return skb->next;
  1250. }
  1251. /**
  1252. * skb_queue_prev - return the prev packet in the queue
  1253. * @list: queue head
  1254. * @skb: current buffer
  1255. *
  1256. * Return the prev packet in @list before @skb. It is only valid to
  1257. * call this if skb_queue_is_first() evaluates to false.
  1258. */
  1259. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  1260. const struct sk_buff *skb)
  1261. {
  1262. /* This BUG_ON may seem severe, but if we just return then we
  1263. * are going to dereference garbage.
  1264. */
  1265. BUG_ON(skb_queue_is_first(list, skb));
  1266. return skb->prev;
  1267. }
  1268. /**
  1269. * skb_get - reference buffer
  1270. * @skb: buffer to reference
  1271. *
  1272. * Makes another reference to a socket buffer and returns a pointer
  1273. * to the buffer.
  1274. */
  1275. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  1276. {
  1277. refcount_inc(&skb->users);
  1278. return skb;
  1279. }
  1280. /*
  1281. * If users == 1, we are the only owner and can avoid redundant atomic changes.
  1282. */
  1283. /**
  1284. * skb_cloned - is the buffer a clone
  1285. * @skb: buffer to check
  1286. *
  1287. * Returns true if the buffer was generated with skb_clone() and is
  1288. * one of multiple shared copies of the buffer. Cloned buffers are
  1289. * shared data so must not be written to under normal circumstances.
  1290. */
  1291. static inline int skb_cloned(const struct sk_buff *skb)
  1292. {
  1293. return skb->cloned &&
  1294. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  1295. }
  1296. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  1297. {
  1298. might_sleep_if(gfpflags_allow_blocking(pri));
  1299. if (skb_cloned(skb))
  1300. return pskb_expand_head(skb, 0, 0, pri);
  1301. return 0;
  1302. }
  1303. /**
  1304. * skb_header_cloned - is the header a clone
  1305. * @skb: buffer to check
  1306. *
  1307. * Returns true if modifying the header part of the buffer requires
  1308. * the data to be copied.
  1309. */
  1310. static inline int skb_header_cloned(const struct sk_buff *skb)
  1311. {
  1312. int dataref;
  1313. if (!skb->cloned)
  1314. return 0;
  1315. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  1316. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  1317. return dataref != 1;
  1318. }
  1319. static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
  1320. {
  1321. might_sleep_if(gfpflags_allow_blocking(pri));
  1322. if (skb_header_cloned(skb))
  1323. return pskb_expand_head(skb, 0, 0, pri);
  1324. return 0;
  1325. }
  1326. /**
  1327. * __skb_header_release - release reference to header
  1328. * @skb: buffer to operate on
  1329. */
  1330. static inline void __skb_header_release(struct sk_buff *skb)
  1331. {
  1332. skb->nohdr = 1;
  1333. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1334. }
  1335. /**
  1336. * skb_shared - is the buffer shared
  1337. * @skb: buffer to check
  1338. *
  1339. * Returns true if more than one person has a reference to this
  1340. * buffer.
  1341. */
  1342. static inline int skb_shared(const struct sk_buff *skb)
  1343. {
  1344. return refcount_read(&skb->users) != 1;
  1345. }
  1346. /**
  1347. * skb_share_check - check if buffer is shared and if so clone it
  1348. * @skb: buffer to check
  1349. * @pri: priority for memory allocation
  1350. *
  1351. * If the buffer is shared the buffer is cloned and the old copy
  1352. * drops a reference. A new clone with a single reference is returned.
  1353. * If the buffer is not shared the original buffer is returned. When
  1354. * being called from interrupt status or with spinlocks held pri must
  1355. * be GFP_ATOMIC.
  1356. *
  1357. * NULL is returned on a memory allocation failure.
  1358. */
  1359. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1360. {
  1361. might_sleep_if(gfpflags_allow_blocking(pri));
  1362. if (skb_shared(skb)) {
  1363. struct sk_buff *nskb = skb_clone(skb, pri);
  1364. if (likely(nskb))
  1365. consume_skb(skb);
  1366. else
  1367. kfree_skb(skb);
  1368. skb = nskb;
  1369. }
  1370. return skb;
  1371. }
  1372. /*
  1373. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1374. * packets to handle cases where we have a local reader and forward
  1375. * and a couple of other messy ones. The normal one is tcpdumping
  1376. * a packet thats being forwarded.
  1377. */
  1378. /**
  1379. * skb_unshare - make a copy of a shared buffer
  1380. * @skb: buffer to check
  1381. * @pri: priority for memory allocation
  1382. *
  1383. * If the socket buffer is a clone then this function creates a new
  1384. * copy of the data, drops a reference count on the old copy and returns
  1385. * the new copy with the reference count at 1. If the buffer is not a clone
  1386. * the original buffer is returned. When called with a spinlock held or
  1387. * from interrupt state @pri must be %GFP_ATOMIC
  1388. *
  1389. * %NULL is returned on a memory allocation failure.
  1390. */
  1391. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1392. gfp_t pri)
  1393. {
  1394. might_sleep_if(gfpflags_allow_blocking(pri));
  1395. if (skb_cloned(skb)) {
  1396. struct sk_buff *nskb = skb_copy(skb, pri);
  1397. /* Free our shared copy */
  1398. if (likely(nskb))
  1399. consume_skb(skb);
  1400. else
  1401. kfree_skb(skb);
  1402. skb = nskb;
  1403. }
  1404. return skb;
  1405. }
  1406. /**
  1407. * skb_peek - peek at the head of an &sk_buff_head
  1408. * @list_: list to peek at
  1409. *
  1410. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1411. * be careful with this one. A peek leaves the buffer on the
  1412. * list and someone else may run off with it. You must hold
  1413. * the appropriate locks or have a private queue to do this.
  1414. *
  1415. * Returns %NULL for an empty list or a pointer to the head element.
  1416. * The reference count is not incremented and the reference is therefore
  1417. * volatile. Use with caution.
  1418. */
  1419. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1420. {
  1421. struct sk_buff *skb = list_->next;
  1422. if (skb == (struct sk_buff *)list_)
  1423. skb = NULL;
  1424. return skb;
  1425. }
  1426. /**
  1427. * skb_peek_next - peek skb following the given one from a queue
  1428. * @skb: skb to start from
  1429. * @list_: list to peek at
  1430. *
  1431. * Returns %NULL when the end of the list is met or a pointer to the
  1432. * next element. The reference count is not incremented and the
  1433. * reference is therefore volatile. Use with caution.
  1434. */
  1435. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1436. const struct sk_buff_head *list_)
  1437. {
  1438. struct sk_buff *next = skb->next;
  1439. if (next == (struct sk_buff *)list_)
  1440. next = NULL;
  1441. return next;
  1442. }
  1443. /**
  1444. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1445. * @list_: list to peek at
  1446. *
  1447. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1448. * be careful with this one. A peek leaves the buffer on the
  1449. * list and someone else may run off with it. You must hold
  1450. * the appropriate locks or have a private queue to do this.
  1451. *
  1452. * Returns %NULL for an empty list or a pointer to the tail element.
  1453. * The reference count is not incremented and the reference is therefore
  1454. * volatile. Use with caution.
  1455. */
  1456. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1457. {
  1458. struct sk_buff *skb = list_->prev;
  1459. if (skb == (struct sk_buff *)list_)
  1460. skb = NULL;
  1461. return skb;
  1462. }
  1463. /**
  1464. * skb_queue_len - get queue length
  1465. * @list_: list to measure
  1466. *
  1467. * Return the length of an &sk_buff queue.
  1468. */
  1469. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1470. {
  1471. return list_->qlen;
  1472. }
  1473. /**
  1474. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1475. * @list: queue to initialize
  1476. *
  1477. * This initializes only the list and queue length aspects of
  1478. * an sk_buff_head object. This allows to initialize the list
  1479. * aspects of an sk_buff_head without reinitializing things like
  1480. * the spinlock. It can also be used for on-stack sk_buff_head
  1481. * objects where the spinlock is known to not be used.
  1482. */
  1483. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1484. {
  1485. list->prev = list->next = (struct sk_buff *)list;
  1486. list->qlen = 0;
  1487. }
  1488. /*
  1489. * This function creates a split out lock class for each invocation;
  1490. * this is needed for now since a whole lot of users of the skb-queue
  1491. * infrastructure in drivers have different locking usage (in hardirq)
  1492. * than the networking core (in softirq only). In the long run either the
  1493. * network layer or drivers should need annotation to consolidate the
  1494. * main types of usage into 3 classes.
  1495. */
  1496. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1497. {
  1498. spin_lock_init(&list->lock);
  1499. __skb_queue_head_init(list);
  1500. }
  1501. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1502. struct lock_class_key *class)
  1503. {
  1504. skb_queue_head_init(list);
  1505. lockdep_set_class(&list->lock, class);
  1506. }
  1507. /*
  1508. * Insert an sk_buff on a list.
  1509. *
  1510. * The "__skb_xxxx()" functions are the non-atomic ones that
  1511. * can only be called with interrupts disabled.
  1512. */
  1513. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1514. struct sk_buff_head *list);
  1515. static inline void __skb_insert(struct sk_buff *newsk,
  1516. struct sk_buff *prev, struct sk_buff *next,
  1517. struct sk_buff_head *list)
  1518. {
  1519. newsk->next = next;
  1520. newsk->prev = prev;
  1521. next->prev = prev->next = newsk;
  1522. list->qlen++;
  1523. }
  1524. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1525. struct sk_buff *prev,
  1526. struct sk_buff *next)
  1527. {
  1528. struct sk_buff *first = list->next;
  1529. struct sk_buff *last = list->prev;
  1530. first->prev = prev;
  1531. prev->next = first;
  1532. last->next = next;
  1533. next->prev = last;
  1534. }
  1535. /**
  1536. * skb_queue_splice - join two skb lists, this is designed for stacks
  1537. * @list: the new list to add
  1538. * @head: the place to add it in the first list
  1539. */
  1540. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1541. struct sk_buff_head *head)
  1542. {
  1543. if (!skb_queue_empty(list)) {
  1544. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1545. head->qlen += list->qlen;
  1546. }
  1547. }
  1548. /**
  1549. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1550. * @list: the new list to add
  1551. * @head: the place to add it in the first list
  1552. *
  1553. * The list at @list is reinitialised
  1554. */
  1555. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1556. struct sk_buff_head *head)
  1557. {
  1558. if (!skb_queue_empty(list)) {
  1559. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1560. head->qlen += list->qlen;
  1561. __skb_queue_head_init(list);
  1562. }
  1563. }
  1564. /**
  1565. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1566. * @list: the new list to add
  1567. * @head: the place to add it in the first list
  1568. */
  1569. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1570. struct sk_buff_head *head)
  1571. {
  1572. if (!skb_queue_empty(list)) {
  1573. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1574. head->qlen += list->qlen;
  1575. }
  1576. }
  1577. /**
  1578. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1579. * @list: the new list to add
  1580. * @head: the place to add it in the first list
  1581. *
  1582. * Each of the lists is a queue.
  1583. * The list at @list is reinitialised
  1584. */
  1585. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1586. struct sk_buff_head *head)
  1587. {
  1588. if (!skb_queue_empty(list)) {
  1589. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1590. head->qlen += list->qlen;
  1591. __skb_queue_head_init(list);
  1592. }
  1593. }
  1594. /**
  1595. * __skb_queue_after - queue a buffer at the list head
  1596. * @list: list to use
  1597. * @prev: place after this buffer
  1598. * @newsk: buffer to queue
  1599. *
  1600. * Queue a buffer int the middle of a list. This function takes no locks
  1601. * and you must therefore hold required locks before calling it.
  1602. *
  1603. * A buffer cannot be placed on two lists at the same time.
  1604. */
  1605. static inline void __skb_queue_after(struct sk_buff_head *list,
  1606. struct sk_buff *prev,
  1607. struct sk_buff *newsk)
  1608. {
  1609. __skb_insert(newsk, prev, prev->next, list);
  1610. }
  1611. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1612. struct sk_buff_head *list);
  1613. static inline void __skb_queue_before(struct sk_buff_head *list,
  1614. struct sk_buff *next,
  1615. struct sk_buff *newsk)
  1616. {
  1617. __skb_insert(newsk, next->prev, next, list);
  1618. }
  1619. /**
  1620. * __skb_queue_head - queue a buffer at the list head
  1621. * @list: list to use
  1622. * @newsk: buffer to queue
  1623. *
  1624. * Queue a buffer at the start of a list. This function takes no locks
  1625. * and you must therefore hold required locks before calling it.
  1626. *
  1627. * A buffer cannot be placed on two lists at the same time.
  1628. */
  1629. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1630. static inline void __skb_queue_head(struct sk_buff_head *list,
  1631. struct sk_buff *newsk)
  1632. {
  1633. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1634. }
  1635. /**
  1636. * __skb_queue_tail - queue a buffer at the list tail
  1637. * @list: list to use
  1638. * @newsk: buffer to queue
  1639. *
  1640. * Queue a buffer at the end of a list. This function takes no locks
  1641. * and you must therefore hold required locks before calling it.
  1642. *
  1643. * A buffer cannot be placed on two lists at the same time.
  1644. */
  1645. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1646. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1647. struct sk_buff *newsk)
  1648. {
  1649. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1650. }
  1651. /*
  1652. * remove sk_buff from list. _Must_ be called atomically, and with
  1653. * the list known..
  1654. */
  1655. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1656. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1657. {
  1658. struct sk_buff *next, *prev;
  1659. list->qlen--;
  1660. next = skb->next;
  1661. prev = skb->prev;
  1662. skb->next = skb->prev = NULL;
  1663. next->prev = prev;
  1664. prev->next = next;
  1665. }
  1666. /**
  1667. * __skb_dequeue - remove from the head of the queue
  1668. * @list: list to dequeue from
  1669. *
  1670. * Remove the head of the list. This function does not take any locks
  1671. * so must be used with appropriate locks held only. The head item is
  1672. * returned or %NULL if the list is empty.
  1673. */
  1674. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1675. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1676. {
  1677. struct sk_buff *skb = skb_peek(list);
  1678. if (skb)
  1679. __skb_unlink(skb, list);
  1680. return skb;
  1681. }
  1682. /**
  1683. * __skb_dequeue_tail - remove from the tail of the queue
  1684. * @list: list to dequeue from
  1685. *
  1686. * Remove the tail of the list. This function does not take any locks
  1687. * so must be used with appropriate locks held only. The tail item is
  1688. * returned or %NULL if the list is empty.
  1689. */
  1690. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1691. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1692. {
  1693. struct sk_buff *skb = skb_peek_tail(list);
  1694. if (skb)
  1695. __skb_unlink(skb, list);
  1696. return skb;
  1697. }
  1698. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1699. {
  1700. return skb->data_len;
  1701. }
  1702. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1703. {
  1704. return skb->len - skb->data_len;
  1705. }
  1706. static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
  1707. {
  1708. unsigned int i, len = 0;
  1709. for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
  1710. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1711. return len;
  1712. }
  1713. static inline unsigned int skb_pagelen(const struct sk_buff *skb)
  1714. {
  1715. return skb_headlen(skb) + __skb_pagelen(skb);
  1716. }
  1717. /**
  1718. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1719. * @skb: buffer containing fragment to be initialised
  1720. * @i: paged fragment index to initialise
  1721. * @page: the page to use for this fragment
  1722. * @off: the offset to the data with @page
  1723. * @size: the length of the data
  1724. *
  1725. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1726. * offset @off within @page.
  1727. *
  1728. * Does not take any additional reference on the fragment.
  1729. */
  1730. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1731. struct page *page, int off, int size)
  1732. {
  1733. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1734. /*
  1735. * Propagate page pfmemalloc to the skb if we can. The problem is
  1736. * that not all callers have unique ownership of the page but rely
  1737. * on page_is_pfmemalloc doing the right thing(tm).
  1738. */
  1739. frag->page.p = page;
  1740. frag->page_offset = off;
  1741. skb_frag_size_set(frag, size);
  1742. page = compound_head(page);
  1743. if (page_is_pfmemalloc(page))
  1744. skb->pfmemalloc = true;
  1745. }
  1746. /**
  1747. * skb_fill_page_desc - initialise a paged fragment in an skb
  1748. * @skb: buffer containing fragment to be initialised
  1749. * @i: paged fragment index to initialise
  1750. * @page: the page to use for this fragment
  1751. * @off: the offset to the data with @page
  1752. * @size: the length of the data
  1753. *
  1754. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1755. * @skb to point to @size bytes at offset @off within @page. In
  1756. * addition updates @skb such that @i is the last fragment.
  1757. *
  1758. * Does not take any additional reference on the fragment.
  1759. */
  1760. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1761. struct page *page, int off, int size)
  1762. {
  1763. __skb_fill_page_desc(skb, i, page, off, size);
  1764. skb_shinfo(skb)->nr_frags = i + 1;
  1765. }
  1766. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1767. int size, unsigned int truesize);
  1768. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1769. unsigned int truesize);
  1770. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1771. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1772. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1773. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1774. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1775. {
  1776. return skb->head + skb->tail;
  1777. }
  1778. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1779. {
  1780. skb->tail = skb->data - skb->head;
  1781. }
  1782. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1783. {
  1784. skb_reset_tail_pointer(skb);
  1785. skb->tail += offset;
  1786. }
  1787. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1788. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1789. {
  1790. return skb->tail;
  1791. }
  1792. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1793. {
  1794. skb->tail = skb->data;
  1795. }
  1796. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1797. {
  1798. skb->tail = skb->data + offset;
  1799. }
  1800. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1801. /*
  1802. * Add data to an sk_buff
  1803. */
  1804. void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1805. void *skb_put(struct sk_buff *skb, unsigned int len);
  1806. static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
  1807. {
  1808. void *tmp = skb_tail_pointer(skb);
  1809. SKB_LINEAR_ASSERT(skb);
  1810. skb->tail += len;
  1811. skb->len += len;
  1812. return tmp;
  1813. }
  1814. static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
  1815. {
  1816. void *tmp = __skb_put(skb, len);
  1817. memset(tmp, 0, len);
  1818. return tmp;
  1819. }
  1820. static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
  1821. unsigned int len)
  1822. {
  1823. void *tmp = __skb_put(skb, len);
  1824. memcpy(tmp, data, len);
  1825. return tmp;
  1826. }
  1827. static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
  1828. {
  1829. *(u8 *)__skb_put(skb, 1) = val;
  1830. }
  1831. static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
  1832. {
  1833. void *tmp = skb_put(skb, len);
  1834. memset(tmp, 0, len);
  1835. return tmp;
  1836. }
  1837. static inline void *skb_put_data(struct sk_buff *skb, const void *data,
  1838. unsigned int len)
  1839. {
  1840. void *tmp = skb_put(skb, len);
  1841. memcpy(tmp, data, len);
  1842. return tmp;
  1843. }
  1844. static inline void skb_put_u8(struct sk_buff *skb, u8 val)
  1845. {
  1846. *(u8 *)skb_put(skb, 1) = val;
  1847. }
  1848. void *skb_push(struct sk_buff *skb, unsigned int len);
  1849. static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
  1850. {
  1851. skb->data -= len;
  1852. skb->len += len;
  1853. return skb->data;
  1854. }
  1855. void *skb_pull(struct sk_buff *skb, unsigned int len);
  1856. static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
  1857. {
  1858. skb->len -= len;
  1859. BUG_ON(skb->len < skb->data_len);
  1860. return skb->data += len;
  1861. }
  1862. static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1863. {
  1864. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1865. }
  1866. void *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1867. static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1868. {
  1869. if (len > skb_headlen(skb) &&
  1870. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1871. return NULL;
  1872. skb->len -= len;
  1873. return skb->data += len;
  1874. }
  1875. static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
  1876. {
  1877. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1878. }
  1879. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1880. {
  1881. if (likely(len <= skb_headlen(skb)))
  1882. return 1;
  1883. if (unlikely(len > skb->len))
  1884. return 0;
  1885. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1886. }
  1887. void skb_condense(struct sk_buff *skb);
  1888. /**
  1889. * skb_headroom - bytes at buffer head
  1890. * @skb: buffer to check
  1891. *
  1892. * Return the number of bytes of free space at the head of an &sk_buff.
  1893. */
  1894. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1895. {
  1896. return skb->data - skb->head;
  1897. }
  1898. /**
  1899. * skb_tailroom - bytes at buffer end
  1900. * @skb: buffer to check
  1901. *
  1902. * Return the number of bytes of free space at the tail of an sk_buff
  1903. */
  1904. static inline int skb_tailroom(const struct sk_buff *skb)
  1905. {
  1906. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1907. }
  1908. /**
  1909. * skb_availroom - bytes at buffer end
  1910. * @skb: buffer to check
  1911. *
  1912. * Return the number of bytes of free space at the tail of an sk_buff
  1913. * allocated by sk_stream_alloc()
  1914. */
  1915. static inline int skb_availroom(const struct sk_buff *skb)
  1916. {
  1917. if (skb_is_nonlinear(skb))
  1918. return 0;
  1919. return skb->end - skb->tail - skb->reserved_tailroom;
  1920. }
  1921. /**
  1922. * skb_reserve - adjust headroom
  1923. * @skb: buffer to alter
  1924. * @len: bytes to move
  1925. *
  1926. * Increase the headroom of an empty &sk_buff by reducing the tail
  1927. * room. This is only allowed for an empty buffer.
  1928. */
  1929. static inline void skb_reserve(struct sk_buff *skb, int len)
  1930. {
  1931. skb->data += len;
  1932. skb->tail += len;
  1933. }
  1934. /**
  1935. * skb_tailroom_reserve - adjust reserved_tailroom
  1936. * @skb: buffer to alter
  1937. * @mtu: maximum amount of headlen permitted
  1938. * @needed_tailroom: minimum amount of reserved_tailroom
  1939. *
  1940. * Set reserved_tailroom so that headlen can be as large as possible but
  1941. * not larger than mtu and tailroom cannot be smaller than
  1942. * needed_tailroom.
  1943. * The required headroom should already have been reserved before using
  1944. * this function.
  1945. */
  1946. static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
  1947. unsigned int needed_tailroom)
  1948. {
  1949. SKB_LINEAR_ASSERT(skb);
  1950. if (mtu < skb_tailroom(skb) - needed_tailroom)
  1951. /* use at most mtu */
  1952. skb->reserved_tailroom = skb_tailroom(skb) - mtu;
  1953. else
  1954. /* use up to all available space */
  1955. skb->reserved_tailroom = needed_tailroom;
  1956. }
  1957. #define ENCAP_TYPE_ETHER 0
  1958. #define ENCAP_TYPE_IPPROTO 1
  1959. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1960. __be16 protocol)
  1961. {
  1962. skb->inner_protocol = protocol;
  1963. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1964. }
  1965. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1966. __u8 ipproto)
  1967. {
  1968. skb->inner_ipproto = ipproto;
  1969. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1970. }
  1971. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1972. {
  1973. skb->inner_mac_header = skb->mac_header;
  1974. skb->inner_network_header = skb->network_header;
  1975. skb->inner_transport_header = skb->transport_header;
  1976. }
  1977. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1978. {
  1979. skb->mac_len = skb->network_header - skb->mac_header;
  1980. }
  1981. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1982. *skb)
  1983. {
  1984. return skb->head + skb->inner_transport_header;
  1985. }
  1986. static inline int skb_inner_transport_offset(const struct sk_buff *skb)
  1987. {
  1988. return skb_inner_transport_header(skb) - skb->data;
  1989. }
  1990. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1991. {
  1992. skb->inner_transport_header = skb->data - skb->head;
  1993. }
  1994. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1995. const int offset)
  1996. {
  1997. skb_reset_inner_transport_header(skb);
  1998. skb->inner_transport_header += offset;
  1999. }
  2000. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  2001. {
  2002. return skb->head + skb->inner_network_header;
  2003. }
  2004. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  2005. {
  2006. skb->inner_network_header = skb->data - skb->head;
  2007. }
  2008. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  2009. const int offset)
  2010. {
  2011. skb_reset_inner_network_header(skb);
  2012. skb->inner_network_header += offset;
  2013. }
  2014. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  2015. {
  2016. return skb->head + skb->inner_mac_header;
  2017. }
  2018. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  2019. {
  2020. skb->inner_mac_header = skb->data - skb->head;
  2021. }
  2022. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  2023. const int offset)
  2024. {
  2025. skb_reset_inner_mac_header(skb);
  2026. skb->inner_mac_header += offset;
  2027. }
  2028. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  2029. {
  2030. return skb->transport_header != (typeof(skb->transport_header))~0U;
  2031. }
  2032. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  2033. {
  2034. return skb->head + skb->transport_header;
  2035. }
  2036. static inline void skb_reset_transport_header(struct sk_buff *skb)
  2037. {
  2038. skb->transport_header = skb->data - skb->head;
  2039. }
  2040. static inline void skb_set_transport_header(struct sk_buff *skb,
  2041. const int offset)
  2042. {
  2043. skb_reset_transport_header(skb);
  2044. skb->transport_header += offset;
  2045. }
  2046. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  2047. {
  2048. return skb->head + skb->network_header;
  2049. }
  2050. static inline void skb_reset_network_header(struct sk_buff *skb)
  2051. {
  2052. skb->network_header = skb->data - skb->head;
  2053. }
  2054. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  2055. {
  2056. skb_reset_network_header(skb);
  2057. skb->network_header += offset;
  2058. }
  2059. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  2060. {
  2061. return skb->head + skb->mac_header;
  2062. }
  2063. static inline int skb_mac_offset(const struct sk_buff *skb)
  2064. {
  2065. return skb_mac_header(skb) - skb->data;
  2066. }
  2067. static inline u32 skb_mac_header_len(const struct sk_buff *skb)
  2068. {
  2069. return skb->network_header - skb->mac_header;
  2070. }
  2071. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  2072. {
  2073. return skb->mac_header != (typeof(skb->mac_header))~0U;
  2074. }
  2075. static inline void skb_reset_mac_header(struct sk_buff *skb)
  2076. {
  2077. skb->mac_header = skb->data - skb->head;
  2078. }
  2079. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  2080. {
  2081. skb_reset_mac_header(skb);
  2082. skb->mac_header += offset;
  2083. }
  2084. static inline void skb_pop_mac_header(struct sk_buff *skb)
  2085. {
  2086. skb->mac_header = skb->network_header;
  2087. }
  2088. static inline void skb_probe_transport_header(struct sk_buff *skb,
  2089. const int offset_hint)
  2090. {
  2091. struct flow_keys_basic keys;
  2092. if (skb_transport_header_was_set(skb))
  2093. return;
  2094. if (skb_flow_dissect_flow_keys_basic(skb, &keys, NULL, 0, 0, 0, 0))
  2095. skb_set_transport_header(skb, keys.control.thoff);
  2096. else
  2097. skb_set_transport_header(skb, offset_hint);
  2098. }
  2099. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  2100. {
  2101. if (skb_mac_header_was_set(skb)) {
  2102. const unsigned char *old_mac = skb_mac_header(skb);
  2103. skb_set_mac_header(skb, -skb->mac_len);
  2104. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  2105. }
  2106. }
  2107. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  2108. {
  2109. return skb->csum_start - skb_headroom(skb);
  2110. }
  2111. static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
  2112. {
  2113. return skb->head + skb->csum_start;
  2114. }
  2115. static inline int skb_transport_offset(const struct sk_buff *skb)
  2116. {
  2117. return skb_transport_header(skb) - skb->data;
  2118. }
  2119. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  2120. {
  2121. return skb->transport_header - skb->network_header;
  2122. }
  2123. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  2124. {
  2125. return skb->inner_transport_header - skb->inner_network_header;
  2126. }
  2127. static inline int skb_network_offset(const struct sk_buff *skb)
  2128. {
  2129. return skb_network_header(skb) - skb->data;
  2130. }
  2131. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  2132. {
  2133. return skb_inner_network_header(skb) - skb->data;
  2134. }
  2135. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  2136. {
  2137. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  2138. }
  2139. /*
  2140. * CPUs often take a performance hit when accessing unaligned memory
  2141. * locations. The actual performance hit varies, it can be small if the
  2142. * hardware handles it or large if we have to take an exception and fix it
  2143. * in software.
  2144. *
  2145. * Since an ethernet header is 14 bytes network drivers often end up with
  2146. * the IP header at an unaligned offset. The IP header can be aligned by
  2147. * shifting the start of the packet by 2 bytes. Drivers should do this
  2148. * with:
  2149. *
  2150. * skb_reserve(skb, NET_IP_ALIGN);
  2151. *
  2152. * The downside to this alignment of the IP header is that the DMA is now
  2153. * unaligned. On some architectures the cost of an unaligned DMA is high
  2154. * and this cost outweighs the gains made by aligning the IP header.
  2155. *
  2156. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  2157. * to be overridden.
  2158. */
  2159. #ifndef NET_IP_ALIGN
  2160. #define NET_IP_ALIGN 2
  2161. #endif
  2162. /*
  2163. * The networking layer reserves some headroom in skb data (via
  2164. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  2165. * the header has to grow. In the default case, if the header has to grow
  2166. * 32 bytes or less we avoid the reallocation.
  2167. *
  2168. * Unfortunately this headroom changes the DMA alignment of the resulting
  2169. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  2170. * on some architectures. An architecture can override this value,
  2171. * perhaps setting it to a cacheline in size (since that will maintain
  2172. * cacheline alignment of the DMA). It must be a power of 2.
  2173. *
  2174. * Various parts of the networking layer expect at least 32 bytes of
  2175. * headroom, you should not reduce this.
  2176. *
  2177. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  2178. * to reduce average number of cache lines per packet.
  2179. * get_rps_cpus() for example only access one 64 bytes aligned block :
  2180. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  2181. */
  2182. #ifndef NET_SKB_PAD
  2183. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  2184. #endif
  2185. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  2186. static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
  2187. {
  2188. if (unlikely(skb_is_nonlinear(skb))) {
  2189. WARN_ON(1);
  2190. return;
  2191. }
  2192. skb->len = len;
  2193. skb_set_tail_pointer(skb, len);
  2194. }
  2195. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  2196. {
  2197. __skb_set_length(skb, len);
  2198. }
  2199. void skb_trim(struct sk_buff *skb, unsigned int len);
  2200. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  2201. {
  2202. if (skb->data_len)
  2203. return ___pskb_trim(skb, len);
  2204. __skb_trim(skb, len);
  2205. return 0;
  2206. }
  2207. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  2208. {
  2209. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  2210. }
  2211. /**
  2212. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  2213. * @skb: buffer to alter
  2214. * @len: new length
  2215. *
  2216. * This is identical to pskb_trim except that the caller knows that
  2217. * the skb is not cloned so we should never get an error due to out-
  2218. * of-memory.
  2219. */
  2220. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  2221. {
  2222. int err = pskb_trim(skb, len);
  2223. BUG_ON(err);
  2224. }
  2225. static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
  2226. {
  2227. unsigned int diff = len - skb->len;
  2228. if (skb_tailroom(skb) < diff) {
  2229. int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
  2230. GFP_ATOMIC);
  2231. if (ret)
  2232. return ret;
  2233. }
  2234. __skb_set_length(skb, len);
  2235. return 0;
  2236. }
  2237. /**
  2238. * skb_orphan - orphan a buffer
  2239. * @skb: buffer to orphan
  2240. *
  2241. * If a buffer currently has an owner then we call the owner's
  2242. * destructor function and make the @skb unowned. The buffer continues
  2243. * to exist but is no longer charged to its former owner.
  2244. */
  2245. static inline void skb_orphan(struct sk_buff *skb)
  2246. {
  2247. if (skb->destructor) {
  2248. skb->destructor(skb);
  2249. skb->destructor = NULL;
  2250. skb->sk = NULL;
  2251. } else {
  2252. BUG_ON(skb->sk);
  2253. }
  2254. }
  2255. /**
  2256. * skb_orphan_frags - orphan the frags contained in a buffer
  2257. * @skb: buffer to orphan frags from
  2258. * @gfp_mask: allocation mask for replacement pages
  2259. *
  2260. * For each frag in the SKB which needs a destructor (i.e. has an
  2261. * owner) create a copy of that frag and release the original
  2262. * page by calling the destructor.
  2263. */
  2264. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  2265. {
  2266. if (likely(!skb_zcopy(skb)))
  2267. return 0;
  2268. if (skb_uarg(skb)->callback == sock_zerocopy_callback)
  2269. return 0;
  2270. return skb_copy_ubufs(skb, gfp_mask);
  2271. }
  2272. /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
  2273. static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
  2274. {
  2275. if (likely(!skb_zcopy(skb)))
  2276. return 0;
  2277. return skb_copy_ubufs(skb, gfp_mask);
  2278. }
  2279. /**
  2280. * __skb_queue_purge - empty a list
  2281. * @list: list to empty
  2282. *
  2283. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  2284. * the list and one reference dropped. This function does not take the
  2285. * list lock and the caller must hold the relevant locks to use it.
  2286. */
  2287. void skb_queue_purge(struct sk_buff_head *list);
  2288. static inline void __skb_queue_purge(struct sk_buff_head *list)
  2289. {
  2290. struct sk_buff *skb;
  2291. while ((skb = __skb_dequeue(list)) != NULL)
  2292. kfree_skb(skb);
  2293. }
  2294. unsigned int skb_rbtree_purge(struct rb_root *root);
  2295. void *netdev_alloc_frag(unsigned int fragsz);
  2296. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  2297. gfp_t gfp_mask);
  2298. /**
  2299. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  2300. * @dev: network device to receive on
  2301. * @length: length to allocate
  2302. *
  2303. * Allocate a new &sk_buff and assign it a usage count of one. The
  2304. * buffer has unspecified headroom built in. Users should allocate
  2305. * the headroom they think they need without accounting for the
  2306. * built in space. The built in space is used for optimisations.
  2307. *
  2308. * %NULL is returned if there is no free memory. Although this function
  2309. * allocates memory it can be called from an interrupt.
  2310. */
  2311. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  2312. unsigned int length)
  2313. {
  2314. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  2315. }
  2316. /* legacy helper around __netdev_alloc_skb() */
  2317. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  2318. gfp_t gfp_mask)
  2319. {
  2320. return __netdev_alloc_skb(NULL, length, gfp_mask);
  2321. }
  2322. /* legacy helper around netdev_alloc_skb() */
  2323. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  2324. {
  2325. return netdev_alloc_skb(NULL, length);
  2326. }
  2327. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  2328. unsigned int length, gfp_t gfp)
  2329. {
  2330. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  2331. if (NET_IP_ALIGN && skb)
  2332. skb_reserve(skb, NET_IP_ALIGN);
  2333. return skb;
  2334. }
  2335. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  2336. unsigned int length)
  2337. {
  2338. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  2339. }
  2340. static inline void skb_free_frag(void *addr)
  2341. {
  2342. page_frag_free(addr);
  2343. }
  2344. void *napi_alloc_frag(unsigned int fragsz);
  2345. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  2346. unsigned int length, gfp_t gfp_mask);
  2347. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  2348. unsigned int length)
  2349. {
  2350. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  2351. }
  2352. void napi_consume_skb(struct sk_buff *skb, int budget);
  2353. void __kfree_skb_flush(void);
  2354. void __kfree_skb_defer(struct sk_buff *skb);
  2355. /**
  2356. * __dev_alloc_pages - allocate page for network Rx
  2357. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2358. * @order: size of the allocation
  2359. *
  2360. * Allocate a new page.
  2361. *
  2362. * %NULL is returned if there is no free memory.
  2363. */
  2364. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  2365. unsigned int order)
  2366. {
  2367. /* This piece of code contains several assumptions.
  2368. * 1. This is for device Rx, therefor a cold page is preferred.
  2369. * 2. The expectation is the user wants a compound page.
  2370. * 3. If requesting a order 0 page it will not be compound
  2371. * due to the check to see if order has a value in prep_new_page
  2372. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  2373. * code in gfp_to_alloc_flags that should be enforcing this.
  2374. */
  2375. gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
  2376. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  2377. }
  2378. static inline struct page *dev_alloc_pages(unsigned int order)
  2379. {
  2380. return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
  2381. }
  2382. /**
  2383. * __dev_alloc_page - allocate a page for network Rx
  2384. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2385. *
  2386. * Allocate a new page.
  2387. *
  2388. * %NULL is returned if there is no free memory.
  2389. */
  2390. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  2391. {
  2392. return __dev_alloc_pages(gfp_mask, 0);
  2393. }
  2394. static inline struct page *dev_alloc_page(void)
  2395. {
  2396. return dev_alloc_pages(0);
  2397. }
  2398. /**
  2399. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  2400. * @page: The page that was allocated from skb_alloc_page
  2401. * @skb: The skb that may need pfmemalloc set
  2402. */
  2403. static inline void skb_propagate_pfmemalloc(struct page *page,
  2404. struct sk_buff *skb)
  2405. {
  2406. if (page_is_pfmemalloc(page))
  2407. skb->pfmemalloc = true;
  2408. }
  2409. /**
  2410. * skb_frag_page - retrieve the page referred to by a paged fragment
  2411. * @frag: the paged fragment
  2412. *
  2413. * Returns the &struct page associated with @frag.
  2414. */
  2415. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2416. {
  2417. return frag->page.p;
  2418. }
  2419. /**
  2420. * __skb_frag_ref - take an addition reference on a paged fragment.
  2421. * @frag: the paged fragment
  2422. *
  2423. * Takes an additional reference on the paged fragment @frag.
  2424. */
  2425. static inline void __skb_frag_ref(skb_frag_t *frag)
  2426. {
  2427. get_page(skb_frag_page(frag));
  2428. }
  2429. /**
  2430. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2431. * @skb: the buffer
  2432. * @f: the fragment offset.
  2433. *
  2434. * Takes an additional reference on the @f'th paged fragment of @skb.
  2435. */
  2436. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2437. {
  2438. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2439. }
  2440. /**
  2441. * __skb_frag_unref - release a reference on a paged fragment.
  2442. * @frag: the paged fragment
  2443. *
  2444. * Releases a reference on the paged fragment @frag.
  2445. */
  2446. static inline void __skb_frag_unref(skb_frag_t *frag)
  2447. {
  2448. put_page(skb_frag_page(frag));
  2449. }
  2450. /**
  2451. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2452. * @skb: the buffer
  2453. * @f: the fragment offset
  2454. *
  2455. * Releases a reference on the @f'th paged fragment of @skb.
  2456. */
  2457. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2458. {
  2459. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2460. }
  2461. /**
  2462. * skb_frag_address - gets the address of the data contained in a paged fragment
  2463. * @frag: the paged fragment buffer
  2464. *
  2465. * Returns the address of the data within @frag. The page must already
  2466. * be mapped.
  2467. */
  2468. static inline void *skb_frag_address(const skb_frag_t *frag)
  2469. {
  2470. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2471. }
  2472. /**
  2473. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2474. * @frag: the paged fragment buffer
  2475. *
  2476. * Returns the address of the data within @frag. Checks that the page
  2477. * is mapped and returns %NULL otherwise.
  2478. */
  2479. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2480. {
  2481. void *ptr = page_address(skb_frag_page(frag));
  2482. if (unlikely(!ptr))
  2483. return NULL;
  2484. return ptr + frag->page_offset;
  2485. }
  2486. /**
  2487. * __skb_frag_set_page - sets the page contained in a paged fragment
  2488. * @frag: the paged fragment
  2489. * @page: the page to set
  2490. *
  2491. * Sets the fragment @frag to contain @page.
  2492. */
  2493. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2494. {
  2495. frag->page.p = page;
  2496. }
  2497. /**
  2498. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2499. * @skb: the buffer
  2500. * @f: the fragment offset
  2501. * @page: the page to set
  2502. *
  2503. * Sets the @f'th fragment of @skb to contain @page.
  2504. */
  2505. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2506. struct page *page)
  2507. {
  2508. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2509. }
  2510. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2511. /**
  2512. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2513. * @dev: the device to map the fragment to
  2514. * @frag: the paged fragment to map
  2515. * @offset: the offset within the fragment (starting at the
  2516. * fragment's own offset)
  2517. * @size: the number of bytes to map
  2518. * @dir: the direction of the mapping (``PCI_DMA_*``)
  2519. *
  2520. * Maps the page associated with @frag to @device.
  2521. */
  2522. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2523. const skb_frag_t *frag,
  2524. size_t offset, size_t size,
  2525. enum dma_data_direction dir)
  2526. {
  2527. return dma_map_page(dev, skb_frag_page(frag),
  2528. frag->page_offset + offset, size, dir);
  2529. }
  2530. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2531. gfp_t gfp_mask)
  2532. {
  2533. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2534. }
  2535. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2536. gfp_t gfp_mask)
  2537. {
  2538. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2539. }
  2540. /**
  2541. * skb_clone_writable - is the header of a clone writable
  2542. * @skb: buffer to check
  2543. * @len: length up to which to write
  2544. *
  2545. * Returns true if modifying the header part of the cloned buffer
  2546. * does not requires the data to be copied.
  2547. */
  2548. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2549. {
  2550. return !skb_header_cloned(skb) &&
  2551. skb_headroom(skb) + len <= skb->hdr_len;
  2552. }
  2553. static inline int skb_try_make_writable(struct sk_buff *skb,
  2554. unsigned int write_len)
  2555. {
  2556. return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
  2557. pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
  2558. }
  2559. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2560. int cloned)
  2561. {
  2562. int delta = 0;
  2563. if (headroom > skb_headroom(skb))
  2564. delta = headroom - skb_headroom(skb);
  2565. if (delta || cloned)
  2566. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2567. GFP_ATOMIC);
  2568. return 0;
  2569. }
  2570. /**
  2571. * skb_cow - copy header of skb when it is required
  2572. * @skb: buffer to cow
  2573. * @headroom: needed headroom
  2574. *
  2575. * If the skb passed lacks sufficient headroom or its data part
  2576. * is shared, data is reallocated. If reallocation fails, an error
  2577. * is returned and original skb is not changed.
  2578. *
  2579. * The result is skb with writable area skb->head...skb->tail
  2580. * and at least @headroom of space at head.
  2581. */
  2582. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2583. {
  2584. return __skb_cow(skb, headroom, skb_cloned(skb));
  2585. }
  2586. /**
  2587. * skb_cow_head - skb_cow but only making the head writable
  2588. * @skb: buffer to cow
  2589. * @headroom: needed headroom
  2590. *
  2591. * This function is identical to skb_cow except that we replace the
  2592. * skb_cloned check by skb_header_cloned. It should be used when
  2593. * you only need to push on some header and do not need to modify
  2594. * the data.
  2595. */
  2596. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2597. {
  2598. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2599. }
  2600. /**
  2601. * skb_padto - pad an skbuff up to a minimal size
  2602. * @skb: buffer to pad
  2603. * @len: minimal length
  2604. *
  2605. * Pads up a buffer to ensure the trailing bytes exist and are
  2606. * blanked. If the buffer already contains sufficient data it
  2607. * is untouched. Otherwise it is extended. Returns zero on
  2608. * success. The skb is freed on error.
  2609. */
  2610. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2611. {
  2612. unsigned int size = skb->len;
  2613. if (likely(size >= len))
  2614. return 0;
  2615. return skb_pad(skb, len - size);
  2616. }
  2617. /**
  2618. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2619. * @skb: buffer to pad
  2620. * @len: minimal length
  2621. * @free_on_error: free buffer on error
  2622. *
  2623. * Pads up a buffer to ensure the trailing bytes exist and are
  2624. * blanked. If the buffer already contains sufficient data it
  2625. * is untouched. Otherwise it is extended. Returns zero on
  2626. * success. The skb is freed on error if @free_on_error is true.
  2627. */
  2628. static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
  2629. bool free_on_error)
  2630. {
  2631. unsigned int size = skb->len;
  2632. if (unlikely(size < len)) {
  2633. len -= size;
  2634. if (__skb_pad(skb, len, free_on_error))
  2635. return -ENOMEM;
  2636. __skb_put(skb, len);
  2637. }
  2638. return 0;
  2639. }
  2640. /**
  2641. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2642. * @skb: buffer to pad
  2643. * @len: minimal length
  2644. *
  2645. * Pads up a buffer to ensure the trailing bytes exist and are
  2646. * blanked. If the buffer already contains sufficient data it
  2647. * is untouched. Otherwise it is extended. Returns zero on
  2648. * success. The skb is freed on error.
  2649. */
  2650. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2651. {
  2652. return __skb_put_padto(skb, len, true);
  2653. }
  2654. static inline int skb_add_data(struct sk_buff *skb,
  2655. struct iov_iter *from, int copy)
  2656. {
  2657. const int off = skb->len;
  2658. if (skb->ip_summed == CHECKSUM_NONE) {
  2659. __wsum csum = 0;
  2660. if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
  2661. &csum, from)) {
  2662. skb->csum = csum_block_add(skb->csum, csum, off);
  2663. return 0;
  2664. }
  2665. } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
  2666. return 0;
  2667. __skb_trim(skb, off);
  2668. return -EFAULT;
  2669. }
  2670. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2671. const struct page *page, int off)
  2672. {
  2673. if (skb_zcopy(skb))
  2674. return false;
  2675. if (i) {
  2676. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2677. return page == skb_frag_page(frag) &&
  2678. off == frag->page_offset + skb_frag_size(frag);
  2679. }
  2680. return false;
  2681. }
  2682. static inline int __skb_linearize(struct sk_buff *skb)
  2683. {
  2684. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2685. }
  2686. /**
  2687. * skb_linearize - convert paged skb to linear one
  2688. * @skb: buffer to linarize
  2689. *
  2690. * If there is no free memory -ENOMEM is returned, otherwise zero
  2691. * is returned and the old skb data released.
  2692. */
  2693. static inline int skb_linearize(struct sk_buff *skb)
  2694. {
  2695. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2696. }
  2697. /**
  2698. * skb_has_shared_frag - can any frag be overwritten
  2699. * @skb: buffer to test
  2700. *
  2701. * Return true if the skb has at least one frag that might be modified
  2702. * by an external entity (as in vmsplice()/sendfile())
  2703. */
  2704. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2705. {
  2706. return skb_is_nonlinear(skb) &&
  2707. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2708. }
  2709. /**
  2710. * skb_linearize_cow - make sure skb is linear and writable
  2711. * @skb: buffer to process
  2712. *
  2713. * If there is no free memory -ENOMEM is returned, otherwise zero
  2714. * is returned and the old skb data released.
  2715. */
  2716. static inline int skb_linearize_cow(struct sk_buff *skb)
  2717. {
  2718. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2719. __skb_linearize(skb) : 0;
  2720. }
  2721. static __always_inline void
  2722. __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2723. unsigned int off)
  2724. {
  2725. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2726. skb->csum = csum_block_sub(skb->csum,
  2727. csum_partial(start, len, 0), off);
  2728. else if (skb->ip_summed == CHECKSUM_PARTIAL &&
  2729. skb_checksum_start_offset(skb) < 0)
  2730. skb->ip_summed = CHECKSUM_NONE;
  2731. }
  2732. /**
  2733. * skb_postpull_rcsum - update checksum for received skb after pull
  2734. * @skb: buffer to update
  2735. * @start: start of data before pull
  2736. * @len: length of data pulled
  2737. *
  2738. * After doing a pull on a received packet, you need to call this to
  2739. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2740. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2741. */
  2742. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2743. const void *start, unsigned int len)
  2744. {
  2745. __skb_postpull_rcsum(skb, start, len, 0);
  2746. }
  2747. static __always_inline void
  2748. __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2749. unsigned int off)
  2750. {
  2751. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2752. skb->csum = csum_block_add(skb->csum,
  2753. csum_partial(start, len, 0), off);
  2754. }
  2755. /**
  2756. * skb_postpush_rcsum - update checksum for received skb after push
  2757. * @skb: buffer to update
  2758. * @start: start of data after push
  2759. * @len: length of data pushed
  2760. *
  2761. * After doing a push on a received packet, you need to call this to
  2762. * update the CHECKSUM_COMPLETE checksum.
  2763. */
  2764. static inline void skb_postpush_rcsum(struct sk_buff *skb,
  2765. const void *start, unsigned int len)
  2766. {
  2767. __skb_postpush_rcsum(skb, start, len, 0);
  2768. }
  2769. void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2770. /**
  2771. * skb_push_rcsum - push skb and update receive checksum
  2772. * @skb: buffer to update
  2773. * @len: length of data pulled
  2774. *
  2775. * This function performs an skb_push on the packet and updates
  2776. * the CHECKSUM_COMPLETE checksum. It should be used on
  2777. * receive path processing instead of skb_push unless you know
  2778. * that the checksum difference is zero (e.g., a valid IP header)
  2779. * or you are setting ip_summed to CHECKSUM_NONE.
  2780. */
  2781. static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
  2782. {
  2783. skb_push(skb, len);
  2784. skb_postpush_rcsum(skb, skb->data, len);
  2785. return skb->data;
  2786. }
  2787. int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
  2788. /**
  2789. * pskb_trim_rcsum - trim received skb and update checksum
  2790. * @skb: buffer to trim
  2791. * @len: new length
  2792. *
  2793. * This is exactly the same as pskb_trim except that it ensures the
  2794. * checksum of received packets are still valid after the operation.
  2795. */
  2796. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2797. {
  2798. if (likely(len >= skb->len))
  2799. return 0;
  2800. return pskb_trim_rcsum_slow(skb, len);
  2801. }
  2802. static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2803. {
  2804. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2805. skb->ip_summed = CHECKSUM_NONE;
  2806. __skb_trim(skb, len);
  2807. return 0;
  2808. }
  2809. static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
  2810. {
  2811. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2812. skb->ip_summed = CHECKSUM_NONE;
  2813. return __skb_grow(skb, len);
  2814. }
  2815. #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
  2816. #define skb_rb_first(root) rb_to_skb(rb_first(root))
  2817. #define skb_rb_last(root) rb_to_skb(rb_last(root))
  2818. #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
  2819. #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
  2820. #define skb_queue_walk(queue, skb) \
  2821. for (skb = (queue)->next; \
  2822. skb != (struct sk_buff *)(queue); \
  2823. skb = skb->next)
  2824. #define skb_queue_walk_safe(queue, skb, tmp) \
  2825. for (skb = (queue)->next, tmp = skb->next; \
  2826. skb != (struct sk_buff *)(queue); \
  2827. skb = tmp, tmp = skb->next)
  2828. #define skb_queue_walk_from(queue, skb) \
  2829. for (; skb != (struct sk_buff *)(queue); \
  2830. skb = skb->next)
  2831. #define skb_rbtree_walk(skb, root) \
  2832. for (skb = skb_rb_first(root); skb != NULL; \
  2833. skb = skb_rb_next(skb))
  2834. #define skb_rbtree_walk_from(skb) \
  2835. for (; skb != NULL; \
  2836. skb = skb_rb_next(skb))
  2837. #define skb_rbtree_walk_from_safe(skb, tmp) \
  2838. for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
  2839. skb = tmp)
  2840. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2841. for (tmp = skb->next; \
  2842. skb != (struct sk_buff *)(queue); \
  2843. skb = tmp, tmp = skb->next)
  2844. #define skb_queue_reverse_walk(queue, skb) \
  2845. for (skb = (queue)->prev; \
  2846. skb != (struct sk_buff *)(queue); \
  2847. skb = skb->prev)
  2848. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2849. for (skb = (queue)->prev, tmp = skb->prev; \
  2850. skb != (struct sk_buff *)(queue); \
  2851. skb = tmp, tmp = skb->prev)
  2852. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2853. for (tmp = skb->prev; \
  2854. skb != (struct sk_buff *)(queue); \
  2855. skb = tmp, tmp = skb->prev)
  2856. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2857. {
  2858. return skb_shinfo(skb)->frag_list != NULL;
  2859. }
  2860. static inline void skb_frag_list_init(struct sk_buff *skb)
  2861. {
  2862. skb_shinfo(skb)->frag_list = NULL;
  2863. }
  2864. #define skb_walk_frags(skb, iter) \
  2865. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2866. int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
  2867. const struct sk_buff *skb);
  2868. struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
  2869. struct sk_buff_head *queue,
  2870. unsigned int flags,
  2871. void (*destructor)(struct sock *sk,
  2872. struct sk_buff *skb),
  2873. int *peeked, int *off, int *err,
  2874. struct sk_buff **last);
  2875. struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
  2876. void (*destructor)(struct sock *sk,
  2877. struct sk_buff *skb),
  2878. int *peeked, int *off, int *err,
  2879. struct sk_buff **last);
  2880. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2881. void (*destructor)(struct sock *sk,
  2882. struct sk_buff *skb),
  2883. int *peeked, int *off, int *err);
  2884. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2885. int *err);
  2886. __poll_t datagram_poll(struct file *file, struct socket *sock,
  2887. struct poll_table_struct *wait);
  2888. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2889. struct iov_iter *to, int size);
  2890. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2891. struct msghdr *msg, int size)
  2892. {
  2893. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2894. }
  2895. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2896. struct msghdr *msg);
  2897. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2898. struct iov_iter *from, int len);
  2899. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2900. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2901. void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
  2902. static inline void skb_free_datagram_locked(struct sock *sk,
  2903. struct sk_buff *skb)
  2904. {
  2905. __skb_free_datagram_locked(sk, skb, 0);
  2906. }
  2907. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2908. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2909. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2910. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2911. int len, __wsum csum);
  2912. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2913. struct pipe_inode_info *pipe, unsigned int len,
  2914. unsigned int flags);
  2915. int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
  2916. int len);
  2917. int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
  2918. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2919. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2920. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2921. int len, int hlen);
  2922. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2923. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2924. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2925. bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
  2926. bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
  2927. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2928. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2929. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2930. int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
  2931. int skb_vlan_pop(struct sk_buff *skb);
  2932. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2933. struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
  2934. gfp_t gfp);
  2935. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2936. {
  2937. return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
  2938. }
  2939. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2940. {
  2941. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2942. }
  2943. struct skb_checksum_ops {
  2944. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2945. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2946. };
  2947. extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
  2948. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2949. __wsum csum, const struct skb_checksum_ops *ops);
  2950. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2951. __wsum csum);
  2952. static inline void * __must_check
  2953. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2954. int len, void *data, int hlen, void *buffer)
  2955. {
  2956. if (hlen - offset >= len)
  2957. return data + offset;
  2958. if (!skb ||
  2959. skb_copy_bits(skb, offset, buffer, len) < 0)
  2960. return NULL;
  2961. return buffer;
  2962. }
  2963. static inline void * __must_check
  2964. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2965. {
  2966. return __skb_header_pointer(skb, offset, len, skb->data,
  2967. skb_headlen(skb), buffer);
  2968. }
  2969. /**
  2970. * skb_needs_linearize - check if we need to linearize a given skb
  2971. * depending on the given device features.
  2972. * @skb: socket buffer to check
  2973. * @features: net device features
  2974. *
  2975. * Returns true if either:
  2976. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2977. * 2. skb is fragmented and the device does not support SG.
  2978. */
  2979. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2980. netdev_features_t features)
  2981. {
  2982. return skb_is_nonlinear(skb) &&
  2983. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2984. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2985. }
  2986. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2987. void *to,
  2988. const unsigned int len)
  2989. {
  2990. memcpy(to, skb->data, len);
  2991. }
  2992. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2993. const int offset, void *to,
  2994. const unsigned int len)
  2995. {
  2996. memcpy(to, skb->data + offset, len);
  2997. }
  2998. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2999. const void *from,
  3000. const unsigned int len)
  3001. {
  3002. memcpy(skb->data, from, len);
  3003. }
  3004. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  3005. const int offset,
  3006. const void *from,
  3007. const unsigned int len)
  3008. {
  3009. memcpy(skb->data + offset, from, len);
  3010. }
  3011. void skb_init(void);
  3012. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  3013. {
  3014. return skb->tstamp;
  3015. }
  3016. /**
  3017. * skb_get_timestamp - get timestamp from a skb
  3018. * @skb: skb to get stamp from
  3019. * @stamp: pointer to struct timeval to store stamp in
  3020. *
  3021. * Timestamps are stored in the skb as offsets to a base timestamp.
  3022. * This function converts the offset back to a struct timeval and stores
  3023. * it in stamp.
  3024. */
  3025. static inline void skb_get_timestamp(const struct sk_buff *skb,
  3026. struct timeval *stamp)
  3027. {
  3028. *stamp = ktime_to_timeval(skb->tstamp);
  3029. }
  3030. static inline void skb_get_timestampns(const struct sk_buff *skb,
  3031. struct timespec *stamp)
  3032. {
  3033. *stamp = ktime_to_timespec(skb->tstamp);
  3034. }
  3035. static inline void __net_timestamp(struct sk_buff *skb)
  3036. {
  3037. skb->tstamp = ktime_get_real();
  3038. }
  3039. static inline ktime_t net_timedelta(ktime_t t)
  3040. {
  3041. return ktime_sub(ktime_get_real(), t);
  3042. }
  3043. static inline ktime_t net_invalid_timestamp(void)
  3044. {
  3045. return 0;
  3046. }
  3047. static inline u8 skb_metadata_len(const struct sk_buff *skb)
  3048. {
  3049. return skb_shinfo(skb)->meta_len;
  3050. }
  3051. static inline void *skb_metadata_end(const struct sk_buff *skb)
  3052. {
  3053. return skb_mac_header(skb);
  3054. }
  3055. static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
  3056. const struct sk_buff *skb_b,
  3057. u8 meta_len)
  3058. {
  3059. const void *a = skb_metadata_end(skb_a);
  3060. const void *b = skb_metadata_end(skb_b);
  3061. /* Using more efficient varaiant than plain call to memcmp(). */
  3062. #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
  3063. u64 diffs = 0;
  3064. switch (meta_len) {
  3065. #define __it(x, op) (x -= sizeof(u##op))
  3066. #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
  3067. case 32: diffs |= __it_diff(a, b, 64);
  3068. case 24: diffs |= __it_diff(a, b, 64);
  3069. case 16: diffs |= __it_diff(a, b, 64);
  3070. case 8: diffs |= __it_diff(a, b, 64);
  3071. break;
  3072. case 28: diffs |= __it_diff(a, b, 64);
  3073. case 20: diffs |= __it_diff(a, b, 64);
  3074. case 12: diffs |= __it_diff(a, b, 64);
  3075. case 4: diffs |= __it_diff(a, b, 32);
  3076. break;
  3077. }
  3078. return diffs;
  3079. #else
  3080. return memcmp(a - meta_len, b - meta_len, meta_len);
  3081. #endif
  3082. }
  3083. static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
  3084. const struct sk_buff *skb_b)
  3085. {
  3086. u8 len_a = skb_metadata_len(skb_a);
  3087. u8 len_b = skb_metadata_len(skb_b);
  3088. if (!(len_a | len_b))
  3089. return false;
  3090. return len_a != len_b ?
  3091. true : __skb_metadata_differs(skb_a, skb_b, len_a);
  3092. }
  3093. static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
  3094. {
  3095. skb_shinfo(skb)->meta_len = meta_len;
  3096. }
  3097. static inline void skb_metadata_clear(struct sk_buff *skb)
  3098. {
  3099. skb_metadata_set(skb, 0);
  3100. }
  3101. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  3102. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  3103. void skb_clone_tx_timestamp(struct sk_buff *skb);
  3104. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  3105. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  3106. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  3107. {
  3108. }
  3109. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  3110. {
  3111. return false;
  3112. }
  3113. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  3114. /**
  3115. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  3116. *
  3117. * PHY drivers may accept clones of transmitted packets for
  3118. * timestamping via their phy_driver.txtstamp method. These drivers
  3119. * must call this function to return the skb back to the stack with a
  3120. * timestamp.
  3121. *
  3122. * @skb: clone of the the original outgoing packet
  3123. * @hwtstamps: hardware time stamps
  3124. *
  3125. */
  3126. void skb_complete_tx_timestamp(struct sk_buff *skb,
  3127. struct skb_shared_hwtstamps *hwtstamps);
  3128. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  3129. struct skb_shared_hwtstamps *hwtstamps,
  3130. struct sock *sk, int tstype);
  3131. /**
  3132. * skb_tstamp_tx - queue clone of skb with send time stamps
  3133. * @orig_skb: the original outgoing packet
  3134. * @hwtstamps: hardware time stamps, may be NULL if not available
  3135. *
  3136. * If the skb has a socket associated, then this function clones the
  3137. * skb (thus sharing the actual data and optional structures), stores
  3138. * the optional hardware time stamping information (if non NULL) or
  3139. * generates a software time stamp (otherwise), then queues the clone
  3140. * to the error queue of the socket. Errors are silently ignored.
  3141. */
  3142. void skb_tstamp_tx(struct sk_buff *orig_skb,
  3143. struct skb_shared_hwtstamps *hwtstamps);
  3144. /**
  3145. * skb_tx_timestamp() - Driver hook for transmit timestamping
  3146. *
  3147. * Ethernet MAC Drivers should call this function in their hard_xmit()
  3148. * function immediately before giving the sk_buff to the MAC hardware.
  3149. *
  3150. * Specifically, one should make absolutely sure that this function is
  3151. * called before TX completion of this packet can trigger. Otherwise
  3152. * the packet could potentially already be freed.
  3153. *
  3154. * @skb: A socket buffer.
  3155. */
  3156. static inline void skb_tx_timestamp(struct sk_buff *skb)
  3157. {
  3158. skb_clone_tx_timestamp(skb);
  3159. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
  3160. skb_tstamp_tx(skb, NULL);
  3161. }
  3162. /**
  3163. * skb_complete_wifi_ack - deliver skb with wifi status
  3164. *
  3165. * @skb: the original outgoing packet
  3166. * @acked: ack status
  3167. *
  3168. */
  3169. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  3170. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  3171. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  3172. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  3173. {
  3174. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  3175. skb->csum_valid ||
  3176. (skb->ip_summed == CHECKSUM_PARTIAL &&
  3177. skb_checksum_start_offset(skb) >= 0));
  3178. }
  3179. /**
  3180. * skb_checksum_complete - Calculate checksum of an entire packet
  3181. * @skb: packet to process
  3182. *
  3183. * This function calculates the checksum over the entire packet plus
  3184. * the value of skb->csum. The latter can be used to supply the
  3185. * checksum of a pseudo header as used by TCP/UDP. It returns the
  3186. * checksum.
  3187. *
  3188. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  3189. * this function can be used to verify that checksum on received
  3190. * packets. In that case the function should return zero if the
  3191. * checksum is correct. In particular, this function will return zero
  3192. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  3193. * hardware has already verified the correctness of the checksum.
  3194. */
  3195. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  3196. {
  3197. return skb_csum_unnecessary(skb) ?
  3198. 0 : __skb_checksum_complete(skb);
  3199. }
  3200. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  3201. {
  3202. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  3203. if (skb->csum_level == 0)
  3204. skb->ip_summed = CHECKSUM_NONE;
  3205. else
  3206. skb->csum_level--;
  3207. }
  3208. }
  3209. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  3210. {
  3211. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  3212. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  3213. skb->csum_level++;
  3214. } else if (skb->ip_summed == CHECKSUM_NONE) {
  3215. skb->ip_summed = CHECKSUM_UNNECESSARY;
  3216. skb->csum_level = 0;
  3217. }
  3218. }
  3219. /* Check if we need to perform checksum complete validation.
  3220. *
  3221. * Returns true if checksum complete is needed, false otherwise
  3222. * (either checksum is unnecessary or zero checksum is allowed).
  3223. */
  3224. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  3225. bool zero_okay,
  3226. __sum16 check)
  3227. {
  3228. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  3229. skb->csum_valid = 1;
  3230. __skb_decr_checksum_unnecessary(skb);
  3231. return false;
  3232. }
  3233. return true;
  3234. }
  3235. /* For small packets <= CHECKSUM_BREAK perform checksum complete directly
  3236. * in checksum_init.
  3237. */
  3238. #define CHECKSUM_BREAK 76
  3239. /* Unset checksum-complete
  3240. *
  3241. * Unset checksum complete can be done when packet is being modified
  3242. * (uncompressed for instance) and checksum-complete value is
  3243. * invalidated.
  3244. */
  3245. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  3246. {
  3247. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3248. skb->ip_summed = CHECKSUM_NONE;
  3249. }
  3250. /* Validate (init) checksum based on checksum complete.
  3251. *
  3252. * Return values:
  3253. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  3254. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  3255. * checksum is stored in skb->csum for use in __skb_checksum_complete
  3256. * non-zero: value of invalid checksum
  3257. *
  3258. */
  3259. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  3260. bool complete,
  3261. __wsum psum)
  3262. {
  3263. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  3264. if (!csum_fold(csum_add(psum, skb->csum))) {
  3265. skb->csum_valid = 1;
  3266. return 0;
  3267. }
  3268. }
  3269. skb->csum = psum;
  3270. if (complete || skb->len <= CHECKSUM_BREAK) {
  3271. __sum16 csum;
  3272. csum = __skb_checksum_complete(skb);
  3273. skb->csum_valid = !csum;
  3274. return csum;
  3275. }
  3276. return 0;
  3277. }
  3278. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  3279. {
  3280. return 0;
  3281. }
  3282. /* Perform checksum validate (init). Note that this is a macro since we only
  3283. * want to calculate the pseudo header which is an input function if necessary.
  3284. * First we try to validate without any computation (checksum unnecessary) and
  3285. * then calculate based on checksum complete calling the function to compute
  3286. * pseudo header.
  3287. *
  3288. * Return values:
  3289. * 0: checksum is validated or try to in skb_checksum_complete
  3290. * non-zero: value of invalid checksum
  3291. */
  3292. #define __skb_checksum_validate(skb, proto, complete, \
  3293. zero_okay, check, compute_pseudo) \
  3294. ({ \
  3295. __sum16 __ret = 0; \
  3296. skb->csum_valid = 0; \
  3297. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  3298. __ret = __skb_checksum_validate_complete(skb, \
  3299. complete, compute_pseudo(skb, proto)); \
  3300. __ret; \
  3301. })
  3302. #define skb_checksum_init(skb, proto, compute_pseudo) \
  3303. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  3304. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  3305. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  3306. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  3307. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  3308. #define skb_checksum_validate_zero_check(skb, proto, check, \
  3309. compute_pseudo) \
  3310. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  3311. #define skb_checksum_simple_validate(skb) \
  3312. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  3313. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  3314. {
  3315. return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
  3316. }
  3317. static inline void __skb_checksum_convert(struct sk_buff *skb,
  3318. __sum16 check, __wsum pseudo)
  3319. {
  3320. skb->csum = ~pseudo;
  3321. skb->ip_summed = CHECKSUM_COMPLETE;
  3322. }
  3323. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  3324. do { \
  3325. if (__skb_checksum_convert_check(skb)) \
  3326. __skb_checksum_convert(skb, check, \
  3327. compute_pseudo(skb, proto)); \
  3328. } while (0)
  3329. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  3330. u16 start, u16 offset)
  3331. {
  3332. skb->ip_summed = CHECKSUM_PARTIAL;
  3333. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  3334. skb->csum_offset = offset - start;
  3335. }
  3336. /* Update skbuf and packet to reflect the remote checksum offload operation.
  3337. * When called, ptr indicates the starting point for skb->csum when
  3338. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  3339. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  3340. */
  3341. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  3342. int start, int offset, bool nopartial)
  3343. {
  3344. __wsum delta;
  3345. if (!nopartial) {
  3346. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  3347. return;
  3348. }
  3349. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  3350. __skb_checksum_complete(skb);
  3351. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  3352. }
  3353. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  3354. /* Adjust skb->csum since we changed the packet */
  3355. skb->csum = csum_add(skb->csum, delta);
  3356. }
  3357. static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
  3358. {
  3359. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  3360. return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
  3361. #else
  3362. return NULL;
  3363. #endif
  3364. }
  3365. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3366. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  3367. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  3368. {
  3369. if (nfct && atomic_dec_and_test(&nfct->use))
  3370. nf_conntrack_destroy(nfct);
  3371. }
  3372. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  3373. {
  3374. if (nfct)
  3375. atomic_inc(&nfct->use);
  3376. }
  3377. #endif
  3378. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3379. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  3380. {
  3381. if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
  3382. kfree(nf_bridge);
  3383. }
  3384. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  3385. {
  3386. if (nf_bridge)
  3387. refcount_inc(&nf_bridge->use);
  3388. }
  3389. #endif /* CONFIG_BRIDGE_NETFILTER */
  3390. static inline void nf_reset(struct sk_buff *skb)
  3391. {
  3392. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3393. nf_conntrack_put(skb_nfct(skb));
  3394. skb->_nfct = 0;
  3395. #endif
  3396. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3397. nf_bridge_put(skb->nf_bridge);
  3398. skb->nf_bridge = NULL;
  3399. #endif
  3400. }
  3401. static inline void nf_reset_trace(struct sk_buff *skb)
  3402. {
  3403. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3404. skb->nf_trace = 0;
  3405. #endif
  3406. }
  3407. static inline void ipvs_reset(struct sk_buff *skb)
  3408. {
  3409. #if IS_ENABLED(CONFIG_IP_VS)
  3410. skb->ipvs_property = 0;
  3411. #endif
  3412. }
  3413. /* Note: This doesn't put any conntrack and bridge info in dst. */
  3414. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  3415. bool copy)
  3416. {
  3417. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3418. dst->_nfct = src->_nfct;
  3419. nf_conntrack_get(skb_nfct(src));
  3420. #endif
  3421. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3422. dst->nf_bridge = src->nf_bridge;
  3423. nf_bridge_get(src->nf_bridge);
  3424. #endif
  3425. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3426. if (copy)
  3427. dst->nf_trace = src->nf_trace;
  3428. #endif
  3429. }
  3430. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  3431. {
  3432. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3433. nf_conntrack_put(skb_nfct(dst));
  3434. #endif
  3435. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3436. nf_bridge_put(dst->nf_bridge);
  3437. #endif
  3438. __nf_copy(dst, src, true);
  3439. }
  3440. #ifdef CONFIG_NETWORK_SECMARK
  3441. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3442. {
  3443. to->secmark = from->secmark;
  3444. }
  3445. static inline void skb_init_secmark(struct sk_buff *skb)
  3446. {
  3447. skb->secmark = 0;
  3448. }
  3449. #else
  3450. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3451. { }
  3452. static inline void skb_init_secmark(struct sk_buff *skb)
  3453. { }
  3454. #endif
  3455. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  3456. {
  3457. return !skb->destructor &&
  3458. #if IS_ENABLED(CONFIG_XFRM)
  3459. !skb->sp &&
  3460. #endif
  3461. !skb_nfct(skb) &&
  3462. !skb->_skb_refdst &&
  3463. !skb_has_frag_list(skb);
  3464. }
  3465. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  3466. {
  3467. skb->queue_mapping = queue_mapping;
  3468. }
  3469. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  3470. {
  3471. return skb->queue_mapping;
  3472. }
  3473. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  3474. {
  3475. to->queue_mapping = from->queue_mapping;
  3476. }
  3477. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  3478. {
  3479. skb->queue_mapping = rx_queue + 1;
  3480. }
  3481. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  3482. {
  3483. return skb->queue_mapping - 1;
  3484. }
  3485. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  3486. {
  3487. return skb->queue_mapping != 0;
  3488. }
  3489. static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
  3490. {
  3491. skb->dst_pending_confirm = val;
  3492. }
  3493. static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
  3494. {
  3495. return skb->dst_pending_confirm != 0;
  3496. }
  3497. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  3498. {
  3499. #ifdef CONFIG_XFRM
  3500. return skb->sp;
  3501. #else
  3502. return NULL;
  3503. #endif
  3504. }
  3505. /* Keeps track of mac header offset relative to skb->head.
  3506. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  3507. * For non-tunnel skb it points to skb_mac_header() and for
  3508. * tunnel skb it points to outer mac header.
  3509. * Keeps track of level of encapsulation of network headers.
  3510. */
  3511. struct skb_gso_cb {
  3512. union {
  3513. int mac_offset;
  3514. int data_offset;
  3515. };
  3516. int encap_level;
  3517. __wsum csum;
  3518. __u16 csum_start;
  3519. };
  3520. #define SKB_SGO_CB_OFFSET 32
  3521. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
  3522. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  3523. {
  3524. return (skb_mac_header(inner_skb) - inner_skb->head) -
  3525. SKB_GSO_CB(inner_skb)->mac_offset;
  3526. }
  3527. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  3528. {
  3529. int new_headroom, headroom;
  3530. int ret;
  3531. headroom = skb_headroom(skb);
  3532. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  3533. if (ret)
  3534. return ret;
  3535. new_headroom = skb_headroom(skb);
  3536. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  3537. return 0;
  3538. }
  3539. static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
  3540. {
  3541. /* Do not update partial checksums if remote checksum is enabled. */
  3542. if (skb->remcsum_offload)
  3543. return;
  3544. SKB_GSO_CB(skb)->csum = res;
  3545. SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
  3546. }
  3547. /* Compute the checksum for a gso segment. First compute the checksum value
  3548. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  3549. * then add in skb->csum (checksum from csum_start to end of packet).
  3550. * skb->csum and csum_start are then updated to reflect the checksum of the
  3551. * resultant packet starting from the transport header-- the resultant checksum
  3552. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  3553. * header.
  3554. */
  3555. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  3556. {
  3557. unsigned char *csum_start = skb_transport_header(skb);
  3558. int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
  3559. __wsum partial = SKB_GSO_CB(skb)->csum;
  3560. SKB_GSO_CB(skb)->csum = res;
  3561. SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
  3562. return csum_fold(csum_partial(csum_start, plen, partial));
  3563. }
  3564. static inline bool skb_is_gso(const struct sk_buff *skb)
  3565. {
  3566. return skb_shinfo(skb)->gso_size;
  3567. }
  3568. /* Note: Should be called only if skb_is_gso(skb) is true */
  3569. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  3570. {
  3571. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  3572. }
  3573. /* Note: Should be called only if skb_is_gso(skb) is true */
  3574. static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
  3575. {
  3576. return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
  3577. }
  3578. static inline void skb_gso_reset(struct sk_buff *skb)
  3579. {
  3580. skb_shinfo(skb)->gso_size = 0;
  3581. skb_shinfo(skb)->gso_segs = 0;
  3582. skb_shinfo(skb)->gso_type = 0;
  3583. }
  3584. static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
  3585. u16 increment)
  3586. {
  3587. if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
  3588. return;
  3589. shinfo->gso_size += increment;
  3590. }
  3591. static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
  3592. u16 decrement)
  3593. {
  3594. if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
  3595. return;
  3596. shinfo->gso_size -= decrement;
  3597. }
  3598. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  3599. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  3600. {
  3601. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  3602. * wanted then gso_type will be set. */
  3603. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  3604. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  3605. unlikely(shinfo->gso_type == 0)) {
  3606. __skb_warn_lro_forwarding(skb);
  3607. return true;
  3608. }
  3609. return false;
  3610. }
  3611. static inline void skb_forward_csum(struct sk_buff *skb)
  3612. {
  3613. /* Unfortunately we don't support this one. Any brave souls? */
  3614. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3615. skb->ip_summed = CHECKSUM_NONE;
  3616. }
  3617. /**
  3618. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3619. * @skb: skb to check
  3620. *
  3621. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3622. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3623. * use this helper, to document places where we make this assertion.
  3624. */
  3625. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3626. {
  3627. #ifdef DEBUG
  3628. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3629. #endif
  3630. }
  3631. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3632. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3633. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3634. unsigned int transport_len,
  3635. __sum16(*skb_chkf)(struct sk_buff *skb));
  3636. /**
  3637. * skb_head_is_locked - Determine if the skb->head is locked down
  3638. * @skb: skb to check
  3639. *
  3640. * The head on skbs build around a head frag can be removed if they are
  3641. * not cloned. This function returns true if the skb head is locked down
  3642. * due to either being allocated via kmalloc, or by being a clone with
  3643. * multiple references to the head.
  3644. */
  3645. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3646. {
  3647. return !skb->head_frag || skb_cloned(skb);
  3648. }
  3649. /* Local Checksum Offload.
  3650. * Compute outer checksum based on the assumption that the
  3651. * inner checksum will be offloaded later.
  3652. * See Documentation/networking/checksum-offloads.txt for
  3653. * explanation of how this works.
  3654. * Fill in outer checksum adjustment (e.g. with sum of outer
  3655. * pseudo-header) before calling.
  3656. * Also ensure that inner checksum is in linear data area.
  3657. */
  3658. static inline __wsum lco_csum(struct sk_buff *skb)
  3659. {
  3660. unsigned char *csum_start = skb_checksum_start(skb);
  3661. unsigned char *l4_hdr = skb_transport_header(skb);
  3662. __wsum partial;
  3663. /* Start with complement of inner checksum adjustment */
  3664. partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
  3665. skb->csum_offset));
  3666. /* Add in checksum of our headers (incl. outer checksum
  3667. * adjustment filled in by caller) and return result.
  3668. */
  3669. return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
  3670. }
  3671. #endif /* __KERNEL__ */
  3672. #endif /* _LINUX_SKBUFF_H */