skbuff.h 111 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963296429652966296729682969297029712972297329742975297629772978297929802981298229832984298529862987298829892990299129922993299429952996299729982999300030013002300330043005300630073008300930103011301230133014301530163017301830193020302130223023302430253026302730283029303030313032303330343035303630373038303930403041304230433044304530463047304830493050305130523053305430553056305730583059306030613062306330643065306630673068306930703071307230733074307530763077307830793080308130823083308430853086308730883089309030913092309330943095309630973098309931003101310231033104310531063107310831093110311131123113311431153116311731183119312031213122312331243125312631273128312931303131313231333134313531363137313831393140314131423143314431453146314731483149315031513152315331543155315631573158315931603161316231633164316531663167316831693170317131723173317431753176317731783179318031813182318331843185318631873188318931903191319231933194319531963197319831993200320132023203320432053206320732083209321032113212321332143215321632173218321932203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278327932803281328232833284328532863287328832893290329132923293329432953296329732983299330033013302330333043305330633073308330933103311331233133314331533163317331833193320332133223323332433253326332733283329333033313332333333343335333633373338333933403341334233433344334533463347334833493350335133523353335433553356335733583359336033613362336333643365336633673368336933703371337233733374337533763377337833793380338133823383338433853386338733883389339033913392339333943395339633973398339934003401340234033404340534063407340834093410341134123413341434153416341734183419342034213422342334243425342634273428342934303431343234333434343534363437343834393440344134423443344434453446344734483449345034513452345334543455345634573458345934603461346234633464346534663467346834693470347134723473347434753476347734783479348034813482348334843485348634873488348934903491349234933494349534963497349834993500350135023503350435053506350735083509351035113512351335143515351635173518351935203521352235233524352535263527352835293530353135323533353435353536353735383539354035413542354335443545354635473548354935503551355235533554355535563557355835593560356135623563356435653566356735683569357035713572357335743575357635773578357935803581358235833584358535863587358835893590359135923593359435953596359735983599360036013602360336043605360636073608360936103611361236133614361536163617361836193620362136223623362436253626362736283629363036313632363336343635363636373638363936403641364236433644364536463647364836493650365136523653365436553656365736583659366036613662366336643665366636673668366936703671367236733674367536763677367836793680368136823683368436853686368736883689369036913692369336943695369636973698369937003701370237033704370537063707370837093710371137123713371437153716371737183719372037213722372337243725372637273728372937303731373237333734373537363737373837393740374137423743374437453746374737483749375037513752375337543755375637573758375937603761376237633764376537663767376837693770377137723773377437753776377737783779378037813782378337843785378637873788378937903791379237933794379537963797379837993800380138023803380438053806380738083809381038113812381338143815381638173818381938203821382238233824382538263827382838293830383138323833383438353836383738383839384038413842384338443845384638473848384938503851385238533854385538563857385838593860386138623863386438653866386738683869387038713872387338743875387638773878387938803881388238833884388538863887
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/kernel.h>
  16. #include <linux/kmemcheck.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/bug.h>
  20. #include <linux/cache.h>
  21. #include <linux/rbtree.h>
  22. #include <linux/socket.h>
  23. #include <linux/atomic.h>
  24. #include <asm/types.h>
  25. #include <linux/spinlock.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #include <linux/rcupdate.h>
  30. #include <linux/hrtimer.h>
  31. #include <linux/dma-mapping.h>
  32. #include <linux/netdev_features.h>
  33. #include <linux/sched.h>
  34. #include <net/flow_dissector.h>
  35. #include <linux/splice.h>
  36. #include <linux/in6.h>
  37. #include <linux/if_packet.h>
  38. #include <net/flow.h>
  39. /* The interface for checksum offload between the stack and networking drivers
  40. * is as follows...
  41. *
  42. * A. IP checksum related features
  43. *
  44. * Drivers advertise checksum offload capabilities in the features of a device.
  45. * From the stack's point of view these are capabilities offered by the driver,
  46. * a driver typically only advertises features that it is capable of offloading
  47. * to its device.
  48. *
  49. * The checksum related features are:
  50. *
  51. * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
  52. * IP (one's complement) checksum for any combination
  53. * of protocols or protocol layering. The checksum is
  54. * computed and set in a packet per the CHECKSUM_PARTIAL
  55. * interface (see below).
  56. *
  57. * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
  58. * TCP or UDP packets over IPv4. These are specifically
  59. * unencapsulated packets of the form IPv4|TCP or
  60. * IPv4|UDP where the Protocol field in the IPv4 header
  61. * is TCP or UDP. The IPv4 header may contain IP options
  62. * This feature cannot be set in features for a device
  63. * with NETIF_F_HW_CSUM also set. This feature is being
  64. * DEPRECATED (see below).
  65. *
  66. * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
  67. * TCP or UDP packets over IPv6. These are specifically
  68. * unencapsulated packets of the form IPv6|TCP or
  69. * IPv4|UDP where the Next Header field in the IPv6
  70. * header is either TCP or UDP. IPv6 extension headers
  71. * are not supported with this feature. This feature
  72. * cannot be set in features for a device with
  73. * NETIF_F_HW_CSUM also set. This feature is being
  74. * DEPRECATED (see below).
  75. *
  76. * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
  77. * This flag is used only used to disable the RX checksum
  78. * feature for a device. The stack will accept receive
  79. * checksum indication in packets received on a device
  80. * regardless of whether NETIF_F_RXCSUM is set.
  81. *
  82. * B. Checksumming of received packets by device. Indication of checksum
  83. * verification is in set skb->ip_summed. Possible values are:
  84. *
  85. * CHECKSUM_NONE:
  86. *
  87. * Device did not checksum this packet e.g. due to lack of capabilities.
  88. * The packet contains full (though not verified) checksum in packet but
  89. * not in skb->csum. Thus, skb->csum is undefined in this case.
  90. *
  91. * CHECKSUM_UNNECESSARY:
  92. *
  93. * The hardware you're dealing with doesn't calculate the full checksum
  94. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  95. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  96. * if their checksums are okay. skb->csum is still undefined in this case
  97. * though. A driver or device must never modify the checksum field in the
  98. * packet even if checksum is verified.
  99. *
  100. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  101. * TCP: IPv6 and IPv4.
  102. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  103. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  104. * may perform further validation in this case.
  105. * GRE: only if the checksum is present in the header.
  106. * SCTP: indicates the CRC in SCTP header has been validated.
  107. *
  108. * skb->csum_level indicates the number of consecutive checksums found in
  109. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  110. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  111. * and a device is able to verify the checksums for UDP (possibly zero),
  112. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  113. * two. If the device were only able to verify the UDP checksum and not
  114. * GRE, either because it doesn't support GRE checksum of because GRE
  115. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  116. * not considered in this case).
  117. *
  118. * CHECKSUM_COMPLETE:
  119. *
  120. * This is the most generic way. The device supplied checksum of the _whole_
  121. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  122. * hardware doesn't need to parse L3/L4 headers to implement this.
  123. *
  124. * Note: Even if device supports only some protocols, but is able to produce
  125. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  126. *
  127. * CHECKSUM_PARTIAL:
  128. *
  129. * A checksum is set up to be offloaded to a device as described in the
  130. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  131. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  132. * on the same host, or it may be set in the input path in GRO or remote
  133. * checksum offload. For the purposes of checksum verification, the checksum
  134. * referred to by skb->csum_start + skb->csum_offset and any preceding
  135. * checksums in the packet are considered verified. Any checksums in the
  136. * packet that are after the checksum being offloaded are not considered to
  137. * be verified.
  138. *
  139. * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
  140. * in the skb->ip_summed for a packet. Values are:
  141. *
  142. * CHECKSUM_PARTIAL:
  143. *
  144. * The driver is required to checksum the packet as seen by hard_start_xmit()
  145. * from skb->csum_start up to the end, and to record/write the checksum at
  146. * offset skb->csum_start + skb->csum_offset. A driver may verify that the
  147. * csum_start and csum_offset values are valid values given the length and
  148. * offset of the packet, however they should not attempt to validate that the
  149. * checksum refers to a legitimate transport layer checksum-- it is the
  150. * purview of the stack to validate that csum_start and csum_offset are set
  151. * correctly.
  152. *
  153. * When the stack requests checksum offload for a packet, the driver MUST
  154. * ensure that the checksum is set correctly. A driver can either offload the
  155. * checksum calculation to the device, or call skb_checksum_help (in the case
  156. * that the device does not support offload for a particular checksum).
  157. *
  158. * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
  159. * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
  160. * checksum offload capability. If a device has limited checksum capabilities
  161. * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
  162. * described above) a helper function can be called to resolve
  163. * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
  164. * function takes a spec argument that describes the protocol layer that is
  165. * supported for checksum offload and can be called for each packet. If a
  166. * packet does not match the specification for offload, skb_checksum_help
  167. * is called to resolve the checksum.
  168. *
  169. * CHECKSUM_NONE:
  170. *
  171. * The skb was already checksummed by the protocol, or a checksum is not
  172. * required.
  173. *
  174. * CHECKSUM_UNNECESSARY:
  175. *
  176. * This has the same meaning on as CHECKSUM_NONE for checksum offload on
  177. * output.
  178. *
  179. * CHECKSUM_COMPLETE:
  180. * Not used in checksum output. If a driver observes a packet with this value
  181. * set in skbuff, if should treat as CHECKSUM_NONE being set.
  182. *
  183. * D. Non-IP checksum (CRC) offloads
  184. *
  185. * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
  186. * offloading the SCTP CRC in a packet. To perform this offload the stack
  187. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  188. * accordingly. Note the there is no indication in the skbuff that the
  189. * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
  190. * both IP checksum offload and SCTP CRC offload must verify which offload
  191. * is configured for a packet presumably by inspecting packet headers.
  192. *
  193. * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
  194. * offloading the FCOE CRC in a packet. To perform this offload the stack
  195. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  196. * accordingly. Note the there is no indication in the skbuff that the
  197. * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
  198. * both IP checksum offload and FCOE CRC offload must verify which offload
  199. * is configured for a packet presumably by inspecting packet headers.
  200. *
  201. * E. Checksumming on output with GSO.
  202. *
  203. * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
  204. * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
  205. * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
  206. * part of the GSO operation is implied. If a checksum is being offloaded
  207. * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
  208. * are set to refer to the outermost checksum being offload (two offloaded
  209. * checksums are possible with UDP encapsulation).
  210. */
  211. /* Don't change this without changing skb_csum_unnecessary! */
  212. #define CHECKSUM_NONE 0
  213. #define CHECKSUM_UNNECESSARY 1
  214. #define CHECKSUM_COMPLETE 2
  215. #define CHECKSUM_PARTIAL 3
  216. /* Maximum value in skb->csum_level */
  217. #define SKB_MAX_CSUM_LEVEL 3
  218. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  219. #define SKB_WITH_OVERHEAD(X) \
  220. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  221. #define SKB_MAX_ORDER(X, ORDER) \
  222. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  223. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  224. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  225. /* return minimum truesize of one skb containing X bytes of data */
  226. #define SKB_TRUESIZE(X) ((X) + \
  227. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  228. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  229. struct net_device;
  230. struct scatterlist;
  231. struct pipe_inode_info;
  232. struct iov_iter;
  233. struct napi_struct;
  234. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  235. struct nf_conntrack {
  236. atomic_t use;
  237. };
  238. #endif
  239. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  240. struct nf_bridge_info {
  241. atomic_t use;
  242. enum {
  243. BRNF_PROTO_UNCHANGED,
  244. BRNF_PROTO_8021Q,
  245. BRNF_PROTO_PPPOE
  246. } orig_proto:8;
  247. u8 pkt_otherhost:1;
  248. u8 in_prerouting:1;
  249. u8 bridged_dnat:1;
  250. __u16 frag_max_size;
  251. struct net_device *physindev;
  252. /* always valid & non-NULL from FORWARD on, for physdev match */
  253. struct net_device *physoutdev;
  254. union {
  255. /* prerouting: detect dnat in orig/reply direction */
  256. __be32 ipv4_daddr;
  257. struct in6_addr ipv6_daddr;
  258. /* after prerouting + nat detected: store original source
  259. * mac since neigh resolution overwrites it, only used while
  260. * skb is out in neigh layer.
  261. */
  262. char neigh_header[8];
  263. };
  264. };
  265. #endif
  266. struct sk_buff_head {
  267. /* These two members must be first. */
  268. struct sk_buff *next;
  269. struct sk_buff *prev;
  270. __u32 qlen;
  271. spinlock_t lock;
  272. };
  273. struct sk_buff;
  274. /* To allow 64K frame to be packed as single skb without frag_list we
  275. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  276. * buffers which do not start on a page boundary.
  277. *
  278. * Since GRO uses frags we allocate at least 16 regardless of page
  279. * size.
  280. */
  281. #if (65536/PAGE_SIZE + 1) < 16
  282. #define MAX_SKB_FRAGS 16UL
  283. #else
  284. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  285. #endif
  286. extern int sysctl_max_skb_frags;
  287. /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
  288. * segment using its current segmentation instead.
  289. */
  290. #define GSO_BY_FRAGS 0xFFFF
  291. typedef struct skb_frag_struct skb_frag_t;
  292. struct skb_frag_struct {
  293. struct {
  294. struct page *p;
  295. } page;
  296. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  297. __u32 page_offset;
  298. __u32 size;
  299. #else
  300. __u16 page_offset;
  301. __u16 size;
  302. #endif
  303. };
  304. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  305. {
  306. return frag->size;
  307. }
  308. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  309. {
  310. frag->size = size;
  311. }
  312. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  313. {
  314. frag->size += delta;
  315. }
  316. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  317. {
  318. frag->size -= delta;
  319. }
  320. #define HAVE_HW_TIME_STAMP
  321. /**
  322. * struct skb_shared_hwtstamps - hardware time stamps
  323. * @hwtstamp: hardware time stamp transformed into duration
  324. * since arbitrary point in time
  325. *
  326. * Software time stamps generated by ktime_get_real() are stored in
  327. * skb->tstamp.
  328. *
  329. * hwtstamps can only be compared against other hwtstamps from
  330. * the same device.
  331. *
  332. * This structure is attached to packets as part of the
  333. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  334. */
  335. struct skb_shared_hwtstamps {
  336. ktime_t hwtstamp;
  337. };
  338. /* Definitions for tx_flags in struct skb_shared_info */
  339. enum {
  340. /* generate hardware time stamp */
  341. SKBTX_HW_TSTAMP = 1 << 0,
  342. /* generate software time stamp when queueing packet to NIC */
  343. SKBTX_SW_TSTAMP = 1 << 1,
  344. /* device driver is going to provide hardware time stamp */
  345. SKBTX_IN_PROGRESS = 1 << 2,
  346. /* device driver supports TX zero-copy buffers */
  347. SKBTX_DEV_ZEROCOPY = 1 << 3,
  348. /* generate wifi status information (where possible) */
  349. SKBTX_WIFI_STATUS = 1 << 4,
  350. /* This indicates at least one fragment might be overwritten
  351. * (as in vmsplice(), sendfile() ...)
  352. * If we need to compute a TX checksum, we'll need to copy
  353. * all frags to avoid possible bad checksum
  354. */
  355. SKBTX_SHARED_FRAG = 1 << 5,
  356. /* generate software time stamp when entering packet scheduling */
  357. SKBTX_SCHED_TSTAMP = 1 << 6,
  358. };
  359. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  360. SKBTX_SCHED_TSTAMP)
  361. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  362. /*
  363. * The callback notifies userspace to release buffers when skb DMA is done in
  364. * lower device, the skb last reference should be 0 when calling this.
  365. * The zerocopy_success argument is true if zero copy transmit occurred,
  366. * false on data copy or out of memory error caused by data copy attempt.
  367. * The ctx field is used to track device context.
  368. * The desc field is used to track userspace buffer index.
  369. */
  370. struct ubuf_info {
  371. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  372. void *ctx;
  373. unsigned long desc;
  374. };
  375. /* This data is invariant across clones and lives at
  376. * the end of the header data, ie. at skb->end.
  377. */
  378. struct skb_shared_info {
  379. unsigned char nr_frags;
  380. __u8 tx_flags;
  381. unsigned short gso_size;
  382. /* Warning: this field is not always filled in (UFO)! */
  383. unsigned short gso_segs;
  384. unsigned short gso_type;
  385. struct sk_buff *frag_list;
  386. struct skb_shared_hwtstamps hwtstamps;
  387. u32 tskey;
  388. __be32 ip6_frag_id;
  389. /*
  390. * Warning : all fields before dataref are cleared in __alloc_skb()
  391. */
  392. atomic_t dataref;
  393. /* Intermediate layers must ensure that destructor_arg
  394. * remains valid until skb destructor */
  395. void * destructor_arg;
  396. /* must be last field, see pskb_expand_head() */
  397. skb_frag_t frags[MAX_SKB_FRAGS];
  398. };
  399. /* We divide dataref into two halves. The higher 16 bits hold references
  400. * to the payload part of skb->data. The lower 16 bits hold references to
  401. * the entire skb->data. A clone of a headerless skb holds the length of
  402. * the header in skb->hdr_len.
  403. *
  404. * All users must obey the rule that the skb->data reference count must be
  405. * greater than or equal to the payload reference count.
  406. *
  407. * Holding a reference to the payload part means that the user does not
  408. * care about modifications to the header part of skb->data.
  409. */
  410. #define SKB_DATAREF_SHIFT 16
  411. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  412. enum {
  413. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  414. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  415. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  416. };
  417. enum {
  418. SKB_GSO_TCPV4 = 1 << 0,
  419. SKB_GSO_UDP = 1 << 1,
  420. /* This indicates the skb is from an untrusted source. */
  421. SKB_GSO_DODGY = 1 << 2,
  422. /* This indicates the tcp segment has CWR set. */
  423. SKB_GSO_TCP_ECN = 1 << 3,
  424. SKB_GSO_TCP_FIXEDID = 1 << 4,
  425. SKB_GSO_TCPV6 = 1 << 5,
  426. SKB_GSO_FCOE = 1 << 6,
  427. SKB_GSO_GRE = 1 << 7,
  428. SKB_GSO_GRE_CSUM = 1 << 8,
  429. SKB_GSO_IPXIP4 = 1 << 9,
  430. SKB_GSO_IPXIP6 = 1 << 10,
  431. SKB_GSO_UDP_TUNNEL = 1 << 11,
  432. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
  433. SKB_GSO_PARTIAL = 1 << 13,
  434. SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
  435. SKB_GSO_SCTP = 1 << 15,
  436. };
  437. #if BITS_PER_LONG > 32
  438. #define NET_SKBUFF_DATA_USES_OFFSET 1
  439. #endif
  440. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  441. typedef unsigned int sk_buff_data_t;
  442. #else
  443. typedef unsigned char *sk_buff_data_t;
  444. #endif
  445. /**
  446. * struct skb_mstamp - multi resolution time stamps
  447. * @stamp_us: timestamp in us resolution
  448. * @stamp_jiffies: timestamp in jiffies
  449. */
  450. struct skb_mstamp {
  451. union {
  452. u64 v64;
  453. struct {
  454. u32 stamp_us;
  455. u32 stamp_jiffies;
  456. };
  457. };
  458. };
  459. /**
  460. * skb_mstamp_get - get current timestamp
  461. * @cl: place to store timestamps
  462. */
  463. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  464. {
  465. u64 val = local_clock();
  466. do_div(val, NSEC_PER_USEC);
  467. cl->stamp_us = (u32)val;
  468. cl->stamp_jiffies = (u32)jiffies;
  469. }
  470. /**
  471. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  472. * @t1: pointer to newest sample
  473. * @t0: pointer to oldest sample
  474. */
  475. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  476. const struct skb_mstamp *t0)
  477. {
  478. s32 delta_us = t1->stamp_us - t0->stamp_us;
  479. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  480. /* If delta_us is negative, this might be because interval is too big,
  481. * or local_clock() drift is too big : fallback using jiffies.
  482. */
  483. if (delta_us <= 0 ||
  484. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  485. delta_us = jiffies_to_usecs(delta_jiffies);
  486. return delta_us;
  487. }
  488. static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
  489. const struct skb_mstamp *t0)
  490. {
  491. s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
  492. if (!diff)
  493. diff = t1->stamp_us - t0->stamp_us;
  494. return diff > 0;
  495. }
  496. /**
  497. * struct sk_buff - socket buffer
  498. * @next: Next buffer in list
  499. * @prev: Previous buffer in list
  500. * @tstamp: Time we arrived/left
  501. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  502. * @sk: Socket we are owned by
  503. * @dev: Device we arrived on/are leaving by
  504. * @cb: Control buffer. Free for use by every layer. Put private vars here
  505. * @_skb_refdst: destination entry (with norefcount bit)
  506. * @sp: the security path, used for xfrm
  507. * @len: Length of actual data
  508. * @data_len: Data length
  509. * @mac_len: Length of link layer header
  510. * @hdr_len: writable header length of cloned skb
  511. * @csum: Checksum (must include start/offset pair)
  512. * @csum_start: Offset from skb->head where checksumming should start
  513. * @csum_offset: Offset from csum_start where checksum should be stored
  514. * @priority: Packet queueing priority
  515. * @ignore_df: allow local fragmentation
  516. * @cloned: Head may be cloned (check refcnt to be sure)
  517. * @ip_summed: Driver fed us an IP checksum
  518. * @nohdr: Payload reference only, must not modify header
  519. * @nfctinfo: Relationship of this skb to the connection
  520. * @pkt_type: Packet class
  521. * @fclone: skbuff clone status
  522. * @ipvs_property: skbuff is owned by ipvs
  523. * @peeked: this packet has been seen already, so stats have been
  524. * done for it, don't do them again
  525. * @nf_trace: netfilter packet trace flag
  526. * @protocol: Packet protocol from driver
  527. * @destructor: Destruct function
  528. * @nfct: Associated connection, if any
  529. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  530. * @skb_iif: ifindex of device we arrived on
  531. * @tc_index: Traffic control index
  532. * @tc_verd: traffic control verdict
  533. * @hash: the packet hash
  534. * @queue_mapping: Queue mapping for multiqueue devices
  535. * @xmit_more: More SKBs are pending for this queue
  536. * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
  537. * @ndisc_nodetype: router type (from link layer)
  538. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  539. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  540. * ports.
  541. * @sw_hash: indicates hash was computed in software stack
  542. * @wifi_acked_valid: wifi_acked was set
  543. * @wifi_acked: whether frame was acked on wifi or not
  544. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  545. * @napi_id: id of the NAPI struct this skb came from
  546. * @secmark: security marking
  547. * @mark: Generic packet mark
  548. * @vlan_proto: vlan encapsulation protocol
  549. * @vlan_tci: vlan tag control information
  550. * @inner_protocol: Protocol (encapsulation)
  551. * @inner_transport_header: Inner transport layer header (encapsulation)
  552. * @inner_network_header: Network layer header (encapsulation)
  553. * @inner_mac_header: Link layer header (encapsulation)
  554. * @transport_header: Transport layer header
  555. * @network_header: Network layer header
  556. * @mac_header: Link layer header
  557. * @tail: Tail pointer
  558. * @end: End pointer
  559. * @head: Head of buffer
  560. * @data: Data head pointer
  561. * @truesize: Buffer size
  562. * @users: User count - see {datagram,tcp}.c
  563. */
  564. struct sk_buff {
  565. union {
  566. struct {
  567. /* These two members must be first. */
  568. struct sk_buff *next;
  569. struct sk_buff *prev;
  570. union {
  571. ktime_t tstamp;
  572. struct skb_mstamp skb_mstamp;
  573. };
  574. };
  575. struct rb_node rbnode; /* used in netem & tcp stack */
  576. };
  577. struct sock *sk;
  578. struct net_device *dev;
  579. /*
  580. * This is the control buffer. It is free to use for every
  581. * layer. Please put your private variables there. If you
  582. * want to keep them across layers you have to do a skb_clone()
  583. * first. This is owned by whoever has the skb queued ATM.
  584. */
  585. char cb[48] __aligned(8);
  586. unsigned long _skb_refdst;
  587. void (*destructor)(struct sk_buff *skb);
  588. #ifdef CONFIG_XFRM
  589. struct sec_path *sp;
  590. #endif
  591. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  592. struct nf_conntrack *nfct;
  593. #endif
  594. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  595. struct nf_bridge_info *nf_bridge;
  596. #endif
  597. unsigned int len,
  598. data_len;
  599. __u16 mac_len,
  600. hdr_len;
  601. /* Following fields are _not_ copied in __copy_skb_header()
  602. * Note that queue_mapping is here mostly to fill a hole.
  603. */
  604. kmemcheck_bitfield_begin(flags1);
  605. __u16 queue_mapping;
  606. /* if you move cloned around you also must adapt those constants */
  607. #ifdef __BIG_ENDIAN_BITFIELD
  608. #define CLONED_MASK (1 << 7)
  609. #else
  610. #define CLONED_MASK 1
  611. #endif
  612. #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
  613. __u8 __cloned_offset[0];
  614. __u8 cloned:1,
  615. nohdr:1,
  616. fclone:2,
  617. peeked:1,
  618. head_frag:1,
  619. xmit_more:1,
  620. pfmemalloc:1;
  621. kmemcheck_bitfield_end(flags1);
  622. /* fields enclosed in headers_start/headers_end are copied
  623. * using a single memcpy() in __copy_skb_header()
  624. */
  625. /* private: */
  626. __u32 headers_start[0];
  627. /* public: */
  628. /* if you move pkt_type around you also must adapt those constants */
  629. #ifdef __BIG_ENDIAN_BITFIELD
  630. #define PKT_TYPE_MAX (7 << 5)
  631. #else
  632. #define PKT_TYPE_MAX 7
  633. #endif
  634. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  635. __u8 __pkt_type_offset[0];
  636. __u8 pkt_type:3;
  637. __u8 ignore_df:1;
  638. __u8 nfctinfo:3;
  639. __u8 nf_trace:1;
  640. __u8 ip_summed:2;
  641. __u8 ooo_okay:1;
  642. __u8 l4_hash:1;
  643. __u8 sw_hash:1;
  644. __u8 wifi_acked_valid:1;
  645. __u8 wifi_acked:1;
  646. __u8 no_fcs:1;
  647. /* Indicates the inner headers are valid in the skbuff. */
  648. __u8 encapsulation:1;
  649. __u8 encap_hdr_csum:1;
  650. __u8 csum_valid:1;
  651. __u8 csum_complete_sw:1;
  652. __u8 csum_level:2;
  653. __u8 csum_bad:1;
  654. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  655. __u8 ndisc_nodetype:2;
  656. #endif
  657. __u8 ipvs_property:1;
  658. __u8 inner_protocol_type:1;
  659. __u8 remcsum_offload:1;
  660. #ifdef CONFIG_NET_SWITCHDEV
  661. __u8 offload_fwd_mark:1;
  662. #endif
  663. /* 2, 4 or 5 bit hole */
  664. #ifdef CONFIG_NET_SCHED
  665. __u16 tc_index; /* traffic control index */
  666. #ifdef CONFIG_NET_CLS_ACT
  667. __u16 tc_verd; /* traffic control verdict */
  668. #endif
  669. #endif
  670. union {
  671. __wsum csum;
  672. struct {
  673. __u16 csum_start;
  674. __u16 csum_offset;
  675. };
  676. };
  677. __u32 priority;
  678. int skb_iif;
  679. __u32 hash;
  680. __be16 vlan_proto;
  681. __u16 vlan_tci;
  682. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  683. union {
  684. unsigned int napi_id;
  685. unsigned int sender_cpu;
  686. };
  687. #endif
  688. #ifdef CONFIG_NETWORK_SECMARK
  689. __u32 secmark;
  690. #endif
  691. union {
  692. __u32 mark;
  693. __u32 reserved_tailroom;
  694. };
  695. union {
  696. __be16 inner_protocol;
  697. __u8 inner_ipproto;
  698. };
  699. __u16 inner_transport_header;
  700. __u16 inner_network_header;
  701. __u16 inner_mac_header;
  702. __be16 protocol;
  703. __u16 transport_header;
  704. __u16 network_header;
  705. __u16 mac_header;
  706. /* private: */
  707. __u32 headers_end[0];
  708. /* public: */
  709. /* These elements must be at the end, see alloc_skb() for details. */
  710. sk_buff_data_t tail;
  711. sk_buff_data_t end;
  712. unsigned char *head,
  713. *data;
  714. unsigned int truesize;
  715. atomic_t users;
  716. };
  717. #ifdef __KERNEL__
  718. /*
  719. * Handling routines are only of interest to the kernel
  720. */
  721. #include <linux/slab.h>
  722. #define SKB_ALLOC_FCLONE 0x01
  723. #define SKB_ALLOC_RX 0x02
  724. #define SKB_ALLOC_NAPI 0x04
  725. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  726. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  727. {
  728. return unlikely(skb->pfmemalloc);
  729. }
  730. /*
  731. * skb might have a dst pointer attached, refcounted or not.
  732. * _skb_refdst low order bit is set if refcount was _not_ taken
  733. */
  734. #define SKB_DST_NOREF 1UL
  735. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  736. /**
  737. * skb_dst - returns skb dst_entry
  738. * @skb: buffer
  739. *
  740. * Returns skb dst_entry, regardless of reference taken or not.
  741. */
  742. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  743. {
  744. /* If refdst was not refcounted, check we still are in a
  745. * rcu_read_lock section
  746. */
  747. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  748. !rcu_read_lock_held() &&
  749. !rcu_read_lock_bh_held());
  750. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  751. }
  752. /**
  753. * skb_dst_set - sets skb dst
  754. * @skb: buffer
  755. * @dst: dst entry
  756. *
  757. * Sets skb dst, assuming a reference was taken on dst and should
  758. * be released by skb_dst_drop()
  759. */
  760. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  761. {
  762. skb->_skb_refdst = (unsigned long)dst;
  763. }
  764. /**
  765. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  766. * @skb: buffer
  767. * @dst: dst entry
  768. *
  769. * Sets skb dst, assuming a reference was not taken on dst.
  770. * If dst entry is cached, we do not take reference and dst_release
  771. * will be avoided by refdst_drop. If dst entry is not cached, we take
  772. * reference, so that last dst_release can destroy the dst immediately.
  773. */
  774. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  775. {
  776. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  777. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  778. }
  779. /**
  780. * skb_dst_is_noref - Test if skb dst isn't refcounted
  781. * @skb: buffer
  782. */
  783. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  784. {
  785. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  786. }
  787. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  788. {
  789. return (struct rtable *)skb_dst(skb);
  790. }
  791. /* For mangling skb->pkt_type from user space side from applications
  792. * such as nft, tc, etc, we only allow a conservative subset of
  793. * possible pkt_types to be set.
  794. */
  795. static inline bool skb_pkt_type_ok(u32 ptype)
  796. {
  797. return ptype <= PACKET_OTHERHOST;
  798. }
  799. void kfree_skb(struct sk_buff *skb);
  800. void kfree_skb_list(struct sk_buff *segs);
  801. void skb_tx_error(struct sk_buff *skb);
  802. void consume_skb(struct sk_buff *skb);
  803. void __kfree_skb(struct sk_buff *skb);
  804. extern struct kmem_cache *skbuff_head_cache;
  805. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  806. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  807. bool *fragstolen, int *delta_truesize);
  808. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  809. int node);
  810. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  811. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  812. static inline struct sk_buff *alloc_skb(unsigned int size,
  813. gfp_t priority)
  814. {
  815. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  816. }
  817. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  818. unsigned long data_len,
  819. int max_page_order,
  820. int *errcode,
  821. gfp_t gfp_mask);
  822. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  823. struct sk_buff_fclones {
  824. struct sk_buff skb1;
  825. struct sk_buff skb2;
  826. atomic_t fclone_ref;
  827. };
  828. /**
  829. * skb_fclone_busy - check if fclone is busy
  830. * @sk: socket
  831. * @skb: buffer
  832. *
  833. * Returns true if skb is a fast clone, and its clone is not freed.
  834. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  835. * so we also check that this didnt happen.
  836. */
  837. static inline bool skb_fclone_busy(const struct sock *sk,
  838. const struct sk_buff *skb)
  839. {
  840. const struct sk_buff_fclones *fclones;
  841. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  842. return skb->fclone == SKB_FCLONE_ORIG &&
  843. atomic_read(&fclones->fclone_ref) > 1 &&
  844. fclones->skb2.sk == sk;
  845. }
  846. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  847. gfp_t priority)
  848. {
  849. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  850. }
  851. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  852. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  853. {
  854. return __alloc_skb_head(priority, -1);
  855. }
  856. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  857. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  858. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  859. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  860. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  861. gfp_t gfp_mask, bool fclone);
  862. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  863. gfp_t gfp_mask)
  864. {
  865. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  866. }
  867. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  868. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  869. unsigned int headroom);
  870. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  871. int newtailroom, gfp_t priority);
  872. int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  873. int offset, int len);
  874. int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
  875. int offset, int len);
  876. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  877. int skb_pad(struct sk_buff *skb, int pad);
  878. #define dev_kfree_skb(a) consume_skb(a)
  879. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  880. int getfrag(void *from, char *to, int offset,
  881. int len, int odd, struct sk_buff *skb),
  882. void *from, int length);
  883. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  884. int offset, size_t size);
  885. struct skb_seq_state {
  886. __u32 lower_offset;
  887. __u32 upper_offset;
  888. __u32 frag_idx;
  889. __u32 stepped_offset;
  890. struct sk_buff *root_skb;
  891. struct sk_buff *cur_skb;
  892. __u8 *frag_data;
  893. };
  894. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  895. unsigned int to, struct skb_seq_state *st);
  896. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  897. struct skb_seq_state *st);
  898. void skb_abort_seq_read(struct skb_seq_state *st);
  899. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  900. unsigned int to, struct ts_config *config);
  901. /*
  902. * Packet hash types specify the type of hash in skb_set_hash.
  903. *
  904. * Hash types refer to the protocol layer addresses which are used to
  905. * construct a packet's hash. The hashes are used to differentiate or identify
  906. * flows of the protocol layer for the hash type. Hash types are either
  907. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  908. *
  909. * Properties of hashes:
  910. *
  911. * 1) Two packets in different flows have different hash values
  912. * 2) Two packets in the same flow should have the same hash value
  913. *
  914. * A hash at a higher layer is considered to be more specific. A driver should
  915. * set the most specific hash possible.
  916. *
  917. * A driver cannot indicate a more specific hash than the layer at which a hash
  918. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  919. *
  920. * A driver may indicate a hash level which is less specific than the
  921. * actual layer the hash was computed on. For instance, a hash computed
  922. * at L4 may be considered an L3 hash. This should only be done if the
  923. * driver can't unambiguously determine that the HW computed the hash at
  924. * the higher layer. Note that the "should" in the second property above
  925. * permits this.
  926. */
  927. enum pkt_hash_types {
  928. PKT_HASH_TYPE_NONE, /* Undefined type */
  929. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  930. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  931. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  932. };
  933. static inline void skb_clear_hash(struct sk_buff *skb)
  934. {
  935. skb->hash = 0;
  936. skb->sw_hash = 0;
  937. skb->l4_hash = 0;
  938. }
  939. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  940. {
  941. if (!skb->l4_hash)
  942. skb_clear_hash(skb);
  943. }
  944. static inline void
  945. __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
  946. {
  947. skb->l4_hash = is_l4;
  948. skb->sw_hash = is_sw;
  949. skb->hash = hash;
  950. }
  951. static inline void
  952. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  953. {
  954. /* Used by drivers to set hash from HW */
  955. __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
  956. }
  957. static inline void
  958. __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
  959. {
  960. __skb_set_hash(skb, hash, true, is_l4);
  961. }
  962. void __skb_get_hash(struct sk_buff *skb);
  963. u32 __skb_get_hash_symmetric(struct sk_buff *skb);
  964. u32 skb_get_poff(const struct sk_buff *skb);
  965. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  966. const struct flow_keys *keys, int hlen);
  967. __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
  968. void *data, int hlen_proto);
  969. static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
  970. int thoff, u8 ip_proto)
  971. {
  972. return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
  973. }
  974. void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  975. const struct flow_dissector_key *key,
  976. unsigned int key_count);
  977. bool __skb_flow_dissect(const struct sk_buff *skb,
  978. struct flow_dissector *flow_dissector,
  979. void *target_container,
  980. void *data, __be16 proto, int nhoff, int hlen,
  981. unsigned int flags);
  982. static inline bool skb_flow_dissect(const struct sk_buff *skb,
  983. struct flow_dissector *flow_dissector,
  984. void *target_container, unsigned int flags)
  985. {
  986. return __skb_flow_dissect(skb, flow_dissector, target_container,
  987. NULL, 0, 0, 0, flags);
  988. }
  989. static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
  990. struct flow_keys *flow,
  991. unsigned int flags)
  992. {
  993. memset(flow, 0, sizeof(*flow));
  994. return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
  995. NULL, 0, 0, 0, flags);
  996. }
  997. static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
  998. void *data, __be16 proto,
  999. int nhoff, int hlen,
  1000. unsigned int flags)
  1001. {
  1002. memset(flow, 0, sizeof(*flow));
  1003. return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
  1004. data, proto, nhoff, hlen, flags);
  1005. }
  1006. static inline __u32 skb_get_hash(struct sk_buff *skb)
  1007. {
  1008. if (!skb->l4_hash && !skb->sw_hash)
  1009. __skb_get_hash(skb);
  1010. return skb->hash;
  1011. }
  1012. __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
  1013. static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
  1014. {
  1015. if (!skb->l4_hash && !skb->sw_hash) {
  1016. struct flow_keys keys;
  1017. __u32 hash = __get_hash_from_flowi6(fl6, &keys);
  1018. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1019. }
  1020. return skb->hash;
  1021. }
  1022. __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
  1023. static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
  1024. {
  1025. if (!skb->l4_hash && !skb->sw_hash) {
  1026. struct flow_keys keys;
  1027. __u32 hash = __get_hash_from_flowi4(fl4, &keys);
  1028. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1029. }
  1030. return skb->hash;
  1031. }
  1032. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  1033. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  1034. {
  1035. return skb->hash;
  1036. }
  1037. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  1038. {
  1039. to->hash = from->hash;
  1040. to->sw_hash = from->sw_hash;
  1041. to->l4_hash = from->l4_hash;
  1042. };
  1043. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1044. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1045. {
  1046. return skb->head + skb->end;
  1047. }
  1048. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1049. {
  1050. return skb->end;
  1051. }
  1052. #else
  1053. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1054. {
  1055. return skb->end;
  1056. }
  1057. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1058. {
  1059. return skb->end - skb->head;
  1060. }
  1061. #endif
  1062. /* Internal */
  1063. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  1064. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  1065. {
  1066. return &skb_shinfo(skb)->hwtstamps;
  1067. }
  1068. /**
  1069. * skb_queue_empty - check if a queue is empty
  1070. * @list: queue head
  1071. *
  1072. * Returns true if the queue is empty, false otherwise.
  1073. */
  1074. static inline int skb_queue_empty(const struct sk_buff_head *list)
  1075. {
  1076. return list->next == (const struct sk_buff *) list;
  1077. }
  1078. /**
  1079. * skb_queue_is_last - check if skb is the last entry in the queue
  1080. * @list: queue head
  1081. * @skb: buffer
  1082. *
  1083. * Returns true if @skb is the last buffer on the list.
  1084. */
  1085. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  1086. const struct sk_buff *skb)
  1087. {
  1088. return skb->next == (const struct sk_buff *) list;
  1089. }
  1090. /**
  1091. * skb_queue_is_first - check if skb is the first entry in the queue
  1092. * @list: queue head
  1093. * @skb: buffer
  1094. *
  1095. * Returns true if @skb is the first buffer on the list.
  1096. */
  1097. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  1098. const struct sk_buff *skb)
  1099. {
  1100. return skb->prev == (const struct sk_buff *) list;
  1101. }
  1102. /**
  1103. * skb_queue_next - return the next packet in the queue
  1104. * @list: queue head
  1105. * @skb: current buffer
  1106. *
  1107. * Return the next packet in @list after @skb. It is only valid to
  1108. * call this if skb_queue_is_last() evaluates to false.
  1109. */
  1110. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  1111. const struct sk_buff *skb)
  1112. {
  1113. /* This BUG_ON may seem severe, but if we just return then we
  1114. * are going to dereference garbage.
  1115. */
  1116. BUG_ON(skb_queue_is_last(list, skb));
  1117. return skb->next;
  1118. }
  1119. /**
  1120. * skb_queue_prev - return the prev packet in the queue
  1121. * @list: queue head
  1122. * @skb: current buffer
  1123. *
  1124. * Return the prev packet in @list before @skb. It is only valid to
  1125. * call this if skb_queue_is_first() evaluates to false.
  1126. */
  1127. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  1128. const struct sk_buff *skb)
  1129. {
  1130. /* This BUG_ON may seem severe, but if we just return then we
  1131. * are going to dereference garbage.
  1132. */
  1133. BUG_ON(skb_queue_is_first(list, skb));
  1134. return skb->prev;
  1135. }
  1136. /**
  1137. * skb_get - reference buffer
  1138. * @skb: buffer to reference
  1139. *
  1140. * Makes another reference to a socket buffer and returns a pointer
  1141. * to the buffer.
  1142. */
  1143. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  1144. {
  1145. atomic_inc(&skb->users);
  1146. return skb;
  1147. }
  1148. /*
  1149. * If users == 1, we are the only owner and are can avoid redundant
  1150. * atomic change.
  1151. */
  1152. /**
  1153. * skb_cloned - is the buffer a clone
  1154. * @skb: buffer to check
  1155. *
  1156. * Returns true if the buffer was generated with skb_clone() and is
  1157. * one of multiple shared copies of the buffer. Cloned buffers are
  1158. * shared data so must not be written to under normal circumstances.
  1159. */
  1160. static inline int skb_cloned(const struct sk_buff *skb)
  1161. {
  1162. return skb->cloned &&
  1163. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  1164. }
  1165. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  1166. {
  1167. might_sleep_if(gfpflags_allow_blocking(pri));
  1168. if (skb_cloned(skb))
  1169. return pskb_expand_head(skb, 0, 0, pri);
  1170. return 0;
  1171. }
  1172. /**
  1173. * skb_header_cloned - is the header a clone
  1174. * @skb: buffer to check
  1175. *
  1176. * Returns true if modifying the header part of the buffer requires
  1177. * the data to be copied.
  1178. */
  1179. static inline int skb_header_cloned(const struct sk_buff *skb)
  1180. {
  1181. int dataref;
  1182. if (!skb->cloned)
  1183. return 0;
  1184. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  1185. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  1186. return dataref != 1;
  1187. }
  1188. static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
  1189. {
  1190. might_sleep_if(gfpflags_allow_blocking(pri));
  1191. if (skb_header_cloned(skb))
  1192. return pskb_expand_head(skb, 0, 0, pri);
  1193. return 0;
  1194. }
  1195. /**
  1196. * skb_header_release - release reference to header
  1197. * @skb: buffer to operate on
  1198. *
  1199. * Drop a reference to the header part of the buffer. This is done
  1200. * by acquiring a payload reference. You must not read from the header
  1201. * part of skb->data after this.
  1202. * Note : Check if you can use __skb_header_release() instead.
  1203. */
  1204. static inline void skb_header_release(struct sk_buff *skb)
  1205. {
  1206. BUG_ON(skb->nohdr);
  1207. skb->nohdr = 1;
  1208. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1209. }
  1210. /**
  1211. * __skb_header_release - release reference to header
  1212. * @skb: buffer to operate on
  1213. *
  1214. * Variant of skb_header_release() assuming skb is private to caller.
  1215. * We can avoid one atomic operation.
  1216. */
  1217. static inline void __skb_header_release(struct sk_buff *skb)
  1218. {
  1219. skb->nohdr = 1;
  1220. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1221. }
  1222. /**
  1223. * skb_shared - is the buffer shared
  1224. * @skb: buffer to check
  1225. *
  1226. * Returns true if more than one person has a reference to this
  1227. * buffer.
  1228. */
  1229. static inline int skb_shared(const struct sk_buff *skb)
  1230. {
  1231. return atomic_read(&skb->users) != 1;
  1232. }
  1233. /**
  1234. * skb_share_check - check if buffer is shared and if so clone it
  1235. * @skb: buffer to check
  1236. * @pri: priority for memory allocation
  1237. *
  1238. * If the buffer is shared the buffer is cloned and the old copy
  1239. * drops a reference. A new clone with a single reference is returned.
  1240. * If the buffer is not shared the original buffer is returned. When
  1241. * being called from interrupt status or with spinlocks held pri must
  1242. * be GFP_ATOMIC.
  1243. *
  1244. * NULL is returned on a memory allocation failure.
  1245. */
  1246. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1247. {
  1248. might_sleep_if(gfpflags_allow_blocking(pri));
  1249. if (skb_shared(skb)) {
  1250. struct sk_buff *nskb = skb_clone(skb, pri);
  1251. if (likely(nskb))
  1252. consume_skb(skb);
  1253. else
  1254. kfree_skb(skb);
  1255. skb = nskb;
  1256. }
  1257. return skb;
  1258. }
  1259. /*
  1260. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1261. * packets to handle cases where we have a local reader and forward
  1262. * and a couple of other messy ones. The normal one is tcpdumping
  1263. * a packet thats being forwarded.
  1264. */
  1265. /**
  1266. * skb_unshare - make a copy of a shared buffer
  1267. * @skb: buffer to check
  1268. * @pri: priority for memory allocation
  1269. *
  1270. * If the socket buffer is a clone then this function creates a new
  1271. * copy of the data, drops a reference count on the old copy and returns
  1272. * the new copy with the reference count at 1. If the buffer is not a clone
  1273. * the original buffer is returned. When called with a spinlock held or
  1274. * from interrupt state @pri must be %GFP_ATOMIC
  1275. *
  1276. * %NULL is returned on a memory allocation failure.
  1277. */
  1278. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1279. gfp_t pri)
  1280. {
  1281. might_sleep_if(gfpflags_allow_blocking(pri));
  1282. if (skb_cloned(skb)) {
  1283. struct sk_buff *nskb = skb_copy(skb, pri);
  1284. /* Free our shared copy */
  1285. if (likely(nskb))
  1286. consume_skb(skb);
  1287. else
  1288. kfree_skb(skb);
  1289. skb = nskb;
  1290. }
  1291. return skb;
  1292. }
  1293. /**
  1294. * skb_peek - peek at the head of an &sk_buff_head
  1295. * @list_: list to peek at
  1296. *
  1297. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1298. * be careful with this one. A peek leaves the buffer on the
  1299. * list and someone else may run off with it. You must hold
  1300. * the appropriate locks or have a private queue to do this.
  1301. *
  1302. * Returns %NULL for an empty list or a pointer to the head element.
  1303. * The reference count is not incremented and the reference is therefore
  1304. * volatile. Use with caution.
  1305. */
  1306. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1307. {
  1308. struct sk_buff *skb = list_->next;
  1309. if (skb == (struct sk_buff *)list_)
  1310. skb = NULL;
  1311. return skb;
  1312. }
  1313. /**
  1314. * skb_peek_next - peek skb following the given one from a queue
  1315. * @skb: skb to start from
  1316. * @list_: list to peek at
  1317. *
  1318. * Returns %NULL when the end of the list is met or a pointer to the
  1319. * next element. The reference count is not incremented and the
  1320. * reference is therefore volatile. Use with caution.
  1321. */
  1322. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1323. const struct sk_buff_head *list_)
  1324. {
  1325. struct sk_buff *next = skb->next;
  1326. if (next == (struct sk_buff *)list_)
  1327. next = NULL;
  1328. return next;
  1329. }
  1330. /**
  1331. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1332. * @list_: list to peek at
  1333. *
  1334. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1335. * be careful with this one. A peek leaves the buffer on the
  1336. * list and someone else may run off with it. You must hold
  1337. * the appropriate locks or have a private queue to do this.
  1338. *
  1339. * Returns %NULL for an empty list or a pointer to the tail element.
  1340. * The reference count is not incremented and the reference is therefore
  1341. * volatile. Use with caution.
  1342. */
  1343. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1344. {
  1345. struct sk_buff *skb = list_->prev;
  1346. if (skb == (struct sk_buff *)list_)
  1347. skb = NULL;
  1348. return skb;
  1349. }
  1350. /**
  1351. * skb_queue_len - get queue length
  1352. * @list_: list to measure
  1353. *
  1354. * Return the length of an &sk_buff queue.
  1355. */
  1356. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1357. {
  1358. return list_->qlen;
  1359. }
  1360. /**
  1361. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1362. * @list: queue to initialize
  1363. *
  1364. * This initializes only the list and queue length aspects of
  1365. * an sk_buff_head object. This allows to initialize the list
  1366. * aspects of an sk_buff_head without reinitializing things like
  1367. * the spinlock. It can also be used for on-stack sk_buff_head
  1368. * objects where the spinlock is known to not be used.
  1369. */
  1370. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1371. {
  1372. list->prev = list->next = (struct sk_buff *)list;
  1373. list->qlen = 0;
  1374. }
  1375. /*
  1376. * This function creates a split out lock class for each invocation;
  1377. * this is needed for now since a whole lot of users of the skb-queue
  1378. * infrastructure in drivers have different locking usage (in hardirq)
  1379. * than the networking core (in softirq only). In the long run either the
  1380. * network layer or drivers should need annotation to consolidate the
  1381. * main types of usage into 3 classes.
  1382. */
  1383. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1384. {
  1385. spin_lock_init(&list->lock);
  1386. __skb_queue_head_init(list);
  1387. }
  1388. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1389. struct lock_class_key *class)
  1390. {
  1391. skb_queue_head_init(list);
  1392. lockdep_set_class(&list->lock, class);
  1393. }
  1394. /*
  1395. * Insert an sk_buff on a list.
  1396. *
  1397. * The "__skb_xxxx()" functions are the non-atomic ones that
  1398. * can only be called with interrupts disabled.
  1399. */
  1400. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1401. struct sk_buff_head *list);
  1402. static inline void __skb_insert(struct sk_buff *newsk,
  1403. struct sk_buff *prev, struct sk_buff *next,
  1404. struct sk_buff_head *list)
  1405. {
  1406. newsk->next = next;
  1407. newsk->prev = prev;
  1408. next->prev = prev->next = newsk;
  1409. list->qlen++;
  1410. }
  1411. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1412. struct sk_buff *prev,
  1413. struct sk_buff *next)
  1414. {
  1415. struct sk_buff *first = list->next;
  1416. struct sk_buff *last = list->prev;
  1417. first->prev = prev;
  1418. prev->next = first;
  1419. last->next = next;
  1420. next->prev = last;
  1421. }
  1422. /**
  1423. * skb_queue_splice - join two skb lists, this is designed for stacks
  1424. * @list: the new list to add
  1425. * @head: the place to add it in the first list
  1426. */
  1427. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1428. struct sk_buff_head *head)
  1429. {
  1430. if (!skb_queue_empty(list)) {
  1431. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1432. head->qlen += list->qlen;
  1433. }
  1434. }
  1435. /**
  1436. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1437. * @list: the new list to add
  1438. * @head: the place to add it in the first list
  1439. *
  1440. * The list at @list is reinitialised
  1441. */
  1442. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1443. struct sk_buff_head *head)
  1444. {
  1445. if (!skb_queue_empty(list)) {
  1446. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1447. head->qlen += list->qlen;
  1448. __skb_queue_head_init(list);
  1449. }
  1450. }
  1451. /**
  1452. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1453. * @list: the new list to add
  1454. * @head: the place to add it in the first list
  1455. */
  1456. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1457. struct sk_buff_head *head)
  1458. {
  1459. if (!skb_queue_empty(list)) {
  1460. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1461. head->qlen += list->qlen;
  1462. }
  1463. }
  1464. /**
  1465. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1466. * @list: the new list to add
  1467. * @head: the place to add it in the first list
  1468. *
  1469. * Each of the lists is a queue.
  1470. * The list at @list is reinitialised
  1471. */
  1472. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1473. struct sk_buff_head *head)
  1474. {
  1475. if (!skb_queue_empty(list)) {
  1476. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1477. head->qlen += list->qlen;
  1478. __skb_queue_head_init(list);
  1479. }
  1480. }
  1481. /**
  1482. * __skb_queue_after - queue a buffer at the list head
  1483. * @list: list to use
  1484. * @prev: place after this buffer
  1485. * @newsk: buffer to queue
  1486. *
  1487. * Queue a buffer int the middle of a list. This function takes no locks
  1488. * and you must therefore hold required locks before calling it.
  1489. *
  1490. * A buffer cannot be placed on two lists at the same time.
  1491. */
  1492. static inline void __skb_queue_after(struct sk_buff_head *list,
  1493. struct sk_buff *prev,
  1494. struct sk_buff *newsk)
  1495. {
  1496. __skb_insert(newsk, prev, prev->next, list);
  1497. }
  1498. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1499. struct sk_buff_head *list);
  1500. static inline void __skb_queue_before(struct sk_buff_head *list,
  1501. struct sk_buff *next,
  1502. struct sk_buff *newsk)
  1503. {
  1504. __skb_insert(newsk, next->prev, next, list);
  1505. }
  1506. /**
  1507. * __skb_queue_head - queue a buffer at the list head
  1508. * @list: list to use
  1509. * @newsk: buffer to queue
  1510. *
  1511. * Queue a buffer at the start of a list. This function takes no locks
  1512. * and you must therefore hold required locks before calling it.
  1513. *
  1514. * A buffer cannot be placed on two lists at the same time.
  1515. */
  1516. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1517. static inline void __skb_queue_head(struct sk_buff_head *list,
  1518. struct sk_buff *newsk)
  1519. {
  1520. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1521. }
  1522. /**
  1523. * __skb_queue_tail - queue a buffer at the list tail
  1524. * @list: list to use
  1525. * @newsk: buffer to queue
  1526. *
  1527. * Queue a buffer at the end of a list. This function takes no locks
  1528. * and you must therefore hold required locks before calling it.
  1529. *
  1530. * A buffer cannot be placed on two lists at the same time.
  1531. */
  1532. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1533. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1534. struct sk_buff *newsk)
  1535. {
  1536. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1537. }
  1538. /*
  1539. * remove sk_buff from list. _Must_ be called atomically, and with
  1540. * the list known..
  1541. */
  1542. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1543. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1544. {
  1545. struct sk_buff *next, *prev;
  1546. list->qlen--;
  1547. next = skb->next;
  1548. prev = skb->prev;
  1549. skb->next = skb->prev = NULL;
  1550. next->prev = prev;
  1551. prev->next = next;
  1552. }
  1553. /**
  1554. * __skb_dequeue - remove from the head of the queue
  1555. * @list: list to dequeue from
  1556. *
  1557. * Remove the head of the list. This function does not take any locks
  1558. * so must be used with appropriate locks held only. The head item is
  1559. * returned or %NULL if the list is empty.
  1560. */
  1561. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1562. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1563. {
  1564. struct sk_buff *skb = skb_peek(list);
  1565. if (skb)
  1566. __skb_unlink(skb, list);
  1567. return skb;
  1568. }
  1569. /**
  1570. * __skb_dequeue_tail - remove from the tail of the queue
  1571. * @list: list to dequeue from
  1572. *
  1573. * Remove the tail of the list. This function does not take any locks
  1574. * so must be used with appropriate locks held only. The tail item is
  1575. * returned or %NULL if the list is empty.
  1576. */
  1577. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1578. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1579. {
  1580. struct sk_buff *skb = skb_peek_tail(list);
  1581. if (skb)
  1582. __skb_unlink(skb, list);
  1583. return skb;
  1584. }
  1585. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1586. {
  1587. return skb->data_len;
  1588. }
  1589. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1590. {
  1591. return skb->len - skb->data_len;
  1592. }
  1593. static inline int skb_pagelen(const struct sk_buff *skb)
  1594. {
  1595. int i, len = 0;
  1596. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1597. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1598. return len + skb_headlen(skb);
  1599. }
  1600. /**
  1601. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1602. * @skb: buffer containing fragment to be initialised
  1603. * @i: paged fragment index to initialise
  1604. * @page: the page to use for this fragment
  1605. * @off: the offset to the data with @page
  1606. * @size: the length of the data
  1607. *
  1608. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1609. * offset @off within @page.
  1610. *
  1611. * Does not take any additional reference on the fragment.
  1612. */
  1613. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1614. struct page *page, int off, int size)
  1615. {
  1616. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1617. /*
  1618. * Propagate page pfmemalloc to the skb if we can. The problem is
  1619. * that not all callers have unique ownership of the page but rely
  1620. * on page_is_pfmemalloc doing the right thing(tm).
  1621. */
  1622. frag->page.p = page;
  1623. frag->page_offset = off;
  1624. skb_frag_size_set(frag, size);
  1625. page = compound_head(page);
  1626. if (page_is_pfmemalloc(page))
  1627. skb->pfmemalloc = true;
  1628. }
  1629. /**
  1630. * skb_fill_page_desc - initialise a paged fragment in an skb
  1631. * @skb: buffer containing fragment to be initialised
  1632. * @i: paged fragment index to initialise
  1633. * @page: the page to use for this fragment
  1634. * @off: the offset to the data with @page
  1635. * @size: the length of the data
  1636. *
  1637. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1638. * @skb to point to @size bytes at offset @off within @page. In
  1639. * addition updates @skb such that @i is the last fragment.
  1640. *
  1641. * Does not take any additional reference on the fragment.
  1642. */
  1643. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1644. struct page *page, int off, int size)
  1645. {
  1646. __skb_fill_page_desc(skb, i, page, off, size);
  1647. skb_shinfo(skb)->nr_frags = i + 1;
  1648. }
  1649. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1650. int size, unsigned int truesize);
  1651. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1652. unsigned int truesize);
  1653. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1654. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1655. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1656. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1657. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1658. {
  1659. return skb->head + skb->tail;
  1660. }
  1661. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1662. {
  1663. skb->tail = skb->data - skb->head;
  1664. }
  1665. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1666. {
  1667. skb_reset_tail_pointer(skb);
  1668. skb->tail += offset;
  1669. }
  1670. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1671. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1672. {
  1673. return skb->tail;
  1674. }
  1675. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1676. {
  1677. skb->tail = skb->data;
  1678. }
  1679. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1680. {
  1681. skb->tail = skb->data + offset;
  1682. }
  1683. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1684. /*
  1685. * Add data to an sk_buff
  1686. */
  1687. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1688. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1689. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1690. {
  1691. unsigned char *tmp = skb_tail_pointer(skb);
  1692. SKB_LINEAR_ASSERT(skb);
  1693. skb->tail += len;
  1694. skb->len += len;
  1695. return tmp;
  1696. }
  1697. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1698. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1699. {
  1700. skb->data -= len;
  1701. skb->len += len;
  1702. return skb->data;
  1703. }
  1704. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1705. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1706. {
  1707. skb->len -= len;
  1708. BUG_ON(skb->len < skb->data_len);
  1709. return skb->data += len;
  1710. }
  1711. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1712. {
  1713. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1714. }
  1715. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1716. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1717. {
  1718. if (len > skb_headlen(skb) &&
  1719. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1720. return NULL;
  1721. skb->len -= len;
  1722. return skb->data += len;
  1723. }
  1724. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1725. {
  1726. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1727. }
  1728. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1729. {
  1730. if (likely(len <= skb_headlen(skb)))
  1731. return 1;
  1732. if (unlikely(len > skb->len))
  1733. return 0;
  1734. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1735. }
  1736. /**
  1737. * skb_headroom - bytes at buffer head
  1738. * @skb: buffer to check
  1739. *
  1740. * Return the number of bytes of free space at the head of an &sk_buff.
  1741. */
  1742. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1743. {
  1744. return skb->data - skb->head;
  1745. }
  1746. /**
  1747. * skb_tailroom - bytes at buffer end
  1748. * @skb: buffer to check
  1749. *
  1750. * Return the number of bytes of free space at the tail of an sk_buff
  1751. */
  1752. static inline int skb_tailroom(const struct sk_buff *skb)
  1753. {
  1754. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1755. }
  1756. /**
  1757. * skb_availroom - bytes at buffer end
  1758. * @skb: buffer to check
  1759. *
  1760. * Return the number of bytes of free space at the tail of an sk_buff
  1761. * allocated by sk_stream_alloc()
  1762. */
  1763. static inline int skb_availroom(const struct sk_buff *skb)
  1764. {
  1765. if (skb_is_nonlinear(skb))
  1766. return 0;
  1767. return skb->end - skb->tail - skb->reserved_tailroom;
  1768. }
  1769. /**
  1770. * skb_reserve - adjust headroom
  1771. * @skb: buffer to alter
  1772. * @len: bytes to move
  1773. *
  1774. * Increase the headroom of an empty &sk_buff by reducing the tail
  1775. * room. This is only allowed for an empty buffer.
  1776. */
  1777. static inline void skb_reserve(struct sk_buff *skb, int len)
  1778. {
  1779. skb->data += len;
  1780. skb->tail += len;
  1781. }
  1782. /**
  1783. * skb_tailroom_reserve - adjust reserved_tailroom
  1784. * @skb: buffer to alter
  1785. * @mtu: maximum amount of headlen permitted
  1786. * @needed_tailroom: minimum amount of reserved_tailroom
  1787. *
  1788. * Set reserved_tailroom so that headlen can be as large as possible but
  1789. * not larger than mtu and tailroom cannot be smaller than
  1790. * needed_tailroom.
  1791. * The required headroom should already have been reserved before using
  1792. * this function.
  1793. */
  1794. static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
  1795. unsigned int needed_tailroom)
  1796. {
  1797. SKB_LINEAR_ASSERT(skb);
  1798. if (mtu < skb_tailroom(skb) - needed_tailroom)
  1799. /* use at most mtu */
  1800. skb->reserved_tailroom = skb_tailroom(skb) - mtu;
  1801. else
  1802. /* use up to all available space */
  1803. skb->reserved_tailroom = needed_tailroom;
  1804. }
  1805. #define ENCAP_TYPE_ETHER 0
  1806. #define ENCAP_TYPE_IPPROTO 1
  1807. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1808. __be16 protocol)
  1809. {
  1810. skb->inner_protocol = protocol;
  1811. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1812. }
  1813. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1814. __u8 ipproto)
  1815. {
  1816. skb->inner_ipproto = ipproto;
  1817. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1818. }
  1819. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1820. {
  1821. skb->inner_mac_header = skb->mac_header;
  1822. skb->inner_network_header = skb->network_header;
  1823. skb->inner_transport_header = skb->transport_header;
  1824. }
  1825. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1826. {
  1827. skb->mac_len = skb->network_header - skb->mac_header;
  1828. }
  1829. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1830. *skb)
  1831. {
  1832. return skb->head + skb->inner_transport_header;
  1833. }
  1834. static inline int skb_inner_transport_offset(const struct sk_buff *skb)
  1835. {
  1836. return skb_inner_transport_header(skb) - skb->data;
  1837. }
  1838. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1839. {
  1840. skb->inner_transport_header = skb->data - skb->head;
  1841. }
  1842. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1843. const int offset)
  1844. {
  1845. skb_reset_inner_transport_header(skb);
  1846. skb->inner_transport_header += offset;
  1847. }
  1848. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1849. {
  1850. return skb->head + skb->inner_network_header;
  1851. }
  1852. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1853. {
  1854. skb->inner_network_header = skb->data - skb->head;
  1855. }
  1856. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1857. const int offset)
  1858. {
  1859. skb_reset_inner_network_header(skb);
  1860. skb->inner_network_header += offset;
  1861. }
  1862. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1863. {
  1864. return skb->head + skb->inner_mac_header;
  1865. }
  1866. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1867. {
  1868. skb->inner_mac_header = skb->data - skb->head;
  1869. }
  1870. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1871. const int offset)
  1872. {
  1873. skb_reset_inner_mac_header(skb);
  1874. skb->inner_mac_header += offset;
  1875. }
  1876. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1877. {
  1878. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1879. }
  1880. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1881. {
  1882. return skb->head + skb->transport_header;
  1883. }
  1884. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1885. {
  1886. skb->transport_header = skb->data - skb->head;
  1887. }
  1888. static inline void skb_set_transport_header(struct sk_buff *skb,
  1889. const int offset)
  1890. {
  1891. skb_reset_transport_header(skb);
  1892. skb->transport_header += offset;
  1893. }
  1894. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1895. {
  1896. return skb->head + skb->network_header;
  1897. }
  1898. static inline void skb_reset_network_header(struct sk_buff *skb)
  1899. {
  1900. skb->network_header = skb->data - skb->head;
  1901. }
  1902. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1903. {
  1904. skb_reset_network_header(skb);
  1905. skb->network_header += offset;
  1906. }
  1907. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1908. {
  1909. return skb->head + skb->mac_header;
  1910. }
  1911. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1912. {
  1913. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1914. }
  1915. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1916. {
  1917. skb->mac_header = skb->data - skb->head;
  1918. }
  1919. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1920. {
  1921. skb_reset_mac_header(skb);
  1922. skb->mac_header += offset;
  1923. }
  1924. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1925. {
  1926. skb->mac_header = skb->network_header;
  1927. }
  1928. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1929. const int offset_hint)
  1930. {
  1931. struct flow_keys keys;
  1932. if (skb_transport_header_was_set(skb))
  1933. return;
  1934. else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
  1935. skb_set_transport_header(skb, keys.control.thoff);
  1936. else
  1937. skb_set_transport_header(skb, offset_hint);
  1938. }
  1939. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1940. {
  1941. if (skb_mac_header_was_set(skb)) {
  1942. const unsigned char *old_mac = skb_mac_header(skb);
  1943. skb_set_mac_header(skb, -skb->mac_len);
  1944. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1945. }
  1946. }
  1947. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1948. {
  1949. return skb->csum_start - skb_headroom(skb);
  1950. }
  1951. static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
  1952. {
  1953. return skb->head + skb->csum_start;
  1954. }
  1955. static inline int skb_transport_offset(const struct sk_buff *skb)
  1956. {
  1957. return skb_transport_header(skb) - skb->data;
  1958. }
  1959. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1960. {
  1961. return skb->transport_header - skb->network_header;
  1962. }
  1963. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1964. {
  1965. return skb->inner_transport_header - skb->inner_network_header;
  1966. }
  1967. static inline int skb_network_offset(const struct sk_buff *skb)
  1968. {
  1969. return skb_network_header(skb) - skb->data;
  1970. }
  1971. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1972. {
  1973. return skb_inner_network_header(skb) - skb->data;
  1974. }
  1975. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1976. {
  1977. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1978. }
  1979. /*
  1980. * CPUs often take a performance hit when accessing unaligned memory
  1981. * locations. The actual performance hit varies, it can be small if the
  1982. * hardware handles it or large if we have to take an exception and fix it
  1983. * in software.
  1984. *
  1985. * Since an ethernet header is 14 bytes network drivers often end up with
  1986. * the IP header at an unaligned offset. The IP header can be aligned by
  1987. * shifting the start of the packet by 2 bytes. Drivers should do this
  1988. * with:
  1989. *
  1990. * skb_reserve(skb, NET_IP_ALIGN);
  1991. *
  1992. * The downside to this alignment of the IP header is that the DMA is now
  1993. * unaligned. On some architectures the cost of an unaligned DMA is high
  1994. * and this cost outweighs the gains made by aligning the IP header.
  1995. *
  1996. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1997. * to be overridden.
  1998. */
  1999. #ifndef NET_IP_ALIGN
  2000. #define NET_IP_ALIGN 2
  2001. #endif
  2002. /*
  2003. * The networking layer reserves some headroom in skb data (via
  2004. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  2005. * the header has to grow. In the default case, if the header has to grow
  2006. * 32 bytes or less we avoid the reallocation.
  2007. *
  2008. * Unfortunately this headroom changes the DMA alignment of the resulting
  2009. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  2010. * on some architectures. An architecture can override this value,
  2011. * perhaps setting it to a cacheline in size (since that will maintain
  2012. * cacheline alignment of the DMA). It must be a power of 2.
  2013. *
  2014. * Various parts of the networking layer expect at least 32 bytes of
  2015. * headroom, you should not reduce this.
  2016. *
  2017. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  2018. * to reduce average number of cache lines per packet.
  2019. * get_rps_cpus() for example only access one 64 bytes aligned block :
  2020. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  2021. */
  2022. #ifndef NET_SKB_PAD
  2023. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  2024. #endif
  2025. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  2026. static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
  2027. {
  2028. if (unlikely(skb_is_nonlinear(skb))) {
  2029. WARN_ON(1);
  2030. return;
  2031. }
  2032. skb->len = len;
  2033. skb_set_tail_pointer(skb, len);
  2034. }
  2035. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  2036. {
  2037. __skb_set_length(skb, len);
  2038. }
  2039. void skb_trim(struct sk_buff *skb, unsigned int len);
  2040. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  2041. {
  2042. if (skb->data_len)
  2043. return ___pskb_trim(skb, len);
  2044. __skb_trim(skb, len);
  2045. return 0;
  2046. }
  2047. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  2048. {
  2049. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  2050. }
  2051. /**
  2052. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  2053. * @skb: buffer to alter
  2054. * @len: new length
  2055. *
  2056. * This is identical to pskb_trim except that the caller knows that
  2057. * the skb is not cloned so we should never get an error due to out-
  2058. * of-memory.
  2059. */
  2060. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  2061. {
  2062. int err = pskb_trim(skb, len);
  2063. BUG_ON(err);
  2064. }
  2065. static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
  2066. {
  2067. unsigned int diff = len - skb->len;
  2068. if (skb_tailroom(skb) < diff) {
  2069. int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
  2070. GFP_ATOMIC);
  2071. if (ret)
  2072. return ret;
  2073. }
  2074. __skb_set_length(skb, len);
  2075. return 0;
  2076. }
  2077. /**
  2078. * skb_orphan - orphan a buffer
  2079. * @skb: buffer to orphan
  2080. *
  2081. * If a buffer currently has an owner then we call the owner's
  2082. * destructor function and make the @skb unowned. The buffer continues
  2083. * to exist but is no longer charged to its former owner.
  2084. */
  2085. static inline void skb_orphan(struct sk_buff *skb)
  2086. {
  2087. if (skb->destructor) {
  2088. skb->destructor(skb);
  2089. skb->destructor = NULL;
  2090. skb->sk = NULL;
  2091. } else {
  2092. BUG_ON(skb->sk);
  2093. }
  2094. }
  2095. /**
  2096. * skb_orphan_frags - orphan the frags contained in a buffer
  2097. * @skb: buffer to orphan frags from
  2098. * @gfp_mask: allocation mask for replacement pages
  2099. *
  2100. * For each frag in the SKB which needs a destructor (i.e. has an
  2101. * owner) create a copy of that frag and release the original
  2102. * page by calling the destructor.
  2103. */
  2104. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  2105. {
  2106. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  2107. return 0;
  2108. return skb_copy_ubufs(skb, gfp_mask);
  2109. }
  2110. /**
  2111. * __skb_queue_purge - empty a list
  2112. * @list: list to empty
  2113. *
  2114. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  2115. * the list and one reference dropped. This function does not take the
  2116. * list lock and the caller must hold the relevant locks to use it.
  2117. */
  2118. void skb_queue_purge(struct sk_buff_head *list);
  2119. static inline void __skb_queue_purge(struct sk_buff_head *list)
  2120. {
  2121. struct sk_buff *skb;
  2122. while ((skb = __skb_dequeue(list)) != NULL)
  2123. kfree_skb(skb);
  2124. }
  2125. void skb_rbtree_purge(struct rb_root *root);
  2126. void *netdev_alloc_frag(unsigned int fragsz);
  2127. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  2128. gfp_t gfp_mask);
  2129. /**
  2130. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  2131. * @dev: network device to receive on
  2132. * @length: length to allocate
  2133. *
  2134. * Allocate a new &sk_buff and assign it a usage count of one. The
  2135. * buffer has unspecified headroom built in. Users should allocate
  2136. * the headroom they think they need without accounting for the
  2137. * built in space. The built in space is used for optimisations.
  2138. *
  2139. * %NULL is returned if there is no free memory. Although this function
  2140. * allocates memory it can be called from an interrupt.
  2141. */
  2142. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  2143. unsigned int length)
  2144. {
  2145. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  2146. }
  2147. /* legacy helper around __netdev_alloc_skb() */
  2148. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  2149. gfp_t gfp_mask)
  2150. {
  2151. return __netdev_alloc_skb(NULL, length, gfp_mask);
  2152. }
  2153. /* legacy helper around netdev_alloc_skb() */
  2154. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  2155. {
  2156. return netdev_alloc_skb(NULL, length);
  2157. }
  2158. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  2159. unsigned int length, gfp_t gfp)
  2160. {
  2161. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  2162. if (NET_IP_ALIGN && skb)
  2163. skb_reserve(skb, NET_IP_ALIGN);
  2164. return skb;
  2165. }
  2166. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  2167. unsigned int length)
  2168. {
  2169. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  2170. }
  2171. static inline void skb_free_frag(void *addr)
  2172. {
  2173. __free_page_frag(addr);
  2174. }
  2175. void *napi_alloc_frag(unsigned int fragsz);
  2176. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  2177. unsigned int length, gfp_t gfp_mask);
  2178. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  2179. unsigned int length)
  2180. {
  2181. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  2182. }
  2183. void napi_consume_skb(struct sk_buff *skb, int budget);
  2184. void __kfree_skb_flush(void);
  2185. void __kfree_skb_defer(struct sk_buff *skb);
  2186. /**
  2187. * __dev_alloc_pages - allocate page for network Rx
  2188. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2189. * @order: size of the allocation
  2190. *
  2191. * Allocate a new page.
  2192. *
  2193. * %NULL is returned if there is no free memory.
  2194. */
  2195. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  2196. unsigned int order)
  2197. {
  2198. /* This piece of code contains several assumptions.
  2199. * 1. This is for device Rx, therefor a cold page is preferred.
  2200. * 2. The expectation is the user wants a compound page.
  2201. * 3. If requesting a order 0 page it will not be compound
  2202. * due to the check to see if order has a value in prep_new_page
  2203. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  2204. * code in gfp_to_alloc_flags that should be enforcing this.
  2205. */
  2206. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  2207. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  2208. }
  2209. static inline struct page *dev_alloc_pages(unsigned int order)
  2210. {
  2211. return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
  2212. }
  2213. /**
  2214. * __dev_alloc_page - allocate a page for network Rx
  2215. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2216. *
  2217. * Allocate a new page.
  2218. *
  2219. * %NULL is returned if there is no free memory.
  2220. */
  2221. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  2222. {
  2223. return __dev_alloc_pages(gfp_mask, 0);
  2224. }
  2225. static inline struct page *dev_alloc_page(void)
  2226. {
  2227. return dev_alloc_pages(0);
  2228. }
  2229. /**
  2230. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  2231. * @page: The page that was allocated from skb_alloc_page
  2232. * @skb: The skb that may need pfmemalloc set
  2233. */
  2234. static inline void skb_propagate_pfmemalloc(struct page *page,
  2235. struct sk_buff *skb)
  2236. {
  2237. if (page_is_pfmemalloc(page))
  2238. skb->pfmemalloc = true;
  2239. }
  2240. /**
  2241. * skb_frag_page - retrieve the page referred to by a paged fragment
  2242. * @frag: the paged fragment
  2243. *
  2244. * Returns the &struct page associated with @frag.
  2245. */
  2246. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2247. {
  2248. return frag->page.p;
  2249. }
  2250. /**
  2251. * __skb_frag_ref - take an addition reference on a paged fragment.
  2252. * @frag: the paged fragment
  2253. *
  2254. * Takes an additional reference on the paged fragment @frag.
  2255. */
  2256. static inline void __skb_frag_ref(skb_frag_t *frag)
  2257. {
  2258. get_page(skb_frag_page(frag));
  2259. }
  2260. /**
  2261. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2262. * @skb: the buffer
  2263. * @f: the fragment offset.
  2264. *
  2265. * Takes an additional reference on the @f'th paged fragment of @skb.
  2266. */
  2267. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2268. {
  2269. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2270. }
  2271. /**
  2272. * __skb_frag_unref - release a reference on a paged fragment.
  2273. * @frag: the paged fragment
  2274. *
  2275. * Releases a reference on the paged fragment @frag.
  2276. */
  2277. static inline void __skb_frag_unref(skb_frag_t *frag)
  2278. {
  2279. put_page(skb_frag_page(frag));
  2280. }
  2281. /**
  2282. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2283. * @skb: the buffer
  2284. * @f: the fragment offset
  2285. *
  2286. * Releases a reference on the @f'th paged fragment of @skb.
  2287. */
  2288. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2289. {
  2290. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2291. }
  2292. /**
  2293. * skb_frag_address - gets the address of the data contained in a paged fragment
  2294. * @frag: the paged fragment buffer
  2295. *
  2296. * Returns the address of the data within @frag. The page must already
  2297. * be mapped.
  2298. */
  2299. static inline void *skb_frag_address(const skb_frag_t *frag)
  2300. {
  2301. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2302. }
  2303. /**
  2304. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2305. * @frag: the paged fragment buffer
  2306. *
  2307. * Returns the address of the data within @frag. Checks that the page
  2308. * is mapped and returns %NULL otherwise.
  2309. */
  2310. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2311. {
  2312. void *ptr = page_address(skb_frag_page(frag));
  2313. if (unlikely(!ptr))
  2314. return NULL;
  2315. return ptr + frag->page_offset;
  2316. }
  2317. /**
  2318. * __skb_frag_set_page - sets the page contained in a paged fragment
  2319. * @frag: the paged fragment
  2320. * @page: the page to set
  2321. *
  2322. * Sets the fragment @frag to contain @page.
  2323. */
  2324. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2325. {
  2326. frag->page.p = page;
  2327. }
  2328. /**
  2329. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2330. * @skb: the buffer
  2331. * @f: the fragment offset
  2332. * @page: the page to set
  2333. *
  2334. * Sets the @f'th fragment of @skb to contain @page.
  2335. */
  2336. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2337. struct page *page)
  2338. {
  2339. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2340. }
  2341. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2342. /**
  2343. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2344. * @dev: the device to map the fragment to
  2345. * @frag: the paged fragment to map
  2346. * @offset: the offset within the fragment (starting at the
  2347. * fragment's own offset)
  2348. * @size: the number of bytes to map
  2349. * @dir: the direction of the mapping (%PCI_DMA_*)
  2350. *
  2351. * Maps the page associated with @frag to @device.
  2352. */
  2353. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2354. const skb_frag_t *frag,
  2355. size_t offset, size_t size,
  2356. enum dma_data_direction dir)
  2357. {
  2358. return dma_map_page(dev, skb_frag_page(frag),
  2359. frag->page_offset + offset, size, dir);
  2360. }
  2361. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2362. gfp_t gfp_mask)
  2363. {
  2364. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2365. }
  2366. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2367. gfp_t gfp_mask)
  2368. {
  2369. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2370. }
  2371. /**
  2372. * skb_clone_writable - is the header of a clone writable
  2373. * @skb: buffer to check
  2374. * @len: length up to which to write
  2375. *
  2376. * Returns true if modifying the header part of the cloned buffer
  2377. * does not requires the data to be copied.
  2378. */
  2379. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2380. {
  2381. return !skb_header_cloned(skb) &&
  2382. skb_headroom(skb) + len <= skb->hdr_len;
  2383. }
  2384. static inline int skb_try_make_writable(struct sk_buff *skb,
  2385. unsigned int write_len)
  2386. {
  2387. return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
  2388. pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
  2389. }
  2390. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2391. int cloned)
  2392. {
  2393. int delta = 0;
  2394. if (headroom > skb_headroom(skb))
  2395. delta = headroom - skb_headroom(skb);
  2396. if (delta || cloned)
  2397. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2398. GFP_ATOMIC);
  2399. return 0;
  2400. }
  2401. /**
  2402. * skb_cow - copy header of skb when it is required
  2403. * @skb: buffer to cow
  2404. * @headroom: needed headroom
  2405. *
  2406. * If the skb passed lacks sufficient headroom or its data part
  2407. * is shared, data is reallocated. If reallocation fails, an error
  2408. * is returned and original skb is not changed.
  2409. *
  2410. * The result is skb with writable area skb->head...skb->tail
  2411. * and at least @headroom of space at head.
  2412. */
  2413. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2414. {
  2415. return __skb_cow(skb, headroom, skb_cloned(skb));
  2416. }
  2417. /**
  2418. * skb_cow_head - skb_cow but only making the head writable
  2419. * @skb: buffer to cow
  2420. * @headroom: needed headroom
  2421. *
  2422. * This function is identical to skb_cow except that we replace the
  2423. * skb_cloned check by skb_header_cloned. It should be used when
  2424. * you only need to push on some header and do not need to modify
  2425. * the data.
  2426. */
  2427. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2428. {
  2429. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2430. }
  2431. /**
  2432. * skb_padto - pad an skbuff up to a minimal size
  2433. * @skb: buffer to pad
  2434. * @len: minimal length
  2435. *
  2436. * Pads up a buffer to ensure the trailing bytes exist and are
  2437. * blanked. If the buffer already contains sufficient data it
  2438. * is untouched. Otherwise it is extended. Returns zero on
  2439. * success. The skb is freed on error.
  2440. */
  2441. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2442. {
  2443. unsigned int size = skb->len;
  2444. if (likely(size >= len))
  2445. return 0;
  2446. return skb_pad(skb, len - size);
  2447. }
  2448. /**
  2449. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2450. * @skb: buffer to pad
  2451. * @len: minimal length
  2452. *
  2453. * Pads up a buffer to ensure the trailing bytes exist and are
  2454. * blanked. If the buffer already contains sufficient data it
  2455. * is untouched. Otherwise it is extended. Returns zero on
  2456. * success. The skb is freed on error.
  2457. */
  2458. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2459. {
  2460. unsigned int size = skb->len;
  2461. if (unlikely(size < len)) {
  2462. len -= size;
  2463. if (skb_pad(skb, len))
  2464. return -ENOMEM;
  2465. __skb_put(skb, len);
  2466. }
  2467. return 0;
  2468. }
  2469. static inline int skb_add_data(struct sk_buff *skb,
  2470. struct iov_iter *from, int copy)
  2471. {
  2472. const int off = skb->len;
  2473. if (skb->ip_summed == CHECKSUM_NONE) {
  2474. __wsum csum = 0;
  2475. if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
  2476. &csum, from) == copy) {
  2477. skb->csum = csum_block_add(skb->csum, csum, off);
  2478. return 0;
  2479. }
  2480. } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
  2481. return 0;
  2482. __skb_trim(skb, off);
  2483. return -EFAULT;
  2484. }
  2485. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2486. const struct page *page, int off)
  2487. {
  2488. if (i) {
  2489. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2490. return page == skb_frag_page(frag) &&
  2491. off == frag->page_offset + skb_frag_size(frag);
  2492. }
  2493. return false;
  2494. }
  2495. static inline int __skb_linearize(struct sk_buff *skb)
  2496. {
  2497. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2498. }
  2499. /**
  2500. * skb_linearize - convert paged skb to linear one
  2501. * @skb: buffer to linarize
  2502. *
  2503. * If there is no free memory -ENOMEM is returned, otherwise zero
  2504. * is returned and the old skb data released.
  2505. */
  2506. static inline int skb_linearize(struct sk_buff *skb)
  2507. {
  2508. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2509. }
  2510. /**
  2511. * skb_has_shared_frag - can any frag be overwritten
  2512. * @skb: buffer to test
  2513. *
  2514. * Return true if the skb has at least one frag that might be modified
  2515. * by an external entity (as in vmsplice()/sendfile())
  2516. */
  2517. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2518. {
  2519. return skb_is_nonlinear(skb) &&
  2520. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2521. }
  2522. /**
  2523. * skb_linearize_cow - make sure skb is linear and writable
  2524. * @skb: buffer to process
  2525. *
  2526. * If there is no free memory -ENOMEM is returned, otherwise zero
  2527. * is returned and the old skb data released.
  2528. */
  2529. static inline int skb_linearize_cow(struct sk_buff *skb)
  2530. {
  2531. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2532. __skb_linearize(skb) : 0;
  2533. }
  2534. static __always_inline void
  2535. __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2536. unsigned int off)
  2537. {
  2538. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2539. skb->csum = csum_block_sub(skb->csum,
  2540. csum_partial(start, len, 0), off);
  2541. else if (skb->ip_summed == CHECKSUM_PARTIAL &&
  2542. skb_checksum_start_offset(skb) < 0)
  2543. skb->ip_summed = CHECKSUM_NONE;
  2544. }
  2545. /**
  2546. * skb_postpull_rcsum - update checksum for received skb after pull
  2547. * @skb: buffer to update
  2548. * @start: start of data before pull
  2549. * @len: length of data pulled
  2550. *
  2551. * After doing a pull on a received packet, you need to call this to
  2552. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2553. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2554. */
  2555. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2556. const void *start, unsigned int len)
  2557. {
  2558. __skb_postpull_rcsum(skb, start, len, 0);
  2559. }
  2560. static __always_inline void
  2561. __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2562. unsigned int off)
  2563. {
  2564. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2565. skb->csum = csum_block_add(skb->csum,
  2566. csum_partial(start, len, 0), off);
  2567. }
  2568. /**
  2569. * skb_postpush_rcsum - update checksum for received skb after push
  2570. * @skb: buffer to update
  2571. * @start: start of data after push
  2572. * @len: length of data pushed
  2573. *
  2574. * After doing a push on a received packet, you need to call this to
  2575. * update the CHECKSUM_COMPLETE checksum.
  2576. */
  2577. static inline void skb_postpush_rcsum(struct sk_buff *skb,
  2578. const void *start, unsigned int len)
  2579. {
  2580. __skb_postpush_rcsum(skb, start, len, 0);
  2581. }
  2582. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2583. /**
  2584. * skb_push_rcsum - push skb and update receive checksum
  2585. * @skb: buffer to update
  2586. * @len: length of data pulled
  2587. *
  2588. * This function performs an skb_push on the packet and updates
  2589. * the CHECKSUM_COMPLETE checksum. It should be used on
  2590. * receive path processing instead of skb_push unless you know
  2591. * that the checksum difference is zero (e.g., a valid IP header)
  2592. * or you are setting ip_summed to CHECKSUM_NONE.
  2593. */
  2594. static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
  2595. unsigned int len)
  2596. {
  2597. skb_push(skb, len);
  2598. skb_postpush_rcsum(skb, skb->data, len);
  2599. return skb->data;
  2600. }
  2601. /**
  2602. * pskb_trim_rcsum - trim received skb and update checksum
  2603. * @skb: buffer to trim
  2604. * @len: new length
  2605. *
  2606. * This is exactly the same as pskb_trim except that it ensures the
  2607. * checksum of received packets are still valid after the operation.
  2608. */
  2609. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2610. {
  2611. if (likely(len >= skb->len))
  2612. return 0;
  2613. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2614. skb->ip_summed = CHECKSUM_NONE;
  2615. return __pskb_trim(skb, len);
  2616. }
  2617. static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2618. {
  2619. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2620. skb->ip_summed = CHECKSUM_NONE;
  2621. __skb_trim(skb, len);
  2622. return 0;
  2623. }
  2624. static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
  2625. {
  2626. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2627. skb->ip_summed = CHECKSUM_NONE;
  2628. return __skb_grow(skb, len);
  2629. }
  2630. #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
  2631. #define skb_queue_walk(queue, skb) \
  2632. for (skb = (queue)->next; \
  2633. skb != (struct sk_buff *)(queue); \
  2634. skb = skb->next)
  2635. #define skb_queue_walk_safe(queue, skb, tmp) \
  2636. for (skb = (queue)->next, tmp = skb->next; \
  2637. skb != (struct sk_buff *)(queue); \
  2638. skb = tmp, tmp = skb->next)
  2639. #define skb_queue_walk_from(queue, skb) \
  2640. for (; skb != (struct sk_buff *)(queue); \
  2641. skb = skb->next)
  2642. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2643. for (tmp = skb->next; \
  2644. skb != (struct sk_buff *)(queue); \
  2645. skb = tmp, tmp = skb->next)
  2646. #define skb_queue_reverse_walk(queue, skb) \
  2647. for (skb = (queue)->prev; \
  2648. skb != (struct sk_buff *)(queue); \
  2649. skb = skb->prev)
  2650. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2651. for (skb = (queue)->prev, tmp = skb->prev; \
  2652. skb != (struct sk_buff *)(queue); \
  2653. skb = tmp, tmp = skb->prev)
  2654. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2655. for (tmp = skb->prev; \
  2656. skb != (struct sk_buff *)(queue); \
  2657. skb = tmp, tmp = skb->prev)
  2658. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2659. {
  2660. return skb_shinfo(skb)->frag_list != NULL;
  2661. }
  2662. static inline void skb_frag_list_init(struct sk_buff *skb)
  2663. {
  2664. skb_shinfo(skb)->frag_list = NULL;
  2665. }
  2666. #define skb_walk_frags(skb, iter) \
  2667. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2668. int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
  2669. const struct sk_buff *skb);
  2670. struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
  2671. int *peeked, int *off, int *err,
  2672. struct sk_buff **last);
  2673. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2674. int *peeked, int *off, int *err);
  2675. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2676. int *err);
  2677. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2678. struct poll_table_struct *wait);
  2679. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2680. struct iov_iter *to, int size);
  2681. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2682. struct msghdr *msg, int size)
  2683. {
  2684. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2685. }
  2686. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2687. struct msghdr *msg);
  2688. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2689. struct iov_iter *from, int len);
  2690. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2691. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2692. void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
  2693. static inline void skb_free_datagram_locked(struct sock *sk,
  2694. struct sk_buff *skb)
  2695. {
  2696. __skb_free_datagram_locked(sk, skb, 0);
  2697. }
  2698. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2699. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2700. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2701. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2702. int len, __wsum csum);
  2703. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2704. struct pipe_inode_info *pipe, unsigned int len,
  2705. unsigned int flags);
  2706. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2707. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2708. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2709. int len, int hlen);
  2710. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2711. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2712. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2713. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2714. bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
  2715. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2716. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2717. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2718. int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
  2719. int skb_vlan_pop(struct sk_buff *skb);
  2720. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2721. struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
  2722. gfp_t gfp);
  2723. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2724. {
  2725. return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2726. }
  2727. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2728. {
  2729. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2730. }
  2731. struct skb_checksum_ops {
  2732. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2733. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2734. };
  2735. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2736. __wsum csum, const struct skb_checksum_ops *ops);
  2737. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2738. __wsum csum);
  2739. static inline void * __must_check
  2740. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2741. int len, void *data, int hlen, void *buffer)
  2742. {
  2743. if (hlen - offset >= len)
  2744. return data + offset;
  2745. if (!skb ||
  2746. skb_copy_bits(skb, offset, buffer, len) < 0)
  2747. return NULL;
  2748. return buffer;
  2749. }
  2750. static inline void * __must_check
  2751. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2752. {
  2753. return __skb_header_pointer(skb, offset, len, skb->data,
  2754. skb_headlen(skb), buffer);
  2755. }
  2756. /**
  2757. * skb_needs_linearize - check if we need to linearize a given skb
  2758. * depending on the given device features.
  2759. * @skb: socket buffer to check
  2760. * @features: net device features
  2761. *
  2762. * Returns true if either:
  2763. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2764. * 2. skb is fragmented and the device does not support SG.
  2765. */
  2766. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2767. netdev_features_t features)
  2768. {
  2769. return skb_is_nonlinear(skb) &&
  2770. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2771. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2772. }
  2773. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2774. void *to,
  2775. const unsigned int len)
  2776. {
  2777. memcpy(to, skb->data, len);
  2778. }
  2779. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2780. const int offset, void *to,
  2781. const unsigned int len)
  2782. {
  2783. memcpy(to, skb->data + offset, len);
  2784. }
  2785. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2786. const void *from,
  2787. const unsigned int len)
  2788. {
  2789. memcpy(skb->data, from, len);
  2790. }
  2791. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2792. const int offset,
  2793. const void *from,
  2794. const unsigned int len)
  2795. {
  2796. memcpy(skb->data + offset, from, len);
  2797. }
  2798. void skb_init(void);
  2799. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2800. {
  2801. return skb->tstamp;
  2802. }
  2803. /**
  2804. * skb_get_timestamp - get timestamp from a skb
  2805. * @skb: skb to get stamp from
  2806. * @stamp: pointer to struct timeval to store stamp in
  2807. *
  2808. * Timestamps are stored in the skb as offsets to a base timestamp.
  2809. * This function converts the offset back to a struct timeval and stores
  2810. * it in stamp.
  2811. */
  2812. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2813. struct timeval *stamp)
  2814. {
  2815. *stamp = ktime_to_timeval(skb->tstamp);
  2816. }
  2817. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2818. struct timespec *stamp)
  2819. {
  2820. *stamp = ktime_to_timespec(skb->tstamp);
  2821. }
  2822. static inline void __net_timestamp(struct sk_buff *skb)
  2823. {
  2824. skb->tstamp = ktime_get_real();
  2825. }
  2826. static inline ktime_t net_timedelta(ktime_t t)
  2827. {
  2828. return ktime_sub(ktime_get_real(), t);
  2829. }
  2830. static inline ktime_t net_invalid_timestamp(void)
  2831. {
  2832. return ktime_set(0, 0);
  2833. }
  2834. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2835. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2836. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2837. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2838. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2839. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2840. {
  2841. }
  2842. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2843. {
  2844. return false;
  2845. }
  2846. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2847. /**
  2848. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2849. *
  2850. * PHY drivers may accept clones of transmitted packets for
  2851. * timestamping via their phy_driver.txtstamp method. These drivers
  2852. * must call this function to return the skb back to the stack with a
  2853. * timestamp.
  2854. *
  2855. * @skb: clone of the the original outgoing packet
  2856. * @hwtstamps: hardware time stamps
  2857. *
  2858. */
  2859. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2860. struct skb_shared_hwtstamps *hwtstamps);
  2861. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2862. struct skb_shared_hwtstamps *hwtstamps,
  2863. struct sock *sk, int tstype);
  2864. /**
  2865. * skb_tstamp_tx - queue clone of skb with send time stamps
  2866. * @orig_skb: the original outgoing packet
  2867. * @hwtstamps: hardware time stamps, may be NULL if not available
  2868. *
  2869. * If the skb has a socket associated, then this function clones the
  2870. * skb (thus sharing the actual data and optional structures), stores
  2871. * the optional hardware time stamping information (if non NULL) or
  2872. * generates a software time stamp (otherwise), then queues the clone
  2873. * to the error queue of the socket. Errors are silently ignored.
  2874. */
  2875. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2876. struct skb_shared_hwtstamps *hwtstamps);
  2877. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2878. {
  2879. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2880. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2881. skb_tstamp_tx(skb, NULL);
  2882. }
  2883. /**
  2884. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2885. *
  2886. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2887. * function immediately before giving the sk_buff to the MAC hardware.
  2888. *
  2889. * Specifically, one should make absolutely sure that this function is
  2890. * called before TX completion of this packet can trigger. Otherwise
  2891. * the packet could potentially already be freed.
  2892. *
  2893. * @skb: A socket buffer.
  2894. */
  2895. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2896. {
  2897. skb_clone_tx_timestamp(skb);
  2898. sw_tx_timestamp(skb);
  2899. }
  2900. /**
  2901. * skb_complete_wifi_ack - deliver skb with wifi status
  2902. *
  2903. * @skb: the original outgoing packet
  2904. * @acked: ack status
  2905. *
  2906. */
  2907. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2908. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2909. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2910. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2911. {
  2912. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2913. skb->csum_valid ||
  2914. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2915. skb_checksum_start_offset(skb) >= 0));
  2916. }
  2917. /**
  2918. * skb_checksum_complete - Calculate checksum of an entire packet
  2919. * @skb: packet to process
  2920. *
  2921. * This function calculates the checksum over the entire packet plus
  2922. * the value of skb->csum. The latter can be used to supply the
  2923. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2924. * checksum.
  2925. *
  2926. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2927. * this function can be used to verify that checksum on received
  2928. * packets. In that case the function should return zero if the
  2929. * checksum is correct. In particular, this function will return zero
  2930. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2931. * hardware has already verified the correctness of the checksum.
  2932. */
  2933. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2934. {
  2935. return skb_csum_unnecessary(skb) ?
  2936. 0 : __skb_checksum_complete(skb);
  2937. }
  2938. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2939. {
  2940. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2941. if (skb->csum_level == 0)
  2942. skb->ip_summed = CHECKSUM_NONE;
  2943. else
  2944. skb->csum_level--;
  2945. }
  2946. }
  2947. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2948. {
  2949. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2950. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2951. skb->csum_level++;
  2952. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2953. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2954. skb->csum_level = 0;
  2955. }
  2956. }
  2957. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2958. {
  2959. /* Mark current checksum as bad (typically called from GRO
  2960. * path). In the case that ip_summed is CHECKSUM_NONE
  2961. * this must be the first checksum encountered in the packet.
  2962. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2963. * checksum after the last one validated. For UDP, a zero
  2964. * checksum can not be marked as bad.
  2965. */
  2966. if (skb->ip_summed == CHECKSUM_NONE ||
  2967. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2968. skb->csum_bad = 1;
  2969. }
  2970. /* Check if we need to perform checksum complete validation.
  2971. *
  2972. * Returns true if checksum complete is needed, false otherwise
  2973. * (either checksum is unnecessary or zero checksum is allowed).
  2974. */
  2975. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2976. bool zero_okay,
  2977. __sum16 check)
  2978. {
  2979. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2980. skb->csum_valid = 1;
  2981. __skb_decr_checksum_unnecessary(skb);
  2982. return false;
  2983. }
  2984. return true;
  2985. }
  2986. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2987. * in checksum_init.
  2988. */
  2989. #define CHECKSUM_BREAK 76
  2990. /* Unset checksum-complete
  2991. *
  2992. * Unset checksum complete can be done when packet is being modified
  2993. * (uncompressed for instance) and checksum-complete value is
  2994. * invalidated.
  2995. */
  2996. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2997. {
  2998. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2999. skb->ip_summed = CHECKSUM_NONE;
  3000. }
  3001. /* Validate (init) checksum based on checksum complete.
  3002. *
  3003. * Return values:
  3004. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  3005. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  3006. * checksum is stored in skb->csum for use in __skb_checksum_complete
  3007. * non-zero: value of invalid checksum
  3008. *
  3009. */
  3010. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  3011. bool complete,
  3012. __wsum psum)
  3013. {
  3014. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  3015. if (!csum_fold(csum_add(psum, skb->csum))) {
  3016. skb->csum_valid = 1;
  3017. return 0;
  3018. }
  3019. } else if (skb->csum_bad) {
  3020. /* ip_summed == CHECKSUM_NONE in this case */
  3021. return (__force __sum16)1;
  3022. }
  3023. skb->csum = psum;
  3024. if (complete || skb->len <= CHECKSUM_BREAK) {
  3025. __sum16 csum;
  3026. csum = __skb_checksum_complete(skb);
  3027. skb->csum_valid = !csum;
  3028. return csum;
  3029. }
  3030. return 0;
  3031. }
  3032. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  3033. {
  3034. return 0;
  3035. }
  3036. /* Perform checksum validate (init). Note that this is a macro since we only
  3037. * want to calculate the pseudo header which is an input function if necessary.
  3038. * First we try to validate without any computation (checksum unnecessary) and
  3039. * then calculate based on checksum complete calling the function to compute
  3040. * pseudo header.
  3041. *
  3042. * Return values:
  3043. * 0: checksum is validated or try to in skb_checksum_complete
  3044. * non-zero: value of invalid checksum
  3045. */
  3046. #define __skb_checksum_validate(skb, proto, complete, \
  3047. zero_okay, check, compute_pseudo) \
  3048. ({ \
  3049. __sum16 __ret = 0; \
  3050. skb->csum_valid = 0; \
  3051. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  3052. __ret = __skb_checksum_validate_complete(skb, \
  3053. complete, compute_pseudo(skb, proto)); \
  3054. __ret; \
  3055. })
  3056. #define skb_checksum_init(skb, proto, compute_pseudo) \
  3057. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  3058. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  3059. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  3060. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  3061. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  3062. #define skb_checksum_validate_zero_check(skb, proto, check, \
  3063. compute_pseudo) \
  3064. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  3065. #define skb_checksum_simple_validate(skb) \
  3066. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  3067. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  3068. {
  3069. return (skb->ip_summed == CHECKSUM_NONE &&
  3070. skb->csum_valid && !skb->csum_bad);
  3071. }
  3072. static inline void __skb_checksum_convert(struct sk_buff *skb,
  3073. __sum16 check, __wsum pseudo)
  3074. {
  3075. skb->csum = ~pseudo;
  3076. skb->ip_summed = CHECKSUM_COMPLETE;
  3077. }
  3078. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  3079. do { \
  3080. if (__skb_checksum_convert_check(skb)) \
  3081. __skb_checksum_convert(skb, check, \
  3082. compute_pseudo(skb, proto)); \
  3083. } while (0)
  3084. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  3085. u16 start, u16 offset)
  3086. {
  3087. skb->ip_summed = CHECKSUM_PARTIAL;
  3088. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  3089. skb->csum_offset = offset - start;
  3090. }
  3091. /* Update skbuf and packet to reflect the remote checksum offload operation.
  3092. * When called, ptr indicates the starting point for skb->csum when
  3093. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  3094. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  3095. */
  3096. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  3097. int start, int offset, bool nopartial)
  3098. {
  3099. __wsum delta;
  3100. if (!nopartial) {
  3101. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  3102. return;
  3103. }
  3104. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  3105. __skb_checksum_complete(skb);
  3106. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  3107. }
  3108. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  3109. /* Adjust skb->csum since we changed the packet */
  3110. skb->csum = csum_add(skb->csum, delta);
  3111. }
  3112. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3113. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  3114. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  3115. {
  3116. if (nfct && atomic_dec_and_test(&nfct->use))
  3117. nf_conntrack_destroy(nfct);
  3118. }
  3119. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  3120. {
  3121. if (nfct)
  3122. atomic_inc(&nfct->use);
  3123. }
  3124. #endif
  3125. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3126. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  3127. {
  3128. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  3129. kfree(nf_bridge);
  3130. }
  3131. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  3132. {
  3133. if (nf_bridge)
  3134. atomic_inc(&nf_bridge->use);
  3135. }
  3136. #endif /* CONFIG_BRIDGE_NETFILTER */
  3137. static inline void nf_reset(struct sk_buff *skb)
  3138. {
  3139. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3140. nf_conntrack_put(skb->nfct);
  3141. skb->nfct = NULL;
  3142. #endif
  3143. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3144. nf_bridge_put(skb->nf_bridge);
  3145. skb->nf_bridge = NULL;
  3146. #endif
  3147. }
  3148. static inline void nf_reset_trace(struct sk_buff *skb)
  3149. {
  3150. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3151. skb->nf_trace = 0;
  3152. #endif
  3153. }
  3154. static inline void ipvs_reset(struct sk_buff *skb)
  3155. {
  3156. #if IS_ENABLED(CONFIG_IP_VS)
  3157. skb->ipvs_property = 0;
  3158. #endif
  3159. }
  3160. /* Note: This doesn't put any conntrack and bridge info in dst. */
  3161. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  3162. bool copy)
  3163. {
  3164. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3165. dst->nfct = src->nfct;
  3166. nf_conntrack_get(src->nfct);
  3167. if (copy)
  3168. dst->nfctinfo = src->nfctinfo;
  3169. #endif
  3170. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3171. dst->nf_bridge = src->nf_bridge;
  3172. nf_bridge_get(src->nf_bridge);
  3173. #endif
  3174. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3175. if (copy)
  3176. dst->nf_trace = src->nf_trace;
  3177. #endif
  3178. }
  3179. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  3180. {
  3181. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3182. nf_conntrack_put(dst->nfct);
  3183. #endif
  3184. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3185. nf_bridge_put(dst->nf_bridge);
  3186. #endif
  3187. __nf_copy(dst, src, true);
  3188. }
  3189. #ifdef CONFIG_NETWORK_SECMARK
  3190. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3191. {
  3192. to->secmark = from->secmark;
  3193. }
  3194. static inline void skb_init_secmark(struct sk_buff *skb)
  3195. {
  3196. skb->secmark = 0;
  3197. }
  3198. #else
  3199. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3200. { }
  3201. static inline void skb_init_secmark(struct sk_buff *skb)
  3202. { }
  3203. #endif
  3204. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  3205. {
  3206. return !skb->destructor &&
  3207. #if IS_ENABLED(CONFIG_XFRM)
  3208. !skb->sp &&
  3209. #endif
  3210. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  3211. !skb->nfct &&
  3212. #endif
  3213. !skb->_skb_refdst &&
  3214. !skb_has_frag_list(skb);
  3215. }
  3216. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  3217. {
  3218. skb->queue_mapping = queue_mapping;
  3219. }
  3220. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  3221. {
  3222. return skb->queue_mapping;
  3223. }
  3224. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  3225. {
  3226. to->queue_mapping = from->queue_mapping;
  3227. }
  3228. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  3229. {
  3230. skb->queue_mapping = rx_queue + 1;
  3231. }
  3232. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  3233. {
  3234. return skb->queue_mapping - 1;
  3235. }
  3236. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  3237. {
  3238. return skb->queue_mapping != 0;
  3239. }
  3240. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  3241. {
  3242. #ifdef CONFIG_XFRM
  3243. return skb->sp;
  3244. #else
  3245. return NULL;
  3246. #endif
  3247. }
  3248. /* Keeps track of mac header offset relative to skb->head.
  3249. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  3250. * For non-tunnel skb it points to skb_mac_header() and for
  3251. * tunnel skb it points to outer mac header.
  3252. * Keeps track of level of encapsulation of network headers.
  3253. */
  3254. struct skb_gso_cb {
  3255. union {
  3256. int mac_offset;
  3257. int data_offset;
  3258. };
  3259. int encap_level;
  3260. __wsum csum;
  3261. __u16 csum_start;
  3262. };
  3263. #define SKB_SGO_CB_OFFSET 32
  3264. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
  3265. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  3266. {
  3267. return (skb_mac_header(inner_skb) - inner_skb->head) -
  3268. SKB_GSO_CB(inner_skb)->mac_offset;
  3269. }
  3270. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  3271. {
  3272. int new_headroom, headroom;
  3273. int ret;
  3274. headroom = skb_headroom(skb);
  3275. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  3276. if (ret)
  3277. return ret;
  3278. new_headroom = skb_headroom(skb);
  3279. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  3280. return 0;
  3281. }
  3282. static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
  3283. {
  3284. /* Do not update partial checksums if remote checksum is enabled. */
  3285. if (skb->remcsum_offload)
  3286. return;
  3287. SKB_GSO_CB(skb)->csum = res;
  3288. SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
  3289. }
  3290. /* Compute the checksum for a gso segment. First compute the checksum value
  3291. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  3292. * then add in skb->csum (checksum from csum_start to end of packet).
  3293. * skb->csum and csum_start are then updated to reflect the checksum of the
  3294. * resultant packet starting from the transport header-- the resultant checksum
  3295. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  3296. * header.
  3297. */
  3298. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  3299. {
  3300. unsigned char *csum_start = skb_transport_header(skb);
  3301. int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
  3302. __wsum partial = SKB_GSO_CB(skb)->csum;
  3303. SKB_GSO_CB(skb)->csum = res;
  3304. SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
  3305. return csum_fold(csum_partial(csum_start, plen, partial));
  3306. }
  3307. static inline bool skb_is_gso(const struct sk_buff *skb)
  3308. {
  3309. return skb_shinfo(skb)->gso_size;
  3310. }
  3311. /* Note: Should be called only if skb_is_gso(skb) is true */
  3312. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  3313. {
  3314. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  3315. }
  3316. static inline void skb_gso_reset(struct sk_buff *skb)
  3317. {
  3318. skb_shinfo(skb)->gso_size = 0;
  3319. skb_shinfo(skb)->gso_segs = 0;
  3320. skb_shinfo(skb)->gso_type = 0;
  3321. }
  3322. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  3323. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  3324. {
  3325. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  3326. * wanted then gso_type will be set. */
  3327. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  3328. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  3329. unlikely(shinfo->gso_type == 0)) {
  3330. __skb_warn_lro_forwarding(skb);
  3331. return true;
  3332. }
  3333. return false;
  3334. }
  3335. static inline void skb_forward_csum(struct sk_buff *skb)
  3336. {
  3337. /* Unfortunately we don't support this one. Any brave souls? */
  3338. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3339. skb->ip_summed = CHECKSUM_NONE;
  3340. }
  3341. /**
  3342. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3343. * @skb: skb to check
  3344. *
  3345. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3346. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3347. * use this helper, to document places where we make this assertion.
  3348. */
  3349. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3350. {
  3351. #ifdef DEBUG
  3352. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3353. #endif
  3354. }
  3355. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3356. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3357. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3358. unsigned int transport_len,
  3359. __sum16(*skb_chkf)(struct sk_buff *skb));
  3360. /**
  3361. * skb_head_is_locked - Determine if the skb->head is locked down
  3362. * @skb: skb to check
  3363. *
  3364. * The head on skbs build around a head frag can be removed if they are
  3365. * not cloned. This function returns true if the skb head is locked down
  3366. * due to either being allocated via kmalloc, or by being a clone with
  3367. * multiple references to the head.
  3368. */
  3369. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3370. {
  3371. return !skb->head_frag || skb_cloned(skb);
  3372. }
  3373. /**
  3374. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3375. *
  3376. * @skb: GSO skb
  3377. *
  3378. * skb_gso_network_seglen is used to determine the real size of the
  3379. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3380. *
  3381. * The MAC/L2 header is not accounted for.
  3382. */
  3383. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3384. {
  3385. unsigned int hdr_len = skb_transport_header(skb) -
  3386. skb_network_header(skb);
  3387. return hdr_len + skb_gso_transport_seglen(skb);
  3388. }
  3389. /* Local Checksum Offload.
  3390. * Compute outer checksum based on the assumption that the
  3391. * inner checksum will be offloaded later.
  3392. * See Documentation/networking/checksum-offloads.txt for
  3393. * explanation of how this works.
  3394. * Fill in outer checksum adjustment (e.g. with sum of outer
  3395. * pseudo-header) before calling.
  3396. * Also ensure that inner checksum is in linear data area.
  3397. */
  3398. static inline __wsum lco_csum(struct sk_buff *skb)
  3399. {
  3400. unsigned char *csum_start = skb_checksum_start(skb);
  3401. unsigned char *l4_hdr = skb_transport_header(skb);
  3402. __wsum partial;
  3403. /* Start with complement of inner checksum adjustment */
  3404. partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
  3405. skb->csum_offset));
  3406. /* Add in checksum of our headers (incl. outer checksum
  3407. * adjustment filled in by caller) and return result.
  3408. */
  3409. return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
  3410. }
  3411. #endif /* __KERNEL__ */
  3412. #endif /* _LINUX_SKBUFF_H */