skbuff.h 116 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963296429652966296729682969297029712972297329742975297629772978297929802981298229832984298529862987298829892990299129922993299429952996299729982999300030013002300330043005300630073008300930103011301230133014301530163017301830193020302130223023302430253026302730283029303030313032303330343035303630373038303930403041304230433044304530463047304830493050305130523053305430553056305730583059306030613062306330643065306630673068306930703071307230733074307530763077307830793080308130823083308430853086308730883089309030913092309330943095309630973098309931003101310231033104310531063107310831093110311131123113311431153116311731183119312031213122312331243125312631273128312931303131313231333134313531363137313831393140314131423143314431453146314731483149315031513152315331543155315631573158315931603161316231633164316531663167316831693170317131723173317431753176317731783179318031813182318331843185318631873188318931903191319231933194319531963197319831993200320132023203320432053206320732083209321032113212321332143215321632173218321932203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278327932803281328232833284328532863287328832893290329132923293329432953296329732983299330033013302330333043305330633073308330933103311331233133314331533163317331833193320332133223323332433253326332733283329333033313332333333343335333633373338333933403341334233433344334533463347334833493350335133523353335433553356335733583359336033613362336333643365336633673368336933703371337233733374337533763377337833793380338133823383338433853386338733883389339033913392339333943395339633973398339934003401340234033404340534063407340834093410341134123413341434153416341734183419342034213422342334243425342634273428342934303431343234333434343534363437343834393440344134423443344434453446344734483449345034513452345334543455345634573458345934603461346234633464346534663467346834693470347134723473347434753476347734783479348034813482348334843485348634873488348934903491349234933494349534963497349834993500350135023503350435053506350735083509351035113512351335143515351635173518351935203521352235233524352535263527352835293530353135323533353435353536353735383539354035413542354335443545354635473548354935503551355235533554355535563557355835593560356135623563356435653566356735683569357035713572357335743575357635773578357935803581358235833584358535863587358835893590359135923593359435953596359735983599360036013602360336043605360636073608360936103611361236133614361536163617361836193620362136223623362436253626362736283629363036313632363336343635363636373638363936403641364236433644364536463647364836493650365136523653365436553656365736583659366036613662366336643665366636673668366936703671367236733674367536763677367836793680368136823683368436853686368736883689369036913692369336943695369636973698369937003701370237033704370537063707370837093710371137123713371437153716371737183719372037213722372337243725372637273728372937303731373237333734373537363737373837393740374137423743374437453746374737483749375037513752375337543755375637573758375937603761376237633764376537663767376837693770377137723773377437753776377737783779378037813782378337843785378637873788378937903791379237933794379537963797379837993800380138023803380438053806380738083809381038113812381338143815381638173818381938203821382238233824382538263827382838293830383138323833383438353836383738383839384038413842384338443845384638473848384938503851385238533854385538563857385838593860386138623863386438653866386738683869387038713872387338743875387638773878387938803881388238833884388538863887388838893890389138923893389438953896389738983899390039013902390339043905390639073908390939103911391239133914391539163917391839193920392139223923392439253926392739283929393039313932393339343935393639373938393939403941394239433944394539463947394839493950395139523953395439553956395739583959396039613962396339643965396639673968396939703971397239733974397539763977397839793980398139823983398439853986398739883989399039913992399339943995399639973998399940004001400240034004400540064007400840094010401140124013401440154016401740184019402040214022402340244025402640274028402940304031403240334034403540364037403840394040404140424043404440454046404740484049405040514052405340544055405640574058405940604061406240634064406540664067406840694070407140724073407440754076
  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/refcount.h>
  24. #include <linux/atomic.h>
  25. #include <asm/types.h>
  26. #include <linux/spinlock.h>
  27. #include <linux/net.h>
  28. #include <linux/textsearch.h>
  29. #include <net/checksum.h>
  30. #include <linux/rcupdate.h>
  31. #include <linux/hrtimer.h>
  32. #include <linux/dma-mapping.h>
  33. #include <linux/netdev_features.h>
  34. #include <linux/sched.h>
  35. #include <linux/sched/clock.h>
  36. #include <net/flow_dissector.h>
  37. #include <linux/splice.h>
  38. #include <linux/in6.h>
  39. #include <linux/if_packet.h>
  40. #include <net/flow.h>
  41. /* The interface for checksum offload between the stack and networking drivers
  42. * is as follows...
  43. *
  44. * A. IP checksum related features
  45. *
  46. * Drivers advertise checksum offload capabilities in the features of a device.
  47. * From the stack's point of view these are capabilities offered by the driver,
  48. * a driver typically only advertises features that it is capable of offloading
  49. * to its device.
  50. *
  51. * The checksum related features are:
  52. *
  53. * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
  54. * IP (one's complement) checksum for any combination
  55. * of protocols or protocol layering. The checksum is
  56. * computed and set in a packet per the CHECKSUM_PARTIAL
  57. * interface (see below).
  58. *
  59. * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
  60. * TCP or UDP packets over IPv4. These are specifically
  61. * unencapsulated packets of the form IPv4|TCP or
  62. * IPv4|UDP where the Protocol field in the IPv4 header
  63. * is TCP or UDP. The IPv4 header may contain IP options
  64. * This feature cannot be set in features for a device
  65. * with NETIF_F_HW_CSUM also set. This feature is being
  66. * DEPRECATED (see below).
  67. *
  68. * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
  69. * TCP or UDP packets over IPv6. These are specifically
  70. * unencapsulated packets of the form IPv6|TCP or
  71. * IPv4|UDP where the Next Header field in the IPv6
  72. * header is either TCP or UDP. IPv6 extension headers
  73. * are not supported with this feature. This feature
  74. * cannot be set in features for a device with
  75. * NETIF_F_HW_CSUM also set. This feature is being
  76. * DEPRECATED (see below).
  77. *
  78. * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
  79. * This flag is used only used to disable the RX checksum
  80. * feature for a device. The stack will accept receive
  81. * checksum indication in packets received on a device
  82. * regardless of whether NETIF_F_RXCSUM is set.
  83. *
  84. * B. Checksumming of received packets by device. Indication of checksum
  85. * verification is in set skb->ip_summed. Possible values are:
  86. *
  87. * CHECKSUM_NONE:
  88. *
  89. * Device did not checksum this packet e.g. due to lack of capabilities.
  90. * The packet contains full (though not verified) checksum in packet but
  91. * not in skb->csum. Thus, skb->csum is undefined in this case.
  92. *
  93. * CHECKSUM_UNNECESSARY:
  94. *
  95. * The hardware you're dealing with doesn't calculate the full checksum
  96. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  97. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  98. * if their checksums are okay. skb->csum is still undefined in this case
  99. * though. A driver or device must never modify the checksum field in the
  100. * packet even if checksum is verified.
  101. *
  102. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  103. * TCP: IPv6 and IPv4.
  104. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  105. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  106. * may perform further validation in this case.
  107. * GRE: only if the checksum is present in the header.
  108. * SCTP: indicates the CRC in SCTP header has been validated.
  109. * FCOE: indicates the CRC in FC frame has been validated.
  110. *
  111. * skb->csum_level indicates the number of consecutive checksums found in
  112. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  113. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  114. * and a device is able to verify the checksums for UDP (possibly zero),
  115. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  116. * two. If the device were only able to verify the UDP checksum and not
  117. * GRE, either because it doesn't support GRE checksum of because GRE
  118. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  119. * not considered in this case).
  120. *
  121. * CHECKSUM_COMPLETE:
  122. *
  123. * This is the most generic way. The device supplied checksum of the _whole_
  124. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  125. * hardware doesn't need to parse L3/L4 headers to implement this.
  126. *
  127. * Notes:
  128. * - Even if device supports only some protocols, but is able to produce
  129. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  130. * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
  131. *
  132. * CHECKSUM_PARTIAL:
  133. *
  134. * A checksum is set up to be offloaded to a device as described in the
  135. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  136. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  137. * on the same host, or it may be set in the input path in GRO or remote
  138. * checksum offload. For the purposes of checksum verification, the checksum
  139. * referred to by skb->csum_start + skb->csum_offset and any preceding
  140. * checksums in the packet are considered verified. Any checksums in the
  141. * packet that are after the checksum being offloaded are not considered to
  142. * be verified.
  143. *
  144. * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
  145. * in the skb->ip_summed for a packet. Values are:
  146. *
  147. * CHECKSUM_PARTIAL:
  148. *
  149. * The driver is required to checksum the packet as seen by hard_start_xmit()
  150. * from skb->csum_start up to the end, and to record/write the checksum at
  151. * offset skb->csum_start + skb->csum_offset. A driver may verify that the
  152. * csum_start and csum_offset values are valid values given the length and
  153. * offset of the packet, however they should not attempt to validate that the
  154. * checksum refers to a legitimate transport layer checksum-- it is the
  155. * purview of the stack to validate that csum_start and csum_offset are set
  156. * correctly.
  157. *
  158. * When the stack requests checksum offload for a packet, the driver MUST
  159. * ensure that the checksum is set correctly. A driver can either offload the
  160. * checksum calculation to the device, or call skb_checksum_help (in the case
  161. * that the device does not support offload for a particular checksum).
  162. *
  163. * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
  164. * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
  165. * checksum offload capability.
  166. * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
  167. * on network device checksumming capabilities: if a packet does not match
  168. * them, skb_checksum_help or skb_crc32c_help (depending on the value of
  169. * csum_not_inet, see item D.) is called to resolve the checksum.
  170. *
  171. * CHECKSUM_NONE:
  172. *
  173. * The skb was already checksummed by the protocol, or a checksum is not
  174. * required.
  175. *
  176. * CHECKSUM_UNNECESSARY:
  177. *
  178. * This has the same meaning on as CHECKSUM_NONE for checksum offload on
  179. * output.
  180. *
  181. * CHECKSUM_COMPLETE:
  182. * Not used in checksum output. If a driver observes a packet with this value
  183. * set in skbuff, if should treat as CHECKSUM_NONE being set.
  184. *
  185. * D. Non-IP checksum (CRC) offloads
  186. *
  187. * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
  188. * offloading the SCTP CRC in a packet. To perform this offload the stack
  189. * will set set csum_start and csum_offset accordingly, set ip_summed to
  190. * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
  191. * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
  192. * A driver that supports both IP checksum offload and SCTP CRC32c offload
  193. * must verify which offload is configured for a packet by testing the
  194. * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
  195. * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
  196. *
  197. * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
  198. * offloading the FCOE CRC in a packet. To perform this offload the stack
  199. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  200. * accordingly. Note the there is no indication in the skbuff that the
  201. * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
  202. * both IP checksum offload and FCOE CRC offload must verify which offload
  203. * is configured for a packet presumably by inspecting packet headers.
  204. *
  205. * E. Checksumming on output with GSO.
  206. *
  207. * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
  208. * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
  209. * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
  210. * part of the GSO operation is implied. If a checksum is being offloaded
  211. * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
  212. * are set to refer to the outermost checksum being offload (two offloaded
  213. * checksums are possible with UDP encapsulation).
  214. */
  215. /* Don't change this without changing skb_csum_unnecessary! */
  216. #define CHECKSUM_NONE 0
  217. #define CHECKSUM_UNNECESSARY 1
  218. #define CHECKSUM_COMPLETE 2
  219. #define CHECKSUM_PARTIAL 3
  220. /* Maximum value in skb->csum_level */
  221. #define SKB_MAX_CSUM_LEVEL 3
  222. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  223. #define SKB_WITH_OVERHEAD(X) \
  224. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  225. #define SKB_MAX_ORDER(X, ORDER) \
  226. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  227. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  228. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  229. /* return minimum truesize of one skb containing X bytes of data */
  230. #define SKB_TRUESIZE(X) ((X) + \
  231. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  232. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  233. struct net_device;
  234. struct scatterlist;
  235. struct pipe_inode_info;
  236. struct iov_iter;
  237. struct napi_struct;
  238. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  239. struct nf_conntrack {
  240. atomic_t use;
  241. };
  242. #endif
  243. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  244. struct nf_bridge_info {
  245. refcount_t use;
  246. enum {
  247. BRNF_PROTO_UNCHANGED,
  248. BRNF_PROTO_8021Q,
  249. BRNF_PROTO_PPPOE
  250. } orig_proto:8;
  251. u8 pkt_otherhost:1;
  252. u8 in_prerouting:1;
  253. u8 bridged_dnat:1;
  254. __u16 frag_max_size;
  255. struct net_device *physindev;
  256. /* always valid & non-NULL from FORWARD on, for physdev match */
  257. struct net_device *physoutdev;
  258. union {
  259. /* prerouting: detect dnat in orig/reply direction */
  260. __be32 ipv4_daddr;
  261. struct in6_addr ipv6_daddr;
  262. /* after prerouting + nat detected: store original source
  263. * mac since neigh resolution overwrites it, only used while
  264. * skb is out in neigh layer.
  265. */
  266. char neigh_header[8];
  267. };
  268. };
  269. #endif
  270. struct sk_buff_head {
  271. /* These two members must be first. */
  272. struct sk_buff *next;
  273. struct sk_buff *prev;
  274. __u32 qlen;
  275. spinlock_t lock;
  276. };
  277. struct sk_buff;
  278. /* To allow 64K frame to be packed as single skb without frag_list we
  279. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  280. * buffers which do not start on a page boundary.
  281. *
  282. * Since GRO uses frags we allocate at least 16 regardless of page
  283. * size.
  284. */
  285. #if (65536/PAGE_SIZE + 1) < 16
  286. #define MAX_SKB_FRAGS 16UL
  287. #else
  288. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  289. #endif
  290. extern int sysctl_max_skb_frags;
  291. /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
  292. * segment using its current segmentation instead.
  293. */
  294. #define GSO_BY_FRAGS 0xFFFF
  295. typedef struct skb_frag_struct skb_frag_t;
  296. struct skb_frag_struct {
  297. struct {
  298. struct page *p;
  299. } page;
  300. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  301. __u32 page_offset;
  302. __u32 size;
  303. #else
  304. __u16 page_offset;
  305. __u16 size;
  306. #endif
  307. };
  308. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  309. {
  310. return frag->size;
  311. }
  312. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  313. {
  314. frag->size = size;
  315. }
  316. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  317. {
  318. frag->size += delta;
  319. }
  320. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  321. {
  322. frag->size -= delta;
  323. }
  324. static inline bool skb_frag_must_loop(struct page *p)
  325. {
  326. #if defined(CONFIG_HIGHMEM)
  327. if (PageHighMem(p))
  328. return true;
  329. #endif
  330. return false;
  331. }
  332. /**
  333. * skb_frag_foreach_page - loop over pages in a fragment
  334. *
  335. * @f: skb frag to operate on
  336. * @f_off: offset from start of f->page.p
  337. * @f_len: length from f_off to loop over
  338. * @p: (temp var) current page
  339. * @p_off: (temp var) offset from start of current page,
  340. * non-zero only on first page.
  341. * @p_len: (temp var) length in current page,
  342. * < PAGE_SIZE only on first and last page.
  343. * @copied: (temp var) length so far, excluding current p_len.
  344. *
  345. * A fragment can hold a compound page, in which case per-page
  346. * operations, notably kmap_atomic, must be called for each
  347. * regular page.
  348. */
  349. #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
  350. for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
  351. p_off = (f_off) & (PAGE_SIZE - 1), \
  352. p_len = skb_frag_must_loop(p) ? \
  353. min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
  354. copied = 0; \
  355. copied < f_len; \
  356. copied += p_len, p++, p_off = 0, \
  357. p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
  358. #define HAVE_HW_TIME_STAMP
  359. /**
  360. * struct skb_shared_hwtstamps - hardware time stamps
  361. * @hwtstamp: hardware time stamp transformed into duration
  362. * since arbitrary point in time
  363. *
  364. * Software time stamps generated by ktime_get_real() are stored in
  365. * skb->tstamp.
  366. *
  367. * hwtstamps can only be compared against other hwtstamps from
  368. * the same device.
  369. *
  370. * This structure is attached to packets as part of the
  371. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  372. */
  373. struct skb_shared_hwtstamps {
  374. ktime_t hwtstamp;
  375. };
  376. /* Definitions for tx_flags in struct skb_shared_info */
  377. enum {
  378. /* generate hardware time stamp */
  379. SKBTX_HW_TSTAMP = 1 << 0,
  380. /* generate software time stamp when queueing packet to NIC */
  381. SKBTX_SW_TSTAMP = 1 << 1,
  382. /* device driver is going to provide hardware time stamp */
  383. SKBTX_IN_PROGRESS = 1 << 2,
  384. /* device driver supports TX zero-copy buffers */
  385. SKBTX_DEV_ZEROCOPY = 1 << 3,
  386. /* generate wifi status information (where possible) */
  387. SKBTX_WIFI_STATUS = 1 << 4,
  388. /* This indicates at least one fragment might be overwritten
  389. * (as in vmsplice(), sendfile() ...)
  390. * If we need to compute a TX checksum, we'll need to copy
  391. * all frags to avoid possible bad checksum
  392. */
  393. SKBTX_SHARED_FRAG = 1 << 5,
  394. /* generate software time stamp when entering packet scheduling */
  395. SKBTX_SCHED_TSTAMP = 1 << 6,
  396. };
  397. #define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
  398. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  399. SKBTX_SCHED_TSTAMP)
  400. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  401. /*
  402. * The callback notifies userspace to release buffers when skb DMA is done in
  403. * lower device, the skb last reference should be 0 when calling this.
  404. * The zerocopy_success argument is true if zero copy transmit occurred,
  405. * false on data copy or out of memory error caused by data copy attempt.
  406. * The ctx field is used to track device context.
  407. * The desc field is used to track userspace buffer index.
  408. */
  409. struct ubuf_info {
  410. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  411. union {
  412. struct {
  413. unsigned long desc;
  414. void *ctx;
  415. };
  416. struct {
  417. u32 id;
  418. u16 len;
  419. u16 zerocopy:1;
  420. u32 bytelen;
  421. };
  422. };
  423. refcount_t refcnt;
  424. struct mmpin {
  425. struct user_struct *user;
  426. unsigned int num_pg;
  427. } mmp;
  428. };
  429. #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
  430. struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
  431. struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
  432. struct ubuf_info *uarg);
  433. static inline void sock_zerocopy_get(struct ubuf_info *uarg)
  434. {
  435. refcount_inc(&uarg->refcnt);
  436. }
  437. void sock_zerocopy_put(struct ubuf_info *uarg);
  438. void sock_zerocopy_put_abort(struct ubuf_info *uarg);
  439. void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
  440. int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
  441. struct msghdr *msg, int len,
  442. struct ubuf_info *uarg);
  443. /* This data is invariant across clones and lives at
  444. * the end of the header data, ie. at skb->end.
  445. */
  446. struct skb_shared_info {
  447. unsigned short _unused;
  448. unsigned char nr_frags;
  449. __u8 tx_flags;
  450. unsigned short gso_size;
  451. /* Warning: this field is not always filled in (UFO)! */
  452. unsigned short gso_segs;
  453. struct sk_buff *frag_list;
  454. struct skb_shared_hwtstamps hwtstamps;
  455. unsigned int gso_type;
  456. u32 tskey;
  457. __be32 ip6_frag_id;
  458. /*
  459. * Warning : all fields before dataref are cleared in __alloc_skb()
  460. */
  461. atomic_t dataref;
  462. /* Intermediate layers must ensure that destructor_arg
  463. * remains valid until skb destructor */
  464. void * destructor_arg;
  465. /* must be last field, see pskb_expand_head() */
  466. skb_frag_t frags[MAX_SKB_FRAGS];
  467. };
  468. /* We divide dataref into two halves. The higher 16 bits hold references
  469. * to the payload part of skb->data. The lower 16 bits hold references to
  470. * the entire skb->data. A clone of a headerless skb holds the length of
  471. * the header in skb->hdr_len.
  472. *
  473. * All users must obey the rule that the skb->data reference count must be
  474. * greater than or equal to the payload reference count.
  475. *
  476. * Holding a reference to the payload part means that the user does not
  477. * care about modifications to the header part of skb->data.
  478. */
  479. #define SKB_DATAREF_SHIFT 16
  480. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  481. enum {
  482. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  483. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  484. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  485. };
  486. enum {
  487. SKB_GSO_TCPV4 = 1 << 0,
  488. /* This indicates the skb is from an untrusted source. */
  489. SKB_GSO_DODGY = 1 << 1,
  490. /* This indicates the tcp segment has CWR set. */
  491. SKB_GSO_TCP_ECN = 1 << 2,
  492. SKB_GSO_TCP_FIXEDID = 1 << 3,
  493. SKB_GSO_TCPV6 = 1 << 4,
  494. SKB_GSO_FCOE = 1 << 5,
  495. SKB_GSO_GRE = 1 << 6,
  496. SKB_GSO_GRE_CSUM = 1 << 7,
  497. SKB_GSO_IPXIP4 = 1 << 8,
  498. SKB_GSO_IPXIP6 = 1 << 9,
  499. SKB_GSO_UDP_TUNNEL = 1 << 10,
  500. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  501. SKB_GSO_PARTIAL = 1 << 12,
  502. SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
  503. SKB_GSO_SCTP = 1 << 14,
  504. SKB_GSO_ESP = 1 << 15,
  505. };
  506. #if BITS_PER_LONG > 32
  507. #define NET_SKBUFF_DATA_USES_OFFSET 1
  508. #endif
  509. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  510. typedef unsigned int sk_buff_data_t;
  511. #else
  512. typedef unsigned char *sk_buff_data_t;
  513. #endif
  514. /**
  515. * struct sk_buff - socket buffer
  516. * @next: Next buffer in list
  517. * @prev: Previous buffer in list
  518. * @tstamp: Time we arrived/left
  519. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  520. * @sk: Socket we are owned by
  521. * @dev: Device we arrived on/are leaving by
  522. * @cb: Control buffer. Free for use by every layer. Put private vars here
  523. * @_skb_refdst: destination entry (with norefcount bit)
  524. * @sp: the security path, used for xfrm
  525. * @len: Length of actual data
  526. * @data_len: Data length
  527. * @mac_len: Length of link layer header
  528. * @hdr_len: writable header length of cloned skb
  529. * @csum: Checksum (must include start/offset pair)
  530. * @csum_start: Offset from skb->head where checksumming should start
  531. * @csum_offset: Offset from csum_start where checksum should be stored
  532. * @priority: Packet queueing priority
  533. * @ignore_df: allow local fragmentation
  534. * @cloned: Head may be cloned (check refcnt to be sure)
  535. * @ip_summed: Driver fed us an IP checksum
  536. * @nohdr: Payload reference only, must not modify header
  537. * @pkt_type: Packet class
  538. * @fclone: skbuff clone status
  539. * @ipvs_property: skbuff is owned by ipvs
  540. * @tc_skip_classify: do not classify packet. set by IFB device
  541. * @tc_at_ingress: used within tc_classify to distinguish in/egress
  542. * @tc_redirected: packet was redirected by a tc action
  543. * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
  544. * @peeked: this packet has been seen already, so stats have been
  545. * done for it, don't do them again
  546. * @nf_trace: netfilter packet trace flag
  547. * @protocol: Packet protocol from driver
  548. * @destructor: Destruct function
  549. * @_nfct: Associated connection, if any (with nfctinfo bits)
  550. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  551. * @skb_iif: ifindex of device we arrived on
  552. * @tc_index: Traffic control index
  553. * @hash: the packet hash
  554. * @queue_mapping: Queue mapping for multiqueue devices
  555. * @xmit_more: More SKBs are pending for this queue
  556. * @ndisc_nodetype: router type (from link layer)
  557. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  558. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  559. * ports.
  560. * @sw_hash: indicates hash was computed in software stack
  561. * @wifi_acked_valid: wifi_acked was set
  562. * @wifi_acked: whether frame was acked on wifi or not
  563. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  564. * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
  565. * @dst_pending_confirm: need to confirm neighbour
  566. * @napi_id: id of the NAPI struct this skb came from
  567. * @secmark: security marking
  568. * @mark: Generic packet mark
  569. * @vlan_proto: vlan encapsulation protocol
  570. * @vlan_tci: vlan tag control information
  571. * @inner_protocol: Protocol (encapsulation)
  572. * @inner_transport_header: Inner transport layer header (encapsulation)
  573. * @inner_network_header: Network layer header (encapsulation)
  574. * @inner_mac_header: Link layer header (encapsulation)
  575. * @transport_header: Transport layer header
  576. * @network_header: Network layer header
  577. * @mac_header: Link layer header
  578. * @tail: Tail pointer
  579. * @end: End pointer
  580. * @head: Head of buffer
  581. * @data: Data head pointer
  582. * @truesize: Buffer size
  583. * @users: User count - see {datagram,tcp}.c
  584. */
  585. struct sk_buff {
  586. union {
  587. struct {
  588. /* These two members must be first. */
  589. struct sk_buff *next;
  590. struct sk_buff *prev;
  591. union {
  592. ktime_t tstamp;
  593. u64 skb_mstamp;
  594. };
  595. };
  596. struct rb_node rbnode; /* used in netem & tcp stack */
  597. };
  598. struct sock *sk;
  599. union {
  600. struct net_device *dev;
  601. /* Some protocols might use this space to store information,
  602. * while device pointer would be NULL.
  603. * UDP receive path is one user.
  604. */
  605. unsigned long dev_scratch;
  606. };
  607. /*
  608. * This is the control buffer. It is free to use for every
  609. * layer. Please put your private variables there. If you
  610. * want to keep them across layers you have to do a skb_clone()
  611. * first. This is owned by whoever has the skb queued ATM.
  612. */
  613. char cb[48] __aligned(8);
  614. unsigned long _skb_refdst;
  615. void (*destructor)(struct sk_buff *skb);
  616. #ifdef CONFIG_XFRM
  617. struct sec_path *sp;
  618. #endif
  619. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  620. unsigned long _nfct;
  621. #endif
  622. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  623. struct nf_bridge_info *nf_bridge;
  624. #endif
  625. unsigned int len,
  626. data_len;
  627. __u16 mac_len,
  628. hdr_len;
  629. /* Following fields are _not_ copied in __copy_skb_header()
  630. * Note that queue_mapping is here mostly to fill a hole.
  631. */
  632. kmemcheck_bitfield_begin(flags1);
  633. __u16 queue_mapping;
  634. /* if you move cloned around you also must adapt those constants */
  635. #ifdef __BIG_ENDIAN_BITFIELD
  636. #define CLONED_MASK (1 << 7)
  637. #else
  638. #define CLONED_MASK 1
  639. #endif
  640. #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
  641. __u8 __cloned_offset[0];
  642. __u8 cloned:1,
  643. nohdr:1,
  644. fclone:2,
  645. peeked:1,
  646. head_frag:1,
  647. xmit_more:1,
  648. __unused:1; /* one bit hole */
  649. kmemcheck_bitfield_end(flags1);
  650. /* fields enclosed in headers_start/headers_end are copied
  651. * using a single memcpy() in __copy_skb_header()
  652. */
  653. /* private: */
  654. __u32 headers_start[0];
  655. /* public: */
  656. /* if you move pkt_type around you also must adapt those constants */
  657. #ifdef __BIG_ENDIAN_BITFIELD
  658. #define PKT_TYPE_MAX (7 << 5)
  659. #else
  660. #define PKT_TYPE_MAX 7
  661. #endif
  662. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  663. __u8 __pkt_type_offset[0];
  664. __u8 pkt_type:3;
  665. __u8 pfmemalloc:1;
  666. __u8 ignore_df:1;
  667. __u8 nf_trace:1;
  668. __u8 ip_summed:2;
  669. __u8 ooo_okay:1;
  670. __u8 l4_hash:1;
  671. __u8 sw_hash:1;
  672. __u8 wifi_acked_valid:1;
  673. __u8 wifi_acked:1;
  674. __u8 no_fcs:1;
  675. /* Indicates the inner headers are valid in the skbuff. */
  676. __u8 encapsulation:1;
  677. __u8 encap_hdr_csum:1;
  678. __u8 csum_valid:1;
  679. __u8 csum_complete_sw:1;
  680. __u8 csum_level:2;
  681. __u8 csum_not_inet:1;
  682. __u8 dst_pending_confirm:1;
  683. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  684. __u8 ndisc_nodetype:2;
  685. #endif
  686. __u8 ipvs_property:1;
  687. __u8 inner_protocol_type:1;
  688. __u8 remcsum_offload:1;
  689. #ifdef CONFIG_NET_SWITCHDEV
  690. __u8 offload_fwd_mark:1;
  691. #endif
  692. #ifdef CONFIG_NET_CLS_ACT
  693. __u8 tc_skip_classify:1;
  694. __u8 tc_at_ingress:1;
  695. __u8 tc_redirected:1;
  696. __u8 tc_from_ingress:1;
  697. #endif
  698. #ifdef CONFIG_NET_SCHED
  699. __u16 tc_index; /* traffic control index */
  700. #endif
  701. union {
  702. __wsum csum;
  703. struct {
  704. __u16 csum_start;
  705. __u16 csum_offset;
  706. };
  707. };
  708. __u32 priority;
  709. int skb_iif;
  710. __u32 hash;
  711. __be16 vlan_proto;
  712. __u16 vlan_tci;
  713. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  714. union {
  715. unsigned int napi_id;
  716. unsigned int sender_cpu;
  717. };
  718. #endif
  719. #ifdef CONFIG_NETWORK_SECMARK
  720. __u32 secmark;
  721. #endif
  722. union {
  723. __u32 mark;
  724. __u32 reserved_tailroom;
  725. };
  726. union {
  727. __be16 inner_protocol;
  728. __u8 inner_ipproto;
  729. };
  730. __u16 inner_transport_header;
  731. __u16 inner_network_header;
  732. __u16 inner_mac_header;
  733. __be16 protocol;
  734. __u16 transport_header;
  735. __u16 network_header;
  736. __u16 mac_header;
  737. /* private: */
  738. __u32 headers_end[0];
  739. /* public: */
  740. /* These elements must be at the end, see alloc_skb() for details. */
  741. sk_buff_data_t tail;
  742. sk_buff_data_t end;
  743. unsigned char *head,
  744. *data;
  745. unsigned int truesize;
  746. refcount_t users;
  747. };
  748. #ifdef __KERNEL__
  749. /*
  750. * Handling routines are only of interest to the kernel
  751. */
  752. #include <linux/slab.h>
  753. #define SKB_ALLOC_FCLONE 0x01
  754. #define SKB_ALLOC_RX 0x02
  755. #define SKB_ALLOC_NAPI 0x04
  756. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  757. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  758. {
  759. return unlikely(skb->pfmemalloc);
  760. }
  761. /*
  762. * skb might have a dst pointer attached, refcounted or not.
  763. * _skb_refdst low order bit is set if refcount was _not_ taken
  764. */
  765. #define SKB_DST_NOREF 1UL
  766. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  767. #define SKB_NFCT_PTRMASK ~(7UL)
  768. /**
  769. * skb_dst - returns skb dst_entry
  770. * @skb: buffer
  771. *
  772. * Returns skb dst_entry, regardless of reference taken or not.
  773. */
  774. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  775. {
  776. /* If refdst was not refcounted, check we still are in a
  777. * rcu_read_lock section
  778. */
  779. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  780. !rcu_read_lock_held() &&
  781. !rcu_read_lock_bh_held());
  782. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  783. }
  784. /**
  785. * skb_dst_set - sets skb dst
  786. * @skb: buffer
  787. * @dst: dst entry
  788. *
  789. * Sets skb dst, assuming a reference was taken on dst and should
  790. * be released by skb_dst_drop()
  791. */
  792. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  793. {
  794. skb->_skb_refdst = (unsigned long)dst;
  795. }
  796. /**
  797. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  798. * @skb: buffer
  799. * @dst: dst entry
  800. *
  801. * Sets skb dst, assuming a reference was not taken on dst.
  802. * If dst entry is cached, we do not take reference and dst_release
  803. * will be avoided by refdst_drop. If dst entry is not cached, we take
  804. * reference, so that last dst_release can destroy the dst immediately.
  805. */
  806. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  807. {
  808. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  809. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  810. }
  811. /**
  812. * skb_dst_is_noref - Test if skb dst isn't refcounted
  813. * @skb: buffer
  814. */
  815. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  816. {
  817. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  818. }
  819. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  820. {
  821. return (struct rtable *)skb_dst(skb);
  822. }
  823. /* For mangling skb->pkt_type from user space side from applications
  824. * such as nft, tc, etc, we only allow a conservative subset of
  825. * possible pkt_types to be set.
  826. */
  827. static inline bool skb_pkt_type_ok(u32 ptype)
  828. {
  829. return ptype <= PACKET_OTHERHOST;
  830. }
  831. static inline unsigned int skb_napi_id(const struct sk_buff *skb)
  832. {
  833. #ifdef CONFIG_NET_RX_BUSY_POLL
  834. return skb->napi_id;
  835. #else
  836. return 0;
  837. #endif
  838. }
  839. /* decrement the reference count and return true if we can free the skb */
  840. static inline bool skb_unref(struct sk_buff *skb)
  841. {
  842. if (unlikely(!skb))
  843. return false;
  844. if (likely(refcount_read(&skb->users) == 1))
  845. smp_rmb();
  846. else if (likely(!refcount_dec_and_test(&skb->users)))
  847. return false;
  848. return true;
  849. }
  850. void skb_release_head_state(struct sk_buff *skb);
  851. void kfree_skb(struct sk_buff *skb);
  852. void kfree_skb_list(struct sk_buff *segs);
  853. void skb_tx_error(struct sk_buff *skb);
  854. void consume_skb(struct sk_buff *skb);
  855. void __consume_stateless_skb(struct sk_buff *skb);
  856. void __kfree_skb(struct sk_buff *skb);
  857. extern struct kmem_cache *skbuff_head_cache;
  858. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  859. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  860. bool *fragstolen, int *delta_truesize);
  861. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  862. int node);
  863. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  864. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  865. static inline struct sk_buff *alloc_skb(unsigned int size,
  866. gfp_t priority)
  867. {
  868. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  869. }
  870. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  871. unsigned long data_len,
  872. int max_page_order,
  873. int *errcode,
  874. gfp_t gfp_mask);
  875. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  876. struct sk_buff_fclones {
  877. struct sk_buff skb1;
  878. struct sk_buff skb2;
  879. refcount_t fclone_ref;
  880. };
  881. /**
  882. * skb_fclone_busy - check if fclone is busy
  883. * @sk: socket
  884. * @skb: buffer
  885. *
  886. * Returns true if skb is a fast clone, and its clone is not freed.
  887. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  888. * so we also check that this didnt happen.
  889. */
  890. static inline bool skb_fclone_busy(const struct sock *sk,
  891. const struct sk_buff *skb)
  892. {
  893. const struct sk_buff_fclones *fclones;
  894. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  895. return skb->fclone == SKB_FCLONE_ORIG &&
  896. refcount_read(&fclones->fclone_ref) > 1 &&
  897. fclones->skb2.sk == sk;
  898. }
  899. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  900. gfp_t priority)
  901. {
  902. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  903. }
  904. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  905. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  906. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  907. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  908. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  909. gfp_t gfp_mask, bool fclone);
  910. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  911. gfp_t gfp_mask)
  912. {
  913. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  914. }
  915. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  916. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  917. unsigned int headroom);
  918. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  919. int newtailroom, gfp_t priority);
  920. int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  921. int offset, int len);
  922. int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
  923. int offset, int len);
  924. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  925. int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
  926. /**
  927. * skb_pad - zero pad the tail of an skb
  928. * @skb: buffer to pad
  929. * @pad: space to pad
  930. *
  931. * Ensure that a buffer is followed by a padding area that is zero
  932. * filled. Used by network drivers which may DMA or transfer data
  933. * beyond the buffer end onto the wire.
  934. *
  935. * May return error in out of memory cases. The skb is freed on error.
  936. */
  937. static inline int skb_pad(struct sk_buff *skb, int pad)
  938. {
  939. return __skb_pad(skb, pad, true);
  940. }
  941. #define dev_kfree_skb(a) consume_skb(a)
  942. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  943. int getfrag(void *from, char *to, int offset,
  944. int len, int odd, struct sk_buff *skb),
  945. void *from, int length);
  946. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  947. int offset, size_t size);
  948. struct skb_seq_state {
  949. __u32 lower_offset;
  950. __u32 upper_offset;
  951. __u32 frag_idx;
  952. __u32 stepped_offset;
  953. struct sk_buff *root_skb;
  954. struct sk_buff *cur_skb;
  955. __u8 *frag_data;
  956. };
  957. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  958. unsigned int to, struct skb_seq_state *st);
  959. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  960. struct skb_seq_state *st);
  961. void skb_abort_seq_read(struct skb_seq_state *st);
  962. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  963. unsigned int to, struct ts_config *config);
  964. /*
  965. * Packet hash types specify the type of hash in skb_set_hash.
  966. *
  967. * Hash types refer to the protocol layer addresses which are used to
  968. * construct a packet's hash. The hashes are used to differentiate or identify
  969. * flows of the protocol layer for the hash type. Hash types are either
  970. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  971. *
  972. * Properties of hashes:
  973. *
  974. * 1) Two packets in different flows have different hash values
  975. * 2) Two packets in the same flow should have the same hash value
  976. *
  977. * A hash at a higher layer is considered to be more specific. A driver should
  978. * set the most specific hash possible.
  979. *
  980. * A driver cannot indicate a more specific hash than the layer at which a hash
  981. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  982. *
  983. * A driver may indicate a hash level which is less specific than the
  984. * actual layer the hash was computed on. For instance, a hash computed
  985. * at L4 may be considered an L3 hash. This should only be done if the
  986. * driver can't unambiguously determine that the HW computed the hash at
  987. * the higher layer. Note that the "should" in the second property above
  988. * permits this.
  989. */
  990. enum pkt_hash_types {
  991. PKT_HASH_TYPE_NONE, /* Undefined type */
  992. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  993. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  994. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  995. };
  996. static inline void skb_clear_hash(struct sk_buff *skb)
  997. {
  998. skb->hash = 0;
  999. skb->sw_hash = 0;
  1000. skb->l4_hash = 0;
  1001. }
  1002. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  1003. {
  1004. if (!skb->l4_hash)
  1005. skb_clear_hash(skb);
  1006. }
  1007. static inline void
  1008. __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
  1009. {
  1010. skb->l4_hash = is_l4;
  1011. skb->sw_hash = is_sw;
  1012. skb->hash = hash;
  1013. }
  1014. static inline void
  1015. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  1016. {
  1017. /* Used by drivers to set hash from HW */
  1018. __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
  1019. }
  1020. static inline void
  1021. __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
  1022. {
  1023. __skb_set_hash(skb, hash, true, is_l4);
  1024. }
  1025. void __skb_get_hash(struct sk_buff *skb);
  1026. u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
  1027. u32 skb_get_poff(const struct sk_buff *skb);
  1028. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  1029. const struct flow_keys *keys, int hlen);
  1030. __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
  1031. void *data, int hlen_proto);
  1032. static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
  1033. int thoff, u8 ip_proto)
  1034. {
  1035. return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
  1036. }
  1037. void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  1038. const struct flow_dissector_key *key,
  1039. unsigned int key_count);
  1040. bool __skb_flow_dissect(const struct sk_buff *skb,
  1041. struct flow_dissector *flow_dissector,
  1042. void *target_container,
  1043. void *data, __be16 proto, int nhoff, int hlen,
  1044. unsigned int flags);
  1045. static inline bool skb_flow_dissect(const struct sk_buff *skb,
  1046. struct flow_dissector *flow_dissector,
  1047. void *target_container, unsigned int flags)
  1048. {
  1049. return __skb_flow_dissect(skb, flow_dissector, target_container,
  1050. NULL, 0, 0, 0, flags);
  1051. }
  1052. static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
  1053. struct flow_keys *flow,
  1054. unsigned int flags)
  1055. {
  1056. memset(flow, 0, sizeof(*flow));
  1057. return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
  1058. NULL, 0, 0, 0, flags);
  1059. }
  1060. static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
  1061. void *data, __be16 proto,
  1062. int nhoff, int hlen,
  1063. unsigned int flags)
  1064. {
  1065. memset(flow, 0, sizeof(*flow));
  1066. return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
  1067. data, proto, nhoff, hlen, flags);
  1068. }
  1069. static inline __u32 skb_get_hash(struct sk_buff *skb)
  1070. {
  1071. if (!skb->l4_hash && !skb->sw_hash)
  1072. __skb_get_hash(skb);
  1073. return skb->hash;
  1074. }
  1075. static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
  1076. {
  1077. if (!skb->l4_hash && !skb->sw_hash) {
  1078. struct flow_keys keys;
  1079. __u32 hash = __get_hash_from_flowi6(fl6, &keys);
  1080. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1081. }
  1082. return skb->hash;
  1083. }
  1084. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  1085. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  1086. {
  1087. return skb->hash;
  1088. }
  1089. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  1090. {
  1091. to->hash = from->hash;
  1092. to->sw_hash = from->sw_hash;
  1093. to->l4_hash = from->l4_hash;
  1094. };
  1095. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1096. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1097. {
  1098. return skb->head + skb->end;
  1099. }
  1100. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1101. {
  1102. return skb->end;
  1103. }
  1104. #else
  1105. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1106. {
  1107. return skb->end;
  1108. }
  1109. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1110. {
  1111. return skb->end - skb->head;
  1112. }
  1113. #endif
  1114. /* Internal */
  1115. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  1116. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  1117. {
  1118. return &skb_shinfo(skb)->hwtstamps;
  1119. }
  1120. static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
  1121. {
  1122. bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
  1123. return is_zcopy ? skb_uarg(skb) : NULL;
  1124. }
  1125. static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg)
  1126. {
  1127. if (skb && uarg && !skb_zcopy(skb)) {
  1128. sock_zerocopy_get(uarg);
  1129. skb_shinfo(skb)->destructor_arg = uarg;
  1130. skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
  1131. }
  1132. }
  1133. /* Release a reference on a zerocopy structure */
  1134. static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
  1135. {
  1136. struct ubuf_info *uarg = skb_zcopy(skb);
  1137. if (uarg) {
  1138. if (uarg->callback == sock_zerocopy_callback) {
  1139. uarg->zerocopy = uarg->zerocopy && zerocopy;
  1140. sock_zerocopy_put(uarg);
  1141. } else {
  1142. uarg->callback(uarg, zerocopy);
  1143. }
  1144. skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
  1145. }
  1146. }
  1147. /* Abort a zerocopy operation and revert zckey on error in send syscall */
  1148. static inline void skb_zcopy_abort(struct sk_buff *skb)
  1149. {
  1150. struct ubuf_info *uarg = skb_zcopy(skb);
  1151. if (uarg) {
  1152. sock_zerocopy_put_abort(uarg);
  1153. skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
  1154. }
  1155. }
  1156. /**
  1157. * skb_queue_empty - check if a queue is empty
  1158. * @list: queue head
  1159. *
  1160. * Returns true if the queue is empty, false otherwise.
  1161. */
  1162. static inline int skb_queue_empty(const struct sk_buff_head *list)
  1163. {
  1164. return list->next == (const struct sk_buff *) list;
  1165. }
  1166. /**
  1167. * skb_queue_is_last - check if skb is the last entry in the queue
  1168. * @list: queue head
  1169. * @skb: buffer
  1170. *
  1171. * Returns true if @skb is the last buffer on the list.
  1172. */
  1173. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  1174. const struct sk_buff *skb)
  1175. {
  1176. return skb->next == (const struct sk_buff *) list;
  1177. }
  1178. /**
  1179. * skb_queue_is_first - check if skb is the first entry in the queue
  1180. * @list: queue head
  1181. * @skb: buffer
  1182. *
  1183. * Returns true if @skb is the first buffer on the list.
  1184. */
  1185. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  1186. const struct sk_buff *skb)
  1187. {
  1188. return skb->prev == (const struct sk_buff *) list;
  1189. }
  1190. /**
  1191. * skb_queue_next - return the next packet in the queue
  1192. * @list: queue head
  1193. * @skb: current buffer
  1194. *
  1195. * Return the next packet in @list after @skb. It is only valid to
  1196. * call this if skb_queue_is_last() evaluates to false.
  1197. */
  1198. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  1199. const struct sk_buff *skb)
  1200. {
  1201. /* This BUG_ON may seem severe, but if we just return then we
  1202. * are going to dereference garbage.
  1203. */
  1204. BUG_ON(skb_queue_is_last(list, skb));
  1205. return skb->next;
  1206. }
  1207. /**
  1208. * skb_queue_prev - return the prev packet in the queue
  1209. * @list: queue head
  1210. * @skb: current buffer
  1211. *
  1212. * Return the prev packet in @list before @skb. It is only valid to
  1213. * call this if skb_queue_is_first() evaluates to false.
  1214. */
  1215. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  1216. const struct sk_buff *skb)
  1217. {
  1218. /* This BUG_ON may seem severe, but if we just return then we
  1219. * are going to dereference garbage.
  1220. */
  1221. BUG_ON(skb_queue_is_first(list, skb));
  1222. return skb->prev;
  1223. }
  1224. /**
  1225. * skb_get - reference buffer
  1226. * @skb: buffer to reference
  1227. *
  1228. * Makes another reference to a socket buffer and returns a pointer
  1229. * to the buffer.
  1230. */
  1231. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  1232. {
  1233. refcount_inc(&skb->users);
  1234. return skb;
  1235. }
  1236. /*
  1237. * If users == 1, we are the only owner and are can avoid redundant
  1238. * atomic change.
  1239. */
  1240. /**
  1241. * skb_cloned - is the buffer a clone
  1242. * @skb: buffer to check
  1243. *
  1244. * Returns true if the buffer was generated with skb_clone() and is
  1245. * one of multiple shared copies of the buffer. Cloned buffers are
  1246. * shared data so must not be written to under normal circumstances.
  1247. */
  1248. static inline int skb_cloned(const struct sk_buff *skb)
  1249. {
  1250. return skb->cloned &&
  1251. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  1252. }
  1253. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  1254. {
  1255. might_sleep_if(gfpflags_allow_blocking(pri));
  1256. if (skb_cloned(skb))
  1257. return pskb_expand_head(skb, 0, 0, pri);
  1258. return 0;
  1259. }
  1260. /**
  1261. * skb_header_cloned - is the header a clone
  1262. * @skb: buffer to check
  1263. *
  1264. * Returns true if modifying the header part of the buffer requires
  1265. * the data to be copied.
  1266. */
  1267. static inline int skb_header_cloned(const struct sk_buff *skb)
  1268. {
  1269. int dataref;
  1270. if (!skb->cloned)
  1271. return 0;
  1272. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  1273. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  1274. return dataref != 1;
  1275. }
  1276. static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
  1277. {
  1278. might_sleep_if(gfpflags_allow_blocking(pri));
  1279. if (skb_header_cloned(skb))
  1280. return pskb_expand_head(skb, 0, 0, pri);
  1281. return 0;
  1282. }
  1283. /**
  1284. * skb_header_release - release reference to header
  1285. * @skb: buffer to operate on
  1286. *
  1287. * Drop a reference to the header part of the buffer. This is done
  1288. * by acquiring a payload reference. You must not read from the header
  1289. * part of skb->data after this.
  1290. * Note : Check if you can use __skb_header_release() instead.
  1291. */
  1292. static inline void skb_header_release(struct sk_buff *skb)
  1293. {
  1294. BUG_ON(skb->nohdr);
  1295. skb->nohdr = 1;
  1296. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1297. }
  1298. /**
  1299. * __skb_header_release - release reference to header
  1300. * @skb: buffer to operate on
  1301. *
  1302. * Variant of skb_header_release() assuming skb is private to caller.
  1303. * We can avoid one atomic operation.
  1304. */
  1305. static inline void __skb_header_release(struct sk_buff *skb)
  1306. {
  1307. skb->nohdr = 1;
  1308. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1309. }
  1310. /**
  1311. * skb_shared - is the buffer shared
  1312. * @skb: buffer to check
  1313. *
  1314. * Returns true if more than one person has a reference to this
  1315. * buffer.
  1316. */
  1317. static inline int skb_shared(const struct sk_buff *skb)
  1318. {
  1319. return refcount_read(&skb->users) != 1;
  1320. }
  1321. /**
  1322. * skb_share_check - check if buffer is shared and if so clone it
  1323. * @skb: buffer to check
  1324. * @pri: priority for memory allocation
  1325. *
  1326. * If the buffer is shared the buffer is cloned and the old copy
  1327. * drops a reference. A new clone with a single reference is returned.
  1328. * If the buffer is not shared the original buffer is returned. When
  1329. * being called from interrupt status or with spinlocks held pri must
  1330. * be GFP_ATOMIC.
  1331. *
  1332. * NULL is returned on a memory allocation failure.
  1333. */
  1334. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1335. {
  1336. might_sleep_if(gfpflags_allow_blocking(pri));
  1337. if (skb_shared(skb)) {
  1338. struct sk_buff *nskb = skb_clone(skb, pri);
  1339. if (likely(nskb))
  1340. consume_skb(skb);
  1341. else
  1342. kfree_skb(skb);
  1343. skb = nskb;
  1344. }
  1345. return skb;
  1346. }
  1347. /*
  1348. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1349. * packets to handle cases where we have a local reader and forward
  1350. * and a couple of other messy ones. The normal one is tcpdumping
  1351. * a packet thats being forwarded.
  1352. */
  1353. /**
  1354. * skb_unshare - make a copy of a shared buffer
  1355. * @skb: buffer to check
  1356. * @pri: priority for memory allocation
  1357. *
  1358. * If the socket buffer is a clone then this function creates a new
  1359. * copy of the data, drops a reference count on the old copy and returns
  1360. * the new copy with the reference count at 1. If the buffer is not a clone
  1361. * the original buffer is returned. When called with a spinlock held or
  1362. * from interrupt state @pri must be %GFP_ATOMIC
  1363. *
  1364. * %NULL is returned on a memory allocation failure.
  1365. */
  1366. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1367. gfp_t pri)
  1368. {
  1369. might_sleep_if(gfpflags_allow_blocking(pri));
  1370. if (skb_cloned(skb)) {
  1371. struct sk_buff *nskb = skb_copy(skb, pri);
  1372. /* Free our shared copy */
  1373. if (likely(nskb))
  1374. consume_skb(skb);
  1375. else
  1376. kfree_skb(skb);
  1377. skb = nskb;
  1378. }
  1379. return skb;
  1380. }
  1381. /**
  1382. * skb_peek - peek at the head of an &sk_buff_head
  1383. * @list_: list to peek at
  1384. *
  1385. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1386. * be careful with this one. A peek leaves the buffer on the
  1387. * list and someone else may run off with it. You must hold
  1388. * the appropriate locks or have a private queue to do this.
  1389. *
  1390. * Returns %NULL for an empty list or a pointer to the head element.
  1391. * The reference count is not incremented and the reference is therefore
  1392. * volatile. Use with caution.
  1393. */
  1394. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1395. {
  1396. struct sk_buff *skb = list_->next;
  1397. if (skb == (struct sk_buff *)list_)
  1398. skb = NULL;
  1399. return skb;
  1400. }
  1401. /**
  1402. * skb_peek_next - peek skb following the given one from a queue
  1403. * @skb: skb to start from
  1404. * @list_: list to peek at
  1405. *
  1406. * Returns %NULL when the end of the list is met or a pointer to the
  1407. * next element. The reference count is not incremented and the
  1408. * reference is therefore volatile. Use with caution.
  1409. */
  1410. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1411. const struct sk_buff_head *list_)
  1412. {
  1413. struct sk_buff *next = skb->next;
  1414. if (next == (struct sk_buff *)list_)
  1415. next = NULL;
  1416. return next;
  1417. }
  1418. /**
  1419. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1420. * @list_: list to peek at
  1421. *
  1422. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1423. * be careful with this one. A peek leaves the buffer on the
  1424. * list and someone else may run off with it. You must hold
  1425. * the appropriate locks or have a private queue to do this.
  1426. *
  1427. * Returns %NULL for an empty list or a pointer to the tail element.
  1428. * The reference count is not incremented and the reference is therefore
  1429. * volatile. Use with caution.
  1430. */
  1431. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1432. {
  1433. struct sk_buff *skb = list_->prev;
  1434. if (skb == (struct sk_buff *)list_)
  1435. skb = NULL;
  1436. return skb;
  1437. }
  1438. /**
  1439. * skb_queue_len - get queue length
  1440. * @list_: list to measure
  1441. *
  1442. * Return the length of an &sk_buff queue.
  1443. */
  1444. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1445. {
  1446. return list_->qlen;
  1447. }
  1448. /**
  1449. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1450. * @list: queue to initialize
  1451. *
  1452. * This initializes only the list and queue length aspects of
  1453. * an sk_buff_head object. This allows to initialize the list
  1454. * aspects of an sk_buff_head without reinitializing things like
  1455. * the spinlock. It can also be used for on-stack sk_buff_head
  1456. * objects where the spinlock is known to not be used.
  1457. */
  1458. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1459. {
  1460. list->prev = list->next = (struct sk_buff *)list;
  1461. list->qlen = 0;
  1462. }
  1463. /*
  1464. * This function creates a split out lock class for each invocation;
  1465. * this is needed for now since a whole lot of users of the skb-queue
  1466. * infrastructure in drivers have different locking usage (in hardirq)
  1467. * than the networking core (in softirq only). In the long run either the
  1468. * network layer or drivers should need annotation to consolidate the
  1469. * main types of usage into 3 classes.
  1470. */
  1471. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1472. {
  1473. spin_lock_init(&list->lock);
  1474. __skb_queue_head_init(list);
  1475. }
  1476. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1477. struct lock_class_key *class)
  1478. {
  1479. skb_queue_head_init(list);
  1480. lockdep_set_class(&list->lock, class);
  1481. }
  1482. /*
  1483. * Insert an sk_buff on a list.
  1484. *
  1485. * The "__skb_xxxx()" functions are the non-atomic ones that
  1486. * can only be called with interrupts disabled.
  1487. */
  1488. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1489. struct sk_buff_head *list);
  1490. static inline void __skb_insert(struct sk_buff *newsk,
  1491. struct sk_buff *prev, struct sk_buff *next,
  1492. struct sk_buff_head *list)
  1493. {
  1494. newsk->next = next;
  1495. newsk->prev = prev;
  1496. next->prev = prev->next = newsk;
  1497. list->qlen++;
  1498. }
  1499. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1500. struct sk_buff *prev,
  1501. struct sk_buff *next)
  1502. {
  1503. struct sk_buff *first = list->next;
  1504. struct sk_buff *last = list->prev;
  1505. first->prev = prev;
  1506. prev->next = first;
  1507. last->next = next;
  1508. next->prev = last;
  1509. }
  1510. /**
  1511. * skb_queue_splice - join two skb lists, this is designed for stacks
  1512. * @list: the new list to add
  1513. * @head: the place to add it in the first list
  1514. */
  1515. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1516. struct sk_buff_head *head)
  1517. {
  1518. if (!skb_queue_empty(list)) {
  1519. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1520. head->qlen += list->qlen;
  1521. }
  1522. }
  1523. /**
  1524. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1525. * @list: the new list to add
  1526. * @head: the place to add it in the first list
  1527. *
  1528. * The list at @list is reinitialised
  1529. */
  1530. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1531. struct sk_buff_head *head)
  1532. {
  1533. if (!skb_queue_empty(list)) {
  1534. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1535. head->qlen += list->qlen;
  1536. __skb_queue_head_init(list);
  1537. }
  1538. }
  1539. /**
  1540. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1541. * @list: the new list to add
  1542. * @head: the place to add it in the first list
  1543. */
  1544. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1545. struct sk_buff_head *head)
  1546. {
  1547. if (!skb_queue_empty(list)) {
  1548. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1549. head->qlen += list->qlen;
  1550. }
  1551. }
  1552. /**
  1553. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1554. * @list: the new list to add
  1555. * @head: the place to add it in the first list
  1556. *
  1557. * Each of the lists is a queue.
  1558. * The list at @list is reinitialised
  1559. */
  1560. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1561. struct sk_buff_head *head)
  1562. {
  1563. if (!skb_queue_empty(list)) {
  1564. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1565. head->qlen += list->qlen;
  1566. __skb_queue_head_init(list);
  1567. }
  1568. }
  1569. /**
  1570. * __skb_queue_after - queue a buffer at the list head
  1571. * @list: list to use
  1572. * @prev: place after this buffer
  1573. * @newsk: buffer to queue
  1574. *
  1575. * Queue a buffer int the middle of a list. This function takes no locks
  1576. * and you must therefore hold required locks before calling it.
  1577. *
  1578. * A buffer cannot be placed on two lists at the same time.
  1579. */
  1580. static inline void __skb_queue_after(struct sk_buff_head *list,
  1581. struct sk_buff *prev,
  1582. struct sk_buff *newsk)
  1583. {
  1584. __skb_insert(newsk, prev, prev->next, list);
  1585. }
  1586. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1587. struct sk_buff_head *list);
  1588. static inline void __skb_queue_before(struct sk_buff_head *list,
  1589. struct sk_buff *next,
  1590. struct sk_buff *newsk)
  1591. {
  1592. __skb_insert(newsk, next->prev, next, list);
  1593. }
  1594. /**
  1595. * __skb_queue_head - queue a buffer at the list head
  1596. * @list: list to use
  1597. * @newsk: buffer to queue
  1598. *
  1599. * Queue a buffer at the start of a list. This function takes no locks
  1600. * and you must therefore hold required locks before calling it.
  1601. *
  1602. * A buffer cannot be placed on two lists at the same time.
  1603. */
  1604. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1605. static inline void __skb_queue_head(struct sk_buff_head *list,
  1606. struct sk_buff *newsk)
  1607. {
  1608. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1609. }
  1610. /**
  1611. * __skb_queue_tail - queue a buffer at the list tail
  1612. * @list: list to use
  1613. * @newsk: buffer to queue
  1614. *
  1615. * Queue a buffer at the end of a list. This function takes no locks
  1616. * and you must therefore hold required locks before calling it.
  1617. *
  1618. * A buffer cannot be placed on two lists at the same time.
  1619. */
  1620. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1621. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1622. struct sk_buff *newsk)
  1623. {
  1624. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1625. }
  1626. /*
  1627. * remove sk_buff from list. _Must_ be called atomically, and with
  1628. * the list known..
  1629. */
  1630. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1631. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1632. {
  1633. struct sk_buff *next, *prev;
  1634. list->qlen--;
  1635. next = skb->next;
  1636. prev = skb->prev;
  1637. skb->next = skb->prev = NULL;
  1638. next->prev = prev;
  1639. prev->next = next;
  1640. }
  1641. /**
  1642. * __skb_dequeue - remove from the head of the queue
  1643. * @list: list to dequeue from
  1644. *
  1645. * Remove the head of the list. This function does not take any locks
  1646. * so must be used with appropriate locks held only. The head item is
  1647. * returned or %NULL if the list is empty.
  1648. */
  1649. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1650. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1651. {
  1652. struct sk_buff *skb = skb_peek(list);
  1653. if (skb)
  1654. __skb_unlink(skb, list);
  1655. return skb;
  1656. }
  1657. /**
  1658. * __skb_dequeue_tail - remove from the tail of the queue
  1659. * @list: list to dequeue from
  1660. *
  1661. * Remove the tail of the list. This function does not take any locks
  1662. * so must be used with appropriate locks held only. The tail item is
  1663. * returned or %NULL if the list is empty.
  1664. */
  1665. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1666. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1667. {
  1668. struct sk_buff *skb = skb_peek_tail(list);
  1669. if (skb)
  1670. __skb_unlink(skb, list);
  1671. return skb;
  1672. }
  1673. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1674. {
  1675. return skb->data_len;
  1676. }
  1677. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1678. {
  1679. return skb->len - skb->data_len;
  1680. }
  1681. static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
  1682. {
  1683. unsigned int i, len = 0;
  1684. for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
  1685. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1686. return len;
  1687. }
  1688. static inline unsigned int skb_pagelen(const struct sk_buff *skb)
  1689. {
  1690. return skb_headlen(skb) + __skb_pagelen(skb);
  1691. }
  1692. /**
  1693. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1694. * @skb: buffer containing fragment to be initialised
  1695. * @i: paged fragment index to initialise
  1696. * @page: the page to use for this fragment
  1697. * @off: the offset to the data with @page
  1698. * @size: the length of the data
  1699. *
  1700. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1701. * offset @off within @page.
  1702. *
  1703. * Does not take any additional reference on the fragment.
  1704. */
  1705. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1706. struct page *page, int off, int size)
  1707. {
  1708. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1709. /*
  1710. * Propagate page pfmemalloc to the skb if we can. The problem is
  1711. * that not all callers have unique ownership of the page but rely
  1712. * on page_is_pfmemalloc doing the right thing(tm).
  1713. */
  1714. frag->page.p = page;
  1715. frag->page_offset = off;
  1716. skb_frag_size_set(frag, size);
  1717. page = compound_head(page);
  1718. if (page_is_pfmemalloc(page))
  1719. skb->pfmemalloc = true;
  1720. }
  1721. /**
  1722. * skb_fill_page_desc - initialise a paged fragment in an skb
  1723. * @skb: buffer containing fragment to be initialised
  1724. * @i: paged fragment index to initialise
  1725. * @page: the page to use for this fragment
  1726. * @off: the offset to the data with @page
  1727. * @size: the length of the data
  1728. *
  1729. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1730. * @skb to point to @size bytes at offset @off within @page. In
  1731. * addition updates @skb such that @i is the last fragment.
  1732. *
  1733. * Does not take any additional reference on the fragment.
  1734. */
  1735. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1736. struct page *page, int off, int size)
  1737. {
  1738. __skb_fill_page_desc(skb, i, page, off, size);
  1739. skb_shinfo(skb)->nr_frags = i + 1;
  1740. }
  1741. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1742. int size, unsigned int truesize);
  1743. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1744. unsigned int truesize);
  1745. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1746. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1747. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1748. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1749. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1750. {
  1751. return skb->head + skb->tail;
  1752. }
  1753. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1754. {
  1755. skb->tail = skb->data - skb->head;
  1756. }
  1757. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1758. {
  1759. skb_reset_tail_pointer(skb);
  1760. skb->tail += offset;
  1761. }
  1762. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1763. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1764. {
  1765. return skb->tail;
  1766. }
  1767. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1768. {
  1769. skb->tail = skb->data;
  1770. }
  1771. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1772. {
  1773. skb->tail = skb->data + offset;
  1774. }
  1775. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1776. /*
  1777. * Add data to an sk_buff
  1778. */
  1779. void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1780. void *skb_put(struct sk_buff *skb, unsigned int len);
  1781. static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
  1782. {
  1783. void *tmp = skb_tail_pointer(skb);
  1784. SKB_LINEAR_ASSERT(skb);
  1785. skb->tail += len;
  1786. skb->len += len;
  1787. return tmp;
  1788. }
  1789. static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
  1790. {
  1791. void *tmp = __skb_put(skb, len);
  1792. memset(tmp, 0, len);
  1793. return tmp;
  1794. }
  1795. static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
  1796. unsigned int len)
  1797. {
  1798. void *tmp = __skb_put(skb, len);
  1799. memcpy(tmp, data, len);
  1800. return tmp;
  1801. }
  1802. static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
  1803. {
  1804. *(u8 *)__skb_put(skb, 1) = val;
  1805. }
  1806. static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
  1807. {
  1808. void *tmp = skb_put(skb, len);
  1809. memset(tmp, 0, len);
  1810. return tmp;
  1811. }
  1812. static inline void *skb_put_data(struct sk_buff *skb, const void *data,
  1813. unsigned int len)
  1814. {
  1815. void *tmp = skb_put(skb, len);
  1816. memcpy(tmp, data, len);
  1817. return tmp;
  1818. }
  1819. static inline void skb_put_u8(struct sk_buff *skb, u8 val)
  1820. {
  1821. *(u8 *)skb_put(skb, 1) = val;
  1822. }
  1823. void *skb_push(struct sk_buff *skb, unsigned int len);
  1824. static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
  1825. {
  1826. skb->data -= len;
  1827. skb->len += len;
  1828. return skb->data;
  1829. }
  1830. void *skb_pull(struct sk_buff *skb, unsigned int len);
  1831. static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
  1832. {
  1833. skb->len -= len;
  1834. BUG_ON(skb->len < skb->data_len);
  1835. return skb->data += len;
  1836. }
  1837. static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1838. {
  1839. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1840. }
  1841. void *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1842. static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1843. {
  1844. if (len > skb_headlen(skb) &&
  1845. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1846. return NULL;
  1847. skb->len -= len;
  1848. return skb->data += len;
  1849. }
  1850. static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
  1851. {
  1852. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1853. }
  1854. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1855. {
  1856. if (likely(len <= skb_headlen(skb)))
  1857. return 1;
  1858. if (unlikely(len > skb->len))
  1859. return 0;
  1860. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1861. }
  1862. void skb_condense(struct sk_buff *skb);
  1863. /**
  1864. * skb_headroom - bytes at buffer head
  1865. * @skb: buffer to check
  1866. *
  1867. * Return the number of bytes of free space at the head of an &sk_buff.
  1868. */
  1869. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1870. {
  1871. return skb->data - skb->head;
  1872. }
  1873. /**
  1874. * skb_tailroom - bytes at buffer end
  1875. * @skb: buffer to check
  1876. *
  1877. * Return the number of bytes of free space at the tail of an sk_buff
  1878. */
  1879. static inline int skb_tailroom(const struct sk_buff *skb)
  1880. {
  1881. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1882. }
  1883. /**
  1884. * skb_availroom - bytes at buffer end
  1885. * @skb: buffer to check
  1886. *
  1887. * Return the number of bytes of free space at the tail of an sk_buff
  1888. * allocated by sk_stream_alloc()
  1889. */
  1890. static inline int skb_availroom(const struct sk_buff *skb)
  1891. {
  1892. if (skb_is_nonlinear(skb))
  1893. return 0;
  1894. return skb->end - skb->tail - skb->reserved_tailroom;
  1895. }
  1896. /**
  1897. * skb_reserve - adjust headroom
  1898. * @skb: buffer to alter
  1899. * @len: bytes to move
  1900. *
  1901. * Increase the headroom of an empty &sk_buff by reducing the tail
  1902. * room. This is only allowed for an empty buffer.
  1903. */
  1904. static inline void skb_reserve(struct sk_buff *skb, int len)
  1905. {
  1906. skb->data += len;
  1907. skb->tail += len;
  1908. }
  1909. /**
  1910. * skb_tailroom_reserve - adjust reserved_tailroom
  1911. * @skb: buffer to alter
  1912. * @mtu: maximum amount of headlen permitted
  1913. * @needed_tailroom: minimum amount of reserved_tailroom
  1914. *
  1915. * Set reserved_tailroom so that headlen can be as large as possible but
  1916. * not larger than mtu and tailroom cannot be smaller than
  1917. * needed_tailroom.
  1918. * The required headroom should already have been reserved before using
  1919. * this function.
  1920. */
  1921. static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
  1922. unsigned int needed_tailroom)
  1923. {
  1924. SKB_LINEAR_ASSERT(skb);
  1925. if (mtu < skb_tailroom(skb) - needed_tailroom)
  1926. /* use at most mtu */
  1927. skb->reserved_tailroom = skb_tailroom(skb) - mtu;
  1928. else
  1929. /* use up to all available space */
  1930. skb->reserved_tailroom = needed_tailroom;
  1931. }
  1932. #define ENCAP_TYPE_ETHER 0
  1933. #define ENCAP_TYPE_IPPROTO 1
  1934. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1935. __be16 protocol)
  1936. {
  1937. skb->inner_protocol = protocol;
  1938. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1939. }
  1940. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1941. __u8 ipproto)
  1942. {
  1943. skb->inner_ipproto = ipproto;
  1944. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1945. }
  1946. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1947. {
  1948. skb->inner_mac_header = skb->mac_header;
  1949. skb->inner_network_header = skb->network_header;
  1950. skb->inner_transport_header = skb->transport_header;
  1951. }
  1952. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1953. {
  1954. skb->mac_len = skb->network_header - skb->mac_header;
  1955. }
  1956. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1957. *skb)
  1958. {
  1959. return skb->head + skb->inner_transport_header;
  1960. }
  1961. static inline int skb_inner_transport_offset(const struct sk_buff *skb)
  1962. {
  1963. return skb_inner_transport_header(skb) - skb->data;
  1964. }
  1965. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1966. {
  1967. skb->inner_transport_header = skb->data - skb->head;
  1968. }
  1969. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1970. const int offset)
  1971. {
  1972. skb_reset_inner_transport_header(skb);
  1973. skb->inner_transport_header += offset;
  1974. }
  1975. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1976. {
  1977. return skb->head + skb->inner_network_header;
  1978. }
  1979. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1980. {
  1981. skb->inner_network_header = skb->data - skb->head;
  1982. }
  1983. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1984. const int offset)
  1985. {
  1986. skb_reset_inner_network_header(skb);
  1987. skb->inner_network_header += offset;
  1988. }
  1989. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1990. {
  1991. return skb->head + skb->inner_mac_header;
  1992. }
  1993. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1994. {
  1995. skb->inner_mac_header = skb->data - skb->head;
  1996. }
  1997. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1998. const int offset)
  1999. {
  2000. skb_reset_inner_mac_header(skb);
  2001. skb->inner_mac_header += offset;
  2002. }
  2003. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  2004. {
  2005. return skb->transport_header != (typeof(skb->transport_header))~0U;
  2006. }
  2007. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  2008. {
  2009. return skb->head + skb->transport_header;
  2010. }
  2011. static inline void skb_reset_transport_header(struct sk_buff *skb)
  2012. {
  2013. skb->transport_header = skb->data - skb->head;
  2014. }
  2015. static inline void skb_set_transport_header(struct sk_buff *skb,
  2016. const int offset)
  2017. {
  2018. skb_reset_transport_header(skb);
  2019. skb->transport_header += offset;
  2020. }
  2021. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  2022. {
  2023. return skb->head + skb->network_header;
  2024. }
  2025. static inline void skb_reset_network_header(struct sk_buff *skb)
  2026. {
  2027. skb->network_header = skb->data - skb->head;
  2028. }
  2029. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  2030. {
  2031. skb_reset_network_header(skb);
  2032. skb->network_header += offset;
  2033. }
  2034. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  2035. {
  2036. return skb->head + skb->mac_header;
  2037. }
  2038. static inline int skb_mac_offset(const struct sk_buff *skb)
  2039. {
  2040. return skb_mac_header(skb) - skb->data;
  2041. }
  2042. static inline u32 skb_mac_header_len(const struct sk_buff *skb)
  2043. {
  2044. return skb->network_header - skb->mac_header;
  2045. }
  2046. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  2047. {
  2048. return skb->mac_header != (typeof(skb->mac_header))~0U;
  2049. }
  2050. static inline void skb_reset_mac_header(struct sk_buff *skb)
  2051. {
  2052. skb->mac_header = skb->data - skb->head;
  2053. }
  2054. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  2055. {
  2056. skb_reset_mac_header(skb);
  2057. skb->mac_header += offset;
  2058. }
  2059. static inline void skb_pop_mac_header(struct sk_buff *skb)
  2060. {
  2061. skb->mac_header = skb->network_header;
  2062. }
  2063. static inline void skb_probe_transport_header(struct sk_buff *skb,
  2064. const int offset_hint)
  2065. {
  2066. struct flow_keys keys;
  2067. if (skb_transport_header_was_set(skb))
  2068. return;
  2069. else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
  2070. skb_set_transport_header(skb, keys.control.thoff);
  2071. else
  2072. skb_set_transport_header(skb, offset_hint);
  2073. }
  2074. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  2075. {
  2076. if (skb_mac_header_was_set(skb)) {
  2077. const unsigned char *old_mac = skb_mac_header(skb);
  2078. skb_set_mac_header(skb, -skb->mac_len);
  2079. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  2080. }
  2081. }
  2082. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  2083. {
  2084. return skb->csum_start - skb_headroom(skb);
  2085. }
  2086. static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
  2087. {
  2088. return skb->head + skb->csum_start;
  2089. }
  2090. static inline int skb_transport_offset(const struct sk_buff *skb)
  2091. {
  2092. return skb_transport_header(skb) - skb->data;
  2093. }
  2094. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  2095. {
  2096. return skb->transport_header - skb->network_header;
  2097. }
  2098. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  2099. {
  2100. return skb->inner_transport_header - skb->inner_network_header;
  2101. }
  2102. static inline int skb_network_offset(const struct sk_buff *skb)
  2103. {
  2104. return skb_network_header(skb) - skb->data;
  2105. }
  2106. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  2107. {
  2108. return skb_inner_network_header(skb) - skb->data;
  2109. }
  2110. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  2111. {
  2112. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  2113. }
  2114. /*
  2115. * CPUs often take a performance hit when accessing unaligned memory
  2116. * locations. The actual performance hit varies, it can be small if the
  2117. * hardware handles it or large if we have to take an exception and fix it
  2118. * in software.
  2119. *
  2120. * Since an ethernet header is 14 bytes network drivers often end up with
  2121. * the IP header at an unaligned offset. The IP header can be aligned by
  2122. * shifting the start of the packet by 2 bytes. Drivers should do this
  2123. * with:
  2124. *
  2125. * skb_reserve(skb, NET_IP_ALIGN);
  2126. *
  2127. * The downside to this alignment of the IP header is that the DMA is now
  2128. * unaligned. On some architectures the cost of an unaligned DMA is high
  2129. * and this cost outweighs the gains made by aligning the IP header.
  2130. *
  2131. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  2132. * to be overridden.
  2133. */
  2134. #ifndef NET_IP_ALIGN
  2135. #define NET_IP_ALIGN 2
  2136. #endif
  2137. /*
  2138. * The networking layer reserves some headroom in skb data (via
  2139. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  2140. * the header has to grow. In the default case, if the header has to grow
  2141. * 32 bytes or less we avoid the reallocation.
  2142. *
  2143. * Unfortunately this headroom changes the DMA alignment of the resulting
  2144. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  2145. * on some architectures. An architecture can override this value,
  2146. * perhaps setting it to a cacheline in size (since that will maintain
  2147. * cacheline alignment of the DMA). It must be a power of 2.
  2148. *
  2149. * Various parts of the networking layer expect at least 32 bytes of
  2150. * headroom, you should not reduce this.
  2151. *
  2152. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  2153. * to reduce average number of cache lines per packet.
  2154. * get_rps_cpus() for example only access one 64 bytes aligned block :
  2155. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  2156. */
  2157. #ifndef NET_SKB_PAD
  2158. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  2159. #endif
  2160. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  2161. static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
  2162. {
  2163. if (unlikely(skb_is_nonlinear(skb))) {
  2164. WARN_ON(1);
  2165. return;
  2166. }
  2167. skb->len = len;
  2168. skb_set_tail_pointer(skb, len);
  2169. }
  2170. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  2171. {
  2172. __skb_set_length(skb, len);
  2173. }
  2174. void skb_trim(struct sk_buff *skb, unsigned int len);
  2175. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  2176. {
  2177. if (skb->data_len)
  2178. return ___pskb_trim(skb, len);
  2179. __skb_trim(skb, len);
  2180. return 0;
  2181. }
  2182. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  2183. {
  2184. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  2185. }
  2186. /**
  2187. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  2188. * @skb: buffer to alter
  2189. * @len: new length
  2190. *
  2191. * This is identical to pskb_trim except that the caller knows that
  2192. * the skb is not cloned so we should never get an error due to out-
  2193. * of-memory.
  2194. */
  2195. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  2196. {
  2197. int err = pskb_trim(skb, len);
  2198. BUG_ON(err);
  2199. }
  2200. static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
  2201. {
  2202. unsigned int diff = len - skb->len;
  2203. if (skb_tailroom(skb) < diff) {
  2204. int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
  2205. GFP_ATOMIC);
  2206. if (ret)
  2207. return ret;
  2208. }
  2209. __skb_set_length(skb, len);
  2210. return 0;
  2211. }
  2212. /**
  2213. * skb_orphan - orphan a buffer
  2214. * @skb: buffer to orphan
  2215. *
  2216. * If a buffer currently has an owner then we call the owner's
  2217. * destructor function and make the @skb unowned. The buffer continues
  2218. * to exist but is no longer charged to its former owner.
  2219. */
  2220. static inline void skb_orphan(struct sk_buff *skb)
  2221. {
  2222. if (skb->destructor) {
  2223. skb->destructor(skb);
  2224. skb->destructor = NULL;
  2225. skb->sk = NULL;
  2226. } else {
  2227. BUG_ON(skb->sk);
  2228. }
  2229. }
  2230. /**
  2231. * skb_orphan_frags - orphan the frags contained in a buffer
  2232. * @skb: buffer to orphan frags from
  2233. * @gfp_mask: allocation mask for replacement pages
  2234. *
  2235. * For each frag in the SKB which needs a destructor (i.e. has an
  2236. * owner) create a copy of that frag and release the original
  2237. * page by calling the destructor.
  2238. */
  2239. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  2240. {
  2241. if (likely(!skb_zcopy(skb)))
  2242. return 0;
  2243. if (skb_uarg(skb)->callback == sock_zerocopy_callback)
  2244. return 0;
  2245. return skb_copy_ubufs(skb, gfp_mask);
  2246. }
  2247. /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
  2248. static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
  2249. {
  2250. if (likely(!skb_zcopy(skb)))
  2251. return 0;
  2252. return skb_copy_ubufs(skb, gfp_mask);
  2253. }
  2254. /**
  2255. * __skb_queue_purge - empty a list
  2256. * @list: list to empty
  2257. *
  2258. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  2259. * the list and one reference dropped. This function does not take the
  2260. * list lock and the caller must hold the relevant locks to use it.
  2261. */
  2262. void skb_queue_purge(struct sk_buff_head *list);
  2263. static inline void __skb_queue_purge(struct sk_buff_head *list)
  2264. {
  2265. struct sk_buff *skb;
  2266. while ((skb = __skb_dequeue(list)) != NULL)
  2267. kfree_skb(skb);
  2268. }
  2269. void skb_rbtree_purge(struct rb_root *root);
  2270. void *netdev_alloc_frag(unsigned int fragsz);
  2271. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  2272. gfp_t gfp_mask);
  2273. /**
  2274. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  2275. * @dev: network device to receive on
  2276. * @length: length to allocate
  2277. *
  2278. * Allocate a new &sk_buff and assign it a usage count of one. The
  2279. * buffer has unspecified headroom built in. Users should allocate
  2280. * the headroom they think they need without accounting for the
  2281. * built in space. The built in space is used for optimisations.
  2282. *
  2283. * %NULL is returned if there is no free memory. Although this function
  2284. * allocates memory it can be called from an interrupt.
  2285. */
  2286. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  2287. unsigned int length)
  2288. {
  2289. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  2290. }
  2291. /* legacy helper around __netdev_alloc_skb() */
  2292. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  2293. gfp_t gfp_mask)
  2294. {
  2295. return __netdev_alloc_skb(NULL, length, gfp_mask);
  2296. }
  2297. /* legacy helper around netdev_alloc_skb() */
  2298. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  2299. {
  2300. return netdev_alloc_skb(NULL, length);
  2301. }
  2302. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  2303. unsigned int length, gfp_t gfp)
  2304. {
  2305. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  2306. if (NET_IP_ALIGN && skb)
  2307. skb_reserve(skb, NET_IP_ALIGN);
  2308. return skb;
  2309. }
  2310. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  2311. unsigned int length)
  2312. {
  2313. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  2314. }
  2315. static inline void skb_free_frag(void *addr)
  2316. {
  2317. page_frag_free(addr);
  2318. }
  2319. void *napi_alloc_frag(unsigned int fragsz);
  2320. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  2321. unsigned int length, gfp_t gfp_mask);
  2322. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  2323. unsigned int length)
  2324. {
  2325. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  2326. }
  2327. void napi_consume_skb(struct sk_buff *skb, int budget);
  2328. void __kfree_skb_flush(void);
  2329. void __kfree_skb_defer(struct sk_buff *skb);
  2330. /**
  2331. * __dev_alloc_pages - allocate page for network Rx
  2332. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2333. * @order: size of the allocation
  2334. *
  2335. * Allocate a new page.
  2336. *
  2337. * %NULL is returned if there is no free memory.
  2338. */
  2339. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  2340. unsigned int order)
  2341. {
  2342. /* This piece of code contains several assumptions.
  2343. * 1. This is for device Rx, therefor a cold page is preferred.
  2344. * 2. The expectation is the user wants a compound page.
  2345. * 3. If requesting a order 0 page it will not be compound
  2346. * due to the check to see if order has a value in prep_new_page
  2347. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  2348. * code in gfp_to_alloc_flags that should be enforcing this.
  2349. */
  2350. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  2351. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  2352. }
  2353. static inline struct page *dev_alloc_pages(unsigned int order)
  2354. {
  2355. return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
  2356. }
  2357. /**
  2358. * __dev_alloc_page - allocate a page for network Rx
  2359. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2360. *
  2361. * Allocate a new page.
  2362. *
  2363. * %NULL is returned if there is no free memory.
  2364. */
  2365. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  2366. {
  2367. return __dev_alloc_pages(gfp_mask, 0);
  2368. }
  2369. static inline struct page *dev_alloc_page(void)
  2370. {
  2371. return dev_alloc_pages(0);
  2372. }
  2373. /**
  2374. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  2375. * @page: The page that was allocated from skb_alloc_page
  2376. * @skb: The skb that may need pfmemalloc set
  2377. */
  2378. static inline void skb_propagate_pfmemalloc(struct page *page,
  2379. struct sk_buff *skb)
  2380. {
  2381. if (page_is_pfmemalloc(page))
  2382. skb->pfmemalloc = true;
  2383. }
  2384. /**
  2385. * skb_frag_page - retrieve the page referred to by a paged fragment
  2386. * @frag: the paged fragment
  2387. *
  2388. * Returns the &struct page associated with @frag.
  2389. */
  2390. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2391. {
  2392. return frag->page.p;
  2393. }
  2394. /**
  2395. * __skb_frag_ref - take an addition reference on a paged fragment.
  2396. * @frag: the paged fragment
  2397. *
  2398. * Takes an additional reference on the paged fragment @frag.
  2399. */
  2400. static inline void __skb_frag_ref(skb_frag_t *frag)
  2401. {
  2402. get_page(skb_frag_page(frag));
  2403. }
  2404. /**
  2405. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2406. * @skb: the buffer
  2407. * @f: the fragment offset.
  2408. *
  2409. * Takes an additional reference on the @f'th paged fragment of @skb.
  2410. */
  2411. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2412. {
  2413. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2414. }
  2415. /**
  2416. * __skb_frag_unref - release a reference on a paged fragment.
  2417. * @frag: the paged fragment
  2418. *
  2419. * Releases a reference on the paged fragment @frag.
  2420. */
  2421. static inline void __skb_frag_unref(skb_frag_t *frag)
  2422. {
  2423. put_page(skb_frag_page(frag));
  2424. }
  2425. /**
  2426. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2427. * @skb: the buffer
  2428. * @f: the fragment offset
  2429. *
  2430. * Releases a reference on the @f'th paged fragment of @skb.
  2431. */
  2432. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2433. {
  2434. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2435. }
  2436. /**
  2437. * skb_frag_address - gets the address of the data contained in a paged fragment
  2438. * @frag: the paged fragment buffer
  2439. *
  2440. * Returns the address of the data within @frag. The page must already
  2441. * be mapped.
  2442. */
  2443. static inline void *skb_frag_address(const skb_frag_t *frag)
  2444. {
  2445. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2446. }
  2447. /**
  2448. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2449. * @frag: the paged fragment buffer
  2450. *
  2451. * Returns the address of the data within @frag. Checks that the page
  2452. * is mapped and returns %NULL otherwise.
  2453. */
  2454. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2455. {
  2456. void *ptr = page_address(skb_frag_page(frag));
  2457. if (unlikely(!ptr))
  2458. return NULL;
  2459. return ptr + frag->page_offset;
  2460. }
  2461. /**
  2462. * __skb_frag_set_page - sets the page contained in a paged fragment
  2463. * @frag: the paged fragment
  2464. * @page: the page to set
  2465. *
  2466. * Sets the fragment @frag to contain @page.
  2467. */
  2468. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2469. {
  2470. frag->page.p = page;
  2471. }
  2472. /**
  2473. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2474. * @skb: the buffer
  2475. * @f: the fragment offset
  2476. * @page: the page to set
  2477. *
  2478. * Sets the @f'th fragment of @skb to contain @page.
  2479. */
  2480. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2481. struct page *page)
  2482. {
  2483. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2484. }
  2485. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2486. /**
  2487. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2488. * @dev: the device to map the fragment to
  2489. * @frag: the paged fragment to map
  2490. * @offset: the offset within the fragment (starting at the
  2491. * fragment's own offset)
  2492. * @size: the number of bytes to map
  2493. * @dir: the direction of the mapping (``PCI_DMA_*``)
  2494. *
  2495. * Maps the page associated with @frag to @device.
  2496. */
  2497. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2498. const skb_frag_t *frag,
  2499. size_t offset, size_t size,
  2500. enum dma_data_direction dir)
  2501. {
  2502. return dma_map_page(dev, skb_frag_page(frag),
  2503. frag->page_offset + offset, size, dir);
  2504. }
  2505. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2506. gfp_t gfp_mask)
  2507. {
  2508. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2509. }
  2510. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2511. gfp_t gfp_mask)
  2512. {
  2513. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2514. }
  2515. /**
  2516. * skb_clone_writable - is the header of a clone writable
  2517. * @skb: buffer to check
  2518. * @len: length up to which to write
  2519. *
  2520. * Returns true if modifying the header part of the cloned buffer
  2521. * does not requires the data to be copied.
  2522. */
  2523. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2524. {
  2525. return !skb_header_cloned(skb) &&
  2526. skb_headroom(skb) + len <= skb->hdr_len;
  2527. }
  2528. static inline int skb_try_make_writable(struct sk_buff *skb,
  2529. unsigned int write_len)
  2530. {
  2531. return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
  2532. pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
  2533. }
  2534. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2535. int cloned)
  2536. {
  2537. int delta = 0;
  2538. if (headroom > skb_headroom(skb))
  2539. delta = headroom - skb_headroom(skb);
  2540. if (delta || cloned)
  2541. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2542. GFP_ATOMIC);
  2543. return 0;
  2544. }
  2545. /**
  2546. * skb_cow - copy header of skb when it is required
  2547. * @skb: buffer to cow
  2548. * @headroom: needed headroom
  2549. *
  2550. * If the skb passed lacks sufficient headroom or its data part
  2551. * is shared, data is reallocated. If reallocation fails, an error
  2552. * is returned and original skb is not changed.
  2553. *
  2554. * The result is skb with writable area skb->head...skb->tail
  2555. * and at least @headroom of space at head.
  2556. */
  2557. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2558. {
  2559. return __skb_cow(skb, headroom, skb_cloned(skb));
  2560. }
  2561. /**
  2562. * skb_cow_head - skb_cow but only making the head writable
  2563. * @skb: buffer to cow
  2564. * @headroom: needed headroom
  2565. *
  2566. * This function is identical to skb_cow except that we replace the
  2567. * skb_cloned check by skb_header_cloned. It should be used when
  2568. * you only need to push on some header and do not need to modify
  2569. * the data.
  2570. */
  2571. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2572. {
  2573. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2574. }
  2575. /**
  2576. * skb_padto - pad an skbuff up to a minimal size
  2577. * @skb: buffer to pad
  2578. * @len: minimal length
  2579. *
  2580. * Pads up a buffer to ensure the trailing bytes exist and are
  2581. * blanked. If the buffer already contains sufficient data it
  2582. * is untouched. Otherwise it is extended. Returns zero on
  2583. * success. The skb is freed on error.
  2584. */
  2585. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2586. {
  2587. unsigned int size = skb->len;
  2588. if (likely(size >= len))
  2589. return 0;
  2590. return skb_pad(skb, len - size);
  2591. }
  2592. /**
  2593. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2594. * @skb: buffer to pad
  2595. * @len: minimal length
  2596. * @free_on_error: free buffer on error
  2597. *
  2598. * Pads up a buffer to ensure the trailing bytes exist and are
  2599. * blanked. If the buffer already contains sufficient data it
  2600. * is untouched. Otherwise it is extended. Returns zero on
  2601. * success. The skb is freed on error if @free_on_error is true.
  2602. */
  2603. static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
  2604. bool free_on_error)
  2605. {
  2606. unsigned int size = skb->len;
  2607. if (unlikely(size < len)) {
  2608. len -= size;
  2609. if (__skb_pad(skb, len, free_on_error))
  2610. return -ENOMEM;
  2611. __skb_put(skb, len);
  2612. }
  2613. return 0;
  2614. }
  2615. /**
  2616. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2617. * @skb: buffer to pad
  2618. * @len: minimal length
  2619. *
  2620. * Pads up a buffer to ensure the trailing bytes exist and are
  2621. * blanked. If the buffer already contains sufficient data it
  2622. * is untouched. Otherwise it is extended. Returns zero on
  2623. * success. The skb is freed on error.
  2624. */
  2625. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2626. {
  2627. return __skb_put_padto(skb, len, true);
  2628. }
  2629. static inline int skb_add_data(struct sk_buff *skb,
  2630. struct iov_iter *from, int copy)
  2631. {
  2632. const int off = skb->len;
  2633. if (skb->ip_summed == CHECKSUM_NONE) {
  2634. __wsum csum = 0;
  2635. if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
  2636. &csum, from)) {
  2637. skb->csum = csum_block_add(skb->csum, csum, off);
  2638. return 0;
  2639. }
  2640. } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
  2641. return 0;
  2642. __skb_trim(skb, off);
  2643. return -EFAULT;
  2644. }
  2645. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2646. const struct page *page, int off)
  2647. {
  2648. if (skb_zcopy(skb))
  2649. return false;
  2650. if (i) {
  2651. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2652. return page == skb_frag_page(frag) &&
  2653. off == frag->page_offset + skb_frag_size(frag);
  2654. }
  2655. return false;
  2656. }
  2657. static inline int __skb_linearize(struct sk_buff *skb)
  2658. {
  2659. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2660. }
  2661. /**
  2662. * skb_linearize - convert paged skb to linear one
  2663. * @skb: buffer to linarize
  2664. *
  2665. * If there is no free memory -ENOMEM is returned, otherwise zero
  2666. * is returned and the old skb data released.
  2667. */
  2668. static inline int skb_linearize(struct sk_buff *skb)
  2669. {
  2670. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2671. }
  2672. /**
  2673. * skb_has_shared_frag - can any frag be overwritten
  2674. * @skb: buffer to test
  2675. *
  2676. * Return true if the skb has at least one frag that might be modified
  2677. * by an external entity (as in vmsplice()/sendfile())
  2678. */
  2679. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2680. {
  2681. return skb_is_nonlinear(skb) &&
  2682. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2683. }
  2684. /**
  2685. * skb_linearize_cow - make sure skb is linear and writable
  2686. * @skb: buffer to process
  2687. *
  2688. * If there is no free memory -ENOMEM is returned, otherwise zero
  2689. * is returned and the old skb data released.
  2690. */
  2691. static inline int skb_linearize_cow(struct sk_buff *skb)
  2692. {
  2693. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2694. __skb_linearize(skb) : 0;
  2695. }
  2696. static __always_inline void
  2697. __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2698. unsigned int off)
  2699. {
  2700. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2701. skb->csum = csum_block_sub(skb->csum,
  2702. csum_partial(start, len, 0), off);
  2703. else if (skb->ip_summed == CHECKSUM_PARTIAL &&
  2704. skb_checksum_start_offset(skb) < 0)
  2705. skb->ip_summed = CHECKSUM_NONE;
  2706. }
  2707. /**
  2708. * skb_postpull_rcsum - update checksum for received skb after pull
  2709. * @skb: buffer to update
  2710. * @start: start of data before pull
  2711. * @len: length of data pulled
  2712. *
  2713. * After doing a pull on a received packet, you need to call this to
  2714. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2715. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2716. */
  2717. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2718. const void *start, unsigned int len)
  2719. {
  2720. __skb_postpull_rcsum(skb, start, len, 0);
  2721. }
  2722. static __always_inline void
  2723. __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2724. unsigned int off)
  2725. {
  2726. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2727. skb->csum = csum_block_add(skb->csum,
  2728. csum_partial(start, len, 0), off);
  2729. }
  2730. /**
  2731. * skb_postpush_rcsum - update checksum for received skb after push
  2732. * @skb: buffer to update
  2733. * @start: start of data after push
  2734. * @len: length of data pushed
  2735. *
  2736. * After doing a push on a received packet, you need to call this to
  2737. * update the CHECKSUM_COMPLETE checksum.
  2738. */
  2739. static inline void skb_postpush_rcsum(struct sk_buff *skb,
  2740. const void *start, unsigned int len)
  2741. {
  2742. __skb_postpush_rcsum(skb, start, len, 0);
  2743. }
  2744. void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2745. /**
  2746. * skb_push_rcsum - push skb and update receive checksum
  2747. * @skb: buffer to update
  2748. * @len: length of data pulled
  2749. *
  2750. * This function performs an skb_push on the packet and updates
  2751. * the CHECKSUM_COMPLETE checksum. It should be used on
  2752. * receive path processing instead of skb_push unless you know
  2753. * that the checksum difference is zero (e.g., a valid IP header)
  2754. * or you are setting ip_summed to CHECKSUM_NONE.
  2755. */
  2756. static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
  2757. {
  2758. skb_push(skb, len);
  2759. skb_postpush_rcsum(skb, skb->data, len);
  2760. return skb->data;
  2761. }
  2762. /**
  2763. * pskb_trim_rcsum - trim received skb and update checksum
  2764. * @skb: buffer to trim
  2765. * @len: new length
  2766. *
  2767. * This is exactly the same as pskb_trim except that it ensures the
  2768. * checksum of received packets are still valid after the operation.
  2769. */
  2770. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2771. {
  2772. if (likely(len >= skb->len))
  2773. return 0;
  2774. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2775. skb->ip_summed = CHECKSUM_NONE;
  2776. return __pskb_trim(skb, len);
  2777. }
  2778. static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2779. {
  2780. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2781. skb->ip_summed = CHECKSUM_NONE;
  2782. __skb_trim(skb, len);
  2783. return 0;
  2784. }
  2785. static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
  2786. {
  2787. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2788. skb->ip_summed = CHECKSUM_NONE;
  2789. return __skb_grow(skb, len);
  2790. }
  2791. #define skb_queue_walk(queue, skb) \
  2792. for (skb = (queue)->next; \
  2793. skb != (struct sk_buff *)(queue); \
  2794. skb = skb->next)
  2795. #define skb_queue_walk_safe(queue, skb, tmp) \
  2796. for (skb = (queue)->next, tmp = skb->next; \
  2797. skb != (struct sk_buff *)(queue); \
  2798. skb = tmp, tmp = skb->next)
  2799. #define skb_queue_walk_from(queue, skb) \
  2800. for (; skb != (struct sk_buff *)(queue); \
  2801. skb = skb->next)
  2802. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2803. for (tmp = skb->next; \
  2804. skb != (struct sk_buff *)(queue); \
  2805. skb = tmp, tmp = skb->next)
  2806. #define skb_queue_reverse_walk(queue, skb) \
  2807. for (skb = (queue)->prev; \
  2808. skb != (struct sk_buff *)(queue); \
  2809. skb = skb->prev)
  2810. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2811. for (skb = (queue)->prev, tmp = skb->prev; \
  2812. skb != (struct sk_buff *)(queue); \
  2813. skb = tmp, tmp = skb->prev)
  2814. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2815. for (tmp = skb->prev; \
  2816. skb != (struct sk_buff *)(queue); \
  2817. skb = tmp, tmp = skb->prev)
  2818. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2819. {
  2820. return skb_shinfo(skb)->frag_list != NULL;
  2821. }
  2822. static inline void skb_frag_list_init(struct sk_buff *skb)
  2823. {
  2824. skb_shinfo(skb)->frag_list = NULL;
  2825. }
  2826. #define skb_walk_frags(skb, iter) \
  2827. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2828. int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
  2829. const struct sk_buff *skb);
  2830. struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
  2831. struct sk_buff_head *queue,
  2832. unsigned int flags,
  2833. void (*destructor)(struct sock *sk,
  2834. struct sk_buff *skb),
  2835. int *peeked, int *off, int *err,
  2836. struct sk_buff **last);
  2837. struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
  2838. void (*destructor)(struct sock *sk,
  2839. struct sk_buff *skb),
  2840. int *peeked, int *off, int *err,
  2841. struct sk_buff **last);
  2842. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2843. void (*destructor)(struct sock *sk,
  2844. struct sk_buff *skb),
  2845. int *peeked, int *off, int *err);
  2846. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2847. int *err);
  2848. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2849. struct poll_table_struct *wait);
  2850. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2851. struct iov_iter *to, int size);
  2852. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2853. struct msghdr *msg, int size)
  2854. {
  2855. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2856. }
  2857. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2858. struct msghdr *msg);
  2859. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2860. struct iov_iter *from, int len);
  2861. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2862. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2863. void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
  2864. static inline void skb_free_datagram_locked(struct sock *sk,
  2865. struct sk_buff *skb)
  2866. {
  2867. __skb_free_datagram_locked(sk, skb, 0);
  2868. }
  2869. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2870. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2871. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2872. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2873. int len, __wsum csum);
  2874. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2875. struct pipe_inode_info *pipe, unsigned int len,
  2876. unsigned int flags);
  2877. int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
  2878. int len);
  2879. int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
  2880. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2881. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2882. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2883. int len, int hlen);
  2884. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2885. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2886. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2887. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2888. bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
  2889. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2890. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2891. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2892. int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
  2893. int skb_vlan_pop(struct sk_buff *skb);
  2894. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2895. struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
  2896. gfp_t gfp);
  2897. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2898. {
  2899. return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
  2900. }
  2901. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2902. {
  2903. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2904. }
  2905. struct skb_checksum_ops {
  2906. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2907. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2908. };
  2909. extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
  2910. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2911. __wsum csum, const struct skb_checksum_ops *ops);
  2912. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2913. __wsum csum);
  2914. static inline void * __must_check
  2915. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2916. int len, void *data, int hlen, void *buffer)
  2917. {
  2918. if (hlen - offset >= len)
  2919. return data + offset;
  2920. if (!skb ||
  2921. skb_copy_bits(skb, offset, buffer, len) < 0)
  2922. return NULL;
  2923. return buffer;
  2924. }
  2925. static inline void * __must_check
  2926. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2927. {
  2928. return __skb_header_pointer(skb, offset, len, skb->data,
  2929. skb_headlen(skb), buffer);
  2930. }
  2931. /**
  2932. * skb_needs_linearize - check if we need to linearize a given skb
  2933. * depending on the given device features.
  2934. * @skb: socket buffer to check
  2935. * @features: net device features
  2936. *
  2937. * Returns true if either:
  2938. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2939. * 2. skb is fragmented and the device does not support SG.
  2940. */
  2941. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2942. netdev_features_t features)
  2943. {
  2944. return skb_is_nonlinear(skb) &&
  2945. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2946. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2947. }
  2948. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2949. void *to,
  2950. const unsigned int len)
  2951. {
  2952. memcpy(to, skb->data, len);
  2953. }
  2954. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2955. const int offset, void *to,
  2956. const unsigned int len)
  2957. {
  2958. memcpy(to, skb->data + offset, len);
  2959. }
  2960. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2961. const void *from,
  2962. const unsigned int len)
  2963. {
  2964. memcpy(skb->data, from, len);
  2965. }
  2966. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2967. const int offset,
  2968. const void *from,
  2969. const unsigned int len)
  2970. {
  2971. memcpy(skb->data + offset, from, len);
  2972. }
  2973. void skb_init(void);
  2974. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2975. {
  2976. return skb->tstamp;
  2977. }
  2978. /**
  2979. * skb_get_timestamp - get timestamp from a skb
  2980. * @skb: skb to get stamp from
  2981. * @stamp: pointer to struct timeval to store stamp in
  2982. *
  2983. * Timestamps are stored in the skb as offsets to a base timestamp.
  2984. * This function converts the offset back to a struct timeval and stores
  2985. * it in stamp.
  2986. */
  2987. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2988. struct timeval *stamp)
  2989. {
  2990. *stamp = ktime_to_timeval(skb->tstamp);
  2991. }
  2992. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2993. struct timespec *stamp)
  2994. {
  2995. *stamp = ktime_to_timespec(skb->tstamp);
  2996. }
  2997. static inline void __net_timestamp(struct sk_buff *skb)
  2998. {
  2999. skb->tstamp = ktime_get_real();
  3000. }
  3001. static inline ktime_t net_timedelta(ktime_t t)
  3002. {
  3003. return ktime_sub(ktime_get_real(), t);
  3004. }
  3005. static inline ktime_t net_invalid_timestamp(void)
  3006. {
  3007. return 0;
  3008. }
  3009. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  3010. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  3011. void skb_clone_tx_timestamp(struct sk_buff *skb);
  3012. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  3013. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  3014. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  3015. {
  3016. }
  3017. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  3018. {
  3019. return false;
  3020. }
  3021. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  3022. /**
  3023. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  3024. *
  3025. * PHY drivers may accept clones of transmitted packets for
  3026. * timestamping via their phy_driver.txtstamp method. These drivers
  3027. * must call this function to return the skb back to the stack with a
  3028. * timestamp.
  3029. *
  3030. * @skb: clone of the the original outgoing packet
  3031. * @hwtstamps: hardware time stamps
  3032. *
  3033. */
  3034. void skb_complete_tx_timestamp(struct sk_buff *skb,
  3035. struct skb_shared_hwtstamps *hwtstamps);
  3036. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  3037. struct skb_shared_hwtstamps *hwtstamps,
  3038. struct sock *sk, int tstype);
  3039. /**
  3040. * skb_tstamp_tx - queue clone of skb with send time stamps
  3041. * @orig_skb: the original outgoing packet
  3042. * @hwtstamps: hardware time stamps, may be NULL if not available
  3043. *
  3044. * If the skb has a socket associated, then this function clones the
  3045. * skb (thus sharing the actual data and optional structures), stores
  3046. * the optional hardware time stamping information (if non NULL) or
  3047. * generates a software time stamp (otherwise), then queues the clone
  3048. * to the error queue of the socket. Errors are silently ignored.
  3049. */
  3050. void skb_tstamp_tx(struct sk_buff *orig_skb,
  3051. struct skb_shared_hwtstamps *hwtstamps);
  3052. /**
  3053. * skb_tx_timestamp() - Driver hook for transmit timestamping
  3054. *
  3055. * Ethernet MAC Drivers should call this function in their hard_xmit()
  3056. * function immediately before giving the sk_buff to the MAC hardware.
  3057. *
  3058. * Specifically, one should make absolutely sure that this function is
  3059. * called before TX completion of this packet can trigger. Otherwise
  3060. * the packet could potentially already be freed.
  3061. *
  3062. * @skb: A socket buffer.
  3063. */
  3064. static inline void skb_tx_timestamp(struct sk_buff *skb)
  3065. {
  3066. skb_clone_tx_timestamp(skb);
  3067. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
  3068. skb_tstamp_tx(skb, NULL);
  3069. }
  3070. /**
  3071. * skb_complete_wifi_ack - deliver skb with wifi status
  3072. *
  3073. * @skb: the original outgoing packet
  3074. * @acked: ack status
  3075. *
  3076. */
  3077. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  3078. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  3079. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  3080. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  3081. {
  3082. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  3083. skb->csum_valid ||
  3084. (skb->ip_summed == CHECKSUM_PARTIAL &&
  3085. skb_checksum_start_offset(skb) >= 0));
  3086. }
  3087. /**
  3088. * skb_checksum_complete - Calculate checksum of an entire packet
  3089. * @skb: packet to process
  3090. *
  3091. * This function calculates the checksum over the entire packet plus
  3092. * the value of skb->csum. The latter can be used to supply the
  3093. * checksum of a pseudo header as used by TCP/UDP. It returns the
  3094. * checksum.
  3095. *
  3096. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  3097. * this function can be used to verify that checksum on received
  3098. * packets. In that case the function should return zero if the
  3099. * checksum is correct. In particular, this function will return zero
  3100. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  3101. * hardware has already verified the correctness of the checksum.
  3102. */
  3103. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  3104. {
  3105. return skb_csum_unnecessary(skb) ?
  3106. 0 : __skb_checksum_complete(skb);
  3107. }
  3108. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  3109. {
  3110. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  3111. if (skb->csum_level == 0)
  3112. skb->ip_summed = CHECKSUM_NONE;
  3113. else
  3114. skb->csum_level--;
  3115. }
  3116. }
  3117. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  3118. {
  3119. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  3120. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  3121. skb->csum_level++;
  3122. } else if (skb->ip_summed == CHECKSUM_NONE) {
  3123. skb->ip_summed = CHECKSUM_UNNECESSARY;
  3124. skb->csum_level = 0;
  3125. }
  3126. }
  3127. /* Check if we need to perform checksum complete validation.
  3128. *
  3129. * Returns true if checksum complete is needed, false otherwise
  3130. * (either checksum is unnecessary or zero checksum is allowed).
  3131. */
  3132. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  3133. bool zero_okay,
  3134. __sum16 check)
  3135. {
  3136. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  3137. skb->csum_valid = 1;
  3138. __skb_decr_checksum_unnecessary(skb);
  3139. return false;
  3140. }
  3141. return true;
  3142. }
  3143. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  3144. * in checksum_init.
  3145. */
  3146. #define CHECKSUM_BREAK 76
  3147. /* Unset checksum-complete
  3148. *
  3149. * Unset checksum complete can be done when packet is being modified
  3150. * (uncompressed for instance) and checksum-complete value is
  3151. * invalidated.
  3152. */
  3153. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  3154. {
  3155. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3156. skb->ip_summed = CHECKSUM_NONE;
  3157. }
  3158. /* Validate (init) checksum based on checksum complete.
  3159. *
  3160. * Return values:
  3161. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  3162. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  3163. * checksum is stored in skb->csum for use in __skb_checksum_complete
  3164. * non-zero: value of invalid checksum
  3165. *
  3166. */
  3167. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  3168. bool complete,
  3169. __wsum psum)
  3170. {
  3171. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  3172. if (!csum_fold(csum_add(psum, skb->csum))) {
  3173. skb->csum_valid = 1;
  3174. return 0;
  3175. }
  3176. }
  3177. skb->csum = psum;
  3178. if (complete || skb->len <= CHECKSUM_BREAK) {
  3179. __sum16 csum;
  3180. csum = __skb_checksum_complete(skb);
  3181. skb->csum_valid = !csum;
  3182. return csum;
  3183. }
  3184. return 0;
  3185. }
  3186. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  3187. {
  3188. return 0;
  3189. }
  3190. /* Perform checksum validate (init). Note that this is a macro since we only
  3191. * want to calculate the pseudo header which is an input function if necessary.
  3192. * First we try to validate without any computation (checksum unnecessary) and
  3193. * then calculate based on checksum complete calling the function to compute
  3194. * pseudo header.
  3195. *
  3196. * Return values:
  3197. * 0: checksum is validated or try to in skb_checksum_complete
  3198. * non-zero: value of invalid checksum
  3199. */
  3200. #define __skb_checksum_validate(skb, proto, complete, \
  3201. zero_okay, check, compute_pseudo) \
  3202. ({ \
  3203. __sum16 __ret = 0; \
  3204. skb->csum_valid = 0; \
  3205. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  3206. __ret = __skb_checksum_validate_complete(skb, \
  3207. complete, compute_pseudo(skb, proto)); \
  3208. __ret; \
  3209. })
  3210. #define skb_checksum_init(skb, proto, compute_pseudo) \
  3211. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  3212. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  3213. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  3214. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  3215. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  3216. #define skb_checksum_validate_zero_check(skb, proto, check, \
  3217. compute_pseudo) \
  3218. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  3219. #define skb_checksum_simple_validate(skb) \
  3220. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  3221. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  3222. {
  3223. return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
  3224. }
  3225. static inline void __skb_checksum_convert(struct sk_buff *skb,
  3226. __sum16 check, __wsum pseudo)
  3227. {
  3228. skb->csum = ~pseudo;
  3229. skb->ip_summed = CHECKSUM_COMPLETE;
  3230. }
  3231. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  3232. do { \
  3233. if (__skb_checksum_convert_check(skb)) \
  3234. __skb_checksum_convert(skb, check, \
  3235. compute_pseudo(skb, proto)); \
  3236. } while (0)
  3237. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  3238. u16 start, u16 offset)
  3239. {
  3240. skb->ip_summed = CHECKSUM_PARTIAL;
  3241. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  3242. skb->csum_offset = offset - start;
  3243. }
  3244. /* Update skbuf and packet to reflect the remote checksum offload operation.
  3245. * When called, ptr indicates the starting point for skb->csum when
  3246. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  3247. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  3248. */
  3249. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  3250. int start, int offset, bool nopartial)
  3251. {
  3252. __wsum delta;
  3253. if (!nopartial) {
  3254. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  3255. return;
  3256. }
  3257. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  3258. __skb_checksum_complete(skb);
  3259. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  3260. }
  3261. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  3262. /* Adjust skb->csum since we changed the packet */
  3263. skb->csum = csum_add(skb->csum, delta);
  3264. }
  3265. static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
  3266. {
  3267. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  3268. return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
  3269. #else
  3270. return NULL;
  3271. #endif
  3272. }
  3273. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3274. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  3275. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  3276. {
  3277. if (nfct && atomic_dec_and_test(&nfct->use))
  3278. nf_conntrack_destroy(nfct);
  3279. }
  3280. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  3281. {
  3282. if (nfct)
  3283. atomic_inc(&nfct->use);
  3284. }
  3285. #endif
  3286. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3287. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  3288. {
  3289. if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
  3290. kfree(nf_bridge);
  3291. }
  3292. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  3293. {
  3294. if (nf_bridge)
  3295. refcount_inc(&nf_bridge->use);
  3296. }
  3297. #endif /* CONFIG_BRIDGE_NETFILTER */
  3298. static inline void nf_reset(struct sk_buff *skb)
  3299. {
  3300. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3301. nf_conntrack_put(skb_nfct(skb));
  3302. skb->_nfct = 0;
  3303. #endif
  3304. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3305. nf_bridge_put(skb->nf_bridge);
  3306. skb->nf_bridge = NULL;
  3307. #endif
  3308. }
  3309. static inline void nf_reset_trace(struct sk_buff *skb)
  3310. {
  3311. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3312. skb->nf_trace = 0;
  3313. #endif
  3314. }
  3315. static inline void ipvs_reset(struct sk_buff *skb)
  3316. {
  3317. #if IS_ENABLED(CONFIG_IP_VS)
  3318. skb->ipvs_property = 0;
  3319. #endif
  3320. }
  3321. /* Note: This doesn't put any conntrack and bridge info in dst. */
  3322. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  3323. bool copy)
  3324. {
  3325. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3326. dst->_nfct = src->_nfct;
  3327. nf_conntrack_get(skb_nfct(src));
  3328. #endif
  3329. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3330. dst->nf_bridge = src->nf_bridge;
  3331. nf_bridge_get(src->nf_bridge);
  3332. #endif
  3333. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3334. if (copy)
  3335. dst->nf_trace = src->nf_trace;
  3336. #endif
  3337. }
  3338. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  3339. {
  3340. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3341. nf_conntrack_put(skb_nfct(dst));
  3342. #endif
  3343. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3344. nf_bridge_put(dst->nf_bridge);
  3345. #endif
  3346. __nf_copy(dst, src, true);
  3347. }
  3348. #ifdef CONFIG_NETWORK_SECMARK
  3349. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3350. {
  3351. to->secmark = from->secmark;
  3352. }
  3353. static inline void skb_init_secmark(struct sk_buff *skb)
  3354. {
  3355. skb->secmark = 0;
  3356. }
  3357. #else
  3358. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3359. { }
  3360. static inline void skb_init_secmark(struct sk_buff *skb)
  3361. { }
  3362. #endif
  3363. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  3364. {
  3365. return !skb->destructor &&
  3366. #if IS_ENABLED(CONFIG_XFRM)
  3367. !skb->sp &&
  3368. #endif
  3369. !skb_nfct(skb) &&
  3370. !skb->_skb_refdst &&
  3371. !skb_has_frag_list(skb);
  3372. }
  3373. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  3374. {
  3375. skb->queue_mapping = queue_mapping;
  3376. }
  3377. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  3378. {
  3379. return skb->queue_mapping;
  3380. }
  3381. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  3382. {
  3383. to->queue_mapping = from->queue_mapping;
  3384. }
  3385. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  3386. {
  3387. skb->queue_mapping = rx_queue + 1;
  3388. }
  3389. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  3390. {
  3391. return skb->queue_mapping - 1;
  3392. }
  3393. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  3394. {
  3395. return skb->queue_mapping != 0;
  3396. }
  3397. static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
  3398. {
  3399. skb->dst_pending_confirm = val;
  3400. }
  3401. static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
  3402. {
  3403. return skb->dst_pending_confirm != 0;
  3404. }
  3405. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  3406. {
  3407. #ifdef CONFIG_XFRM
  3408. return skb->sp;
  3409. #else
  3410. return NULL;
  3411. #endif
  3412. }
  3413. /* Keeps track of mac header offset relative to skb->head.
  3414. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  3415. * For non-tunnel skb it points to skb_mac_header() and for
  3416. * tunnel skb it points to outer mac header.
  3417. * Keeps track of level of encapsulation of network headers.
  3418. */
  3419. struct skb_gso_cb {
  3420. union {
  3421. int mac_offset;
  3422. int data_offset;
  3423. };
  3424. int encap_level;
  3425. __wsum csum;
  3426. __u16 csum_start;
  3427. };
  3428. #define SKB_SGO_CB_OFFSET 32
  3429. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
  3430. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  3431. {
  3432. return (skb_mac_header(inner_skb) - inner_skb->head) -
  3433. SKB_GSO_CB(inner_skb)->mac_offset;
  3434. }
  3435. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  3436. {
  3437. int new_headroom, headroom;
  3438. int ret;
  3439. headroom = skb_headroom(skb);
  3440. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  3441. if (ret)
  3442. return ret;
  3443. new_headroom = skb_headroom(skb);
  3444. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  3445. return 0;
  3446. }
  3447. static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
  3448. {
  3449. /* Do not update partial checksums if remote checksum is enabled. */
  3450. if (skb->remcsum_offload)
  3451. return;
  3452. SKB_GSO_CB(skb)->csum = res;
  3453. SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
  3454. }
  3455. /* Compute the checksum for a gso segment. First compute the checksum value
  3456. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  3457. * then add in skb->csum (checksum from csum_start to end of packet).
  3458. * skb->csum and csum_start are then updated to reflect the checksum of the
  3459. * resultant packet starting from the transport header-- the resultant checksum
  3460. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  3461. * header.
  3462. */
  3463. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  3464. {
  3465. unsigned char *csum_start = skb_transport_header(skb);
  3466. int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
  3467. __wsum partial = SKB_GSO_CB(skb)->csum;
  3468. SKB_GSO_CB(skb)->csum = res;
  3469. SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
  3470. return csum_fold(csum_partial(csum_start, plen, partial));
  3471. }
  3472. static inline bool skb_is_gso(const struct sk_buff *skb)
  3473. {
  3474. return skb_shinfo(skb)->gso_size;
  3475. }
  3476. /* Note: Should be called only if skb_is_gso(skb) is true */
  3477. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  3478. {
  3479. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  3480. }
  3481. static inline void skb_gso_reset(struct sk_buff *skb)
  3482. {
  3483. skb_shinfo(skb)->gso_size = 0;
  3484. skb_shinfo(skb)->gso_segs = 0;
  3485. skb_shinfo(skb)->gso_type = 0;
  3486. }
  3487. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  3488. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  3489. {
  3490. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  3491. * wanted then gso_type will be set. */
  3492. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  3493. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  3494. unlikely(shinfo->gso_type == 0)) {
  3495. __skb_warn_lro_forwarding(skb);
  3496. return true;
  3497. }
  3498. return false;
  3499. }
  3500. static inline void skb_forward_csum(struct sk_buff *skb)
  3501. {
  3502. /* Unfortunately we don't support this one. Any brave souls? */
  3503. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3504. skb->ip_summed = CHECKSUM_NONE;
  3505. }
  3506. /**
  3507. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3508. * @skb: skb to check
  3509. *
  3510. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3511. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3512. * use this helper, to document places where we make this assertion.
  3513. */
  3514. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3515. {
  3516. #ifdef DEBUG
  3517. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3518. #endif
  3519. }
  3520. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3521. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3522. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3523. unsigned int transport_len,
  3524. __sum16(*skb_chkf)(struct sk_buff *skb));
  3525. /**
  3526. * skb_head_is_locked - Determine if the skb->head is locked down
  3527. * @skb: skb to check
  3528. *
  3529. * The head on skbs build around a head frag can be removed if they are
  3530. * not cloned. This function returns true if the skb head is locked down
  3531. * due to either being allocated via kmalloc, or by being a clone with
  3532. * multiple references to the head.
  3533. */
  3534. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3535. {
  3536. return !skb->head_frag || skb_cloned(skb);
  3537. }
  3538. /**
  3539. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3540. *
  3541. * @skb: GSO skb
  3542. *
  3543. * skb_gso_network_seglen is used to determine the real size of the
  3544. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3545. *
  3546. * The MAC/L2 header is not accounted for.
  3547. */
  3548. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3549. {
  3550. unsigned int hdr_len = skb_transport_header(skb) -
  3551. skb_network_header(skb);
  3552. return hdr_len + skb_gso_transport_seglen(skb);
  3553. }
  3554. /* Local Checksum Offload.
  3555. * Compute outer checksum based on the assumption that the
  3556. * inner checksum will be offloaded later.
  3557. * See Documentation/networking/checksum-offloads.txt for
  3558. * explanation of how this works.
  3559. * Fill in outer checksum adjustment (e.g. with sum of outer
  3560. * pseudo-header) before calling.
  3561. * Also ensure that inner checksum is in linear data area.
  3562. */
  3563. static inline __wsum lco_csum(struct sk_buff *skb)
  3564. {
  3565. unsigned char *csum_start = skb_checksum_start(skb);
  3566. unsigned char *l4_hdr = skb_transport_header(skb);
  3567. __wsum partial;
  3568. /* Start with complement of inner checksum adjustment */
  3569. partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
  3570. skb->csum_offset));
  3571. /* Add in checksum of our headers (incl. outer checksum
  3572. * adjustment filled in by caller) and return result.
  3573. */
  3574. return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
  3575. }
  3576. #endif /* __KERNEL__ */
  3577. #endif /* _LINUX_SKBUFF_H */