skbuff.h 111 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963296429652966296729682969297029712972297329742975297629772978297929802981298229832984298529862987298829892990299129922993299429952996299729982999300030013002300330043005300630073008300930103011301230133014301530163017301830193020302130223023302430253026302730283029303030313032303330343035303630373038303930403041304230433044304530463047304830493050305130523053305430553056305730583059306030613062306330643065306630673068306930703071307230733074307530763077307830793080308130823083308430853086308730883089309030913092309330943095309630973098309931003101310231033104310531063107310831093110311131123113311431153116311731183119312031213122312331243125312631273128312931303131313231333134313531363137313831393140314131423143314431453146314731483149315031513152315331543155315631573158315931603161316231633164316531663167316831693170317131723173317431753176317731783179318031813182318331843185318631873188318931903191319231933194319531963197319831993200320132023203320432053206320732083209321032113212321332143215321632173218321932203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278327932803281328232833284328532863287328832893290329132923293329432953296329732983299330033013302330333043305330633073308330933103311331233133314331533163317331833193320332133223323332433253326332733283329333033313332333333343335333633373338333933403341334233433344334533463347334833493350335133523353335433553356335733583359336033613362336333643365336633673368336933703371337233733374337533763377337833793380338133823383338433853386338733883389339033913392339333943395339633973398339934003401340234033404340534063407340834093410341134123413341434153416341734183419342034213422342334243425342634273428342934303431343234333434343534363437343834393440344134423443344434453446344734483449345034513452345334543455345634573458345934603461346234633464346534663467346834693470347134723473347434753476347734783479348034813482348334843485348634873488348934903491349234933494349534963497349834993500350135023503350435053506350735083509351035113512351335143515351635173518351935203521352235233524352535263527352835293530353135323533353435353536353735383539354035413542354335443545354635473548354935503551355235533554355535563557355835593560356135623563356435653566356735683569357035713572357335743575357635773578357935803581358235833584358535863587358835893590359135923593359435953596359735983599360036013602360336043605360636073608360936103611361236133614361536163617361836193620362136223623362436253626362736283629363036313632363336343635363636373638363936403641364236433644364536463647364836493650365136523653365436553656365736583659366036613662366336643665366636673668366936703671367236733674367536763677367836793680368136823683368436853686368736883689369036913692369336943695369636973698369937003701370237033704370537063707370837093710371137123713371437153716371737183719372037213722372337243725372637273728372937303731373237333734373537363737373837393740374137423743374437453746374737483749375037513752375337543755375637573758375937603761376237633764376537663767376837693770377137723773377437753776377737783779378037813782378337843785378637873788378937903791379237933794379537963797379837993800380138023803380438053806380738083809381038113812381338143815381638173818381938203821382238233824382538263827382838293830383138323833383438353836383738383839384038413842384338443845384638473848384938503851385238533854385538563857385838593860386138623863386438653866386738683869
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/kernel.h>
  16. #include <linux/kmemcheck.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/bug.h>
  20. #include <linux/cache.h>
  21. #include <linux/rbtree.h>
  22. #include <linux/socket.h>
  23. #include <linux/atomic.h>
  24. #include <asm/types.h>
  25. #include <linux/spinlock.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #include <linux/rcupdate.h>
  30. #include <linux/hrtimer.h>
  31. #include <linux/dma-mapping.h>
  32. #include <linux/netdev_features.h>
  33. #include <linux/sched.h>
  34. #include <linux/sched/clock.h>
  35. #include <net/flow_dissector.h>
  36. #include <linux/splice.h>
  37. #include <linux/in6.h>
  38. #include <linux/if_packet.h>
  39. #include <net/flow.h>
  40. /* The interface for checksum offload between the stack and networking drivers
  41. * is as follows...
  42. *
  43. * A. IP checksum related features
  44. *
  45. * Drivers advertise checksum offload capabilities in the features of a device.
  46. * From the stack's point of view these are capabilities offered by the driver,
  47. * a driver typically only advertises features that it is capable of offloading
  48. * to its device.
  49. *
  50. * The checksum related features are:
  51. *
  52. * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
  53. * IP (one's complement) checksum for any combination
  54. * of protocols or protocol layering. The checksum is
  55. * computed and set in a packet per the CHECKSUM_PARTIAL
  56. * interface (see below).
  57. *
  58. * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
  59. * TCP or UDP packets over IPv4. These are specifically
  60. * unencapsulated packets of the form IPv4|TCP or
  61. * IPv4|UDP where the Protocol field in the IPv4 header
  62. * is TCP or UDP. The IPv4 header may contain IP options
  63. * This feature cannot be set in features for a device
  64. * with NETIF_F_HW_CSUM also set. This feature is being
  65. * DEPRECATED (see below).
  66. *
  67. * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
  68. * TCP or UDP packets over IPv6. These are specifically
  69. * unencapsulated packets of the form IPv6|TCP or
  70. * IPv4|UDP where the Next Header field in the IPv6
  71. * header is either TCP or UDP. IPv6 extension headers
  72. * are not supported with this feature. This feature
  73. * cannot be set in features for a device with
  74. * NETIF_F_HW_CSUM also set. This feature is being
  75. * DEPRECATED (see below).
  76. *
  77. * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
  78. * This flag is used only used to disable the RX checksum
  79. * feature for a device. The stack will accept receive
  80. * checksum indication in packets received on a device
  81. * regardless of whether NETIF_F_RXCSUM is set.
  82. *
  83. * B. Checksumming of received packets by device. Indication of checksum
  84. * verification is in set skb->ip_summed. Possible values are:
  85. *
  86. * CHECKSUM_NONE:
  87. *
  88. * Device did not checksum this packet e.g. due to lack of capabilities.
  89. * The packet contains full (though not verified) checksum in packet but
  90. * not in skb->csum. Thus, skb->csum is undefined in this case.
  91. *
  92. * CHECKSUM_UNNECESSARY:
  93. *
  94. * The hardware you're dealing with doesn't calculate the full checksum
  95. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  96. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  97. * if their checksums are okay. skb->csum is still undefined in this case
  98. * though. A driver or device must never modify the checksum field in the
  99. * packet even if checksum is verified.
  100. *
  101. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  102. * TCP: IPv6 and IPv4.
  103. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  104. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  105. * may perform further validation in this case.
  106. * GRE: only if the checksum is present in the header.
  107. * SCTP: indicates the CRC in SCTP header has been validated.
  108. * FCOE: indicates the CRC in FC frame has been validated.
  109. *
  110. * skb->csum_level indicates the number of consecutive checksums found in
  111. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  112. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  113. * and a device is able to verify the checksums for UDP (possibly zero),
  114. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  115. * two. If the device were only able to verify the UDP checksum and not
  116. * GRE, either because it doesn't support GRE checksum of because GRE
  117. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  118. * not considered in this case).
  119. *
  120. * CHECKSUM_COMPLETE:
  121. *
  122. * This is the most generic way. The device supplied checksum of the _whole_
  123. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  124. * hardware doesn't need to parse L3/L4 headers to implement this.
  125. *
  126. * Notes:
  127. * - Even if device supports only some protocols, but is able to produce
  128. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  129. * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
  130. *
  131. * CHECKSUM_PARTIAL:
  132. *
  133. * A checksum is set up to be offloaded to a device as described in the
  134. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  135. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  136. * on the same host, or it may be set in the input path in GRO or remote
  137. * checksum offload. For the purposes of checksum verification, the checksum
  138. * referred to by skb->csum_start + skb->csum_offset and any preceding
  139. * checksums in the packet are considered verified. Any checksums in the
  140. * packet that are after the checksum being offloaded are not considered to
  141. * be verified.
  142. *
  143. * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
  144. * in the skb->ip_summed for a packet. Values are:
  145. *
  146. * CHECKSUM_PARTIAL:
  147. *
  148. * The driver is required to checksum the packet as seen by hard_start_xmit()
  149. * from skb->csum_start up to the end, and to record/write the checksum at
  150. * offset skb->csum_start + skb->csum_offset. A driver may verify that the
  151. * csum_start and csum_offset values are valid values given the length and
  152. * offset of the packet, however they should not attempt to validate that the
  153. * checksum refers to a legitimate transport layer checksum-- it is the
  154. * purview of the stack to validate that csum_start and csum_offset are set
  155. * correctly.
  156. *
  157. * When the stack requests checksum offload for a packet, the driver MUST
  158. * ensure that the checksum is set correctly. A driver can either offload the
  159. * checksum calculation to the device, or call skb_checksum_help (in the case
  160. * that the device does not support offload for a particular checksum).
  161. *
  162. * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
  163. * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
  164. * checksum offload capability.
  165. * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
  166. * on network device checksumming capabilities: if a packet does not match
  167. * them, skb_checksum_help or skb_crc32c_help (depending on the value of
  168. * csum_not_inet, see item D.) is called to resolve the checksum.
  169. *
  170. * CHECKSUM_NONE:
  171. *
  172. * The skb was already checksummed by the protocol, or a checksum is not
  173. * required.
  174. *
  175. * CHECKSUM_UNNECESSARY:
  176. *
  177. * This has the same meaning on as CHECKSUM_NONE for checksum offload on
  178. * output.
  179. *
  180. * CHECKSUM_COMPLETE:
  181. * Not used in checksum output. If a driver observes a packet with this value
  182. * set in skbuff, if should treat as CHECKSUM_NONE being set.
  183. *
  184. * D. Non-IP checksum (CRC) offloads
  185. *
  186. * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
  187. * offloading the SCTP CRC in a packet. To perform this offload the stack
  188. * will set set csum_start and csum_offset accordingly, set ip_summed to
  189. * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
  190. * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
  191. * A driver that supports both IP checksum offload and SCTP CRC32c offload
  192. * must verify which offload is configured for a packet by testing the
  193. * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
  194. * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
  195. *
  196. * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
  197. * offloading the FCOE CRC in a packet. To perform this offload the stack
  198. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  199. * accordingly. Note the there is no indication in the skbuff that the
  200. * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
  201. * both IP checksum offload and FCOE CRC offload must verify which offload
  202. * is configured for a packet presumably by inspecting packet headers.
  203. *
  204. * E. Checksumming on output with GSO.
  205. *
  206. * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
  207. * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
  208. * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
  209. * part of the GSO operation is implied. If a checksum is being offloaded
  210. * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
  211. * are set to refer to the outermost checksum being offload (two offloaded
  212. * checksums are possible with UDP encapsulation).
  213. */
  214. /* Don't change this without changing skb_csum_unnecessary! */
  215. #define CHECKSUM_NONE 0
  216. #define CHECKSUM_UNNECESSARY 1
  217. #define CHECKSUM_COMPLETE 2
  218. #define CHECKSUM_PARTIAL 3
  219. /* Maximum value in skb->csum_level */
  220. #define SKB_MAX_CSUM_LEVEL 3
  221. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  222. #define SKB_WITH_OVERHEAD(X) \
  223. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  224. #define SKB_MAX_ORDER(X, ORDER) \
  225. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  226. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  227. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  228. /* return minimum truesize of one skb containing X bytes of data */
  229. #define SKB_TRUESIZE(X) ((X) + \
  230. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  231. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  232. struct net_device;
  233. struct scatterlist;
  234. struct pipe_inode_info;
  235. struct iov_iter;
  236. struct napi_struct;
  237. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  238. struct nf_conntrack {
  239. atomic_t use;
  240. };
  241. #endif
  242. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  243. struct nf_bridge_info {
  244. atomic_t use;
  245. enum {
  246. BRNF_PROTO_UNCHANGED,
  247. BRNF_PROTO_8021Q,
  248. BRNF_PROTO_PPPOE
  249. } orig_proto:8;
  250. u8 pkt_otherhost:1;
  251. u8 in_prerouting:1;
  252. u8 bridged_dnat:1;
  253. __u16 frag_max_size;
  254. struct net_device *physindev;
  255. /* always valid & non-NULL from FORWARD on, for physdev match */
  256. struct net_device *physoutdev;
  257. union {
  258. /* prerouting: detect dnat in orig/reply direction */
  259. __be32 ipv4_daddr;
  260. struct in6_addr ipv6_daddr;
  261. /* after prerouting + nat detected: store original source
  262. * mac since neigh resolution overwrites it, only used while
  263. * skb is out in neigh layer.
  264. */
  265. char neigh_header[8];
  266. };
  267. };
  268. #endif
  269. struct sk_buff_head {
  270. /* These two members must be first. */
  271. struct sk_buff *next;
  272. struct sk_buff *prev;
  273. __u32 qlen;
  274. spinlock_t lock;
  275. };
  276. struct sk_buff;
  277. /* To allow 64K frame to be packed as single skb without frag_list we
  278. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  279. * buffers which do not start on a page boundary.
  280. *
  281. * Since GRO uses frags we allocate at least 16 regardless of page
  282. * size.
  283. */
  284. #if (65536/PAGE_SIZE + 1) < 16
  285. #define MAX_SKB_FRAGS 16UL
  286. #else
  287. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  288. #endif
  289. extern int sysctl_max_skb_frags;
  290. /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
  291. * segment using its current segmentation instead.
  292. */
  293. #define GSO_BY_FRAGS 0xFFFF
  294. typedef struct skb_frag_struct skb_frag_t;
  295. struct skb_frag_struct {
  296. struct {
  297. struct page *p;
  298. } page;
  299. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  300. __u32 page_offset;
  301. __u32 size;
  302. #else
  303. __u16 page_offset;
  304. __u16 size;
  305. #endif
  306. };
  307. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  308. {
  309. return frag->size;
  310. }
  311. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  312. {
  313. frag->size = size;
  314. }
  315. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  316. {
  317. frag->size += delta;
  318. }
  319. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  320. {
  321. frag->size -= delta;
  322. }
  323. #define HAVE_HW_TIME_STAMP
  324. /**
  325. * struct skb_shared_hwtstamps - hardware time stamps
  326. * @hwtstamp: hardware time stamp transformed into duration
  327. * since arbitrary point in time
  328. *
  329. * Software time stamps generated by ktime_get_real() are stored in
  330. * skb->tstamp.
  331. *
  332. * hwtstamps can only be compared against other hwtstamps from
  333. * the same device.
  334. *
  335. * This structure is attached to packets as part of the
  336. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  337. */
  338. struct skb_shared_hwtstamps {
  339. ktime_t hwtstamp;
  340. };
  341. /* Definitions for tx_flags in struct skb_shared_info */
  342. enum {
  343. /* generate hardware time stamp */
  344. SKBTX_HW_TSTAMP = 1 << 0,
  345. /* generate software time stamp when queueing packet to NIC */
  346. SKBTX_SW_TSTAMP = 1 << 1,
  347. /* device driver is going to provide hardware time stamp */
  348. SKBTX_IN_PROGRESS = 1 << 2,
  349. /* device driver supports TX zero-copy buffers */
  350. SKBTX_DEV_ZEROCOPY = 1 << 3,
  351. /* generate wifi status information (where possible) */
  352. SKBTX_WIFI_STATUS = 1 << 4,
  353. /* This indicates at least one fragment might be overwritten
  354. * (as in vmsplice(), sendfile() ...)
  355. * If we need to compute a TX checksum, we'll need to copy
  356. * all frags to avoid possible bad checksum
  357. */
  358. SKBTX_SHARED_FRAG = 1 << 5,
  359. /* generate software time stamp when entering packet scheduling */
  360. SKBTX_SCHED_TSTAMP = 1 << 6,
  361. };
  362. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  363. SKBTX_SCHED_TSTAMP)
  364. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  365. /*
  366. * The callback notifies userspace to release buffers when skb DMA is done in
  367. * lower device, the skb last reference should be 0 when calling this.
  368. * The zerocopy_success argument is true if zero copy transmit occurred,
  369. * false on data copy or out of memory error caused by data copy attempt.
  370. * The ctx field is used to track device context.
  371. * The desc field is used to track userspace buffer index.
  372. */
  373. struct ubuf_info {
  374. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  375. void *ctx;
  376. unsigned long desc;
  377. };
  378. /* This data is invariant across clones and lives at
  379. * the end of the header data, ie. at skb->end.
  380. */
  381. struct skb_shared_info {
  382. unsigned short _unused;
  383. unsigned char nr_frags;
  384. __u8 tx_flags;
  385. unsigned short gso_size;
  386. /* Warning: this field is not always filled in (UFO)! */
  387. unsigned short gso_segs;
  388. struct sk_buff *frag_list;
  389. struct skb_shared_hwtstamps hwtstamps;
  390. unsigned int gso_type;
  391. u32 tskey;
  392. __be32 ip6_frag_id;
  393. /*
  394. * Warning : all fields before dataref are cleared in __alloc_skb()
  395. */
  396. atomic_t dataref;
  397. /* Intermediate layers must ensure that destructor_arg
  398. * remains valid until skb destructor */
  399. void * destructor_arg;
  400. /* must be last field, see pskb_expand_head() */
  401. skb_frag_t frags[MAX_SKB_FRAGS];
  402. };
  403. /* We divide dataref into two halves. The higher 16 bits hold references
  404. * to the payload part of skb->data. The lower 16 bits hold references to
  405. * the entire skb->data. A clone of a headerless skb holds the length of
  406. * the header in skb->hdr_len.
  407. *
  408. * All users must obey the rule that the skb->data reference count must be
  409. * greater than or equal to the payload reference count.
  410. *
  411. * Holding a reference to the payload part means that the user does not
  412. * care about modifications to the header part of skb->data.
  413. */
  414. #define SKB_DATAREF_SHIFT 16
  415. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  416. enum {
  417. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  418. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  419. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  420. };
  421. enum {
  422. SKB_GSO_TCPV4 = 1 << 0,
  423. SKB_GSO_UDP = 1 << 1,
  424. /* This indicates the skb is from an untrusted source. */
  425. SKB_GSO_DODGY = 1 << 2,
  426. /* This indicates the tcp segment has CWR set. */
  427. SKB_GSO_TCP_ECN = 1 << 3,
  428. SKB_GSO_TCP_FIXEDID = 1 << 4,
  429. SKB_GSO_TCPV6 = 1 << 5,
  430. SKB_GSO_FCOE = 1 << 6,
  431. SKB_GSO_GRE = 1 << 7,
  432. SKB_GSO_GRE_CSUM = 1 << 8,
  433. SKB_GSO_IPXIP4 = 1 << 9,
  434. SKB_GSO_IPXIP6 = 1 << 10,
  435. SKB_GSO_UDP_TUNNEL = 1 << 11,
  436. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
  437. SKB_GSO_PARTIAL = 1 << 13,
  438. SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
  439. SKB_GSO_SCTP = 1 << 15,
  440. SKB_GSO_ESP = 1 << 16,
  441. };
  442. #if BITS_PER_LONG > 32
  443. #define NET_SKBUFF_DATA_USES_OFFSET 1
  444. #endif
  445. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  446. typedef unsigned int sk_buff_data_t;
  447. #else
  448. typedef unsigned char *sk_buff_data_t;
  449. #endif
  450. /**
  451. * struct sk_buff - socket buffer
  452. * @next: Next buffer in list
  453. * @prev: Previous buffer in list
  454. * @tstamp: Time we arrived/left
  455. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  456. * @sk: Socket we are owned by
  457. * @dev: Device we arrived on/are leaving by
  458. * @cb: Control buffer. Free for use by every layer. Put private vars here
  459. * @_skb_refdst: destination entry (with norefcount bit)
  460. * @sp: the security path, used for xfrm
  461. * @len: Length of actual data
  462. * @data_len: Data length
  463. * @mac_len: Length of link layer header
  464. * @hdr_len: writable header length of cloned skb
  465. * @csum: Checksum (must include start/offset pair)
  466. * @csum_start: Offset from skb->head where checksumming should start
  467. * @csum_offset: Offset from csum_start where checksum should be stored
  468. * @priority: Packet queueing priority
  469. * @ignore_df: allow local fragmentation
  470. * @cloned: Head may be cloned (check refcnt to be sure)
  471. * @ip_summed: Driver fed us an IP checksum
  472. * @nohdr: Payload reference only, must not modify header
  473. * @pkt_type: Packet class
  474. * @fclone: skbuff clone status
  475. * @ipvs_property: skbuff is owned by ipvs
  476. * @tc_skip_classify: do not classify packet. set by IFB device
  477. * @tc_at_ingress: used within tc_classify to distinguish in/egress
  478. * @tc_redirected: packet was redirected by a tc action
  479. * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
  480. * @peeked: this packet has been seen already, so stats have been
  481. * done for it, don't do them again
  482. * @nf_trace: netfilter packet trace flag
  483. * @protocol: Packet protocol from driver
  484. * @destructor: Destruct function
  485. * @_nfct: Associated connection, if any (with nfctinfo bits)
  486. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  487. * @skb_iif: ifindex of device we arrived on
  488. * @tc_index: Traffic control index
  489. * @hash: the packet hash
  490. * @queue_mapping: Queue mapping for multiqueue devices
  491. * @xmit_more: More SKBs are pending for this queue
  492. * @ndisc_nodetype: router type (from link layer)
  493. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  494. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  495. * ports.
  496. * @sw_hash: indicates hash was computed in software stack
  497. * @wifi_acked_valid: wifi_acked was set
  498. * @wifi_acked: whether frame was acked on wifi or not
  499. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  500. * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
  501. * @dst_pending_confirm: need to confirm neighbour
  502. * @napi_id: id of the NAPI struct this skb came from
  503. * @secmark: security marking
  504. * @mark: Generic packet mark
  505. * @vlan_proto: vlan encapsulation protocol
  506. * @vlan_tci: vlan tag control information
  507. * @inner_protocol: Protocol (encapsulation)
  508. * @inner_transport_header: Inner transport layer header (encapsulation)
  509. * @inner_network_header: Network layer header (encapsulation)
  510. * @inner_mac_header: Link layer header (encapsulation)
  511. * @transport_header: Transport layer header
  512. * @network_header: Network layer header
  513. * @mac_header: Link layer header
  514. * @tail: Tail pointer
  515. * @end: End pointer
  516. * @head: Head of buffer
  517. * @data: Data head pointer
  518. * @truesize: Buffer size
  519. * @users: User count - see {datagram,tcp}.c
  520. */
  521. struct sk_buff {
  522. union {
  523. struct {
  524. /* These two members must be first. */
  525. struct sk_buff *next;
  526. struct sk_buff *prev;
  527. union {
  528. ktime_t tstamp;
  529. u64 skb_mstamp;
  530. };
  531. };
  532. struct rb_node rbnode; /* used in netem & tcp stack */
  533. };
  534. struct sock *sk;
  535. union {
  536. struct net_device *dev;
  537. /* Some protocols might use this space to store information,
  538. * while device pointer would be NULL.
  539. * UDP receive path is one user.
  540. */
  541. unsigned long dev_scratch;
  542. };
  543. /*
  544. * This is the control buffer. It is free to use for every
  545. * layer. Please put your private variables there. If you
  546. * want to keep them across layers you have to do a skb_clone()
  547. * first. This is owned by whoever has the skb queued ATM.
  548. */
  549. char cb[48] __aligned(8);
  550. unsigned long _skb_refdst;
  551. void (*destructor)(struct sk_buff *skb);
  552. #ifdef CONFIG_XFRM
  553. struct sec_path *sp;
  554. #endif
  555. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  556. unsigned long _nfct;
  557. #endif
  558. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  559. struct nf_bridge_info *nf_bridge;
  560. #endif
  561. unsigned int len,
  562. data_len;
  563. __u16 mac_len,
  564. hdr_len;
  565. /* Following fields are _not_ copied in __copy_skb_header()
  566. * Note that queue_mapping is here mostly to fill a hole.
  567. */
  568. kmemcheck_bitfield_begin(flags1);
  569. __u16 queue_mapping;
  570. /* if you move cloned around you also must adapt those constants */
  571. #ifdef __BIG_ENDIAN_BITFIELD
  572. #define CLONED_MASK (1 << 7)
  573. #else
  574. #define CLONED_MASK 1
  575. #endif
  576. #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
  577. __u8 __cloned_offset[0];
  578. __u8 cloned:1,
  579. nohdr:1,
  580. fclone:2,
  581. peeked:1,
  582. head_frag:1,
  583. xmit_more:1,
  584. __unused:1; /* one bit hole */
  585. kmemcheck_bitfield_end(flags1);
  586. /* fields enclosed in headers_start/headers_end are copied
  587. * using a single memcpy() in __copy_skb_header()
  588. */
  589. /* private: */
  590. __u32 headers_start[0];
  591. /* public: */
  592. /* if you move pkt_type around you also must adapt those constants */
  593. #ifdef __BIG_ENDIAN_BITFIELD
  594. #define PKT_TYPE_MAX (7 << 5)
  595. #else
  596. #define PKT_TYPE_MAX 7
  597. #endif
  598. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  599. __u8 __pkt_type_offset[0];
  600. __u8 pkt_type:3;
  601. __u8 pfmemalloc:1;
  602. __u8 ignore_df:1;
  603. __u8 nf_trace:1;
  604. __u8 ip_summed:2;
  605. __u8 ooo_okay:1;
  606. __u8 l4_hash:1;
  607. __u8 sw_hash:1;
  608. __u8 wifi_acked_valid:1;
  609. __u8 wifi_acked:1;
  610. __u8 no_fcs:1;
  611. /* Indicates the inner headers are valid in the skbuff. */
  612. __u8 encapsulation:1;
  613. __u8 encap_hdr_csum:1;
  614. __u8 csum_valid:1;
  615. __u8 csum_complete_sw:1;
  616. __u8 csum_level:2;
  617. __u8 csum_not_inet:1;
  618. __u8 dst_pending_confirm:1;
  619. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  620. __u8 ndisc_nodetype:2;
  621. #endif
  622. __u8 ipvs_property:1;
  623. __u8 inner_protocol_type:1;
  624. __u8 remcsum_offload:1;
  625. #ifdef CONFIG_NET_SWITCHDEV
  626. __u8 offload_fwd_mark:1;
  627. #endif
  628. #ifdef CONFIG_NET_CLS_ACT
  629. __u8 tc_skip_classify:1;
  630. __u8 tc_at_ingress:1;
  631. __u8 tc_redirected:1;
  632. __u8 tc_from_ingress:1;
  633. #endif
  634. #ifdef CONFIG_NET_SCHED
  635. __u16 tc_index; /* traffic control index */
  636. #endif
  637. union {
  638. __wsum csum;
  639. struct {
  640. __u16 csum_start;
  641. __u16 csum_offset;
  642. };
  643. };
  644. __u32 priority;
  645. int skb_iif;
  646. __u32 hash;
  647. __be16 vlan_proto;
  648. __u16 vlan_tci;
  649. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  650. union {
  651. unsigned int napi_id;
  652. unsigned int sender_cpu;
  653. };
  654. #endif
  655. #ifdef CONFIG_NETWORK_SECMARK
  656. __u32 secmark;
  657. #endif
  658. union {
  659. __u32 mark;
  660. __u32 reserved_tailroom;
  661. };
  662. union {
  663. __be16 inner_protocol;
  664. __u8 inner_ipproto;
  665. };
  666. __u16 inner_transport_header;
  667. __u16 inner_network_header;
  668. __u16 inner_mac_header;
  669. __be16 protocol;
  670. __u16 transport_header;
  671. __u16 network_header;
  672. __u16 mac_header;
  673. /* private: */
  674. __u32 headers_end[0];
  675. /* public: */
  676. /* These elements must be at the end, see alloc_skb() for details. */
  677. sk_buff_data_t tail;
  678. sk_buff_data_t end;
  679. unsigned char *head,
  680. *data;
  681. unsigned int truesize;
  682. atomic_t users;
  683. };
  684. #ifdef __KERNEL__
  685. /*
  686. * Handling routines are only of interest to the kernel
  687. */
  688. #include <linux/slab.h>
  689. #define SKB_ALLOC_FCLONE 0x01
  690. #define SKB_ALLOC_RX 0x02
  691. #define SKB_ALLOC_NAPI 0x04
  692. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  693. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  694. {
  695. return unlikely(skb->pfmemalloc);
  696. }
  697. /*
  698. * skb might have a dst pointer attached, refcounted or not.
  699. * _skb_refdst low order bit is set if refcount was _not_ taken
  700. */
  701. #define SKB_DST_NOREF 1UL
  702. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  703. #define SKB_NFCT_PTRMASK ~(7UL)
  704. /**
  705. * skb_dst - returns skb dst_entry
  706. * @skb: buffer
  707. *
  708. * Returns skb dst_entry, regardless of reference taken or not.
  709. */
  710. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  711. {
  712. /* If refdst was not refcounted, check we still are in a
  713. * rcu_read_lock section
  714. */
  715. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  716. !rcu_read_lock_held() &&
  717. !rcu_read_lock_bh_held());
  718. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  719. }
  720. /**
  721. * skb_dst_set - sets skb dst
  722. * @skb: buffer
  723. * @dst: dst entry
  724. *
  725. * Sets skb dst, assuming a reference was taken on dst and should
  726. * be released by skb_dst_drop()
  727. */
  728. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  729. {
  730. skb->_skb_refdst = (unsigned long)dst;
  731. }
  732. /**
  733. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  734. * @skb: buffer
  735. * @dst: dst entry
  736. *
  737. * Sets skb dst, assuming a reference was not taken on dst.
  738. * If dst entry is cached, we do not take reference and dst_release
  739. * will be avoided by refdst_drop. If dst entry is not cached, we take
  740. * reference, so that last dst_release can destroy the dst immediately.
  741. */
  742. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  743. {
  744. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  745. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  746. }
  747. /**
  748. * skb_dst_is_noref - Test if skb dst isn't refcounted
  749. * @skb: buffer
  750. */
  751. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  752. {
  753. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  754. }
  755. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  756. {
  757. return (struct rtable *)skb_dst(skb);
  758. }
  759. /* For mangling skb->pkt_type from user space side from applications
  760. * such as nft, tc, etc, we only allow a conservative subset of
  761. * possible pkt_types to be set.
  762. */
  763. static inline bool skb_pkt_type_ok(u32 ptype)
  764. {
  765. return ptype <= PACKET_OTHERHOST;
  766. }
  767. static inline unsigned int skb_napi_id(const struct sk_buff *skb)
  768. {
  769. #ifdef CONFIG_NET_RX_BUSY_POLL
  770. return skb->napi_id;
  771. #else
  772. return 0;
  773. #endif
  774. }
  775. /* decrement the reference count and return true if we can free the skb */
  776. static inline bool skb_unref(struct sk_buff *skb)
  777. {
  778. if (unlikely(!skb))
  779. return false;
  780. if (likely(atomic_read(&skb->users) == 1))
  781. smp_rmb();
  782. else if (likely(!atomic_dec_and_test(&skb->users)))
  783. return false;
  784. return true;
  785. }
  786. void kfree_skb(struct sk_buff *skb);
  787. void kfree_skb_list(struct sk_buff *segs);
  788. void skb_tx_error(struct sk_buff *skb);
  789. void consume_skb(struct sk_buff *skb);
  790. void __kfree_skb(struct sk_buff *skb);
  791. extern struct kmem_cache *skbuff_head_cache;
  792. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  793. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  794. bool *fragstolen, int *delta_truesize);
  795. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  796. int node);
  797. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  798. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  799. static inline struct sk_buff *alloc_skb(unsigned int size,
  800. gfp_t priority)
  801. {
  802. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  803. }
  804. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  805. unsigned long data_len,
  806. int max_page_order,
  807. int *errcode,
  808. gfp_t gfp_mask);
  809. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  810. struct sk_buff_fclones {
  811. struct sk_buff skb1;
  812. struct sk_buff skb2;
  813. atomic_t fclone_ref;
  814. };
  815. /**
  816. * skb_fclone_busy - check if fclone is busy
  817. * @sk: socket
  818. * @skb: buffer
  819. *
  820. * Returns true if skb is a fast clone, and its clone is not freed.
  821. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  822. * so we also check that this didnt happen.
  823. */
  824. static inline bool skb_fclone_busy(const struct sock *sk,
  825. const struct sk_buff *skb)
  826. {
  827. const struct sk_buff_fclones *fclones;
  828. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  829. return skb->fclone == SKB_FCLONE_ORIG &&
  830. atomic_read(&fclones->fclone_ref) > 1 &&
  831. fclones->skb2.sk == sk;
  832. }
  833. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  834. gfp_t priority)
  835. {
  836. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  837. }
  838. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  839. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  840. {
  841. return __alloc_skb_head(priority, -1);
  842. }
  843. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  844. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  845. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  846. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  847. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  848. gfp_t gfp_mask, bool fclone);
  849. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  850. gfp_t gfp_mask)
  851. {
  852. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  853. }
  854. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  855. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  856. unsigned int headroom);
  857. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  858. int newtailroom, gfp_t priority);
  859. int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  860. int offset, int len);
  861. int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
  862. int offset, int len);
  863. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  864. int skb_pad(struct sk_buff *skb, int pad);
  865. #define dev_kfree_skb(a) consume_skb(a)
  866. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  867. int getfrag(void *from, char *to, int offset,
  868. int len, int odd, struct sk_buff *skb),
  869. void *from, int length);
  870. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  871. int offset, size_t size);
  872. struct skb_seq_state {
  873. __u32 lower_offset;
  874. __u32 upper_offset;
  875. __u32 frag_idx;
  876. __u32 stepped_offset;
  877. struct sk_buff *root_skb;
  878. struct sk_buff *cur_skb;
  879. __u8 *frag_data;
  880. };
  881. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  882. unsigned int to, struct skb_seq_state *st);
  883. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  884. struct skb_seq_state *st);
  885. void skb_abort_seq_read(struct skb_seq_state *st);
  886. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  887. unsigned int to, struct ts_config *config);
  888. /*
  889. * Packet hash types specify the type of hash in skb_set_hash.
  890. *
  891. * Hash types refer to the protocol layer addresses which are used to
  892. * construct a packet's hash. The hashes are used to differentiate or identify
  893. * flows of the protocol layer for the hash type. Hash types are either
  894. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  895. *
  896. * Properties of hashes:
  897. *
  898. * 1) Two packets in different flows have different hash values
  899. * 2) Two packets in the same flow should have the same hash value
  900. *
  901. * A hash at a higher layer is considered to be more specific. A driver should
  902. * set the most specific hash possible.
  903. *
  904. * A driver cannot indicate a more specific hash than the layer at which a hash
  905. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  906. *
  907. * A driver may indicate a hash level which is less specific than the
  908. * actual layer the hash was computed on. For instance, a hash computed
  909. * at L4 may be considered an L3 hash. This should only be done if the
  910. * driver can't unambiguously determine that the HW computed the hash at
  911. * the higher layer. Note that the "should" in the second property above
  912. * permits this.
  913. */
  914. enum pkt_hash_types {
  915. PKT_HASH_TYPE_NONE, /* Undefined type */
  916. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  917. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  918. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  919. };
  920. static inline void skb_clear_hash(struct sk_buff *skb)
  921. {
  922. skb->hash = 0;
  923. skb->sw_hash = 0;
  924. skb->l4_hash = 0;
  925. }
  926. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  927. {
  928. if (!skb->l4_hash)
  929. skb_clear_hash(skb);
  930. }
  931. static inline void
  932. __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
  933. {
  934. skb->l4_hash = is_l4;
  935. skb->sw_hash = is_sw;
  936. skb->hash = hash;
  937. }
  938. static inline void
  939. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  940. {
  941. /* Used by drivers to set hash from HW */
  942. __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
  943. }
  944. static inline void
  945. __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
  946. {
  947. __skb_set_hash(skb, hash, true, is_l4);
  948. }
  949. void __skb_get_hash(struct sk_buff *skb);
  950. u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
  951. u32 skb_get_poff(const struct sk_buff *skb);
  952. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  953. const struct flow_keys *keys, int hlen);
  954. __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
  955. void *data, int hlen_proto);
  956. static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
  957. int thoff, u8 ip_proto)
  958. {
  959. return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
  960. }
  961. void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  962. const struct flow_dissector_key *key,
  963. unsigned int key_count);
  964. bool __skb_flow_dissect(const struct sk_buff *skb,
  965. struct flow_dissector *flow_dissector,
  966. void *target_container,
  967. void *data, __be16 proto, int nhoff, int hlen,
  968. unsigned int flags);
  969. static inline bool skb_flow_dissect(const struct sk_buff *skb,
  970. struct flow_dissector *flow_dissector,
  971. void *target_container, unsigned int flags)
  972. {
  973. return __skb_flow_dissect(skb, flow_dissector, target_container,
  974. NULL, 0, 0, 0, flags);
  975. }
  976. static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
  977. struct flow_keys *flow,
  978. unsigned int flags)
  979. {
  980. memset(flow, 0, sizeof(*flow));
  981. return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
  982. NULL, 0, 0, 0, flags);
  983. }
  984. static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
  985. void *data, __be16 proto,
  986. int nhoff, int hlen,
  987. unsigned int flags)
  988. {
  989. memset(flow, 0, sizeof(*flow));
  990. return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
  991. data, proto, nhoff, hlen, flags);
  992. }
  993. static inline __u32 skb_get_hash(struct sk_buff *skb)
  994. {
  995. if (!skb->l4_hash && !skb->sw_hash)
  996. __skb_get_hash(skb);
  997. return skb->hash;
  998. }
  999. __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
  1000. static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
  1001. {
  1002. if (!skb->l4_hash && !skb->sw_hash) {
  1003. struct flow_keys keys;
  1004. __u32 hash = __get_hash_from_flowi6(fl6, &keys);
  1005. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1006. }
  1007. return skb->hash;
  1008. }
  1009. __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
  1010. static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
  1011. {
  1012. if (!skb->l4_hash && !skb->sw_hash) {
  1013. struct flow_keys keys;
  1014. __u32 hash = __get_hash_from_flowi4(fl4, &keys);
  1015. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1016. }
  1017. return skb->hash;
  1018. }
  1019. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  1020. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  1021. {
  1022. return skb->hash;
  1023. }
  1024. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  1025. {
  1026. to->hash = from->hash;
  1027. to->sw_hash = from->sw_hash;
  1028. to->l4_hash = from->l4_hash;
  1029. };
  1030. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1031. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1032. {
  1033. return skb->head + skb->end;
  1034. }
  1035. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1036. {
  1037. return skb->end;
  1038. }
  1039. #else
  1040. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1041. {
  1042. return skb->end;
  1043. }
  1044. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1045. {
  1046. return skb->end - skb->head;
  1047. }
  1048. #endif
  1049. /* Internal */
  1050. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  1051. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  1052. {
  1053. return &skb_shinfo(skb)->hwtstamps;
  1054. }
  1055. /**
  1056. * skb_queue_empty - check if a queue is empty
  1057. * @list: queue head
  1058. *
  1059. * Returns true if the queue is empty, false otherwise.
  1060. */
  1061. static inline int skb_queue_empty(const struct sk_buff_head *list)
  1062. {
  1063. return list->next == (const struct sk_buff *) list;
  1064. }
  1065. /**
  1066. * skb_queue_is_last - check if skb is the last entry in the queue
  1067. * @list: queue head
  1068. * @skb: buffer
  1069. *
  1070. * Returns true if @skb is the last buffer on the list.
  1071. */
  1072. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  1073. const struct sk_buff *skb)
  1074. {
  1075. return skb->next == (const struct sk_buff *) list;
  1076. }
  1077. /**
  1078. * skb_queue_is_first - check if skb is the first entry in the queue
  1079. * @list: queue head
  1080. * @skb: buffer
  1081. *
  1082. * Returns true if @skb is the first buffer on the list.
  1083. */
  1084. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  1085. const struct sk_buff *skb)
  1086. {
  1087. return skb->prev == (const struct sk_buff *) list;
  1088. }
  1089. /**
  1090. * skb_queue_next - return the next packet in the queue
  1091. * @list: queue head
  1092. * @skb: current buffer
  1093. *
  1094. * Return the next packet in @list after @skb. It is only valid to
  1095. * call this if skb_queue_is_last() evaluates to false.
  1096. */
  1097. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  1098. const struct sk_buff *skb)
  1099. {
  1100. /* This BUG_ON may seem severe, but if we just return then we
  1101. * are going to dereference garbage.
  1102. */
  1103. BUG_ON(skb_queue_is_last(list, skb));
  1104. return skb->next;
  1105. }
  1106. /**
  1107. * skb_queue_prev - return the prev packet in the queue
  1108. * @list: queue head
  1109. * @skb: current buffer
  1110. *
  1111. * Return the prev packet in @list before @skb. It is only valid to
  1112. * call this if skb_queue_is_first() evaluates to false.
  1113. */
  1114. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  1115. const struct sk_buff *skb)
  1116. {
  1117. /* This BUG_ON may seem severe, but if we just return then we
  1118. * are going to dereference garbage.
  1119. */
  1120. BUG_ON(skb_queue_is_first(list, skb));
  1121. return skb->prev;
  1122. }
  1123. /**
  1124. * skb_get - reference buffer
  1125. * @skb: buffer to reference
  1126. *
  1127. * Makes another reference to a socket buffer and returns a pointer
  1128. * to the buffer.
  1129. */
  1130. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  1131. {
  1132. atomic_inc(&skb->users);
  1133. return skb;
  1134. }
  1135. /*
  1136. * If users == 1, we are the only owner and are can avoid redundant
  1137. * atomic change.
  1138. */
  1139. /**
  1140. * skb_cloned - is the buffer a clone
  1141. * @skb: buffer to check
  1142. *
  1143. * Returns true if the buffer was generated with skb_clone() and is
  1144. * one of multiple shared copies of the buffer. Cloned buffers are
  1145. * shared data so must not be written to under normal circumstances.
  1146. */
  1147. static inline int skb_cloned(const struct sk_buff *skb)
  1148. {
  1149. return skb->cloned &&
  1150. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  1151. }
  1152. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  1153. {
  1154. might_sleep_if(gfpflags_allow_blocking(pri));
  1155. if (skb_cloned(skb))
  1156. return pskb_expand_head(skb, 0, 0, pri);
  1157. return 0;
  1158. }
  1159. /**
  1160. * skb_header_cloned - is the header a clone
  1161. * @skb: buffer to check
  1162. *
  1163. * Returns true if modifying the header part of the buffer requires
  1164. * the data to be copied.
  1165. */
  1166. static inline int skb_header_cloned(const struct sk_buff *skb)
  1167. {
  1168. int dataref;
  1169. if (!skb->cloned)
  1170. return 0;
  1171. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  1172. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  1173. return dataref != 1;
  1174. }
  1175. static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
  1176. {
  1177. might_sleep_if(gfpflags_allow_blocking(pri));
  1178. if (skb_header_cloned(skb))
  1179. return pskb_expand_head(skb, 0, 0, pri);
  1180. return 0;
  1181. }
  1182. /**
  1183. * skb_header_release - release reference to header
  1184. * @skb: buffer to operate on
  1185. *
  1186. * Drop a reference to the header part of the buffer. This is done
  1187. * by acquiring a payload reference. You must not read from the header
  1188. * part of skb->data after this.
  1189. * Note : Check if you can use __skb_header_release() instead.
  1190. */
  1191. static inline void skb_header_release(struct sk_buff *skb)
  1192. {
  1193. BUG_ON(skb->nohdr);
  1194. skb->nohdr = 1;
  1195. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1196. }
  1197. /**
  1198. * __skb_header_release - release reference to header
  1199. * @skb: buffer to operate on
  1200. *
  1201. * Variant of skb_header_release() assuming skb is private to caller.
  1202. * We can avoid one atomic operation.
  1203. */
  1204. static inline void __skb_header_release(struct sk_buff *skb)
  1205. {
  1206. skb->nohdr = 1;
  1207. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1208. }
  1209. /**
  1210. * skb_shared - is the buffer shared
  1211. * @skb: buffer to check
  1212. *
  1213. * Returns true if more than one person has a reference to this
  1214. * buffer.
  1215. */
  1216. static inline int skb_shared(const struct sk_buff *skb)
  1217. {
  1218. return atomic_read(&skb->users) != 1;
  1219. }
  1220. /**
  1221. * skb_share_check - check if buffer is shared and if so clone it
  1222. * @skb: buffer to check
  1223. * @pri: priority for memory allocation
  1224. *
  1225. * If the buffer is shared the buffer is cloned and the old copy
  1226. * drops a reference. A new clone with a single reference is returned.
  1227. * If the buffer is not shared the original buffer is returned. When
  1228. * being called from interrupt status or with spinlocks held pri must
  1229. * be GFP_ATOMIC.
  1230. *
  1231. * NULL is returned on a memory allocation failure.
  1232. */
  1233. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1234. {
  1235. might_sleep_if(gfpflags_allow_blocking(pri));
  1236. if (skb_shared(skb)) {
  1237. struct sk_buff *nskb = skb_clone(skb, pri);
  1238. if (likely(nskb))
  1239. consume_skb(skb);
  1240. else
  1241. kfree_skb(skb);
  1242. skb = nskb;
  1243. }
  1244. return skb;
  1245. }
  1246. /*
  1247. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1248. * packets to handle cases where we have a local reader and forward
  1249. * and a couple of other messy ones. The normal one is tcpdumping
  1250. * a packet thats being forwarded.
  1251. */
  1252. /**
  1253. * skb_unshare - make a copy of a shared buffer
  1254. * @skb: buffer to check
  1255. * @pri: priority for memory allocation
  1256. *
  1257. * If the socket buffer is a clone then this function creates a new
  1258. * copy of the data, drops a reference count on the old copy and returns
  1259. * the new copy with the reference count at 1. If the buffer is not a clone
  1260. * the original buffer is returned. When called with a spinlock held or
  1261. * from interrupt state @pri must be %GFP_ATOMIC
  1262. *
  1263. * %NULL is returned on a memory allocation failure.
  1264. */
  1265. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1266. gfp_t pri)
  1267. {
  1268. might_sleep_if(gfpflags_allow_blocking(pri));
  1269. if (skb_cloned(skb)) {
  1270. struct sk_buff *nskb = skb_copy(skb, pri);
  1271. /* Free our shared copy */
  1272. if (likely(nskb))
  1273. consume_skb(skb);
  1274. else
  1275. kfree_skb(skb);
  1276. skb = nskb;
  1277. }
  1278. return skb;
  1279. }
  1280. /**
  1281. * skb_peek - peek at the head of an &sk_buff_head
  1282. * @list_: list to peek at
  1283. *
  1284. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1285. * be careful with this one. A peek leaves the buffer on the
  1286. * list and someone else may run off with it. You must hold
  1287. * the appropriate locks or have a private queue to do this.
  1288. *
  1289. * Returns %NULL for an empty list or a pointer to the head element.
  1290. * The reference count is not incremented and the reference is therefore
  1291. * volatile. Use with caution.
  1292. */
  1293. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1294. {
  1295. struct sk_buff *skb = list_->next;
  1296. if (skb == (struct sk_buff *)list_)
  1297. skb = NULL;
  1298. return skb;
  1299. }
  1300. /**
  1301. * skb_peek_next - peek skb following the given one from a queue
  1302. * @skb: skb to start from
  1303. * @list_: list to peek at
  1304. *
  1305. * Returns %NULL when the end of the list is met or a pointer to the
  1306. * next element. The reference count is not incremented and the
  1307. * reference is therefore volatile. Use with caution.
  1308. */
  1309. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1310. const struct sk_buff_head *list_)
  1311. {
  1312. struct sk_buff *next = skb->next;
  1313. if (next == (struct sk_buff *)list_)
  1314. next = NULL;
  1315. return next;
  1316. }
  1317. /**
  1318. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1319. * @list_: list to peek at
  1320. *
  1321. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1322. * be careful with this one. A peek leaves the buffer on the
  1323. * list and someone else may run off with it. You must hold
  1324. * the appropriate locks or have a private queue to do this.
  1325. *
  1326. * Returns %NULL for an empty list or a pointer to the tail element.
  1327. * The reference count is not incremented and the reference is therefore
  1328. * volatile. Use with caution.
  1329. */
  1330. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1331. {
  1332. struct sk_buff *skb = list_->prev;
  1333. if (skb == (struct sk_buff *)list_)
  1334. skb = NULL;
  1335. return skb;
  1336. }
  1337. /**
  1338. * skb_queue_len - get queue length
  1339. * @list_: list to measure
  1340. *
  1341. * Return the length of an &sk_buff queue.
  1342. */
  1343. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1344. {
  1345. return list_->qlen;
  1346. }
  1347. /**
  1348. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1349. * @list: queue to initialize
  1350. *
  1351. * This initializes only the list and queue length aspects of
  1352. * an sk_buff_head object. This allows to initialize the list
  1353. * aspects of an sk_buff_head without reinitializing things like
  1354. * the spinlock. It can also be used for on-stack sk_buff_head
  1355. * objects where the spinlock is known to not be used.
  1356. */
  1357. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1358. {
  1359. list->prev = list->next = (struct sk_buff *)list;
  1360. list->qlen = 0;
  1361. }
  1362. /*
  1363. * This function creates a split out lock class for each invocation;
  1364. * this is needed for now since a whole lot of users of the skb-queue
  1365. * infrastructure in drivers have different locking usage (in hardirq)
  1366. * than the networking core (in softirq only). In the long run either the
  1367. * network layer or drivers should need annotation to consolidate the
  1368. * main types of usage into 3 classes.
  1369. */
  1370. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1371. {
  1372. spin_lock_init(&list->lock);
  1373. __skb_queue_head_init(list);
  1374. }
  1375. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1376. struct lock_class_key *class)
  1377. {
  1378. skb_queue_head_init(list);
  1379. lockdep_set_class(&list->lock, class);
  1380. }
  1381. /*
  1382. * Insert an sk_buff on a list.
  1383. *
  1384. * The "__skb_xxxx()" functions are the non-atomic ones that
  1385. * can only be called with interrupts disabled.
  1386. */
  1387. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1388. struct sk_buff_head *list);
  1389. static inline void __skb_insert(struct sk_buff *newsk,
  1390. struct sk_buff *prev, struct sk_buff *next,
  1391. struct sk_buff_head *list)
  1392. {
  1393. newsk->next = next;
  1394. newsk->prev = prev;
  1395. next->prev = prev->next = newsk;
  1396. list->qlen++;
  1397. }
  1398. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1399. struct sk_buff *prev,
  1400. struct sk_buff *next)
  1401. {
  1402. struct sk_buff *first = list->next;
  1403. struct sk_buff *last = list->prev;
  1404. first->prev = prev;
  1405. prev->next = first;
  1406. last->next = next;
  1407. next->prev = last;
  1408. }
  1409. /**
  1410. * skb_queue_splice - join two skb lists, this is designed for stacks
  1411. * @list: the new list to add
  1412. * @head: the place to add it in the first list
  1413. */
  1414. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1415. struct sk_buff_head *head)
  1416. {
  1417. if (!skb_queue_empty(list)) {
  1418. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1419. head->qlen += list->qlen;
  1420. }
  1421. }
  1422. /**
  1423. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1424. * @list: the new list to add
  1425. * @head: the place to add it in the first list
  1426. *
  1427. * The list at @list is reinitialised
  1428. */
  1429. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1430. struct sk_buff_head *head)
  1431. {
  1432. if (!skb_queue_empty(list)) {
  1433. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1434. head->qlen += list->qlen;
  1435. __skb_queue_head_init(list);
  1436. }
  1437. }
  1438. /**
  1439. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1440. * @list: the new list to add
  1441. * @head: the place to add it in the first list
  1442. */
  1443. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1444. struct sk_buff_head *head)
  1445. {
  1446. if (!skb_queue_empty(list)) {
  1447. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1448. head->qlen += list->qlen;
  1449. }
  1450. }
  1451. /**
  1452. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1453. * @list: the new list to add
  1454. * @head: the place to add it in the first list
  1455. *
  1456. * Each of the lists is a queue.
  1457. * The list at @list is reinitialised
  1458. */
  1459. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1460. struct sk_buff_head *head)
  1461. {
  1462. if (!skb_queue_empty(list)) {
  1463. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1464. head->qlen += list->qlen;
  1465. __skb_queue_head_init(list);
  1466. }
  1467. }
  1468. /**
  1469. * __skb_queue_after - queue a buffer at the list head
  1470. * @list: list to use
  1471. * @prev: place after this buffer
  1472. * @newsk: buffer to queue
  1473. *
  1474. * Queue a buffer int the middle of a list. This function takes no locks
  1475. * and you must therefore hold required locks before calling it.
  1476. *
  1477. * A buffer cannot be placed on two lists at the same time.
  1478. */
  1479. static inline void __skb_queue_after(struct sk_buff_head *list,
  1480. struct sk_buff *prev,
  1481. struct sk_buff *newsk)
  1482. {
  1483. __skb_insert(newsk, prev, prev->next, list);
  1484. }
  1485. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1486. struct sk_buff_head *list);
  1487. static inline void __skb_queue_before(struct sk_buff_head *list,
  1488. struct sk_buff *next,
  1489. struct sk_buff *newsk)
  1490. {
  1491. __skb_insert(newsk, next->prev, next, list);
  1492. }
  1493. /**
  1494. * __skb_queue_head - queue a buffer at the list head
  1495. * @list: list to use
  1496. * @newsk: buffer to queue
  1497. *
  1498. * Queue a buffer at the start of a list. This function takes no locks
  1499. * and you must therefore hold required locks before calling it.
  1500. *
  1501. * A buffer cannot be placed on two lists at the same time.
  1502. */
  1503. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1504. static inline void __skb_queue_head(struct sk_buff_head *list,
  1505. struct sk_buff *newsk)
  1506. {
  1507. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1508. }
  1509. /**
  1510. * __skb_queue_tail - queue a buffer at the list tail
  1511. * @list: list to use
  1512. * @newsk: buffer to queue
  1513. *
  1514. * Queue a buffer at the end of a list. This function takes no locks
  1515. * and you must therefore hold required locks before calling it.
  1516. *
  1517. * A buffer cannot be placed on two lists at the same time.
  1518. */
  1519. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1520. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1521. struct sk_buff *newsk)
  1522. {
  1523. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1524. }
  1525. /*
  1526. * remove sk_buff from list. _Must_ be called atomically, and with
  1527. * the list known..
  1528. */
  1529. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1530. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1531. {
  1532. struct sk_buff *next, *prev;
  1533. list->qlen--;
  1534. next = skb->next;
  1535. prev = skb->prev;
  1536. skb->next = skb->prev = NULL;
  1537. next->prev = prev;
  1538. prev->next = next;
  1539. }
  1540. /**
  1541. * __skb_dequeue - remove from the head of the queue
  1542. * @list: list to dequeue from
  1543. *
  1544. * Remove the head of the list. This function does not take any locks
  1545. * so must be used with appropriate locks held only. The head item is
  1546. * returned or %NULL if the list is empty.
  1547. */
  1548. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1549. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1550. {
  1551. struct sk_buff *skb = skb_peek(list);
  1552. if (skb)
  1553. __skb_unlink(skb, list);
  1554. return skb;
  1555. }
  1556. /**
  1557. * __skb_dequeue_tail - remove from the tail of the queue
  1558. * @list: list to dequeue from
  1559. *
  1560. * Remove the tail of the list. This function does not take any locks
  1561. * so must be used with appropriate locks held only. The tail item is
  1562. * returned or %NULL if the list is empty.
  1563. */
  1564. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1565. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1566. {
  1567. struct sk_buff *skb = skb_peek_tail(list);
  1568. if (skb)
  1569. __skb_unlink(skb, list);
  1570. return skb;
  1571. }
  1572. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1573. {
  1574. return skb->data_len;
  1575. }
  1576. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1577. {
  1578. return skb->len - skb->data_len;
  1579. }
  1580. static inline unsigned int skb_pagelen(const struct sk_buff *skb)
  1581. {
  1582. unsigned int i, len = 0;
  1583. for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
  1584. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1585. return len + skb_headlen(skb);
  1586. }
  1587. /**
  1588. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1589. * @skb: buffer containing fragment to be initialised
  1590. * @i: paged fragment index to initialise
  1591. * @page: the page to use for this fragment
  1592. * @off: the offset to the data with @page
  1593. * @size: the length of the data
  1594. *
  1595. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1596. * offset @off within @page.
  1597. *
  1598. * Does not take any additional reference on the fragment.
  1599. */
  1600. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1601. struct page *page, int off, int size)
  1602. {
  1603. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1604. /*
  1605. * Propagate page pfmemalloc to the skb if we can. The problem is
  1606. * that not all callers have unique ownership of the page but rely
  1607. * on page_is_pfmemalloc doing the right thing(tm).
  1608. */
  1609. frag->page.p = page;
  1610. frag->page_offset = off;
  1611. skb_frag_size_set(frag, size);
  1612. page = compound_head(page);
  1613. if (page_is_pfmemalloc(page))
  1614. skb->pfmemalloc = true;
  1615. }
  1616. /**
  1617. * skb_fill_page_desc - initialise a paged fragment in an skb
  1618. * @skb: buffer containing fragment to be initialised
  1619. * @i: paged fragment index to initialise
  1620. * @page: the page to use for this fragment
  1621. * @off: the offset to the data with @page
  1622. * @size: the length of the data
  1623. *
  1624. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1625. * @skb to point to @size bytes at offset @off within @page. In
  1626. * addition updates @skb such that @i is the last fragment.
  1627. *
  1628. * Does not take any additional reference on the fragment.
  1629. */
  1630. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1631. struct page *page, int off, int size)
  1632. {
  1633. __skb_fill_page_desc(skb, i, page, off, size);
  1634. skb_shinfo(skb)->nr_frags = i + 1;
  1635. }
  1636. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1637. int size, unsigned int truesize);
  1638. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1639. unsigned int truesize);
  1640. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1641. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1642. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1643. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1644. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1645. {
  1646. return skb->head + skb->tail;
  1647. }
  1648. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1649. {
  1650. skb->tail = skb->data - skb->head;
  1651. }
  1652. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1653. {
  1654. skb_reset_tail_pointer(skb);
  1655. skb->tail += offset;
  1656. }
  1657. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1658. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1659. {
  1660. return skb->tail;
  1661. }
  1662. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1663. {
  1664. skb->tail = skb->data;
  1665. }
  1666. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1667. {
  1668. skb->tail = skb->data + offset;
  1669. }
  1670. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1671. /*
  1672. * Add data to an sk_buff
  1673. */
  1674. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1675. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1676. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1677. {
  1678. unsigned char *tmp = skb_tail_pointer(skb);
  1679. SKB_LINEAR_ASSERT(skb);
  1680. skb->tail += len;
  1681. skb->len += len;
  1682. return tmp;
  1683. }
  1684. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1685. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1686. {
  1687. skb->data -= len;
  1688. skb->len += len;
  1689. return skb->data;
  1690. }
  1691. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1692. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1693. {
  1694. skb->len -= len;
  1695. BUG_ON(skb->len < skb->data_len);
  1696. return skb->data += len;
  1697. }
  1698. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1699. {
  1700. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1701. }
  1702. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1703. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1704. {
  1705. if (len > skb_headlen(skb) &&
  1706. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1707. return NULL;
  1708. skb->len -= len;
  1709. return skb->data += len;
  1710. }
  1711. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1712. {
  1713. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1714. }
  1715. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1716. {
  1717. if (likely(len <= skb_headlen(skb)))
  1718. return 1;
  1719. if (unlikely(len > skb->len))
  1720. return 0;
  1721. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1722. }
  1723. void skb_condense(struct sk_buff *skb);
  1724. /**
  1725. * skb_headroom - bytes at buffer head
  1726. * @skb: buffer to check
  1727. *
  1728. * Return the number of bytes of free space at the head of an &sk_buff.
  1729. */
  1730. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1731. {
  1732. return skb->data - skb->head;
  1733. }
  1734. /**
  1735. * skb_tailroom - bytes at buffer end
  1736. * @skb: buffer to check
  1737. *
  1738. * Return the number of bytes of free space at the tail of an sk_buff
  1739. */
  1740. static inline int skb_tailroom(const struct sk_buff *skb)
  1741. {
  1742. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1743. }
  1744. /**
  1745. * skb_availroom - bytes at buffer end
  1746. * @skb: buffer to check
  1747. *
  1748. * Return the number of bytes of free space at the tail of an sk_buff
  1749. * allocated by sk_stream_alloc()
  1750. */
  1751. static inline int skb_availroom(const struct sk_buff *skb)
  1752. {
  1753. if (skb_is_nonlinear(skb))
  1754. return 0;
  1755. return skb->end - skb->tail - skb->reserved_tailroom;
  1756. }
  1757. /**
  1758. * skb_reserve - adjust headroom
  1759. * @skb: buffer to alter
  1760. * @len: bytes to move
  1761. *
  1762. * Increase the headroom of an empty &sk_buff by reducing the tail
  1763. * room. This is only allowed for an empty buffer.
  1764. */
  1765. static inline void skb_reserve(struct sk_buff *skb, int len)
  1766. {
  1767. skb->data += len;
  1768. skb->tail += len;
  1769. }
  1770. /**
  1771. * skb_tailroom_reserve - adjust reserved_tailroom
  1772. * @skb: buffer to alter
  1773. * @mtu: maximum amount of headlen permitted
  1774. * @needed_tailroom: minimum amount of reserved_tailroom
  1775. *
  1776. * Set reserved_tailroom so that headlen can be as large as possible but
  1777. * not larger than mtu and tailroom cannot be smaller than
  1778. * needed_tailroom.
  1779. * The required headroom should already have been reserved before using
  1780. * this function.
  1781. */
  1782. static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
  1783. unsigned int needed_tailroom)
  1784. {
  1785. SKB_LINEAR_ASSERT(skb);
  1786. if (mtu < skb_tailroom(skb) - needed_tailroom)
  1787. /* use at most mtu */
  1788. skb->reserved_tailroom = skb_tailroom(skb) - mtu;
  1789. else
  1790. /* use up to all available space */
  1791. skb->reserved_tailroom = needed_tailroom;
  1792. }
  1793. #define ENCAP_TYPE_ETHER 0
  1794. #define ENCAP_TYPE_IPPROTO 1
  1795. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1796. __be16 protocol)
  1797. {
  1798. skb->inner_protocol = protocol;
  1799. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1800. }
  1801. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1802. __u8 ipproto)
  1803. {
  1804. skb->inner_ipproto = ipproto;
  1805. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1806. }
  1807. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1808. {
  1809. skb->inner_mac_header = skb->mac_header;
  1810. skb->inner_network_header = skb->network_header;
  1811. skb->inner_transport_header = skb->transport_header;
  1812. }
  1813. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1814. {
  1815. skb->mac_len = skb->network_header - skb->mac_header;
  1816. }
  1817. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1818. *skb)
  1819. {
  1820. return skb->head + skb->inner_transport_header;
  1821. }
  1822. static inline int skb_inner_transport_offset(const struct sk_buff *skb)
  1823. {
  1824. return skb_inner_transport_header(skb) - skb->data;
  1825. }
  1826. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1827. {
  1828. skb->inner_transport_header = skb->data - skb->head;
  1829. }
  1830. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1831. const int offset)
  1832. {
  1833. skb_reset_inner_transport_header(skb);
  1834. skb->inner_transport_header += offset;
  1835. }
  1836. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1837. {
  1838. return skb->head + skb->inner_network_header;
  1839. }
  1840. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1841. {
  1842. skb->inner_network_header = skb->data - skb->head;
  1843. }
  1844. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1845. const int offset)
  1846. {
  1847. skb_reset_inner_network_header(skb);
  1848. skb->inner_network_header += offset;
  1849. }
  1850. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1851. {
  1852. return skb->head + skb->inner_mac_header;
  1853. }
  1854. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1855. {
  1856. skb->inner_mac_header = skb->data - skb->head;
  1857. }
  1858. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1859. const int offset)
  1860. {
  1861. skb_reset_inner_mac_header(skb);
  1862. skb->inner_mac_header += offset;
  1863. }
  1864. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1865. {
  1866. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1867. }
  1868. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1869. {
  1870. return skb->head + skb->transport_header;
  1871. }
  1872. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1873. {
  1874. skb->transport_header = skb->data - skb->head;
  1875. }
  1876. static inline void skb_set_transport_header(struct sk_buff *skb,
  1877. const int offset)
  1878. {
  1879. skb_reset_transport_header(skb);
  1880. skb->transport_header += offset;
  1881. }
  1882. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1883. {
  1884. return skb->head + skb->network_header;
  1885. }
  1886. static inline void skb_reset_network_header(struct sk_buff *skb)
  1887. {
  1888. skb->network_header = skb->data - skb->head;
  1889. }
  1890. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1891. {
  1892. skb_reset_network_header(skb);
  1893. skb->network_header += offset;
  1894. }
  1895. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1896. {
  1897. return skb->head + skb->mac_header;
  1898. }
  1899. static inline int skb_mac_offset(const struct sk_buff *skb)
  1900. {
  1901. return skb_mac_header(skb) - skb->data;
  1902. }
  1903. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1904. {
  1905. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1906. }
  1907. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1908. {
  1909. skb->mac_header = skb->data - skb->head;
  1910. }
  1911. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1912. {
  1913. skb_reset_mac_header(skb);
  1914. skb->mac_header += offset;
  1915. }
  1916. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1917. {
  1918. skb->mac_header = skb->network_header;
  1919. }
  1920. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1921. const int offset_hint)
  1922. {
  1923. struct flow_keys keys;
  1924. if (skb_transport_header_was_set(skb))
  1925. return;
  1926. else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
  1927. skb_set_transport_header(skb, keys.control.thoff);
  1928. else
  1929. skb_set_transport_header(skb, offset_hint);
  1930. }
  1931. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1932. {
  1933. if (skb_mac_header_was_set(skb)) {
  1934. const unsigned char *old_mac = skb_mac_header(skb);
  1935. skb_set_mac_header(skb, -skb->mac_len);
  1936. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1937. }
  1938. }
  1939. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1940. {
  1941. return skb->csum_start - skb_headroom(skb);
  1942. }
  1943. static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
  1944. {
  1945. return skb->head + skb->csum_start;
  1946. }
  1947. static inline int skb_transport_offset(const struct sk_buff *skb)
  1948. {
  1949. return skb_transport_header(skb) - skb->data;
  1950. }
  1951. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1952. {
  1953. return skb->transport_header - skb->network_header;
  1954. }
  1955. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1956. {
  1957. return skb->inner_transport_header - skb->inner_network_header;
  1958. }
  1959. static inline int skb_network_offset(const struct sk_buff *skb)
  1960. {
  1961. return skb_network_header(skb) - skb->data;
  1962. }
  1963. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1964. {
  1965. return skb_inner_network_header(skb) - skb->data;
  1966. }
  1967. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1968. {
  1969. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1970. }
  1971. /*
  1972. * CPUs often take a performance hit when accessing unaligned memory
  1973. * locations. The actual performance hit varies, it can be small if the
  1974. * hardware handles it or large if we have to take an exception and fix it
  1975. * in software.
  1976. *
  1977. * Since an ethernet header is 14 bytes network drivers often end up with
  1978. * the IP header at an unaligned offset. The IP header can be aligned by
  1979. * shifting the start of the packet by 2 bytes. Drivers should do this
  1980. * with:
  1981. *
  1982. * skb_reserve(skb, NET_IP_ALIGN);
  1983. *
  1984. * The downside to this alignment of the IP header is that the DMA is now
  1985. * unaligned. On some architectures the cost of an unaligned DMA is high
  1986. * and this cost outweighs the gains made by aligning the IP header.
  1987. *
  1988. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1989. * to be overridden.
  1990. */
  1991. #ifndef NET_IP_ALIGN
  1992. #define NET_IP_ALIGN 2
  1993. #endif
  1994. /*
  1995. * The networking layer reserves some headroom in skb data (via
  1996. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1997. * the header has to grow. In the default case, if the header has to grow
  1998. * 32 bytes or less we avoid the reallocation.
  1999. *
  2000. * Unfortunately this headroom changes the DMA alignment of the resulting
  2001. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  2002. * on some architectures. An architecture can override this value,
  2003. * perhaps setting it to a cacheline in size (since that will maintain
  2004. * cacheline alignment of the DMA). It must be a power of 2.
  2005. *
  2006. * Various parts of the networking layer expect at least 32 bytes of
  2007. * headroom, you should not reduce this.
  2008. *
  2009. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  2010. * to reduce average number of cache lines per packet.
  2011. * get_rps_cpus() for example only access one 64 bytes aligned block :
  2012. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  2013. */
  2014. #ifndef NET_SKB_PAD
  2015. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  2016. #endif
  2017. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  2018. static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
  2019. {
  2020. if (unlikely(skb_is_nonlinear(skb))) {
  2021. WARN_ON(1);
  2022. return;
  2023. }
  2024. skb->len = len;
  2025. skb_set_tail_pointer(skb, len);
  2026. }
  2027. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  2028. {
  2029. __skb_set_length(skb, len);
  2030. }
  2031. void skb_trim(struct sk_buff *skb, unsigned int len);
  2032. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  2033. {
  2034. if (skb->data_len)
  2035. return ___pskb_trim(skb, len);
  2036. __skb_trim(skb, len);
  2037. return 0;
  2038. }
  2039. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  2040. {
  2041. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  2042. }
  2043. /**
  2044. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  2045. * @skb: buffer to alter
  2046. * @len: new length
  2047. *
  2048. * This is identical to pskb_trim except that the caller knows that
  2049. * the skb is not cloned so we should never get an error due to out-
  2050. * of-memory.
  2051. */
  2052. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  2053. {
  2054. int err = pskb_trim(skb, len);
  2055. BUG_ON(err);
  2056. }
  2057. static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
  2058. {
  2059. unsigned int diff = len - skb->len;
  2060. if (skb_tailroom(skb) < diff) {
  2061. int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
  2062. GFP_ATOMIC);
  2063. if (ret)
  2064. return ret;
  2065. }
  2066. __skb_set_length(skb, len);
  2067. return 0;
  2068. }
  2069. /**
  2070. * skb_orphan - orphan a buffer
  2071. * @skb: buffer to orphan
  2072. *
  2073. * If a buffer currently has an owner then we call the owner's
  2074. * destructor function and make the @skb unowned. The buffer continues
  2075. * to exist but is no longer charged to its former owner.
  2076. */
  2077. static inline void skb_orphan(struct sk_buff *skb)
  2078. {
  2079. if (skb->destructor) {
  2080. skb->destructor(skb);
  2081. skb->destructor = NULL;
  2082. skb->sk = NULL;
  2083. } else {
  2084. BUG_ON(skb->sk);
  2085. }
  2086. }
  2087. /**
  2088. * skb_orphan_frags - orphan the frags contained in a buffer
  2089. * @skb: buffer to orphan frags from
  2090. * @gfp_mask: allocation mask for replacement pages
  2091. *
  2092. * For each frag in the SKB which needs a destructor (i.e. has an
  2093. * owner) create a copy of that frag and release the original
  2094. * page by calling the destructor.
  2095. */
  2096. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  2097. {
  2098. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  2099. return 0;
  2100. return skb_copy_ubufs(skb, gfp_mask);
  2101. }
  2102. /**
  2103. * __skb_queue_purge - empty a list
  2104. * @list: list to empty
  2105. *
  2106. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  2107. * the list and one reference dropped. This function does not take the
  2108. * list lock and the caller must hold the relevant locks to use it.
  2109. */
  2110. void skb_queue_purge(struct sk_buff_head *list);
  2111. static inline void __skb_queue_purge(struct sk_buff_head *list)
  2112. {
  2113. struct sk_buff *skb;
  2114. while ((skb = __skb_dequeue(list)) != NULL)
  2115. kfree_skb(skb);
  2116. }
  2117. void skb_rbtree_purge(struct rb_root *root);
  2118. void *netdev_alloc_frag(unsigned int fragsz);
  2119. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  2120. gfp_t gfp_mask);
  2121. /**
  2122. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  2123. * @dev: network device to receive on
  2124. * @length: length to allocate
  2125. *
  2126. * Allocate a new &sk_buff and assign it a usage count of one. The
  2127. * buffer has unspecified headroom built in. Users should allocate
  2128. * the headroom they think they need without accounting for the
  2129. * built in space. The built in space is used for optimisations.
  2130. *
  2131. * %NULL is returned if there is no free memory. Although this function
  2132. * allocates memory it can be called from an interrupt.
  2133. */
  2134. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  2135. unsigned int length)
  2136. {
  2137. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  2138. }
  2139. /* legacy helper around __netdev_alloc_skb() */
  2140. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  2141. gfp_t gfp_mask)
  2142. {
  2143. return __netdev_alloc_skb(NULL, length, gfp_mask);
  2144. }
  2145. /* legacy helper around netdev_alloc_skb() */
  2146. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  2147. {
  2148. return netdev_alloc_skb(NULL, length);
  2149. }
  2150. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  2151. unsigned int length, gfp_t gfp)
  2152. {
  2153. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  2154. if (NET_IP_ALIGN && skb)
  2155. skb_reserve(skb, NET_IP_ALIGN);
  2156. return skb;
  2157. }
  2158. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  2159. unsigned int length)
  2160. {
  2161. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  2162. }
  2163. static inline void skb_free_frag(void *addr)
  2164. {
  2165. page_frag_free(addr);
  2166. }
  2167. void *napi_alloc_frag(unsigned int fragsz);
  2168. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  2169. unsigned int length, gfp_t gfp_mask);
  2170. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  2171. unsigned int length)
  2172. {
  2173. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  2174. }
  2175. void napi_consume_skb(struct sk_buff *skb, int budget);
  2176. void __kfree_skb_flush(void);
  2177. void __kfree_skb_defer(struct sk_buff *skb);
  2178. /**
  2179. * __dev_alloc_pages - allocate page for network Rx
  2180. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2181. * @order: size of the allocation
  2182. *
  2183. * Allocate a new page.
  2184. *
  2185. * %NULL is returned if there is no free memory.
  2186. */
  2187. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  2188. unsigned int order)
  2189. {
  2190. /* This piece of code contains several assumptions.
  2191. * 1. This is for device Rx, therefor a cold page is preferred.
  2192. * 2. The expectation is the user wants a compound page.
  2193. * 3. If requesting a order 0 page it will not be compound
  2194. * due to the check to see if order has a value in prep_new_page
  2195. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  2196. * code in gfp_to_alloc_flags that should be enforcing this.
  2197. */
  2198. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  2199. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  2200. }
  2201. static inline struct page *dev_alloc_pages(unsigned int order)
  2202. {
  2203. return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
  2204. }
  2205. /**
  2206. * __dev_alloc_page - allocate a page for network Rx
  2207. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2208. *
  2209. * Allocate a new page.
  2210. *
  2211. * %NULL is returned if there is no free memory.
  2212. */
  2213. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  2214. {
  2215. return __dev_alloc_pages(gfp_mask, 0);
  2216. }
  2217. static inline struct page *dev_alloc_page(void)
  2218. {
  2219. return dev_alloc_pages(0);
  2220. }
  2221. /**
  2222. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  2223. * @page: The page that was allocated from skb_alloc_page
  2224. * @skb: The skb that may need pfmemalloc set
  2225. */
  2226. static inline void skb_propagate_pfmemalloc(struct page *page,
  2227. struct sk_buff *skb)
  2228. {
  2229. if (page_is_pfmemalloc(page))
  2230. skb->pfmemalloc = true;
  2231. }
  2232. /**
  2233. * skb_frag_page - retrieve the page referred to by a paged fragment
  2234. * @frag: the paged fragment
  2235. *
  2236. * Returns the &struct page associated with @frag.
  2237. */
  2238. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2239. {
  2240. return frag->page.p;
  2241. }
  2242. /**
  2243. * __skb_frag_ref - take an addition reference on a paged fragment.
  2244. * @frag: the paged fragment
  2245. *
  2246. * Takes an additional reference on the paged fragment @frag.
  2247. */
  2248. static inline void __skb_frag_ref(skb_frag_t *frag)
  2249. {
  2250. get_page(skb_frag_page(frag));
  2251. }
  2252. /**
  2253. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2254. * @skb: the buffer
  2255. * @f: the fragment offset.
  2256. *
  2257. * Takes an additional reference on the @f'th paged fragment of @skb.
  2258. */
  2259. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2260. {
  2261. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2262. }
  2263. /**
  2264. * __skb_frag_unref - release a reference on a paged fragment.
  2265. * @frag: the paged fragment
  2266. *
  2267. * Releases a reference on the paged fragment @frag.
  2268. */
  2269. static inline void __skb_frag_unref(skb_frag_t *frag)
  2270. {
  2271. put_page(skb_frag_page(frag));
  2272. }
  2273. /**
  2274. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2275. * @skb: the buffer
  2276. * @f: the fragment offset
  2277. *
  2278. * Releases a reference on the @f'th paged fragment of @skb.
  2279. */
  2280. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2281. {
  2282. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2283. }
  2284. /**
  2285. * skb_frag_address - gets the address of the data contained in a paged fragment
  2286. * @frag: the paged fragment buffer
  2287. *
  2288. * Returns the address of the data within @frag. The page must already
  2289. * be mapped.
  2290. */
  2291. static inline void *skb_frag_address(const skb_frag_t *frag)
  2292. {
  2293. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2294. }
  2295. /**
  2296. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2297. * @frag: the paged fragment buffer
  2298. *
  2299. * Returns the address of the data within @frag. Checks that the page
  2300. * is mapped and returns %NULL otherwise.
  2301. */
  2302. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2303. {
  2304. void *ptr = page_address(skb_frag_page(frag));
  2305. if (unlikely(!ptr))
  2306. return NULL;
  2307. return ptr + frag->page_offset;
  2308. }
  2309. /**
  2310. * __skb_frag_set_page - sets the page contained in a paged fragment
  2311. * @frag: the paged fragment
  2312. * @page: the page to set
  2313. *
  2314. * Sets the fragment @frag to contain @page.
  2315. */
  2316. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2317. {
  2318. frag->page.p = page;
  2319. }
  2320. /**
  2321. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2322. * @skb: the buffer
  2323. * @f: the fragment offset
  2324. * @page: the page to set
  2325. *
  2326. * Sets the @f'th fragment of @skb to contain @page.
  2327. */
  2328. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2329. struct page *page)
  2330. {
  2331. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2332. }
  2333. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2334. /**
  2335. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2336. * @dev: the device to map the fragment to
  2337. * @frag: the paged fragment to map
  2338. * @offset: the offset within the fragment (starting at the
  2339. * fragment's own offset)
  2340. * @size: the number of bytes to map
  2341. * @dir: the direction of the mapping (%PCI_DMA_*)
  2342. *
  2343. * Maps the page associated with @frag to @device.
  2344. */
  2345. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2346. const skb_frag_t *frag,
  2347. size_t offset, size_t size,
  2348. enum dma_data_direction dir)
  2349. {
  2350. return dma_map_page(dev, skb_frag_page(frag),
  2351. frag->page_offset + offset, size, dir);
  2352. }
  2353. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2354. gfp_t gfp_mask)
  2355. {
  2356. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2357. }
  2358. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2359. gfp_t gfp_mask)
  2360. {
  2361. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2362. }
  2363. /**
  2364. * skb_clone_writable - is the header of a clone writable
  2365. * @skb: buffer to check
  2366. * @len: length up to which to write
  2367. *
  2368. * Returns true if modifying the header part of the cloned buffer
  2369. * does not requires the data to be copied.
  2370. */
  2371. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2372. {
  2373. return !skb_header_cloned(skb) &&
  2374. skb_headroom(skb) + len <= skb->hdr_len;
  2375. }
  2376. static inline int skb_try_make_writable(struct sk_buff *skb,
  2377. unsigned int write_len)
  2378. {
  2379. return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
  2380. pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
  2381. }
  2382. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2383. int cloned)
  2384. {
  2385. int delta = 0;
  2386. if (headroom > skb_headroom(skb))
  2387. delta = headroom - skb_headroom(skb);
  2388. if (delta || cloned)
  2389. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2390. GFP_ATOMIC);
  2391. return 0;
  2392. }
  2393. /**
  2394. * skb_cow - copy header of skb when it is required
  2395. * @skb: buffer to cow
  2396. * @headroom: needed headroom
  2397. *
  2398. * If the skb passed lacks sufficient headroom or its data part
  2399. * is shared, data is reallocated. If reallocation fails, an error
  2400. * is returned and original skb is not changed.
  2401. *
  2402. * The result is skb with writable area skb->head...skb->tail
  2403. * and at least @headroom of space at head.
  2404. */
  2405. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2406. {
  2407. return __skb_cow(skb, headroom, skb_cloned(skb));
  2408. }
  2409. /**
  2410. * skb_cow_head - skb_cow but only making the head writable
  2411. * @skb: buffer to cow
  2412. * @headroom: needed headroom
  2413. *
  2414. * This function is identical to skb_cow except that we replace the
  2415. * skb_cloned check by skb_header_cloned. It should be used when
  2416. * you only need to push on some header and do not need to modify
  2417. * the data.
  2418. */
  2419. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2420. {
  2421. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2422. }
  2423. /**
  2424. * skb_padto - pad an skbuff up to a minimal size
  2425. * @skb: buffer to pad
  2426. * @len: minimal length
  2427. *
  2428. * Pads up a buffer to ensure the trailing bytes exist and are
  2429. * blanked. If the buffer already contains sufficient data it
  2430. * is untouched. Otherwise it is extended. Returns zero on
  2431. * success. The skb is freed on error.
  2432. */
  2433. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2434. {
  2435. unsigned int size = skb->len;
  2436. if (likely(size >= len))
  2437. return 0;
  2438. return skb_pad(skb, len - size);
  2439. }
  2440. /**
  2441. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2442. * @skb: buffer to pad
  2443. * @len: minimal length
  2444. *
  2445. * Pads up a buffer to ensure the trailing bytes exist and are
  2446. * blanked. If the buffer already contains sufficient data it
  2447. * is untouched. Otherwise it is extended. Returns zero on
  2448. * success. The skb is freed on error.
  2449. */
  2450. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2451. {
  2452. unsigned int size = skb->len;
  2453. if (unlikely(size < len)) {
  2454. len -= size;
  2455. if (skb_pad(skb, len))
  2456. return -ENOMEM;
  2457. __skb_put(skb, len);
  2458. }
  2459. return 0;
  2460. }
  2461. static inline int skb_add_data(struct sk_buff *skb,
  2462. struct iov_iter *from, int copy)
  2463. {
  2464. const int off = skb->len;
  2465. if (skb->ip_summed == CHECKSUM_NONE) {
  2466. __wsum csum = 0;
  2467. if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
  2468. &csum, from)) {
  2469. skb->csum = csum_block_add(skb->csum, csum, off);
  2470. return 0;
  2471. }
  2472. } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
  2473. return 0;
  2474. __skb_trim(skb, off);
  2475. return -EFAULT;
  2476. }
  2477. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2478. const struct page *page, int off)
  2479. {
  2480. if (i) {
  2481. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2482. return page == skb_frag_page(frag) &&
  2483. off == frag->page_offset + skb_frag_size(frag);
  2484. }
  2485. return false;
  2486. }
  2487. static inline int __skb_linearize(struct sk_buff *skb)
  2488. {
  2489. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2490. }
  2491. /**
  2492. * skb_linearize - convert paged skb to linear one
  2493. * @skb: buffer to linarize
  2494. *
  2495. * If there is no free memory -ENOMEM is returned, otherwise zero
  2496. * is returned and the old skb data released.
  2497. */
  2498. static inline int skb_linearize(struct sk_buff *skb)
  2499. {
  2500. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2501. }
  2502. /**
  2503. * skb_has_shared_frag - can any frag be overwritten
  2504. * @skb: buffer to test
  2505. *
  2506. * Return true if the skb has at least one frag that might be modified
  2507. * by an external entity (as in vmsplice()/sendfile())
  2508. */
  2509. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2510. {
  2511. return skb_is_nonlinear(skb) &&
  2512. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2513. }
  2514. /**
  2515. * skb_linearize_cow - make sure skb is linear and writable
  2516. * @skb: buffer to process
  2517. *
  2518. * If there is no free memory -ENOMEM is returned, otherwise zero
  2519. * is returned and the old skb data released.
  2520. */
  2521. static inline int skb_linearize_cow(struct sk_buff *skb)
  2522. {
  2523. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2524. __skb_linearize(skb) : 0;
  2525. }
  2526. static __always_inline void
  2527. __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2528. unsigned int off)
  2529. {
  2530. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2531. skb->csum = csum_block_sub(skb->csum,
  2532. csum_partial(start, len, 0), off);
  2533. else if (skb->ip_summed == CHECKSUM_PARTIAL &&
  2534. skb_checksum_start_offset(skb) < 0)
  2535. skb->ip_summed = CHECKSUM_NONE;
  2536. }
  2537. /**
  2538. * skb_postpull_rcsum - update checksum for received skb after pull
  2539. * @skb: buffer to update
  2540. * @start: start of data before pull
  2541. * @len: length of data pulled
  2542. *
  2543. * After doing a pull on a received packet, you need to call this to
  2544. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2545. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2546. */
  2547. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2548. const void *start, unsigned int len)
  2549. {
  2550. __skb_postpull_rcsum(skb, start, len, 0);
  2551. }
  2552. static __always_inline void
  2553. __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2554. unsigned int off)
  2555. {
  2556. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2557. skb->csum = csum_block_add(skb->csum,
  2558. csum_partial(start, len, 0), off);
  2559. }
  2560. /**
  2561. * skb_postpush_rcsum - update checksum for received skb after push
  2562. * @skb: buffer to update
  2563. * @start: start of data after push
  2564. * @len: length of data pushed
  2565. *
  2566. * After doing a push on a received packet, you need to call this to
  2567. * update the CHECKSUM_COMPLETE checksum.
  2568. */
  2569. static inline void skb_postpush_rcsum(struct sk_buff *skb,
  2570. const void *start, unsigned int len)
  2571. {
  2572. __skb_postpush_rcsum(skb, start, len, 0);
  2573. }
  2574. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2575. /**
  2576. * skb_push_rcsum - push skb and update receive checksum
  2577. * @skb: buffer to update
  2578. * @len: length of data pulled
  2579. *
  2580. * This function performs an skb_push on the packet and updates
  2581. * the CHECKSUM_COMPLETE checksum. It should be used on
  2582. * receive path processing instead of skb_push unless you know
  2583. * that the checksum difference is zero (e.g., a valid IP header)
  2584. * or you are setting ip_summed to CHECKSUM_NONE.
  2585. */
  2586. static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
  2587. unsigned int len)
  2588. {
  2589. skb_push(skb, len);
  2590. skb_postpush_rcsum(skb, skb->data, len);
  2591. return skb->data;
  2592. }
  2593. /**
  2594. * pskb_trim_rcsum - trim received skb and update checksum
  2595. * @skb: buffer to trim
  2596. * @len: new length
  2597. *
  2598. * This is exactly the same as pskb_trim except that it ensures the
  2599. * checksum of received packets are still valid after the operation.
  2600. */
  2601. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2602. {
  2603. if (likely(len >= skb->len))
  2604. return 0;
  2605. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2606. skb->ip_summed = CHECKSUM_NONE;
  2607. return __pskb_trim(skb, len);
  2608. }
  2609. static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2610. {
  2611. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2612. skb->ip_summed = CHECKSUM_NONE;
  2613. __skb_trim(skb, len);
  2614. return 0;
  2615. }
  2616. static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
  2617. {
  2618. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2619. skb->ip_summed = CHECKSUM_NONE;
  2620. return __skb_grow(skb, len);
  2621. }
  2622. #define skb_queue_walk(queue, skb) \
  2623. for (skb = (queue)->next; \
  2624. skb != (struct sk_buff *)(queue); \
  2625. skb = skb->next)
  2626. #define skb_queue_walk_safe(queue, skb, tmp) \
  2627. for (skb = (queue)->next, tmp = skb->next; \
  2628. skb != (struct sk_buff *)(queue); \
  2629. skb = tmp, tmp = skb->next)
  2630. #define skb_queue_walk_from(queue, skb) \
  2631. for (; skb != (struct sk_buff *)(queue); \
  2632. skb = skb->next)
  2633. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2634. for (tmp = skb->next; \
  2635. skb != (struct sk_buff *)(queue); \
  2636. skb = tmp, tmp = skb->next)
  2637. #define skb_queue_reverse_walk(queue, skb) \
  2638. for (skb = (queue)->prev; \
  2639. skb != (struct sk_buff *)(queue); \
  2640. skb = skb->prev)
  2641. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2642. for (skb = (queue)->prev, tmp = skb->prev; \
  2643. skb != (struct sk_buff *)(queue); \
  2644. skb = tmp, tmp = skb->prev)
  2645. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2646. for (tmp = skb->prev; \
  2647. skb != (struct sk_buff *)(queue); \
  2648. skb = tmp, tmp = skb->prev)
  2649. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2650. {
  2651. return skb_shinfo(skb)->frag_list != NULL;
  2652. }
  2653. static inline void skb_frag_list_init(struct sk_buff *skb)
  2654. {
  2655. skb_shinfo(skb)->frag_list = NULL;
  2656. }
  2657. #define skb_walk_frags(skb, iter) \
  2658. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2659. int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
  2660. const struct sk_buff *skb);
  2661. struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
  2662. struct sk_buff_head *queue,
  2663. unsigned int flags,
  2664. void (*destructor)(struct sock *sk,
  2665. struct sk_buff *skb),
  2666. int *peeked, int *off, int *err,
  2667. struct sk_buff **last);
  2668. struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
  2669. void (*destructor)(struct sock *sk,
  2670. struct sk_buff *skb),
  2671. int *peeked, int *off, int *err,
  2672. struct sk_buff **last);
  2673. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2674. void (*destructor)(struct sock *sk,
  2675. struct sk_buff *skb),
  2676. int *peeked, int *off, int *err);
  2677. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2678. int *err);
  2679. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2680. struct poll_table_struct *wait);
  2681. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2682. struct iov_iter *to, int size);
  2683. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2684. struct msghdr *msg, int size)
  2685. {
  2686. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2687. }
  2688. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2689. struct msghdr *msg);
  2690. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2691. struct iov_iter *from, int len);
  2692. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2693. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2694. void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
  2695. static inline void skb_free_datagram_locked(struct sock *sk,
  2696. struct sk_buff *skb)
  2697. {
  2698. __skb_free_datagram_locked(sk, skb, 0);
  2699. }
  2700. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2701. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2702. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2703. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2704. int len, __wsum csum);
  2705. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2706. struct pipe_inode_info *pipe, unsigned int len,
  2707. unsigned int flags);
  2708. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2709. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2710. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2711. int len, int hlen);
  2712. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2713. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2714. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2715. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2716. bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
  2717. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2718. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2719. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2720. int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
  2721. int skb_vlan_pop(struct sk_buff *skb);
  2722. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2723. struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
  2724. gfp_t gfp);
  2725. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2726. {
  2727. return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
  2728. }
  2729. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2730. {
  2731. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2732. }
  2733. struct skb_checksum_ops {
  2734. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2735. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2736. };
  2737. extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
  2738. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2739. __wsum csum, const struct skb_checksum_ops *ops);
  2740. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2741. __wsum csum);
  2742. static inline void * __must_check
  2743. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2744. int len, void *data, int hlen, void *buffer)
  2745. {
  2746. if (hlen - offset >= len)
  2747. return data + offset;
  2748. if (!skb ||
  2749. skb_copy_bits(skb, offset, buffer, len) < 0)
  2750. return NULL;
  2751. return buffer;
  2752. }
  2753. static inline void * __must_check
  2754. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2755. {
  2756. return __skb_header_pointer(skb, offset, len, skb->data,
  2757. skb_headlen(skb), buffer);
  2758. }
  2759. /**
  2760. * skb_needs_linearize - check if we need to linearize a given skb
  2761. * depending on the given device features.
  2762. * @skb: socket buffer to check
  2763. * @features: net device features
  2764. *
  2765. * Returns true if either:
  2766. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2767. * 2. skb is fragmented and the device does not support SG.
  2768. */
  2769. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2770. netdev_features_t features)
  2771. {
  2772. return skb_is_nonlinear(skb) &&
  2773. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2774. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2775. }
  2776. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2777. void *to,
  2778. const unsigned int len)
  2779. {
  2780. memcpy(to, skb->data, len);
  2781. }
  2782. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2783. const int offset, void *to,
  2784. const unsigned int len)
  2785. {
  2786. memcpy(to, skb->data + offset, len);
  2787. }
  2788. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2789. const void *from,
  2790. const unsigned int len)
  2791. {
  2792. memcpy(skb->data, from, len);
  2793. }
  2794. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2795. const int offset,
  2796. const void *from,
  2797. const unsigned int len)
  2798. {
  2799. memcpy(skb->data + offset, from, len);
  2800. }
  2801. void skb_init(void);
  2802. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2803. {
  2804. return skb->tstamp;
  2805. }
  2806. /**
  2807. * skb_get_timestamp - get timestamp from a skb
  2808. * @skb: skb to get stamp from
  2809. * @stamp: pointer to struct timeval to store stamp in
  2810. *
  2811. * Timestamps are stored in the skb as offsets to a base timestamp.
  2812. * This function converts the offset back to a struct timeval and stores
  2813. * it in stamp.
  2814. */
  2815. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2816. struct timeval *stamp)
  2817. {
  2818. *stamp = ktime_to_timeval(skb->tstamp);
  2819. }
  2820. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2821. struct timespec *stamp)
  2822. {
  2823. *stamp = ktime_to_timespec(skb->tstamp);
  2824. }
  2825. static inline void __net_timestamp(struct sk_buff *skb)
  2826. {
  2827. skb->tstamp = ktime_get_real();
  2828. }
  2829. static inline ktime_t net_timedelta(ktime_t t)
  2830. {
  2831. return ktime_sub(ktime_get_real(), t);
  2832. }
  2833. static inline ktime_t net_invalid_timestamp(void)
  2834. {
  2835. return 0;
  2836. }
  2837. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2838. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2839. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2840. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2841. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2842. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2843. {
  2844. }
  2845. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2846. {
  2847. return false;
  2848. }
  2849. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2850. /**
  2851. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2852. *
  2853. * PHY drivers may accept clones of transmitted packets for
  2854. * timestamping via their phy_driver.txtstamp method. These drivers
  2855. * must call this function to return the skb back to the stack with a
  2856. * timestamp.
  2857. *
  2858. * @skb: clone of the the original outgoing packet
  2859. * @hwtstamps: hardware time stamps
  2860. *
  2861. */
  2862. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2863. struct skb_shared_hwtstamps *hwtstamps);
  2864. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2865. struct skb_shared_hwtstamps *hwtstamps,
  2866. struct sock *sk, int tstype);
  2867. /**
  2868. * skb_tstamp_tx - queue clone of skb with send time stamps
  2869. * @orig_skb: the original outgoing packet
  2870. * @hwtstamps: hardware time stamps, may be NULL if not available
  2871. *
  2872. * If the skb has a socket associated, then this function clones the
  2873. * skb (thus sharing the actual data and optional structures), stores
  2874. * the optional hardware time stamping information (if non NULL) or
  2875. * generates a software time stamp (otherwise), then queues the clone
  2876. * to the error queue of the socket. Errors are silently ignored.
  2877. */
  2878. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2879. struct skb_shared_hwtstamps *hwtstamps);
  2880. /**
  2881. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2882. *
  2883. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2884. * function immediately before giving the sk_buff to the MAC hardware.
  2885. *
  2886. * Specifically, one should make absolutely sure that this function is
  2887. * called before TX completion of this packet can trigger. Otherwise
  2888. * the packet could potentially already be freed.
  2889. *
  2890. * @skb: A socket buffer.
  2891. */
  2892. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2893. {
  2894. skb_clone_tx_timestamp(skb);
  2895. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
  2896. skb_tstamp_tx(skb, NULL);
  2897. }
  2898. /**
  2899. * skb_complete_wifi_ack - deliver skb with wifi status
  2900. *
  2901. * @skb: the original outgoing packet
  2902. * @acked: ack status
  2903. *
  2904. */
  2905. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2906. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2907. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2908. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2909. {
  2910. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2911. skb->csum_valid ||
  2912. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2913. skb_checksum_start_offset(skb) >= 0));
  2914. }
  2915. /**
  2916. * skb_checksum_complete - Calculate checksum of an entire packet
  2917. * @skb: packet to process
  2918. *
  2919. * This function calculates the checksum over the entire packet plus
  2920. * the value of skb->csum. The latter can be used to supply the
  2921. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2922. * checksum.
  2923. *
  2924. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2925. * this function can be used to verify that checksum on received
  2926. * packets. In that case the function should return zero if the
  2927. * checksum is correct. In particular, this function will return zero
  2928. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2929. * hardware has already verified the correctness of the checksum.
  2930. */
  2931. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2932. {
  2933. return skb_csum_unnecessary(skb) ?
  2934. 0 : __skb_checksum_complete(skb);
  2935. }
  2936. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2937. {
  2938. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2939. if (skb->csum_level == 0)
  2940. skb->ip_summed = CHECKSUM_NONE;
  2941. else
  2942. skb->csum_level--;
  2943. }
  2944. }
  2945. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2946. {
  2947. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2948. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2949. skb->csum_level++;
  2950. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2951. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2952. skb->csum_level = 0;
  2953. }
  2954. }
  2955. /* Check if we need to perform checksum complete validation.
  2956. *
  2957. * Returns true if checksum complete is needed, false otherwise
  2958. * (either checksum is unnecessary or zero checksum is allowed).
  2959. */
  2960. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2961. bool zero_okay,
  2962. __sum16 check)
  2963. {
  2964. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2965. skb->csum_valid = 1;
  2966. __skb_decr_checksum_unnecessary(skb);
  2967. return false;
  2968. }
  2969. return true;
  2970. }
  2971. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2972. * in checksum_init.
  2973. */
  2974. #define CHECKSUM_BREAK 76
  2975. /* Unset checksum-complete
  2976. *
  2977. * Unset checksum complete can be done when packet is being modified
  2978. * (uncompressed for instance) and checksum-complete value is
  2979. * invalidated.
  2980. */
  2981. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2982. {
  2983. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2984. skb->ip_summed = CHECKSUM_NONE;
  2985. }
  2986. /* Validate (init) checksum based on checksum complete.
  2987. *
  2988. * Return values:
  2989. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2990. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2991. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2992. * non-zero: value of invalid checksum
  2993. *
  2994. */
  2995. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2996. bool complete,
  2997. __wsum psum)
  2998. {
  2999. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  3000. if (!csum_fold(csum_add(psum, skb->csum))) {
  3001. skb->csum_valid = 1;
  3002. return 0;
  3003. }
  3004. }
  3005. skb->csum = psum;
  3006. if (complete || skb->len <= CHECKSUM_BREAK) {
  3007. __sum16 csum;
  3008. csum = __skb_checksum_complete(skb);
  3009. skb->csum_valid = !csum;
  3010. return csum;
  3011. }
  3012. return 0;
  3013. }
  3014. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  3015. {
  3016. return 0;
  3017. }
  3018. /* Perform checksum validate (init). Note that this is a macro since we only
  3019. * want to calculate the pseudo header which is an input function if necessary.
  3020. * First we try to validate without any computation (checksum unnecessary) and
  3021. * then calculate based on checksum complete calling the function to compute
  3022. * pseudo header.
  3023. *
  3024. * Return values:
  3025. * 0: checksum is validated or try to in skb_checksum_complete
  3026. * non-zero: value of invalid checksum
  3027. */
  3028. #define __skb_checksum_validate(skb, proto, complete, \
  3029. zero_okay, check, compute_pseudo) \
  3030. ({ \
  3031. __sum16 __ret = 0; \
  3032. skb->csum_valid = 0; \
  3033. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  3034. __ret = __skb_checksum_validate_complete(skb, \
  3035. complete, compute_pseudo(skb, proto)); \
  3036. __ret; \
  3037. })
  3038. #define skb_checksum_init(skb, proto, compute_pseudo) \
  3039. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  3040. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  3041. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  3042. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  3043. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  3044. #define skb_checksum_validate_zero_check(skb, proto, check, \
  3045. compute_pseudo) \
  3046. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  3047. #define skb_checksum_simple_validate(skb) \
  3048. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  3049. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  3050. {
  3051. return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
  3052. }
  3053. static inline void __skb_checksum_convert(struct sk_buff *skb,
  3054. __sum16 check, __wsum pseudo)
  3055. {
  3056. skb->csum = ~pseudo;
  3057. skb->ip_summed = CHECKSUM_COMPLETE;
  3058. }
  3059. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  3060. do { \
  3061. if (__skb_checksum_convert_check(skb)) \
  3062. __skb_checksum_convert(skb, check, \
  3063. compute_pseudo(skb, proto)); \
  3064. } while (0)
  3065. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  3066. u16 start, u16 offset)
  3067. {
  3068. skb->ip_summed = CHECKSUM_PARTIAL;
  3069. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  3070. skb->csum_offset = offset - start;
  3071. }
  3072. /* Update skbuf and packet to reflect the remote checksum offload operation.
  3073. * When called, ptr indicates the starting point for skb->csum when
  3074. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  3075. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  3076. */
  3077. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  3078. int start, int offset, bool nopartial)
  3079. {
  3080. __wsum delta;
  3081. if (!nopartial) {
  3082. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  3083. return;
  3084. }
  3085. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  3086. __skb_checksum_complete(skb);
  3087. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  3088. }
  3089. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  3090. /* Adjust skb->csum since we changed the packet */
  3091. skb->csum = csum_add(skb->csum, delta);
  3092. }
  3093. static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
  3094. {
  3095. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  3096. return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
  3097. #else
  3098. return NULL;
  3099. #endif
  3100. }
  3101. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3102. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  3103. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  3104. {
  3105. if (nfct && atomic_dec_and_test(&nfct->use))
  3106. nf_conntrack_destroy(nfct);
  3107. }
  3108. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  3109. {
  3110. if (nfct)
  3111. atomic_inc(&nfct->use);
  3112. }
  3113. #endif
  3114. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3115. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  3116. {
  3117. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  3118. kfree(nf_bridge);
  3119. }
  3120. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  3121. {
  3122. if (nf_bridge)
  3123. atomic_inc(&nf_bridge->use);
  3124. }
  3125. #endif /* CONFIG_BRIDGE_NETFILTER */
  3126. static inline void nf_reset(struct sk_buff *skb)
  3127. {
  3128. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3129. nf_conntrack_put(skb_nfct(skb));
  3130. skb->_nfct = 0;
  3131. #endif
  3132. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3133. nf_bridge_put(skb->nf_bridge);
  3134. skb->nf_bridge = NULL;
  3135. #endif
  3136. }
  3137. static inline void nf_reset_trace(struct sk_buff *skb)
  3138. {
  3139. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3140. skb->nf_trace = 0;
  3141. #endif
  3142. }
  3143. /* Note: This doesn't put any conntrack and bridge info in dst. */
  3144. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  3145. bool copy)
  3146. {
  3147. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3148. dst->_nfct = src->_nfct;
  3149. nf_conntrack_get(skb_nfct(src));
  3150. #endif
  3151. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3152. dst->nf_bridge = src->nf_bridge;
  3153. nf_bridge_get(src->nf_bridge);
  3154. #endif
  3155. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3156. if (copy)
  3157. dst->nf_trace = src->nf_trace;
  3158. #endif
  3159. }
  3160. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  3161. {
  3162. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3163. nf_conntrack_put(skb_nfct(dst));
  3164. #endif
  3165. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3166. nf_bridge_put(dst->nf_bridge);
  3167. #endif
  3168. __nf_copy(dst, src, true);
  3169. }
  3170. #ifdef CONFIG_NETWORK_SECMARK
  3171. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3172. {
  3173. to->secmark = from->secmark;
  3174. }
  3175. static inline void skb_init_secmark(struct sk_buff *skb)
  3176. {
  3177. skb->secmark = 0;
  3178. }
  3179. #else
  3180. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3181. { }
  3182. static inline void skb_init_secmark(struct sk_buff *skb)
  3183. { }
  3184. #endif
  3185. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  3186. {
  3187. return !skb->destructor &&
  3188. #if IS_ENABLED(CONFIG_XFRM)
  3189. !skb->sp &&
  3190. #endif
  3191. !skb_nfct(skb) &&
  3192. !skb->_skb_refdst &&
  3193. !skb_has_frag_list(skb);
  3194. }
  3195. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  3196. {
  3197. skb->queue_mapping = queue_mapping;
  3198. }
  3199. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  3200. {
  3201. return skb->queue_mapping;
  3202. }
  3203. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  3204. {
  3205. to->queue_mapping = from->queue_mapping;
  3206. }
  3207. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  3208. {
  3209. skb->queue_mapping = rx_queue + 1;
  3210. }
  3211. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  3212. {
  3213. return skb->queue_mapping - 1;
  3214. }
  3215. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  3216. {
  3217. return skb->queue_mapping != 0;
  3218. }
  3219. static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
  3220. {
  3221. skb->dst_pending_confirm = val;
  3222. }
  3223. static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
  3224. {
  3225. return skb->dst_pending_confirm != 0;
  3226. }
  3227. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  3228. {
  3229. #ifdef CONFIG_XFRM
  3230. return skb->sp;
  3231. #else
  3232. return NULL;
  3233. #endif
  3234. }
  3235. /* Keeps track of mac header offset relative to skb->head.
  3236. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  3237. * For non-tunnel skb it points to skb_mac_header() and for
  3238. * tunnel skb it points to outer mac header.
  3239. * Keeps track of level of encapsulation of network headers.
  3240. */
  3241. struct skb_gso_cb {
  3242. union {
  3243. int mac_offset;
  3244. int data_offset;
  3245. };
  3246. int encap_level;
  3247. __wsum csum;
  3248. __u16 csum_start;
  3249. };
  3250. #define SKB_SGO_CB_OFFSET 32
  3251. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
  3252. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  3253. {
  3254. return (skb_mac_header(inner_skb) - inner_skb->head) -
  3255. SKB_GSO_CB(inner_skb)->mac_offset;
  3256. }
  3257. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  3258. {
  3259. int new_headroom, headroom;
  3260. int ret;
  3261. headroom = skb_headroom(skb);
  3262. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  3263. if (ret)
  3264. return ret;
  3265. new_headroom = skb_headroom(skb);
  3266. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  3267. return 0;
  3268. }
  3269. static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
  3270. {
  3271. /* Do not update partial checksums if remote checksum is enabled. */
  3272. if (skb->remcsum_offload)
  3273. return;
  3274. SKB_GSO_CB(skb)->csum = res;
  3275. SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
  3276. }
  3277. /* Compute the checksum for a gso segment. First compute the checksum value
  3278. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  3279. * then add in skb->csum (checksum from csum_start to end of packet).
  3280. * skb->csum and csum_start are then updated to reflect the checksum of the
  3281. * resultant packet starting from the transport header-- the resultant checksum
  3282. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  3283. * header.
  3284. */
  3285. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  3286. {
  3287. unsigned char *csum_start = skb_transport_header(skb);
  3288. int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
  3289. __wsum partial = SKB_GSO_CB(skb)->csum;
  3290. SKB_GSO_CB(skb)->csum = res;
  3291. SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
  3292. return csum_fold(csum_partial(csum_start, plen, partial));
  3293. }
  3294. static inline bool skb_is_gso(const struct sk_buff *skb)
  3295. {
  3296. return skb_shinfo(skb)->gso_size;
  3297. }
  3298. /* Note: Should be called only if skb_is_gso(skb) is true */
  3299. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  3300. {
  3301. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  3302. }
  3303. static inline void skb_gso_reset(struct sk_buff *skb)
  3304. {
  3305. skb_shinfo(skb)->gso_size = 0;
  3306. skb_shinfo(skb)->gso_segs = 0;
  3307. skb_shinfo(skb)->gso_type = 0;
  3308. }
  3309. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  3310. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  3311. {
  3312. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  3313. * wanted then gso_type will be set. */
  3314. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  3315. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  3316. unlikely(shinfo->gso_type == 0)) {
  3317. __skb_warn_lro_forwarding(skb);
  3318. return true;
  3319. }
  3320. return false;
  3321. }
  3322. static inline void skb_forward_csum(struct sk_buff *skb)
  3323. {
  3324. /* Unfortunately we don't support this one. Any brave souls? */
  3325. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3326. skb->ip_summed = CHECKSUM_NONE;
  3327. }
  3328. /**
  3329. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3330. * @skb: skb to check
  3331. *
  3332. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3333. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3334. * use this helper, to document places where we make this assertion.
  3335. */
  3336. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3337. {
  3338. #ifdef DEBUG
  3339. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3340. #endif
  3341. }
  3342. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3343. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3344. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3345. unsigned int transport_len,
  3346. __sum16(*skb_chkf)(struct sk_buff *skb));
  3347. /**
  3348. * skb_head_is_locked - Determine if the skb->head is locked down
  3349. * @skb: skb to check
  3350. *
  3351. * The head on skbs build around a head frag can be removed if they are
  3352. * not cloned. This function returns true if the skb head is locked down
  3353. * due to either being allocated via kmalloc, or by being a clone with
  3354. * multiple references to the head.
  3355. */
  3356. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3357. {
  3358. return !skb->head_frag || skb_cloned(skb);
  3359. }
  3360. /**
  3361. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3362. *
  3363. * @skb: GSO skb
  3364. *
  3365. * skb_gso_network_seglen is used to determine the real size of the
  3366. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3367. *
  3368. * The MAC/L2 header is not accounted for.
  3369. */
  3370. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3371. {
  3372. unsigned int hdr_len = skb_transport_header(skb) -
  3373. skb_network_header(skb);
  3374. return hdr_len + skb_gso_transport_seglen(skb);
  3375. }
  3376. /* Local Checksum Offload.
  3377. * Compute outer checksum based on the assumption that the
  3378. * inner checksum will be offloaded later.
  3379. * See Documentation/networking/checksum-offloads.txt for
  3380. * explanation of how this works.
  3381. * Fill in outer checksum adjustment (e.g. with sum of outer
  3382. * pseudo-header) before calling.
  3383. * Also ensure that inner checksum is in linear data area.
  3384. */
  3385. static inline __wsum lco_csum(struct sk_buff *skb)
  3386. {
  3387. unsigned char *csum_start = skb_checksum_start(skb);
  3388. unsigned char *l4_hdr = skb_transport_header(skb);
  3389. __wsum partial;
  3390. /* Start with complement of inner checksum adjustment */
  3391. partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
  3392. skb->csum_offset));
  3393. /* Add in checksum of our headers (incl. outer checksum
  3394. * adjustment filled in by caller) and return result.
  3395. */
  3396. return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
  3397. }
  3398. #endif /* __KERNEL__ */
  3399. #endif /* _LINUX_SKBUFF_H */