skbuff.h 108 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472347334743475347634773478347934803481348234833484348534863487348834893490349134923493349434953496349734983499350035013502350335043505350635073508350935103511351235133514351535163517351835193520352135223523352435253526352735283529353035313532353335343535353635373538353935403541354235433544354535463547354835493550355135523553355435553556355735583559356035613562356335643565356635673568356935703571357235733574357535763577357835793580358135823583358435853586358735883589359035913592359335943595359635973598359936003601360236033604360536063607360836093610361136123613361436153616361736183619362036213622362336243625362636273628362936303631363236333634363536363637363836393640364136423643364436453646364736483649365036513652365336543655365636573658365936603661366236633664366536663667366836693670367136723673367436753676367736783679368036813682368336843685368636873688368936903691369236933694369536963697369836993700370137023703370437053706370737083709371037113712371337143715371637173718371937203721372237233724372537263727372837293730373137323733373437353736373737383739374037413742374337443745374637473748374937503751375237533754375537563757375837593760376137623763376437653766376737683769377037713772377337743775377637773778377937803781378237833784
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/kernel.h>
  16. #include <linux/kmemcheck.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/bug.h>
  20. #include <linux/cache.h>
  21. #include <linux/rbtree.h>
  22. #include <linux/socket.h>
  23. #include <linux/atomic.h>
  24. #include <asm/types.h>
  25. #include <linux/spinlock.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #include <linux/rcupdate.h>
  30. #include <linux/hrtimer.h>
  31. #include <linux/dma-mapping.h>
  32. #include <linux/netdev_features.h>
  33. #include <linux/sched.h>
  34. #include <net/flow_dissector.h>
  35. #include <linux/splice.h>
  36. #include <linux/in6.h>
  37. #include <net/flow.h>
  38. /* The interface for checksum offload between the stack and networking drivers
  39. * is as follows...
  40. *
  41. * A. IP checksum related features
  42. *
  43. * Drivers advertise checksum offload capabilities in the features of a device.
  44. * From the stack's point of view these are capabilities offered by the driver,
  45. * a driver typically only advertises features that it is capable of offloading
  46. * to its device.
  47. *
  48. * The checksum related features are:
  49. *
  50. * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
  51. * IP (one's complement) checksum for any combination
  52. * of protocols or protocol layering. The checksum is
  53. * computed and set in a packet per the CHECKSUM_PARTIAL
  54. * interface (see below).
  55. *
  56. * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
  57. * TCP or UDP packets over IPv4. These are specifically
  58. * unencapsulated packets of the form IPv4|TCP or
  59. * IPv4|UDP where the Protocol field in the IPv4 header
  60. * is TCP or UDP. The IPv4 header may contain IP options
  61. * This feature cannot be set in features for a device
  62. * with NETIF_F_HW_CSUM also set. This feature is being
  63. * DEPRECATED (see below).
  64. *
  65. * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
  66. * TCP or UDP packets over IPv6. These are specifically
  67. * unencapsulated packets of the form IPv6|TCP or
  68. * IPv4|UDP where the Next Header field in the IPv6
  69. * header is either TCP or UDP. IPv6 extension headers
  70. * are not supported with this feature. This feature
  71. * cannot be set in features for a device with
  72. * NETIF_F_HW_CSUM also set. This feature is being
  73. * DEPRECATED (see below).
  74. *
  75. * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
  76. * This flag is used only used to disable the RX checksum
  77. * feature for a device. The stack will accept receive
  78. * checksum indication in packets received on a device
  79. * regardless of whether NETIF_F_RXCSUM is set.
  80. *
  81. * B. Checksumming of received packets by device. Indication of checksum
  82. * verification is in set skb->ip_summed. Possible values are:
  83. *
  84. * CHECKSUM_NONE:
  85. *
  86. * Device did not checksum this packet e.g. due to lack of capabilities.
  87. * The packet contains full (though not verified) checksum in packet but
  88. * not in skb->csum. Thus, skb->csum is undefined in this case.
  89. *
  90. * CHECKSUM_UNNECESSARY:
  91. *
  92. * The hardware you're dealing with doesn't calculate the full checksum
  93. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  94. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  95. * if their checksums are okay. skb->csum is still undefined in this case
  96. * though. A driver or device must never modify the checksum field in the
  97. * packet even if checksum is verified.
  98. *
  99. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  100. * TCP: IPv6 and IPv4.
  101. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  102. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  103. * may perform further validation in this case.
  104. * GRE: only if the checksum is present in the header.
  105. * SCTP: indicates the CRC in SCTP header has been validated.
  106. *
  107. * skb->csum_level indicates the number of consecutive checksums found in
  108. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  109. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  110. * and a device is able to verify the checksums for UDP (possibly zero),
  111. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  112. * two. If the device were only able to verify the UDP checksum and not
  113. * GRE, either because it doesn't support GRE checksum of because GRE
  114. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  115. * not considered in this case).
  116. *
  117. * CHECKSUM_COMPLETE:
  118. *
  119. * This is the most generic way. The device supplied checksum of the _whole_
  120. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  121. * hardware doesn't need to parse L3/L4 headers to implement this.
  122. *
  123. * Note: Even if device supports only some protocols, but is able to produce
  124. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  125. *
  126. * CHECKSUM_PARTIAL:
  127. *
  128. * A checksum is set up to be offloaded to a device as described in the
  129. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  130. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  131. * on the same host, or it may be set in the input path in GRO or remote
  132. * checksum offload. For the purposes of checksum verification, the checksum
  133. * referred to by skb->csum_start + skb->csum_offset and any preceding
  134. * checksums in the packet are considered verified. Any checksums in the
  135. * packet that are after the checksum being offloaded are not considered to
  136. * be verified.
  137. *
  138. * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
  139. * in the skb->ip_summed for a packet. Values are:
  140. *
  141. * CHECKSUM_PARTIAL:
  142. *
  143. * The driver is required to checksum the packet as seen by hard_start_xmit()
  144. * from skb->csum_start up to the end, and to record/write the checksum at
  145. * offset skb->csum_start + skb->csum_offset. A driver may verify that the
  146. * csum_start and csum_offset values are valid values given the length and
  147. * offset of the packet, however they should not attempt to validate that the
  148. * checksum refers to a legitimate transport layer checksum-- it is the
  149. * purview of the stack to validate that csum_start and csum_offset are set
  150. * correctly.
  151. *
  152. * When the stack requests checksum offload for a packet, the driver MUST
  153. * ensure that the checksum is set correctly. A driver can either offload the
  154. * checksum calculation to the device, or call skb_checksum_help (in the case
  155. * that the device does not support offload for a particular checksum).
  156. *
  157. * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
  158. * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
  159. * checksum offload capability. If a device has limited checksum capabilities
  160. * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
  161. * described above) a helper function can be called to resolve
  162. * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
  163. * function takes a spec argument that describes the protocol layer that is
  164. * supported for checksum offload and can be called for each packet. If a
  165. * packet does not match the specification for offload, skb_checksum_help
  166. * is called to resolve the checksum.
  167. *
  168. * CHECKSUM_NONE:
  169. *
  170. * The skb was already checksummed by the protocol, or a checksum is not
  171. * required.
  172. *
  173. * CHECKSUM_UNNECESSARY:
  174. *
  175. * This has the same meaning on as CHECKSUM_NONE for checksum offload on
  176. * output.
  177. *
  178. * CHECKSUM_COMPLETE:
  179. * Not used in checksum output. If a driver observes a packet with this value
  180. * set in skbuff, if should treat as CHECKSUM_NONE being set.
  181. *
  182. * D. Non-IP checksum (CRC) offloads
  183. *
  184. * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
  185. * offloading the SCTP CRC in a packet. To perform this offload the stack
  186. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  187. * accordingly. Note the there is no indication in the skbuff that the
  188. * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
  189. * both IP checksum offload and SCTP CRC offload must verify which offload
  190. * is configured for a packet presumably by inspecting packet headers.
  191. *
  192. * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
  193. * offloading the FCOE CRC in a packet. To perform this offload the stack
  194. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  195. * accordingly. Note the there is no indication in the skbuff that the
  196. * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
  197. * both IP checksum offload and FCOE CRC offload must verify which offload
  198. * is configured for a packet presumably by inspecting packet headers.
  199. *
  200. * E. Checksumming on output with GSO.
  201. *
  202. * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
  203. * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
  204. * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
  205. * part of the GSO operation is implied. If a checksum is being offloaded
  206. * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
  207. * are set to refer to the outermost checksum being offload (two offloaded
  208. * checksums are possible with UDP encapsulation).
  209. */
  210. /* Don't change this without changing skb_csum_unnecessary! */
  211. #define CHECKSUM_NONE 0
  212. #define CHECKSUM_UNNECESSARY 1
  213. #define CHECKSUM_COMPLETE 2
  214. #define CHECKSUM_PARTIAL 3
  215. /* Maximum value in skb->csum_level */
  216. #define SKB_MAX_CSUM_LEVEL 3
  217. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  218. #define SKB_WITH_OVERHEAD(X) \
  219. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  220. #define SKB_MAX_ORDER(X, ORDER) \
  221. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  222. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  223. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  224. /* return minimum truesize of one skb containing X bytes of data */
  225. #define SKB_TRUESIZE(X) ((X) + \
  226. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  227. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  228. struct net_device;
  229. struct scatterlist;
  230. struct pipe_inode_info;
  231. struct iov_iter;
  232. struct napi_struct;
  233. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  234. struct nf_conntrack {
  235. atomic_t use;
  236. };
  237. #endif
  238. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  239. struct nf_bridge_info {
  240. atomic_t use;
  241. enum {
  242. BRNF_PROTO_UNCHANGED,
  243. BRNF_PROTO_8021Q,
  244. BRNF_PROTO_PPPOE
  245. } orig_proto:8;
  246. u8 pkt_otherhost:1;
  247. u8 in_prerouting:1;
  248. u8 bridged_dnat:1;
  249. __u16 frag_max_size;
  250. struct net_device *physindev;
  251. /* always valid & non-NULL from FORWARD on, for physdev match */
  252. struct net_device *physoutdev;
  253. union {
  254. /* prerouting: detect dnat in orig/reply direction */
  255. __be32 ipv4_daddr;
  256. struct in6_addr ipv6_daddr;
  257. /* after prerouting + nat detected: store original source
  258. * mac since neigh resolution overwrites it, only used while
  259. * skb is out in neigh layer.
  260. */
  261. char neigh_header[8];
  262. };
  263. };
  264. #endif
  265. struct sk_buff_head {
  266. /* These two members must be first. */
  267. struct sk_buff *next;
  268. struct sk_buff *prev;
  269. __u32 qlen;
  270. spinlock_t lock;
  271. };
  272. struct sk_buff;
  273. /* To allow 64K frame to be packed as single skb without frag_list we
  274. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  275. * buffers which do not start on a page boundary.
  276. *
  277. * Since GRO uses frags we allocate at least 16 regardless of page
  278. * size.
  279. */
  280. #if (65536/PAGE_SIZE + 1) < 16
  281. #define MAX_SKB_FRAGS 16UL
  282. #else
  283. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  284. #endif
  285. extern int sysctl_max_skb_frags;
  286. /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
  287. * segment using its current segmentation instead.
  288. */
  289. #define GSO_BY_FRAGS 0xFFFF
  290. typedef struct skb_frag_struct skb_frag_t;
  291. struct skb_frag_struct {
  292. struct {
  293. struct page *p;
  294. } page;
  295. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  296. __u32 page_offset;
  297. __u32 size;
  298. #else
  299. __u16 page_offset;
  300. __u16 size;
  301. #endif
  302. };
  303. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  304. {
  305. return frag->size;
  306. }
  307. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  308. {
  309. frag->size = size;
  310. }
  311. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  312. {
  313. frag->size += delta;
  314. }
  315. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  316. {
  317. frag->size -= delta;
  318. }
  319. #define HAVE_HW_TIME_STAMP
  320. /**
  321. * struct skb_shared_hwtstamps - hardware time stamps
  322. * @hwtstamp: hardware time stamp transformed into duration
  323. * since arbitrary point in time
  324. *
  325. * Software time stamps generated by ktime_get_real() are stored in
  326. * skb->tstamp.
  327. *
  328. * hwtstamps can only be compared against other hwtstamps from
  329. * the same device.
  330. *
  331. * This structure is attached to packets as part of the
  332. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  333. */
  334. struct skb_shared_hwtstamps {
  335. ktime_t hwtstamp;
  336. };
  337. /* Definitions for tx_flags in struct skb_shared_info */
  338. enum {
  339. /* generate hardware time stamp */
  340. SKBTX_HW_TSTAMP = 1 << 0,
  341. /* generate software time stamp when queueing packet to NIC */
  342. SKBTX_SW_TSTAMP = 1 << 1,
  343. /* device driver is going to provide hardware time stamp */
  344. SKBTX_IN_PROGRESS = 1 << 2,
  345. /* device driver supports TX zero-copy buffers */
  346. SKBTX_DEV_ZEROCOPY = 1 << 3,
  347. /* generate wifi status information (where possible) */
  348. SKBTX_WIFI_STATUS = 1 << 4,
  349. /* This indicates at least one fragment might be overwritten
  350. * (as in vmsplice(), sendfile() ...)
  351. * If we need to compute a TX checksum, we'll need to copy
  352. * all frags to avoid possible bad checksum
  353. */
  354. SKBTX_SHARED_FRAG = 1 << 5,
  355. /* generate software time stamp when entering packet scheduling */
  356. SKBTX_SCHED_TSTAMP = 1 << 6,
  357. };
  358. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  359. SKBTX_SCHED_TSTAMP)
  360. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  361. /*
  362. * The callback notifies userspace to release buffers when skb DMA is done in
  363. * lower device, the skb last reference should be 0 when calling this.
  364. * The zerocopy_success argument is true if zero copy transmit occurred,
  365. * false on data copy or out of memory error caused by data copy attempt.
  366. * The ctx field is used to track device context.
  367. * The desc field is used to track userspace buffer index.
  368. */
  369. struct ubuf_info {
  370. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  371. void *ctx;
  372. unsigned long desc;
  373. };
  374. /* This data is invariant across clones and lives at
  375. * the end of the header data, ie. at skb->end.
  376. */
  377. struct skb_shared_info {
  378. unsigned char nr_frags;
  379. __u8 tx_flags;
  380. unsigned short gso_size;
  381. /* Warning: this field is not always filled in (UFO)! */
  382. unsigned short gso_segs;
  383. unsigned short gso_type;
  384. struct sk_buff *frag_list;
  385. struct skb_shared_hwtstamps hwtstamps;
  386. u32 tskey;
  387. __be32 ip6_frag_id;
  388. /*
  389. * Warning : all fields before dataref are cleared in __alloc_skb()
  390. */
  391. atomic_t dataref;
  392. /* Intermediate layers must ensure that destructor_arg
  393. * remains valid until skb destructor */
  394. void * destructor_arg;
  395. /* must be last field, see pskb_expand_head() */
  396. skb_frag_t frags[MAX_SKB_FRAGS];
  397. };
  398. /* We divide dataref into two halves. The higher 16 bits hold references
  399. * to the payload part of skb->data. The lower 16 bits hold references to
  400. * the entire skb->data. A clone of a headerless skb holds the length of
  401. * the header in skb->hdr_len.
  402. *
  403. * All users must obey the rule that the skb->data reference count must be
  404. * greater than or equal to the payload reference count.
  405. *
  406. * Holding a reference to the payload part means that the user does not
  407. * care about modifications to the header part of skb->data.
  408. */
  409. #define SKB_DATAREF_SHIFT 16
  410. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  411. enum {
  412. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  413. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  414. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  415. };
  416. enum {
  417. SKB_GSO_TCPV4 = 1 << 0,
  418. SKB_GSO_UDP = 1 << 1,
  419. /* This indicates the skb is from an untrusted source. */
  420. SKB_GSO_DODGY = 1 << 2,
  421. /* This indicates the tcp segment has CWR set. */
  422. SKB_GSO_TCP_ECN = 1 << 3,
  423. SKB_GSO_TCP_FIXEDID = 1 << 4,
  424. SKB_GSO_TCPV6 = 1 << 5,
  425. SKB_GSO_FCOE = 1 << 6,
  426. SKB_GSO_GRE = 1 << 7,
  427. SKB_GSO_GRE_CSUM = 1 << 8,
  428. SKB_GSO_IPXIP4 = 1 << 9,
  429. SKB_GSO_IPXIP6 = 1 << 10,
  430. SKB_GSO_UDP_TUNNEL = 1 << 11,
  431. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
  432. SKB_GSO_PARTIAL = 1 << 13,
  433. SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
  434. };
  435. #if BITS_PER_LONG > 32
  436. #define NET_SKBUFF_DATA_USES_OFFSET 1
  437. #endif
  438. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  439. typedef unsigned int sk_buff_data_t;
  440. #else
  441. typedef unsigned char *sk_buff_data_t;
  442. #endif
  443. /**
  444. * struct skb_mstamp - multi resolution time stamps
  445. * @stamp_us: timestamp in us resolution
  446. * @stamp_jiffies: timestamp in jiffies
  447. */
  448. struct skb_mstamp {
  449. union {
  450. u64 v64;
  451. struct {
  452. u32 stamp_us;
  453. u32 stamp_jiffies;
  454. };
  455. };
  456. };
  457. /**
  458. * skb_mstamp_get - get current timestamp
  459. * @cl: place to store timestamps
  460. */
  461. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  462. {
  463. u64 val = local_clock();
  464. do_div(val, NSEC_PER_USEC);
  465. cl->stamp_us = (u32)val;
  466. cl->stamp_jiffies = (u32)jiffies;
  467. }
  468. /**
  469. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  470. * @t1: pointer to newest sample
  471. * @t0: pointer to oldest sample
  472. */
  473. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  474. const struct skb_mstamp *t0)
  475. {
  476. s32 delta_us = t1->stamp_us - t0->stamp_us;
  477. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  478. /* If delta_us is negative, this might be because interval is too big,
  479. * or local_clock() drift is too big : fallback using jiffies.
  480. */
  481. if (delta_us <= 0 ||
  482. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  483. delta_us = jiffies_to_usecs(delta_jiffies);
  484. return delta_us;
  485. }
  486. static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
  487. const struct skb_mstamp *t0)
  488. {
  489. s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
  490. if (!diff)
  491. diff = t1->stamp_us - t0->stamp_us;
  492. return diff > 0;
  493. }
  494. /**
  495. * struct sk_buff - socket buffer
  496. * @next: Next buffer in list
  497. * @prev: Previous buffer in list
  498. * @tstamp: Time we arrived/left
  499. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  500. * @sk: Socket we are owned by
  501. * @dev: Device we arrived on/are leaving by
  502. * @cb: Control buffer. Free for use by every layer. Put private vars here
  503. * @_skb_refdst: destination entry (with norefcount bit)
  504. * @sp: the security path, used for xfrm
  505. * @len: Length of actual data
  506. * @data_len: Data length
  507. * @mac_len: Length of link layer header
  508. * @hdr_len: writable header length of cloned skb
  509. * @csum: Checksum (must include start/offset pair)
  510. * @csum_start: Offset from skb->head where checksumming should start
  511. * @csum_offset: Offset from csum_start where checksum should be stored
  512. * @priority: Packet queueing priority
  513. * @ignore_df: allow local fragmentation
  514. * @cloned: Head may be cloned (check refcnt to be sure)
  515. * @ip_summed: Driver fed us an IP checksum
  516. * @nohdr: Payload reference only, must not modify header
  517. * @nfctinfo: Relationship of this skb to the connection
  518. * @pkt_type: Packet class
  519. * @fclone: skbuff clone status
  520. * @ipvs_property: skbuff is owned by ipvs
  521. * @peeked: this packet has been seen already, so stats have been
  522. * done for it, don't do them again
  523. * @nf_trace: netfilter packet trace flag
  524. * @protocol: Packet protocol from driver
  525. * @destructor: Destruct function
  526. * @nfct: Associated connection, if any
  527. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  528. * @skb_iif: ifindex of device we arrived on
  529. * @tc_index: Traffic control index
  530. * @tc_verd: traffic control verdict
  531. * @hash: the packet hash
  532. * @queue_mapping: Queue mapping for multiqueue devices
  533. * @xmit_more: More SKBs are pending for this queue
  534. * @ndisc_nodetype: router type (from link layer)
  535. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  536. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  537. * ports.
  538. * @sw_hash: indicates hash was computed in software stack
  539. * @wifi_acked_valid: wifi_acked was set
  540. * @wifi_acked: whether frame was acked on wifi or not
  541. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  542. * @napi_id: id of the NAPI struct this skb came from
  543. * @secmark: security marking
  544. * @offload_fwd_mark: fwding offload mark
  545. * @mark: Generic packet mark
  546. * @vlan_proto: vlan encapsulation protocol
  547. * @vlan_tci: vlan tag control information
  548. * @inner_protocol: Protocol (encapsulation)
  549. * @inner_transport_header: Inner transport layer header (encapsulation)
  550. * @inner_network_header: Network layer header (encapsulation)
  551. * @inner_mac_header: Link layer header (encapsulation)
  552. * @transport_header: Transport layer header
  553. * @network_header: Network layer header
  554. * @mac_header: Link layer header
  555. * @tail: Tail pointer
  556. * @end: End pointer
  557. * @head: Head of buffer
  558. * @data: Data head pointer
  559. * @truesize: Buffer size
  560. * @users: User count - see {datagram,tcp}.c
  561. */
  562. struct sk_buff {
  563. union {
  564. struct {
  565. /* These two members must be first. */
  566. struct sk_buff *next;
  567. struct sk_buff *prev;
  568. union {
  569. ktime_t tstamp;
  570. struct skb_mstamp skb_mstamp;
  571. };
  572. };
  573. struct rb_node rbnode; /* used in netem & tcp stack */
  574. };
  575. struct sock *sk;
  576. struct net_device *dev;
  577. /*
  578. * This is the control buffer. It is free to use for every
  579. * layer. Please put your private variables there. If you
  580. * want to keep them across layers you have to do a skb_clone()
  581. * first. This is owned by whoever has the skb queued ATM.
  582. */
  583. char cb[48] __aligned(8);
  584. unsigned long _skb_refdst;
  585. void (*destructor)(struct sk_buff *skb);
  586. #ifdef CONFIG_XFRM
  587. struct sec_path *sp;
  588. #endif
  589. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  590. struct nf_conntrack *nfct;
  591. #endif
  592. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  593. struct nf_bridge_info *nf_bridge;
  594. #endif
  595. unsigned int len,
  596. data_len;
  597. __u16 mac_len,
  598. hdr_len;
  599. /* Following fields are _not_ copied in __copy_skb_header()
  600. * Note that queue_mapping is here mostly to fill a hole.
  601. */
  602. kmemcheck_bitfield_begin(flags1);
  603. __u16 queue_mapping;
  604. __u8 cloned:1,
  605. nohdr:1,
  606. fclone:2,
  607. peeked:1,
  608. head_frag:1,
  609. xmit_more:1;
  610. /* one bit hole */
  611. kmemcheck_bitfield_end(flags1);
  612. /* fields enclosed in headers_start/headers_end are copied
  613. * using a single memcpy() in __copy_skb_header()
  614. */
  615. /* private: */
  616. __u32 headers_start[0];
  617. /* public: */
  618. /* if you move pkt_type around you also must adapt those constants */
  619. #ifdef __BIG_ENDIAN_BITFIELD
  620. #define PKT_TYPE_MAX (7 << 5)
  621. #else
  622. #define PKT_TYPE_MAX 7
  623. #endif
  624. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  625. __u8 __pkt_type_offset[0];
  626. __u8 pkt_type:3;
  627. __u8 pfmemalloc:1;
  628. __u8 ignore_df:1;
  629. __u8 nfctinfo:3;
  630. __u8 nf_trace:1;
  631. __u8 ip_summed:2;
  632. __u8 ooo_okay:1;
  633. __u8 l4_hash:1;
  634. __u8 sw_hash:1;
  635. __u8 wifi_acked_valid:1;
  636. __u8 wifi_acked:1;
  637. __u8 no_fcs:1;
  638. /* Indicates the inner headers are valid in the skbuff. */
  639. __u8 encapsulation:1;
  640. __u8 encap_hdr_csum:1;
  641. __u8 csum_valid:1;
  642. __u8 csum_complete_sw:1;
  643. __u8 csum_level:2;
  644. __u8 csum_bad:1;
  645. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  646. __u8 ndisc_nodetype:2;
  647. #endif
  648. __u8 ipvs_property:1;
  649. __u8 inner_protocol_type:1;
  650. __u8 remcsum_offload:1;
  651. /* 3 or 5 bit hole */
  652. #ifdef CONFIG_NET_SCHED
  653. __u16 tc_index; /* traffic control index */
  654. #ifdef CONFIG_NET_CLS_ACT
  655. __u16 tc_verd; /* traffic control verdict */
  656. #endif
  657. #endif
  658. union {
  659. __wsum csum;
  660. struct {
  661. __u16 csum_start;
  662. __u16 csum_offset;
  663. };
  664. };
  665. __u32 priority;
  666. int skb_iif;
  667. __u32 hash;
  668. __be16 vlan_proto;
  669. __u16 vlan_tci;
  670. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  671. union {
  672. unsigned int napi_id;
  673. unsigned int sender_cpu;
  674. };
  675. #endif
  676. union {
  677. #ifdef CONFIG_NETWORK_SECMARK
  678. __u32 secmark;
  679. #endif
  680. #ifdef CONFIG_NET_SWITCHDEV
  681. __u32 offload_fwd_mark;
  682. #endif
  683. };
  684. union {
  685. __u32 mark;
  686. __u32 reserved_tailroom;
  687. };
  688. union {
  689. __be16 inner_protocol;
  690. __u8 inner_ipproto;
  691. };
  692. __u16 inner_transport_header;
  693. __u16 inner_network_header;
  694. __u16 inner_mac_header;
  695. __be16 protocol;
  696. __u16 transport_header;
  697. __u16 network_header;
  698. __u16 mac_header;
  699. /* private: */
  700. __u32 headers_end[0];
  701. /* public: */
  702. /* These elements must be at the end, see alloc_skb() for details. */
  703. sk_buff_data_t tail;
  704. sk_buff_data_t end;
  705. unsigned char *head,
  706. *data;
  707. unsigned int truesize;
  708. atomic_t users;
  709. };
  710. #ifdef __KERNEL__
  711. /*
  712. * Handling routines are only of interest to the kernel
  713. */
  714. #include <linux/slab.h>
  715. #define SKB_ALLOC_FCLONE 0x01
  716. #define SKB_ALLOC_RX 0x02
  717. #define SKB_ALLOC_NAPI 0x04
  718. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  719. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  720. {
  721. return unlikely(skb->pfmemalloc);
  722. }
  723. /*
  724. * skb might have a dst pointer attached, refcounted or not.
  725. * _skb_refdst low order bit is set if refcount was _not_ taken
  726. */
  727. #define SKB_DST_NOREF 1UL
  728. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  729. /**
  730. * skb_dst - returns skb dst_entry
  731. * @skb: buffer
  732. *
  733. * Returns skb dst_entry, regardless of reference taken or not.
  734. */
  735. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  736. {
  737. /* If refdst was not refcounted, check we still are in a
  738. * rcu_read_lock section
  739. */
  740. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  741. !rcu_read_lock_held() &&
  742. !rcu_read_lock_bh_held());
  743. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  744. }
  745. /**
  746. * skb_dst_set - sets skb dst
  747. * @skb: buffer
  748. * @dst: dst entry
  749. *
  750. * Sets skb dst, assuming a reference was taken on dst and should
  751. * be released by skb_dst_drop()
  752. */
  753. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  754. {
  755. skb->_skb_refdst = (unsigned long)dst;
  756. }
  757. /**
  758. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  759. * @skb: buffer
  760. * @dst: dst entry
  761. *
  762. * Sets skb dst, assuming a reference was not taken on dst.
  763. * If dst entry is cached, we do not take reference and dst_release
  764. * will be avoided by refdst_drop. If dst entry is not cached, we take
  765. * reference, so that last dst_release can destroy the dst immediately.
  766. */
  767. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  768. {
  769. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  770. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  771. }
  772. /**
  773. * skb_dst_is_noref - Test if skb dst isn't refcounted
  774. * @skb: buffer
  775. */
  776. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  777. {
  778. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  779. }
  780. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  781. {
  782. return (struct rtable *)skb_dst(skb);
  783. }
  784. void kfree_skb(struct sk_buff *skb);
  785. void kfree_skb_list(struct sk_buff *segs);
  786. void skb_tx_error(struct sk_buff *skb);
  787. void consume_skb(struct sk_buff *skb);
  788. void __kfree_skb(struct sk_buff *skb);
  789. extern struct kmem_cache *skbuff_head_cache;
  790. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  791. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  792. bool *fragstolen, int *delta_truesize);
  793. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  794. int node);
  795. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  796. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  797. static inline struct sk_buff *alloc_skb(unsigned int size,
  798. gfp_t priority)
  799. {
  800. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  801. }
  802. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  803. unsigned long data_len,
  804. int max_page_order,
  805. int *errcode,
  806. gfp_t gfp_mask);
  807. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  808. struct sk_buff_fclones {
  809. struct sk_buff skb1;
  810. struct sk_buff skb2;
  811. atomic_t fclone_ref;
  812. };
  813. /**
  814. * skb_fclone_busy - check if fclone is busy
  815. * @skb: buffer
  816. *
  817. * Returns true if skb is a fast clone, and its clone is not freed.
  818. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  819. * so we also check that this didnt happen.
  820. */
  821. static inline bool skb_fclone_busy(const struct sock *sk,
  822. const struct sk_buff *skb)
  823. {
  824. const struct sk_buff_fclones *fclones;
  825. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  826. return skb->fclone == SKB_FCLONE_ORIG &&
  827. atomic_read(&fclones->fclone_ref) > 1 &&
  828. fclones->skb2.sk == sk;
  829. }
  830. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  831. gfp_t priority)
  832. {
  833. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  834. }
  835. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  836. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  837. {
  838. return __alloc_skb_head(priority, -1);
  839. }
  840. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  841. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  842. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  843. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  844. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  845. gfp_t gfp_mask, bool fclone);
  846. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  847. gfp_t gfp_mask)
  848. {
  849. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  850. }
  851. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  852. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  853. unsigned int headroom);
  854. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  855. int newtailroom, gfp_t priority);
  856. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  857. int offset, int len);
  858. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  859. int len);
  860. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  861. int skb_pad(struct sk_buff *skb, int pad);
  862. #define dev_kfree_skb(a) consume_skb(a)
  863. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  864. int getfrag(void *from, char *to, int offset,
  865. int len, int odd, struct sk_buff *skb),
  866. void *from, int length);
  867. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  868. int offset, size_t size);
  869. struct skb_seq_state {
  870. __u32 lower_offset;
  871. __u32 upper_offset;
  872. __u32 frag_idx;
  873. __u32 stepped_offset;
  874. struct sk_buff *root_skb;
  875. struct sk_buff *cur_skb;
  876. __u8 *frag_data;
  877. };
  878. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  879. unsigned int to, struct skb_seq_state *st);
  880. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  881. struct skb_seq_state *st);
  882. void skb_abort_seq_read(struct skb_seq_state *st);
  883. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  884. unsigned int to, struct ts_config *config);
  885. /*
  886. * Packet hash types specify the type of hash in skb_set_hash.
  887. *
  888. * Hash types refer to the protocol layer addresses which are used to
  889. * construct a packet's hash. The hashes are used to differentiate or identify
  890. * flows of the protocol layer for the hash type. Hash types are either
  891. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  892. *
  893. * Properties of hashes:
  894. *
  895. * 1) Two packets in different flows have different hash values
  896. * 2) Two packets in the same flow should have the same hash value
  897. *
  898. * A hash at a higher layer is considered to be more specific. A driver should
  899. * set the most specific hash possible.
  900. *
  901. * A driver cannot indicate a more specific hash than the layer at which a hash
  902. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  903. *
  904. * A driver may indicate a hash level which is less specific than the
  905. * actual layer the hash was computed on. For instance, a hash computed
  906. * at L4 may be considered an L3 hash. This should only be done if the
  907. * driver can't unambiguously determine that the HW computed the hash at
  908. * the higher layer. Note that the "should" in the second property above
  909. * permits this.
  910. */
  911. enum pkt_hash_types {
  912. PKT_HASH_TYPE_NONE, /* Undefined type */
  913. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  914. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  915. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  916. };
  917. static inline void skb_clear_hash(struct sk_buff *skb)
  918. {
  919. skb->hash = 0;
  920. skb->sw_hash = 0;
  921. skb->l4_hash = 0;
  922. }
  923. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  924. {
  925. if (!skb->l4_hash)
  926. skb_clear_hash(skb);
  927. }
  928. static inline void
  929. __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
  930. {
  931. skb->l4_hash = is_l4;
  932. skb->sw_hash = is_sw;
  933. skb->hash = hash;
  934. }
  935. static inline void
  936. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  937. {
  938. /* Used by drivers to set hash from HW */
  939. __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
  940. }
  941. static inline void
  942. __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
  943. {
  944. __skb_set_hash(skb, hash, true, is_l4);
  945. }
  946. void __skb_get_hash(struct sk_buff *skb);
  947. u32 skb_get_poff(const struct sk_buff *skb);
  948. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  949. const struct flow_keys *keys, int hlen);
  950. __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
  951. void *data, int hlen_proto);
  952. static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
  953. int thoff, u8 ip_proto)
  954. {
  955. return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
  956. }
  957. void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  958. const struct flow_dissector_key *key,
  959. unsigned int key_count);
  960. bool __skb_flow_dissect(const struct sk_buff *skb,
  961. struct flow_dissector *flow_dissector,
  962. void *target_container,
  963. void *data, __be16 proto, int nhoff, int hlen,
  964. unsigned int flags);
  965. static inline bool skb_flow_dissect(const struct sk_buff *skb,
  966. struct flow_dissector *flow_dissector,
  967. void *target_container, unsigned int flags)
  968. {
  969. return __skb_flow_dissect(skb, flow_dissector, target_container,
  970. NULL, 0, 0, 0, flags);
  971. }
  972. static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
  973. struct flow_keys *flow,
  974. unsigned int flags)
  975. {
  976. memset(flow, 0, sizeof(*flow));
  977. return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
  978. NULL, 0, 0, 0, flags);
  979. }
  980. static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
  981. void *data, __be16 proto,
  982. int nhoff, int hlen,
  983. unsigned int flags)
  984. {
  985. memset(flow, 0, sizeof(*flow));
  986. return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
  987. data, proto, nhoff, hlen, flags);
  988. }
  989. static inline __u32 skb_get_hash(struct sk_buff *skb)
  990. {
  991. if (!skb->l4_hash && !skb->sw_hash)
  992. __skb_get_hash(skb);
  993. return skb->hash;
  994. }
  995. __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
  996. static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
  997. {
  998. if (!skb->l4_hash && !skb->sw_hash) {
  999. struct flow_keys keys;
  1000. __u32 hash = __get_hash_from_flowi6(fl6, &keys);
  1001. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1002. }
  1003. return skb->hash;
  1004. }
  1005. __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
  1006. static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
  1007. {
  1008. if (!skb->l4_hash && !skb->sw_hash) {
  1009. struct flow_keys keys;
  1010. __u32 hash = __get_hash_from_flowi4(fl4, &keys);
  1011. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1012. }
  1013. return skb->hash;
  1014. }
  1015. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  1016. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  1017. {
  1018. return skb->hash;
  1019. }
  1020. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  1021. {
  1022. to->hash = from->hash;
  1023. to->sw_hash = from->sw_hash;
  1024. to->l4_hash = from->l4_hash;
  1025. };
  1026. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1027. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1028. {
  1029. return skb->head + skb->end;
  1030. }
  1031. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1032. {
  1033. return skb->end;
  1034. }
  1035. #else
  1036. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1037. {
  1038. return skb->end;
  1039. }
  1040. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1041. {
  1042. return skb->end - skb->head;
  1043. }
  1044. #endif
  1045. /* Internal */
  1046. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  1047. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  1048. {
  1049. return &skb_shinfo(skb)->hwtstamps;
  1050. }
  1051. /**
  1052. * skb_queue_empty - check if a queue is empty
  1053. * @list: queue head
  1054. *
  1055. * Returns true if the queue is empty, false otherwise.
  1056. */
  1057. static inline int skb_queue_empty(const struct sk_buff_head *list)
  1058. {
  1059. return list->next == (const struct sk_buff *) list;
  1060. }
  1061. /**
  1062. * skb_queue_is_last - check if skb is the last entry in the queue
  1063. * @list: queue head
  1064. * @skb: buffer
  1065. *
  1066. * Returns true if @skb is the last buffer on the list.
  1067. */
  1068. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  1069. const struct sk_buff *skb)
  1070. {
  1071. return skb->next == (const struct sk_buff *) list;
  1072. }
  1073. /**
  1074. * skb_queue_is_first - check if skb is the first entry in the queue
  1075. * @list: queue head
  1076. * @skb: buffer
  1077. *
  1078. * Returns true if @skb is the first buffer on the list.
  1079. */
  1080. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  1081. const struct sk_buff *skb)
  1082. {
  1083. return skb->prev == (const struct sk_buff *) list;
  1084. }
  1085. /**
  1086. * skb_queue_next - return the next packet in the queue
  1087. * @list: queue head
  1088. * @skb: current buffer
  1089. *
  1090. * Return the next packet in @list after @skb. It is only valid to
  1091. * call this if skb_queue_is_last() evaluates to false.
  1092. */
  1093. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  1094. const struct sk_buff *skb)
  1095. {
  1096. /* This BUG_ON may seem severe, but if we just return then we
  1097. * are going to dereference garbage.
  1098. */
  1099. BUG_ON(skb_queue_is_last(list, skb));
  1100. return skb->next;
  1101. }
  1102. /**
  1103. * skb_queue_prev - return the prev packet in the queue
  1104. * @list: queue head
  1105. * @skb: current buffer
  1106. *
  1107. * Return the prev packet in @list before @skb. It is only valid to
  1108. * call this if skb_queue_is_first() evaluates to false.
  1109. */
  1110. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  1111. const struct sk_buff *skb)
  1112. {
  1113. /* This BUG_ON may seem severe, but if we just return then we
  1114. * are going to dereference garbage.
  1115. */
  1116. BUG_ON(skb_queue_is_first(list, skb));
  1117. return skb->prev;
  1118. }
  1119. /**
  1120. * skb_get - reference buffer
  1121. * @skb: buffer to reference
  1122. *
  1123. * Makes another reference to a socket buffer and returns a pointer
  1124. * to the buffer.
  1125. */
  1126. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  1127. {
  1128. atomic_inc(&skb->users);
  1129. return skb;
  1130. }
  1131. /*
  1132. * If users == 1, we are the only owner and are can avoid redundant
  1133. * atomic change.
  1134. */
  1135. /**
  1136. * skb_cloned - is the buffer a clone
  1137. * @skb: buffer to check
  1138. *
  1139. * Returns true if the buffer was generated with skb_clone() and is
  1140. * one of multiple shared copies of the buffer. Cloned buffers are
  1141. * shared data so must not be written to under normal circumstances.
  1142. */
  1143. static inline int skb_cloned(const struct sk_buff *skb)
  1144. {
  1145. return skb->cloned &&
  1146. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  1147. }
  1148. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  1149. {
  1150. might_sleep_if(gfpflags_allow_blocking(pri));
  1151. if (skb_cloned(skb))
  1152. return pskb_expand_head(skb, 0, 0, pri);
  1153. return 0;
  1154. }
  1155. /**
  1156. * skb_header_cloned - is the header a clone
  1157. * @skb: buffer to check
  1158. *
  1159. * Returns true if modifying the header part of the buffer requires
  1160. * the data to be copied.
  1161. */
  1162. static inline int skb_header_cloned(const struct sk_buff *skb)
  1163. {
  1164. int dataref;
  1165. if (!skb->cloned)
  1166. return 0;
  1167. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  1168. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  1169. return dataref != 1;
  1170. }
  1171. static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
  1172. {
  1173. might_sleep_if(gfpflags_allow_blocking(pri));
  1174. if (skb_header_cloned(skb))
  1175. return pskb_expand_head(skb, 0, 0, pri);
  1176. return 0;
  1177. }
  1178. /**
  1179. * skb_header_release - release reference to header
  1180. * @skb: buffer to operate on
  1181. *
  1182. * Drop a reference to the header part of the buffer. This is done
  1183. * by acquiring a payload reference. You must not read from the header
  1184. * part of skb->data after this.
  1185. * Note : Check if you can use __skb_header_release() instead.
  1186. */
  1187. static inline void skb_header_release(struct sk_buff *skb)
  1188. {
  1189. BUG_ON(skb->nohdr);
  1190. skb->nohdr = 1;
  1191. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1192. }
  1193. /**
  1194. * __skb_header_release - release reference to header
  1195. * @skb: buffer to operate on
  1196. *
  1197. * Variant of skb_header_release() assuming skb is private to caller.
  1198. * We can avoid one atomic operation.
  1199. */
  1200. static inline void __skb_header_release(struct sk_buff *skb)
  1201. {
  1202. skb->nohdr = 1;
  1203. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1204. }
  1205. /**
  1206. * skb_shared - is the buffer shared
  1207. * @skb: buffer to check
  1208. *
  1209. * Returns true if more than one person has a reference to this
  1210. * buffer.
  1211. */
  1212. static inline int skb_shared(const struct sk_buff *skb)
  1213. {
  1214. return atomic_read(&skb->users) != 1;
  1215. }
  1216. /**
  1217. * skb_share_check - check if buffer is shared and if so clone it
  1218. * @skb: buffer to check
  1219. * @pri: priority for memory allocation
  1220. *
  1221. * If the buffer is shared the buffer is cloned and the old copy
  1222. * drops a reference. A new clone with a single reference is returned.
  1223. * If the buffer is not shared the original buffer is returned. When
  1224. * being called from interrupt status or with spinlocks held pri must
  1225. * be GFP_ATOMIC.
  1226. *
  1227. * NULL is returned on a memory allocation failure.
  1228. */
  1229. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1230. {
  1231. might_sleep_if(gfpflags_allow_blocking(pri));
  1232. if (skb_shared(skb)) {
  1233. struct sk_buff *nskb = skb_clone(skb, pri);
  1234. if (likely(nskb))
  1235. consume_skb(skb);
  1236. else
  1237. kfree_skb(skb);
  1238. skb = nskb;
  1239. }
  1240. return skb;
  1241. }
  1242. /*
  1243. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1244. * packets to handle cases where we have a local reader and forward
  1245. * and a couple of other messy ones. The normal one is tcpdumping
  1246. * a packet thats being forwarded.
  1247. */
  1248. /**
  1249. * skb_unshare - make a copy of a shared buffer
  1250. * @skb: buffer to check
  1251. * @pri: priority for memory allocation
  1252. *
  1253. * If the socket buffer is a clone then this function creates a new
  1254. * copy of the data, drops a reference count on the old copy and returns
  1255. * the new copy with the reference count at 1. If the buffer is not a clone
  1256. * the original buffer is returned. When called with a spinlock held or
  1257. * from interrupt state @pri must be %GFP_ATOMIC
  1258. *
  1259. * %NULL is returned on a memory allocation failure.
  1260. */
  1261. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1262. gfp_t pri)
  1263. {
  1264. might_sleep_if(gfpflags_allow_blocking(pri));
  1265. if (skb_cloned(skb)) {
  1266. struct sk_buff *nskb = skb_copy(skb, pri);
  1267. /* Free our shared copy */
  1268. if (likely(nskb))
  1269. consume_skb(skb);
  1270. else
  1271. kfree_skb(skb);
  1272. skb = nskb;
  1273. }
  1274. return skb;
  1275. }
  1276. /**
  1277. * skb_peek - peek at the head of an &sk_buff_head
  1278. * @list_: list to peek at
  1279. *
  1280. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1281. * be careful with this one. A peek leaves the buffer on the
  1282. * list and someone else may run off with it. You must hold
  1283. * the appropriate locks or have a private queue to do this.
  1284. *
  1285. * Returns %NULL for an empty list or a pointer to the head element.
  1286. * The reference count is not incremented and the reference is therefore
  1287. * volatile. Use with caution.
  1288. */
  1289. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1290. {
  1291. struct sk_buff *skb = list_->next;
  1292. if (skb == (struct sk_buff *)list_)
  1293. skb = NULL;
  1294. return skb;
  1295. }
  1296. /**
  1297. * skb_peek_next - peek skb following the given one from a queue
  1298. * @skb: skb to start from
  1299. * @list_: list to peek at
  1300. *
  1301. * Returns %NULL when the end of the list is met or a pointer to the
  1302. * next element. The reference count is not incremented and the
  1303. * reference is therefore volatile. Use with caution.
  1304. */
  1305. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1306. const struct sk_buff_head *list_)
  1307. {
  1308. struct sk_buff *next = skb->next;
  1309. if (next == (struct sk_buff *)list_)
  1310. next = NULL;
  1311. return next;
  1312. }
  1313. /**
  1314. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1315. * @list_: list to peek at
  1316. *
  1317. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1318. * be careful with this one. A peek leaves the buffer on the
  1319. * list and someone else may run off with it. You must hold
  1320. * the appropriate locks or have a private queue to do this.
  1321. *
  1322. * Returns %NULL for an empty list or a pointer to the tail element.
  1323. * The reference count is not incremented and the reference is therefore
  1324. * volatile. Use with caution.
  1325. */
  1326. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1327. {
  1328. struct sk_buff *skb = list_->prev;
  1329. if (skb == (struct sk_buff *)list_)
  1330. skb = NULL;
  1331. return skb;
  1332. }
  1333. /**
  1334. * skb_queue_len - get queue length
  1335. * @list_: list to measure
  1336. *
  1337. * Return the length of an &sk_buff queue.
  1338. */
  1339. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1340. {
  1341. return list_->qlen;
  1342. }
  1343. /**
  1344. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1345. * @list: queue to initialize
  1346. *
  1347. * This initializes only the list and queue length aspects of
  1348. * an sk_buff_head object. This allows to initialize the list
  1349. * aspects of an sk_buff_head without reinitializing things like
  1350. * the spinlock. It can also be used for on-stack sk_buff_head
  1351. * objects where the spinlock is known to not be used.
  1352. */
  1353. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1354. {
  1355. list->prev = list->next = (struct sk_buff *)list;
  1356. list->qlen = 0;
  1357. }
  1358. /*
  1359. * This function creates a split out lock class for each invocation;
  1360. * this is needed for now since a whole lot of users of the skb-queue
  1361. * infrastructure in drivers have different locking usage (in hardirq)
  1362. * than the networking core (in softirq only). In the long run either the
  1363. * network layer or drivers should need annotation to consolidate the
  1364. * main types of usage into 3 classes.
  1365. */
  1366. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1367. {
  1368. spin_lock_init(&list->lock);
  1369. __skb_queue_head_init(list);
  1370. }
  1371. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1372. struct lock_class_key *class)
  1373. {
  1374. skb_queue_head_init(list);
  1375. lockdep_set_class(&list->lock, class);
  1376. }
  1377. /*
  1378. * Insert an sk_buff on a list.
  1379. *
  1380. * The "__skb_xxxx()" functions are the non-atomic ones that
  1381. * can only be called with interrupts disabled.
  1382. */
  1383. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1384. struct sk_buff_head *list);
  1385. static inline void __skb_insert(struct sk_buff *newsk,
  1386. struct sk_buff *prev, struct sk_buff *next,
  1387. struct sk_buff_head *list)
  1388. {
  1389. newsk->next = next;
  1390. newsk->prev = prev;
  1391. next->prev = prev->next = newsk;
  1392. list->qlen++;
  1393. }
  1394. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1395. struct sk_buff *prev,
  1396. struct sk_buff *next)
  1397. {
  1398. struct sk_buff *first = list->next;
  1399. struct sk_buff *last = list->prev;
  1400. first->prev = prev;
  1401. prev->next = first;
  1402. last->next = next;
  1403. next->prev = last;
  1404. }
  1405. /**
  1406. * skb_queue_splice - join two skb lists, this is designed for stacks
  1407. * @list: the new list to add
  1408. * @head: the place to add it in the first list
  1409. */
  1410. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1411. struct sk_buff_head *head)
  1412. {
  1413. if (!skb_queue_empty(list)) {
  1414. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1415. head->qlen += list->qlen;
  1416. }
  1417. }
  1418. /**
  1419. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1420. * @list: the new list to add
  1421. * @head: the place to add it in the first list
  1422. *
  1423. * The list at @list is reinitialised
  1424. */
  1425. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1426. struct sk_buff_head *head)
  1427. {
  1428. if (!skb_queue_empty(list)) {
  1429. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1430. head->qlen += list->qlen;
  1431. __skb_queue_head_init(list);
  1432. }
  1433. }
  1434. /**
  1435. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1436. * @list: the new list to add
  1437. * @head: the place to add it in the first list
  1438. */
  1439. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1440. struct sk_buff_head *head)
  1441. {
  1442. if (!skb_queue_empty(list)) {
  1443. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1444. head->qlen += list->qlen;
  1445. }
  1446. }
  1447. /**
  1448. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1449. * @list: the new list to add
  1450. * @head: the place to add it in the first list
  1451. *
  1452. * Each of the lists is a queue.
  1453. * The list at @list is reinitialised
  1454. */
  1455. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1456. struct sk_buff_head *head)
  1457. {
  1458. if (!skb_queue_empty(list)) {
  1459. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1460. head->qlen += list->qlen;
  1461. __skb_queue_head_init(list);
  1462. }
  1463. }
  1464. /**
  1465. * __skb_queue_after - queue a buffer at the list head
  1466. * @list: list to use
  1467. * @prev: place after this buffer
  1468. * @newsk: buffer to queue
  1469. *
  1470. * Queue a buffer int the middle of a list. This function takes no locks
  1471. * and you must therefore hold required locks before calling it.
  1472. *
  1473. * A buffer cannot be placed on two lists at the same time.
  1474. */
  1475. static inline void __skb_queue_after(struct sk_buff_head *list,
  1476. struct sk_buff *prev,
  1477. struct sk_buff *newsk)
  1478. {
  1479. __skb_insert(newsk, prev, prev->next, list);
  1480. }
  1481. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1482. struct sk_buff_head *list);
  1483. static inline void __skb_queue_before(struct sk_buff_head *list,
  1484. struct sk_buff *next,
  1485. struct sk_buff *newsk)
  1486. {
  1487. __skb_insert(newsk, next->prev, next, list);
  1488. }
  1489. /**
  1490. * __skb_queue_head - queue a buffer at the list head
  1491. * @list: list to use
  1492. * @newsk: buffer to queue
  1493. *
  1494. * Queue a buffer at the start of a list. This function takes no locks
  1495. * and you must therefore hold required locks before calling it.
  1496. *
  1497. * A buffer cannot be placed on two lists at the same time.
  1498. */
  1499. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1500. static inline void __skb_queue_head(struct sk_buff_head *list,
  1501. struct sk_buff *newsk)
  1502. {
  1503. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1504. }
  1505. /**
  1506. * __skb_queue_tail - queue a buffer at the list tail
  1507. * @list: list to use
  1508. * @newsk: buffer to queue
  1509. *
  1510. * Queue a buffer at the end of a list. This function takes no locks
  1511. * and you must therefore hold required locks before calling it.
  1512. *
  1513. * A buffer cannot be placed on two lists at the same time.
  1514. */
  1515. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1516. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1517. struct sk_buff *newsk)
  1518. {
  1519. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1520. }
  1521. /*
  1522. * remove sk_buff from list. _Must_ be called atomically, and with
  1523. * the list known..
  1524. */
  1525. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1526. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1527. {
  1528. struct sk_buff *next, *prev;
  1529. list->qlen--;
  1530. next = skb->next;
  1531. prev = skb->prev;
  1532. skb->next = skb->prev = NULL;
  1533. next->prev = prev;
  1534. prev->next = next;
  1535. }
  1536. /**
  1537. * __skb_dequeue - remove from the head of the queue
  1538. * @list: list to dequeue from
  1539. *
  1540. * Remove the head of the list. This function does not take any locks
  1541. * so must be used with appropriate locks held only. The head item is
  1542. * returned or %NULL if the list is empty.
  1543. */
  1544. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1545. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1546. {
  1547. struct sk_buff *skb = skb_peek(list);
  1548. if (skb)
  1549. __skb_unlink(skb, list);
  1550. return skb;
  1551. }
  1552. /**
  1553. * __skb_dequeue_tail - remove from the tail of the queue
  1554. * @list: list to dequeue from
  1555. *
  1556. * Remove the tail of the list. This function does not take any locks
  1557. * so must be used with appropriate locks held only. The tail item is
  1558. * returned or %NULL if the list is empty.
  1559. */
  1560. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1561. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1562. {
  1563. struct sk_buff *skb = skb_peek_tail(list);
  1564. if (skb)
  1565. __skb_unlink(skb, list);
  1566. return skb;
  1567. }
  1568. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1569. {
  1570. return skb->data_len;
  1571. }
  1572. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1573. {
  1574. return skb->len - skb->data_len;
  1575. }
  1576. static inline int skb_pagelen(const struct sk_buff *skb)
  1577. {
  1578. int i, len = 0;
  1579. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1580. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1581. return len + skb_headlen(skb);
  1582. }
  1583. /**
  1584. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1585. * @skb: buffer containing fragment to be initialised
  1586. * @i: paged fragment index to initialise
  1587. * @page: the page to use for this fragment
  1588. * @off: the offset to the data with @page
  1589. * @size: the length of the data
  1590. *
  1591. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1592. * offset @off within @page.
  1593. *
  1594. * Does not take any additional reference on the fragment.
  1595. */
  1596. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1597. struct page *page, int off, int size)
  1598. {
  1599. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1600. /*
  1601. * Propagate page pfmemalloc to the skb if we can. The problem is
  1602. * that not all callers have unique ownership of the page but rely
  1603. * on page_is_pfmemalloc doing the right thing(tm).
  1604. */
  1605. frag->page.p = page;
  1606. frag->page_offset = off;
  1607. skb_frag_size_set(frag, size);
  1608. page = compound_head(page);
  1609. if (page_is_pfmemalloc(page))
  1610. skb->pfmemalloc = true;
  1611. }
  1612. /**
  1613. * skb_fill_page_desc - initialise a paged fragment in an skb
  1614. * @skb: buffer containing fragment to be initialised
  1615. * @i: paged fragment index to initialise
  1616. * @page: the page to use for this fragment
  1617. * @off: the offset to the data with @page
  1618. * @size: the length of the data
  1619. *
  1620. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1621. * @skb to point to @size bytes at offset @off within @page. In
  1622. * addition updates @skb such that @i is the last fragment.
  1623. *
  1624. * Does not take any additional reference on the fragment.
  1625. */
  1626. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1627. struct page *page, int off, int size)
  1628. {
  1629. __skb_fill_page_desc(skb, i, page, off, size);
  1630. skb_shinfo(skb)->nr_frags = i + 1;
  1631. }
  1632. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1633. int size, unsigned int truesize);
  1634. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1635. unsigned int truesize);
  1636. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1637. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1638. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1639. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1640. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1641. {
  1642. return skb->head + skb->tail;
  1643. }
  1644. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1645. {
  1646. skb->tail = skb->data - skb->head;
  1647. }
  1648. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1649. {
  1650. skb_reset_tail_pointer(skb);
  1651. skb->tail += offset;
  1652. }
  1653. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1654. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1655. {
  1656. return skb->tail;
  1657. }
  1658. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1659. {
  1660. skb->tail = skb->data;
  1661. }
  1662. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1663. {
  1664. skb->tail = skb->data + offset;
  1665. }
  1666. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1667. /*
  1668. * Add data to an sk_buff
  1669. */
  1670. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1671. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1672. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1673. {
  1674. unsigned char *tmp = skb_tail_pointer(skb);
  1675. SKB_LINEAR_ASSERT(skb);
  1676. skb->tail += len;
  1677. skb->len += len;
  1678. return tmp;
  1679. }
  1680. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1681. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1682. {
  1683. skb->data -= len;
  1684. skb->len += len;
  1685. return skb->data;
  1686. }
  1687. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1688. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1689. {
  1690. skb->len -= len;
  1691. BUG_ON(skb->len < skb->data_len);
  1692. return skb->data += len;
  1693. }
  1694. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1695. {
  1696. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1697. }
  1698. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1699. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1700. {
  1701. if (len > skb_headlen(skb) &&
  1702. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1703. return NULL;
  1704. skb->len -= len;
  1705. return skb->data += len;
  1706. }
  1707. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1708. {
  1709. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1710. }
  1711. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1712. {
  1713. if (likely(len <= skb_headlen(skb)))
  1714. return 1;
  1715. if (unlikely(len > skb->len))
  1716. return 0;
  1717. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1718. }
  1719. /**
  1720. * skb_headroom - bytes at buffer head
  1721. * @skb: buffer to check
  1722. *
  1723. * Return the number of bytes of free space at the head of an &sk_buff.
  1724. */
  1725. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1726. {
  1727. return skb->data - skb->head;
  1728. }
  1729. /**
  1730. * skb_tailroom - bytes at buffer end
  1731. * @skb: buffer to check
  1732. *
  1733. * Return the number of bytes of free space at the tail of an sk_buff
  1734. */
  1735. static inline int skb_tailroom(const struct sk_buff *skb)
  1736. {
  1737. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1738. }
  1739. /**
  1740. * skb_availroom - bytes at buffer end
  1741. * @skb: buffer to check
  1742. *
  1743. * Return the number of bytes of free space at the tail of an sk_buff
  1744. * allocated by sk_stream_alloc()
  1745. */
  1746. static inline int skb_availroom(const struct sk_buff *skb)
  1747. {
  1748. if (skb_is_nonlinear(skb))
  1749. return 0;
  1750. return skb->end - skb->tail - skb->reserved_tailroom;
  1751. }
  1752. /**
  1753. * skb_reserve - adjust headroom
  1754. * @skb: buffer to alter
  1755. * @len: bytes to move
  1756. *
  1757. * Increase the headroom of an empty &sk_buff by reducing the tail
  1758. * room. This is only allowed for an empty buffer.
  1759. */
  1760. static inline void skb_reserve(struct sk_buff *skb, int len)
  1761. {
  1762. skb->data += len;
  1763. skb->tail += len;
  1764. }
  1765. /**
  1766. * skb_tailroom_reserve - adjust reserved_tailroom
  1767. * @skb: buffer to alter
  1768. * @mtu: maximum amount of headlen permitted
  1769. * @needed_tailroom: minimum amount of reserved_tailroom
  1770. *
  1771. * Set reserved_tailroom so that headlen can be as large as possible but
  1772. * not larger than mtu and tailroom cannot be smaller than
  1773. * needed_tailroom.
  1774. * The required headroom should already have been reserved before using
  1775. * this function.
  1776. */
  1777. static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
  1778. unsigned int needed_tailroom)
  1779. {
  1780. SKB_LINEAR_ASSERT(skb);
  1781. if (mtu < skb_tailroom(skb) - needed_tailroom)
  1782. /* use at most mtu */
  1783. skb->reserved_tailroom = skb_tailroom(skb) - mtu;
  1784. else
  1785. /* use up to all available space */
  1786. skb->reserved_tailroom = needed_tailroom;
  1787. }
  1788. #define ENCAP_TYPE_ETHER 0
  1789. #define ENCAP_TYPE_IPPROTO 1
  1790. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1791. __be16 protocol)
  1792. {
  1793. skb->inner_protocol = protocol;
  1794. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1795. }
  1796. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1797. __u8 ipproto)
  1798. {
  1799. skb->inner_ipproto = ipproto;
  1800. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1801. }
  1802. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1803. {
  1804. skb->inner_mac_header = skb->mac_header;
  1805. skb->inner_network_header = skb->network_header;
  1806. skb->inner_transport_header = skb->transport_header;
  1807. }
  1808. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1809. {
  1810. skb->mac_len = skb->network_header - skb->mac_header;
  1811. }
  1812. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1813. *skb)
  1814. {
  1815. return skb->head + skb->inner_transport_header;
  1816. }
  1817. static inline int skb_inner_transport_offset(const struct sk_buff *skb)
  1818. {
  1819. return skb_inner_transport_header(skb) - skb->data;
  1820. }
  1821. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1822. {
  1823. skb->inner_transport_header = skb->data - skb->head;
  1824. }
  1825. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1826. const int offset)
  1827. {
  1828. skb_reset_inner_transport_header(skb);
  1829. skb->inner_transport_header += offset;
  1830. }
  1831. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1832. {
  1833. return skb->head + skb->inner_network_header;
  1834. }
  1835. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1836. {
  1837. skb->inner_network_header = skb->data - skb->head;
  1838. }
  1839. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1840. const int offset)
  1841. {
  1842. skb_reset_inner_network_header(skb);
  1843. skb->inner_network_header += offset;
  1844. }
  1845. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1846. {
  1847. return skb->head + skb->inner_mac_header;
  1848. }
  1849. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1850. {
  1851. skb->inner_mac_header = skb->data - skb->head;
  1852. }
  1853. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1854. const int offset)
  1855. {
  1856. skb_reset_inner_mac_header(skb);
  1857. skb->inner_mac_header += offset;
  1858. }
  1859. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1860. {
  1861. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1862. }
  1863. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1864. {
  1865. return skb->head + skb->transport_header;
  1866. }
  1867. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1868. {
  1869. skb->transport_header = skb->data - skb->head;
  1870. }
  1871. static inline void skb_set_transport_header(struct sk_buff *skb,
  1872. const int offset)
  1873. {
  1874. skb_reset_transport_header(skb);
  1875. skb->transport_header += offset;
  1876. }
  1877. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1878. {
  1879. return skb->head + skb->network_header;
  1880. }
  1881. static inline void skb_reset_network_header(struct sk_buff *skb)
  1882. {
  1883. skb->network_header = skb->data - skb->head;
  1884. }
  1885. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1886. {
  1887. skb_reset_network_header(skb);
  1888. skb->network_header += offset;
  1889. }
  1890. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1891. {
  1892. return skb->head + skb->mac_header;
  1893. }
  1894. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1895. {
  1896. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1897. }
  1898. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1899. {
  1900. skb->mac_header = skb->data - skb->head;
  1901. }
  1902. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1903. {
  1904. skb_reset_mac_header(skb);
  1905. skb->mac_header += offset;
  1906. }
  1907. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1908. {
  1909. skb->mac_header = skb->network_header;
  1910. }
  1911. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1912. const int offset_hint)
  1913. {
  1914. struct flow_keys keys;
  1915. if (skb_transport_header_was_set(skb))
  1916. return;
  1917. else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
  1918. skb_set_transport_header(skb, keys.control.thoff);
  1919. else
  1920. skb_set_transport_header(skb, offset_hint);
  1921. }
  1922. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1923. {
  1924. if (skb_mac_header_was_set(skb)) {
  1925. const unsigned char *old_mac = skb_mac_header(skb);
  1926. skb_set_mac_header(skb, -skb->mac_len);
  1927. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1928. }
  1929. }
  1930. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1931. {
  1932. return skb->csum_start - skb_headroom(skb);
  1933. }
  1934. static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
  1935. {
  1936. return skb->head + skb->csum_start;
  1937. }
  1938. static inline int skb_transport_offset(const struct sk_buff *skb)
  1939. {
  1940. return skb_transport_header(skb) - skb->data;
  1941. }
  1942. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1943. {
  1944. return skb->transport_header - skb->network_header;
  1945. }
  1946. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1947. {
  1948. return skb->inner_transport_header - skb->inner_network_header;
  1949. }
  1950. static inline int skb_network_offset(const struct sk_buff *skb)
  1951. {
  1952. return skb_network_header(skb) - skb->data;
  1953. }
  1954. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1955. {
  1956. return skb_inner_network_header(skb) - skb->data;
  1957. }
  1958. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1959. {
  1960. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1961. }
  1962. /*
  1963. * CPUs often take a performance hit when accessing unaligned memory
  1964. * locations. The actual performance hit varies, it can be small if the
  1965. * hardware handles it or large if we have to take an exception and fix it
  1966. * in software.
  1967. *
  1968. * Since an ethernet header is 14 bytes network drivers often end up with
  1969. * the IP header at an unaligned offset. The IP header can be aligned by
  1970. * shifting the start of the packet by 2 bytes. Drivers should do this
  1971. * with:
  1972. *
  1973. * skb_reserve(skb, NET_IP_ALIGN);
  1974. *
  1975. * The downside to this alignment of the IP header is that the DMA is now
  1976. * unaligned. On some architectures the cost of an unaligned DMA is high
  1977. * and this cost outweighs the gains made by aligning the IP header.
  1978. *
  1979. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1980. * to be overridden.
  1981. */
  1982. #ifndef NET_IP_ALIGN
  1983. #define NET_IP_ALIGN 2
  1984. #endif
  1985. /*
  1986. * The networking layer reserves some headroom in skb data (via
  1987. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1988. * the header has to grow. In the default case, if the header has to grow
  1989. * 32 bytes or less we avoid the reallocation.
  1990. *
  1991. * Unfortunately this headroom changes the DMA alignment of the resulting
  1992. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1993. * on some architectures. An architecture can override this value,
  1994. * perhaps setting it to a cacheline in size (since that will maintain
  1995. * cacheline alignment of the DMA). It must be a power of 2.
  1996. *
  1997. * Various parts of the networking layer expect at least 32 bytes of
  1998. * headroom, you should not reduce this.
  1999. *
  2000. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  2001. * to reduce average number of cache lines per packet.
  2002. * get_rps_cpus() for example only access one 64 bytes aligned block :
  2003. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  2004. */
  2005. #ifndef NET_SKB_PAD
  2006. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  2007. #endif
  2008. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  2009. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  2010. {
  2011. if (unlikely(skb_is_nonlinear(skb))) {
  2012. WARN_ON(1);
  2013. return;
  2014. }
  2015. skb->len = len;
  2016. skb_set_tail_pointer(skb, len);
  2017. }
  2018. void skb_trim(struct sk_buff *skb, unsigned int len);
  2019. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  2020. {
  2021. if (skb->data_len)
  2022. return ___pskb_trim(skb, len);
  2023. __skb_trim(skb, len);
  2024. return 0;
  2025. }
  2026. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  2027. {
  2028. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  2029. }
  2030. /**
  2031. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  2032. * @skb: buffer to alter
  2033. * @len: new length
  2034. *
  2035. * This is identical to pskb_trim except that the caller knows that
  2036. * the skb is not cloned so we should never get an error due to out-
  2037. * of-memory.
  2038. */
  2039. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  2040. {
  2041. int err = pskb_trim(skb, len);
  2042. BUG_ON(err);
  2043. }
  2044. /**
  2045. * skb_orphan - orphan a buffer
  2046. * @skb: buffer to orphan
  2047. *
  2048. * If a buffer currently has an owner then we call the owner's
  2049. * destructor function and make the @skb unowned. The buffer continues
  2050. * to exist but is no longer charged to its former owner.
  2051. */
  2052. static inline void skb_orphan(struct sk_buff *skb)
  2053. {
  2054. if (skb->destructor) {
  2055. skb->destructor(skb);
  2056. skb->destructor = NULL;
  2057. skb->sk = NULL;
  2058. } else {
  2059. BUG_ON(skb->sk);
  2060. }
  2061. }
  2062. /**
  2063. * skb_orphan_frags - orphan the frags contained in a buffer
  2064. * @skb: buffer to orphan frags from
  2065. * @gfp_mask: allocation mask for replacement pages
  2066. *
  2067. * For each frag in the SKB which needs a destructor (i.e. has an
  2068. * owner) create a copy of that frag and release the original
  2069. * page by calling the destructor.
  2070. */
  2071. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  2072. {
  2073. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  2074. return 0;
  2075. return skb_copy_ubufs(skb, gfp_mask);
  2076. }
  2077. /**
  2078. * __skb_queue_purge - empty a list
  2079. * @list: list to empty
  2080. *
  2081. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  2082. * the list and one reference dropped. This function does not take the
  2083. * list lock and the caller must hold the relevant locks to use it.
  2084. */
  2085. void skb_queue_purge(struct sk_buff_head *list);
  2086. static inline void __skb_queue_purge(struct sk_buff_head *list)
  2087. {
  2088. struct sk_buff *skb;
  2089. while ((skb = __skb_dequeue(list)) != NULL)
  2090. kfree_skb(skb);
  2091. }
  2092. void *netdev_alloc_frag(unsigned int fragsz);
  2093. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  2094. gfp_t gfp_mask);
  2095. /**
  2096. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  2097. * @dev: network device to receive on
  2098. * @length: length to allocate
  2099. *
  2100. * Allocate a new &sk_buff and assign it a usage count of one. The
  2101. * buffer has unspecified headroom built in. Users should allocate
  2102. * the headroom they think they need without accounting for the
  2103. * built in space. The built in space is used for optimisations.
  2104. *
  2105. * %NULL is returned if there is no free memory. Although this function
  2106. * allocates memory it can be called from an interrupt.
  2107. */
  2108. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  2109. unsigned int length)
  2110. {
  2111. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  2112. }
  2113. /* legacy helper around __netdev_alloc_skb() */
  2114. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  2115. gfp_t gfp_mask)
  2116. {
  2117. return __netdev_alloc_skb(NULL, length, gfp_mask);
  2118. }
  2119. /* legacy helper around netdev_alloc_skb() */
  2120. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  2121. {
  2122. return netdev_alloc_skb(NULL, length);
  2123. }
  2124. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  2125. unsigned int length, gfp_t gfp)
  2126. {
  2127. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  2128. if (NET_IP_ALIGN && skb)
  2129. skb_reserve(skb, NET_IP_ALIGN);
  2130. return skb;
  2131. }
  2132. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  2133. unsigned int length)
  2134. {
  2135. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  2136. }
  2137. static inline void skb_free_frag(void *addr)
  2138. {
  2139. __free_page_frag(addr);
  2140. }
  2141. void *napi_alloc_frag(unsigned int fragsz);
  2142. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  2143. unsigned int length, gfp_t gfp_mask);
  2144. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  2145. unsigned int length)
  2146. {
  2147. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  2148. }
  2149. void napi_consume_skb(struct sk_buff *skb, int budget);
  2150. void __kfree_skb_flush(void);
  2151. void __kfree_skb_defer(struct sk_buff *skb);
  2152. /**
  2153. * __dev_alloc_pages - allocate page for network Rx
  2154. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2155. * @order: size of the allocation
  2156. *
  2157. * Allocate a new page.
  2158. *
  2159. * %NULL is returned if there is no free memory.
  2160. */
  2161. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  2162. unsigned int order)
  2163. {
  2164. /* This piece of code contains several assumptions.
  2165. * 1. This is for device Rx, therefor a cold page is preferred.
  2166. * 2. The expectation is the user wants a compound page.
  2167. * 3. If requesting a order 0 page it will not be compound
  2168. * due to the check to see if order has a value in prep_new_page
  2169. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  2170. * code in gfp_to_alloc_flags that should be enforcing this.
  2171. */
  2172. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  2173. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  2174. }
  2175. static inline struct page *dev_alloc_pages(unsigned int order)
  2176. {
  2177. return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
  2178. }
  2179. /**
  2180. * __dev_alloc_page - allocate a page for network Rx
  2181. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  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_page(gfp_t gfp_mask)
  2188. {
  2189. return __dev_alloc_pages(gfp_mask, 0);
  2190. }
  2191. static inline struct page *dev_alloc_page(void)
  2192. {
  2193. return dev_alloc_pages(0);
  2194. }
  2195. /**
  2196. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  2197. * @page: The page that was allocated from skb_alloc_page
  2198. * @skb: The skb that may need pfmemalloc set
  2199. */
  2200. static inline void skb_propagate_pfmemalloc(struct page *page,
  2201. struct sk_buff *skb)
  2202. {
  2203. if (page_is_pfmemalloc(page))
  2204. skb->pfmemalloc = true;
  2205. }
  2206. /**
  2207. * skb_frag_page - retrieve the page referred to by a paged fragment
  2208. * @frag: the paged fragment
  2209. *
  2210. * Returns the &struct page associated with @frag.
  2211. */
  2212. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2213. {
  2214. return frag->page.p;
  2215. }
  2216. /**
  2217. * __skb_frag_ref - take an addition reference on a paged fragment.
  2218. * @frag: the paged fragment
  2219. *
  2220. * Takes an additional reference on the paged fragment @frag.
  2221. */
  2222. static inline void __skb_frag_ref(skb_frag_t *frag)
  2223. {
  2224. get_page(skb_frag_page(frag));
  2225. }
  2226. /**
  2227. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2228. * @skb: the buffer
  2229. * @f: the fragment offset.
  2230. *
  2231. * Takes an additional reference on the @f'th paged fragment of @skb.
  2232. */
  2233. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2234. {
  2235. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2236. }
  2237. /**
  2238. * __skb_frag_unref - release a reference on a paged fragment.
  2239. * @frag: the paged fragment
  2240. *
  2241. * Releases a reference on the paged fragment @frag.
  2242. */
  2243. static inline void __skb_frag_unref(skb_frag_t *frag)
  2244. {
  2245. put_page(skb_frag_page(frag));
  2246. }
  2247. /**
  2248. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2249. * @skb: the buffer
  2250. * @f: the fragment offset
  2251. *
  2252. * Releases a reference on the @f'th paged fragment of @skb.
  2253. */
  2254. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2255. {
  2256. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2257. }
  2258. /**
  2259. * skb_frag_address - gets the address of the data contained in a paged fragment
  2260. * @frag: the paged fragment buffer
  2261. *
  2262. * Returns the address of the data within @frag. The page must already
  2263. * be mapped.
  2264. */
  2265. static inline void *skb_frag_address(const skb_frag_t *frag)
  2266. {
  2267. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2268. }
  2269. /**
  2270. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2271. * @frag: the paged fragment buffer
  2272. *
  2273. * Returns the address of the data within @frag. Checks that the page
  2274. * is mapped and returns %NULL otherwise.
  2275. */
  2276. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2277. {
  2278. void *ptr = page_address(skb_frag_page(frag));
  2279. if (unlikely(!ptr))
  2280. return NULL;
  2281. return ptr + frag->page_offset;
  2282. }
  2283. /**
  2284. * __skb_frag_set_page - sets the page contained in a paged fragment
  2285. * @frag: the paged fragment
  2286. * @page: the page to set
  2287. *
  2288. * Sets the fragment @frag to contain @page.
  2289. */
  2290. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2291. {
  2292. frag->page.p = page;
  2293. }
  2294. /**
  2295. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2296. * @skb: the buffer
  2297. * @f: the fragment offset
  2298. * @page: the page to set
  2299. *
  2300. * Sets the @f'th fragment of @skb to contain @page.
  2301. */
  2302. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2303. struct page *page)
  2304. {
  2305. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2306. }
  2307. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2308. /**
  2309. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2310. * @dev: the device to map the fragment to
  2311. * @frag: the paged fragment to map
  2312. * @offset: the offset within the fragment (starting at the
  2313. * fragment's own offset)
  2314. * @size: the number of bytes to map
  2315. * @dir: the direction of the mapping (%PCI_DMA_*)
  2316. *
  2317. * Maps the page associated with @frag to @device.
  2318. */
  2319. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2320. const skb_frag_t *frag,
  2321. size_t offset, size_t size,
  2322. enum dma_data_direction dir)
  2323. {
  2324. return dma_map_page(dev, skb_frag_page(frag),
  2325. frag->page_offset + offset, size, dir);
  2326. }
  2327. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2328. gfp_t gfp_mask)
  2329. {
  2330. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2331. }
  2332. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2333. gfp_t gfp_mask)
  2334. {
  2335. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2336. }
  2337. /**
  2338. * skb_clone_writable - is the header of a clone writable
  2339. * @skb: buffer to check
  2340. * @len: length up to which to write
  2341. *
  2342. * Returns true if modifying the header part of the cloned buffer
  2343. * does not requires the data to be copied.
  2344. */
  2345. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2346. {
  2347. return !skb_header_cloned(skb) &&
  2348. skb_headroom(skb) + len <= skb->hdr_len;
  2349. }
  2350. static inline int skb_try_make_writable(struct sk_buff *skb,
  2351. unsigned int write_len)
  2352. {
  2353. return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
  2354. pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
  2355. }
  2356. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2357. int cloned)
  2358. {
  2359. int delta = 0;
  2360. if (headroom > skb_headroom(skb))
  2361. delta = headroom - skb_headroom(skb);
  2362. if (delta || cloned)
  2363. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2364. GFP_ATOMIC);
  2365. return 0;
  2366. }
  2367. /**
  2368. * skb_cow - copy header of skb when it is required
  2369. * @skb: buffer to cow
  2370. * @headroom: needed headroom
  2371. *
  2372. * If the skb passed lacks sufficient headroom or its data part
  2373. * is shared, data is reallocated. If reallocation fails, an error
  2374. * is returned and original skb is not changed.
  2375. *
  2376. * The result is skb with writable area skb->head...skb->tail
  2377. * and at least @headroom of space at head.
  2378. */
  2379. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2380. {
  2381. return __skb_cow(skb, headroom, skb_cloned(skb));
  2382. }
  2383. /**
  2384. * skb_cow_head - skb_cow but only making the head writable
  2385. * @skb: buffer to cow
  2386. * @headroom: needed headroom
  2387. *
  2388. * This function is identical to skb_cow except that we replace the
  2389. * skb_cloned check by skb_header_cloned. It should be used when
  2390. * you only need to push on some header and do not need to modify
  2391. * the data.
  2392. */
  2393. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2394. {
  2395. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2396. }
  2397. /**
  2398. * skb_padto - pad an skbuff up to a minimal size
  2399. * @skb: buffer to pad
  2400. * @len: minimal length
  2401. *
  2402. * Pads up a buffer to ensure the trailing bytes exist and are
  2403. * blanked. If the buffer already contains sufficient data it
  2404. * is untouched. Otherwise it is extended. Returns zero on
  2405. * success. The skb is freed on error.
  2406. */
  2407. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2408. {
  2409. unsigned int size = skb->len;
  2410. if (likely(size >= len))
  2411. return 0;
  2412. return skb_pad(skb, len - size);
  2413. }
  2414. /**
  2415. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2416. * @skb: buffer to pad
  2417. * @len: minimal length
  2418. *
  2419. * Pads up a buffer to ensure the trailing bytes exist and are
  2420. * blanked. If the buffer already contains sufficient data it
  2421. * is untouched. Otherwise it is extended. Returns zero on
  2422. * success. The skb is freed on error.
  2423. */
  2424. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2425. {
  2426. unsigned int size = skb->len;
  2427. if (unlikely(size < len)) {
  2428. len -= size;
  2429. if (skb_pad(skb, len))
  2430. return -ENOMEM;
  2431. __skb_put(skb, len);
  2432. }
  2433. return 0;
  2434. }
  2435. static inline int skb_add_data(struct sk_buff *skb,
  2436. struct iov_iter *from, int copy)
  2437. {
  2438. const int off = skb->len;
  2439. if (skb->ip_summed == CHECKSUM_NONE) {
  2440. __wsum csum = 0;
  2441. if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
  2442. &csum, from) == copy) {
  2443. skb->csum = csum_block_add(skb->csum, csum, off);
  2444. return 0;
  2445. }
  2446. } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
  2447. return 0;
  2448. __skb_trim(skb, off);
  2449. return -EFAULT;
  2450. }
  2451. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2452. const struct page *page, int off)
  2453. {
  2454. if (i) {
  2455. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2456. return page == skb_frag_page(frag) &&
  2457. off == frag->page_offset + skb_frag_size(frag);
  2458. }
  2459. return false;
  2460. }
  2461. static inline int __skb_linearize(struct sk_buff *skb)
  2462. {
  2463. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2464. }
  2465. /**
  2466. * skb_linearize - convert paged skb to linear one
  2467. * @skb: buffer to linarize
  2468. *
  2469. * If there is no free memory -ENOMEM is returned, otherwise zero
  2470. * is returned and the old skb data released.
  2471. */
  2472. static inline int skb_linearize(struct sk_buff *skb)
  2473. {
  2474. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2475. }
  2476. /**
  2477. * skb_has_shared_frag - can any frag be overwritten
  2478. * @skb: buffer to test
  2479. *
  2480. * Return true if the skb has at least one frag that might be modified
  2481. * by an external entity (as in vmsplice()/sendfile())
  2482. */
  2483. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2484. {
  2485. return skb_is_nonlinear(skb) &&
  2486. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2487. }
  2488. /**
  2489. * skb_linearize_cow - make sure skb is linear and writable
  2490. * @skb: buffer to process
  2491. *
  2492. * If there is no free memory -ENOMEM is returned, otherwise zero
  2493. * is returned and the old skb data released.
  2494. */
  2495. static inline int skb_linearize_cow(struct sk_buff *skb)
  2496. {
  2497. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2498. __skb_linearize(skb) : 0;
  2499. }
  2500. /**
  2501. * skb_postpull_rcsum - update checksum for received skb after pull
  2502. * @skb: buffer to update
  2503. * @start: start of data before pull
  2504. * @len: length of data pulled
  2505. *
  2506. * After doing a pull on a received packet, you need to call this to
  2507. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2508. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2509. */
  2510. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2511. const void *start, unsigned int len)
  2512. {
  2513. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2514. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2515. else if (skb->ip_summed == CHECKSUM_PARTIAL &&
  2516. skb_checksum_start_offset(skb) < 0)
  2517. skb->ip_summed = CHECKSUM_NONE;
  2518. }
  2519. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2520. static inline void skb_postpush_rcsum(struct sk_buff *skb,
  2521. const void *start, unsigned int len)
  2522. {
  2523. /* For performing the reverse operation to skb_postpull_rcsum(),
  2524. * we can instead of ...
  2525. *
  2526. * skb->csum = csum_add(skb->csum, csum_partial(start, len, 0));
  2527. *
  2528. * ... just use this equivalent version here to save a few
  2529. * instructions. Feeding csum of 0 in csum_partial() and later
  2530. * on adding skb->csum is equivalent to feed skb->csum in the
  2531. * first place.
  2532. */
  2533. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2534. skb->csum = csum_partial(start, len, skb->csum);
  2535. }
  2536. /**
  2537. * pskb_trim_rcsum - trim received skb and update checksum
  2538. * @skb: buffer to trim
  2539. * @len: new length
  2540. *
  2541. * This is exactly the same as pskb_trim except that it ensures the
  2542. * checksum of received packets are still valid after the operation.
  2543. */
  2544. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2545. {
  2546. if (likely(len >= skb->len))
  2547. return 0;
  2548. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2549. skb->ip_summed = CHECKSUM_NONE;
  2550. return __pskb_trim(skb, len);
  2551. }
  2552. #define skb_queue_walk(queue, skb) \
  2553. for (skb = (queue)->next; \
  2554. skb != (struct sk_buff *)(queue); \
  2555. skb = skb->next)
  2556. #define skb_queue_walk_safe(queue, skb, tmp) \
  2557. for (skb = (queue)->next, tmp = skb->next; \
  2558. skb != (struct sk_buff *)(queue); \
  2559. skb = tmp, tmp = skb->next)
  2560. #define skb_queue_walk_from(queue, skb) \
  2561. for (; skb != (struct sk_buff *)(queue); \
  2562. skb = skb->next)
  2563. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2564. for (tmp = skb->next; \
  2565. skb != (struct sk_buff *)(queue); \
  2566. skb = tmp, tmp = skb->next)
  2567. #define skb_queue_reverse_walk(queue, skb) \
  2568. for (skb = (queue)->prev; \
  2569. skb != (struct sk_buff *)(queue); \
  2570. skb = skb->prev)
  2571. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2572. for (skb = (queue)->prev, tmp = skb->prev; \
  2573. skb != (struct sk_buff *)(queue); \
  2574. skb = tmp, tmp = skb->prev)
  2575. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2576. for (tmp = skb->prev; \
  2577. skb != (struct sk_buff *)(queue); \
  2578. skb = tmp, tmp = skb->prev)
  2579. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2580. {
  2581. return skb_shinfo(skb)->frag_list != NULL;
  2582. }
  2583. static inline void skb_frag_list_init(struct sk_buff *skb)
  2584. {
  2585. skb_shinfo(skb)->frag_list = NULL;
  2586. }
  2587. #define skb_walk_frags(skb, iter) \
  2588. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2589. int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
  2590. const struct sk_buff *skb);
  2591. struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
  2592. int *peeked, int *off, int *err,
  2593. struct sk_buff **last);
  2594. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2595. int *peeked, int *off, int *err);
  2596. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2597. int *err);
  2598. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2599. struct poll_table_struct *wait);
  2600. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2601. struct iov_iter *to, int size);
  2602. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2603. struct msghdr *msg, int size)
  2604. {
  2605. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2606. }
  2607. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2608. struct msghdr *msg);
  2609. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2610. struct iov_iter *from, int len);
  2611. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2612. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2613. void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
  2614. static inline void skb_free_datagram_locked(struct sock *sk,
  2615. struct sk_buff *skb)
  2616. {
  2617. __skb_free_datagram_locked(sk, skb, 0);
  2618. }
  2619. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2620. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2621. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2622. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2623. int len, __wsum csum);
  2624. ssize_t skb_socket_splice(struct sock *sk,
  2625. struct pipe_inode_info *pipe,
  2626. struct splice_pipe_desc *spd);
  2627. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2628. struct pipe_inode_info *pipe, unsigned int len,
  2629. unsigned int flags,
  2630. ssize_t (*splice_cb)(struct sock *,
  2631. struct pipe_inode_info *,
  2632. struct splice_pipe_desc *));
  2633. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2634. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2635. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2636. int len, int hlen);
  2637. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2638. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2639. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2640. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2641. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2642. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2643. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2644. int skb_vlan_pop(struct sk_buff *skb);
  2645. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2646. struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
  2647. gfp_t gfp);
  2648. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2649. {
  2650. return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2651. }
  2652. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2653. {
  2654. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2655. }
  2656. struct skb_checksum_ops {
  2657. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2658. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2659. };
  2660. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2661. __wsum csum, const struct skb_checksum_ops *ops);
  2662. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2663. __wsum csum);
  2664. static inline void * __must_check
  2665. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2666. int len, void *data, int hlen, void *buffer)
  2667. {
  2668. if (hlen - offset >= len)
  2669. return data + offset;
  2670. if (!skb ||
  2671. skb_copy_bits(skb, offset, buffer, len) < 0)
  2672. return NULL;
  2673. return buffer;
  2674. }
  2675. static inline void * __must_check
  2676. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2677. {
  2678. return __skb_header_pointer(skb, offset, len, skb->data,
  2679. skb_headlen(skb), buffer);
  2680. }
  2681. /**
  2682. * skb_needs_linearize - check if we need to linearize a given skb
  2683. * depending on the given device features.
  2684. * @skb: socket buffer to check
  2685. * @features: net device features
  2686. *
  2687. * Returns true if either:
  2688. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2689. * 2. skb is fragmented and the device does not support SG.
  2690. */
  2691. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2692. netdev_features_t features)
  2693. {
  2694. return skb_is_nonlinear(skb) &&
  2695. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2696. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2697. }
  2698. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2699. void *to,
  2700. const unsigned int len)
  2701. {
  2702. memcpy(to, skb->data, len);
  2703. }
  2704. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2705. const int offset, void *to,
  2706. const unsigned int len)
  2707. {
  2708. memcpy(to, skb->data + offset, len);
  2709. }
  2710. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2711. const void *from,
  2712. const unsigned int len)
  2713. {
  2714. memcpy(skb->data, from, len);
  2715. }
  2716. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2717. const int offset,
  2718. const void *from,
  2719. const unsigned int len)
  2720. {
  2721. memcpy(skb->data + offset, from, len);
  2722. }
  2723. void skb_init(void);
  2724. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2725. {
  2726. return skb->tstamp;
  2727. }
  2728. /**
  2729. * skb_get_timestamp - get timestamp from a skb
  2730. * @skb: skb to get stamp from
  2731. * @stamp: pointer to struct timeval to store stamp in
  2732. *
  2733. * Timestamps are stored in the skb as offsets to a base timestamp.
  2734. * This function converts the offset back to a struct timeval and stores
  2735. * it in stamp.
  2736. */
  2737. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2738. struct timeval *stamp)
  2739. {
  2740. *stamp = ktime_to_timeval(skb->tstamp);
  2741. }
  2742. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2743. struct timespec *stamp)
  2744. {
  2745. *stamp = ktime_to_timespec(skb->tstamp);
  2746. }
  2747. static inline void __net_timestamp(struct sk_buff *skb)
  2748. {
  2749. skb->tstamp = ktime_get_real();
  2750. }
  2751. static inline ktime_t net_timedelta(ktime_t t)
  2752. {
  2753. return ktime_sub(ktime_get_real(), t);
  2754. }
  2755. static inline ktime_t net_invalid_timestamp(void)
  2756. {
  2757. return ktime_set(0, 0);
  2758. }
  2759. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2760. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2761. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2762. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2763. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2764. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2765. {
  2766. }
  2767. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2768. {
  2769. return false;
  2770. }
  2771. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2772. /**
  2773. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2774. *
  2775. * PHY drivers may accept clones of transmitted packets for
  2776. * timestamping via their phy_driver.txtstamp method. These drivers
  2777. * must call this function to return the skb back to the stack with a
  2778. * timestamp.
  2779. *
  2780. * @skb: clone of the the original outgoing packet
  2781. * @hwtstamps: hardware time stamps
  2782. *
  2783. */
  2784. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2785. struct skb_shared_hwtstamps *hwtstamps);
  2786. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2787. struct skb_shared_hwtstamps *hwtstamps,
  2788. struct sock *sk, int tstype);
  2789. /**
  2790. * skb_tstamp_tx - queue clone of skb with send time stamps
  2791. * @orig_skb: the original outgoing packet
  2792. * @hwtstamps: hardware time stamps, may be NULL if not available
  2793. *
  2794. * If the skb has a socket associated, then this function clones the
  2795. * skb (thus sharing the actual data and optional structures), stores
  2796. * the optional hardware time stamping information (if non NULL) or
  2797. * generates a software time stamp (otherwise), then queues the clone
  2798. * to the error queue of the socket. Errors are silently ignored.
  2799. */
  2800. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2801. struct skb_shared_hwtstamps *hwtstamps);
  2802. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2803. {
  2804. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2805. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2806. skb_tstamp_tx(skb, NULL);
  2807. }
  2808. /**
  2809. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2810. *
  2811. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2812. * function immediately before giving the sk_buff to the MAC hardware.
  2813. *
  2814. * Specifically, one should make absolutely sure that this function is
  2815. * called before TX completion of this packet can trigger. Otherwise
  2816. * the packet could potentially already be freed.
  2817. *
  2818. * @skb: A socket buffer.
  2819. */
  2820. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2821. {
  2822. skb_clone_tx_timestamp(skb);
  2823. sw_tx_timestamp(skb);
  2824. }
  2825. /**
  2826. * skb_complete_wifi_ack - deliver skb with wifi status
  2827. *
  2828. * @skb: the original outgoing packet
  2829. * @acked: ack status
  2830. *
  2831. */
  2832. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2833. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2834. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2835. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2836. {
  2837. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2838. skb->csum_valid ||
  2839. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2840. skb_checksum_start_offset(skb) >= 0));
  2841. }
  2842. /**
  2843. * skb_checksum_complete - Calculate checksum of an entire packet
  2844. * @skb: packet to process
  2845. *
  2846. * This function calculates the checksum over the entire packet plus
  2847. * the value of skb->csum. The latter can be used to supply the
  2848. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2849. * checksum.
  2850. *
  2851. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2852. * this function can be used to verify that checksum on received
  2853. * packets. In that case the function should return zero if the
  2854. * checksum is correct. In particular, this function will return zero
  2855. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2856. * hardware has already verified the correctness of the checksum.
  2857. */
  2858. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2859. {
  2860. return skb_csum_unnecessary(skb) ?
  2861. 0 : __skb_checksum_complete(skb);
  2862. }
  2863. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2864. {
  2865. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2866. if (skb->csum_level == 0)
  2867. skb->ip_summed = CHECKSUM_NONE;
  2868. else
  2869. skb->csum_level--;
  2870. }
  2871. }
  2872. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2873. {
  2874. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2875. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2876. skb->csum_level++;
  2877. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2878. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2879. skb->csum_level = 0;
  2880. }
  2881. }
  2882. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2883. {
  2884. /* Mark current checksum as bad (typically called from GRO
  2885. * path). In the case that ip_summed is CHECKSUM_NONE
  2886. * this must be the first checksum encountered in the packet.
  2887. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2888. * checksum after the last one validated. For UDP, a zero
  2889. * checksum can not be marked as bad.
  2890. */
  2891. if (skb->ip_summed == CHECKSUM_NONE ||
  2892. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2893. skb->csum_bad = 1;
  2894. }
  2895. /* Check if we need to perform checksum complete validation.
  2896. *
  2897. * Returns true if checksum complete is needed, false otherwise
  2898. * (either checksum is unnecessary or zero checksum is allowed).
  2899. */
  2900. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2901. bool zero_okay,
  2902. __sum16 check)
  2903. {
  2904. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2905. skb->csum_valid = 1;
  2906. __skb_decr_checksum_unnecessary(skb);
  2907. return false;
  2908. }
  2909. return true;
  2910. }
  2911. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2912. * in checksum_init.
  2913. */
  2914. #define CHECKSUM_BREAK 76
  2915. /* Unset checksum-complete
  2916. *
  2917. * Unset checksum complete can be done when packet is being modified
  2918. * (uncompressed for instance) and checksum-complete value is
  2919. * invalidated.
  2920. */
  2921. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2922. {
  2923. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2924. skb->ip_summed = CHECKSUM_NONE;
  2925. }
  2926. /* Validate (init) checksum based on checksum complete.
  2927. *
  2928. * Return values:
  2929. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2930. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2931. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2932. * non-zero: value of invalid checksum
  2933. *
  2934. */
  2935. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2936. bool complete,
  2937. __wsum psum)
  2938. {
  2939. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2940. if (!csum_fold(csum_add(psum, skb->csum))) {
  2941. skb->csum_valid = 1;
  2942. return 0;
  2943. }
  2944. } else if (skb->csum_bad) {
  2945. /* ip_summed == CHECKSUM_NONE in this case */
  2946. return (__force __sum16)1;
  2947. }
  2948. skb->csum = psum;
  2949. if (complete || skb->len <= CHECKSUM_BREAK) {
  2950. __sum16 csum;
  2951. csum = __skb_checksum_complete(skb);
  2952. skb->csum_valid = !csum;
  2953. return csum;
  2954. }
  2955. return 0;
  2956. }
  2957. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2958. {
  2959. return 0;
  2960. }
  2961. /* Perform checksum validate (init). Note that this is a macro since we only
  2962. * want to calculate the pseudo header which is an input function if necessary.
  2963. * First we try to validate without any computation (checksum unnecessary) and
  2964. * then calculate based on checksum complete calling the function to compute
  2965. * pseudo header.
  2966. *
  2967. * Return values:
  2968. * 0: checksum is validated or try to in skb_checksum_complete
  2969. * non-zero: value of invalid checksum
  2970. */
  2971. #define __skb_checksum_validate(skb, proto, complete, \
  2972. zero_okay, check, compute_pseudo) \
  2973. ({ \
  2974. __sum16 __ret = 0; \
  2975. skb->csum_valid = 0; \
  2976. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2977. __ret = __skb_checksum_validate_complete(skb, \
  2978. complete, compute_pseudo(skb, proto)); \
  2979. __ret; \
  2980. })
  2981. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2982. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2983. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2984. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2985. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2986. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2987. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2988. compute_pseudo) \
  2989. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  2990. #define skb_checksum_simple_validate(skb) \
  2991. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2992. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2993. {
  2994. return (skb->ip_summed == CHECKSUM_NONE &&
  2995. skb->csum_valid && !skb->csum_bad);
  2996. }
  2997. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2998. __sum16 check, __wsum pseudo)
  2999. {
  3000. skb->csum = ~pseudo;
  3001. skb->ip_summed = CHECKSUM_COMPLETE;
  3002. }
  3003. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  3004. do { \
  3005. if (__skb_checksum_convert_check(skb)) \
  3006. __skb_checksum_convert(skb, check, \
  3007. compute_pseudo(skb, proto)); \
  3008. } while (0)
  3009. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  3010. u16 start, u16 offset)
  3011. {
  3012. skb->ip_summed = CHECKSUM_PARTIAL;
  3013. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  3014. skb->csum_offset = offset - start;
  3015. }
  3016. /* Update skbuf and packet to reflect the remote checksum offload operation.
  3017. * When called, ptr indicates the starting point for skb->csum when
  3018. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  3019. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  3020. */
  3021. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  3022. int start, int offset, bool nopartial)
  3023. {
  3024. __wsum delta;
  3025. if (!nopartial) {
  3026. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  3027. return;
  3028. }
  3029. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  3030. __skb_checksum_complete(skb);
  3031. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  3032. }
  3033. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  3034. /* Adjust skb->csum since we changed the packet */
  3035. skb->csum = csum_add(skb->csum, delta);
  3036. }
  3037. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3038. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  3039. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  3040. {
  3041. if (nfct && atomic_dec_and_test(&nfct->use))
  3042. nf_conntrack_destroy(nfct);
  3043. }
  3044. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  3045. {
  3046. if (nfct)
  3047. atomic_inc(&nfct->use);
  3048. }
  3049. #endif
  3050. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3051. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  3052. {
  3053. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  3054. kfree(nf_bridge);
  3055. }
  3056. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  3057. {
  3058. if (nf_bridge)
  3059. atomic_inc(&nf_bridge->use);
  3060. }
  3061. #endif /* CONFIG_BRIDGE_NETFILTER */
  3062. static inline void nf_reset(struct sk_buff *skb)
  3063. {
  3064. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3065. nf_conntrack_put(skb->nfct);
  3066. skb->nfct = NULL;
  3067. #endif
  3068. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3069. nf_bridge_put(skb->nf_bridge);
  3070. skb->nf_bridge = NULL;
  3071. #endif
  3072. }
  3073. static inline void nf_reset_trace(struct sk_buff *skb)
  3074. {
  3075. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3076. skb->nf_trace = 0;
  3077. #endif
  3078. }
  3079. /* Note: This doesn't put any conntrack and bridge info in dst. */
  3080. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  3081. bool copy)
  3082. {
  3083. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3084. dst->nfct = src->nfct;
  3085. nf_conntrack_get(src->nfct);
  3086. if (copy)
  3087. dst->nfctinfo = src->nfctinfo;
  3088. #endif
  3089. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3090. dst->nf_bridge = src->nf_bridge;
  3091. nf_bridge_get(src->nf_bridge);
  3092. #endif
  3093. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3094. if (copy)
  3095. dst->nf_trace = src->nf_trace;
  3096. #endif
  3097. }
  3098. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  3099. {
  3100. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3101. nf_conntrack_put(dst->nfct);
  3102. #endif
  3103. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3104. nf_bridge_put(dst->nf_bridge);
  3105. #endif
  3106. __nf_copy(dst, src, true);
  3107. }
  3108. #ifdef CONFIG_NETWORK_SECMARK
  3109. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3110. {
  3111. to->secmark = from->secmark;
  3112. }
  3113. static inline void skb_init_secmark(struct sk_buff *skb)
  3114. {
  3115. skb->secmark = 0;
  3116. }
  3117. #else
  3118. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3119. { }
  3120. static inline void skb_init_secmark(struct sk_buff *skb)
  3121. { }
  3122. #endif
  3123. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  3124. {
  3125. return !skb->destructor &&
  3126. #if IS_ENABLED(CONFIG_XFRM)
  3127. !skb->sp &&
  3128. #endif
  3129. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  3130. !skb->nfct &&
  3131. #endif
  3132. !skb->_skb_refdst &&
  3133. !skb_has_frag_list(skb);
  3134. }
  3135. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  3136. {
  3137. skb->queue_mapping = queue_mapping;
  3138. }
  3139. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  3140. {
  3141. return skb->queue_mapping;
  3142. }
  3143. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  3144. {
  3145. to->queue_mapping = from->queue_mapping;
  3146. }
  3147. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  3148. {
  3149. skb->queue_mapping = rx_queue + 1;
  3150. }
  3151. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  3152. {
  3153. return skb->queue_mapping - 1;
  3154. }
  3155. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  3156. {
  3157. return skb->queue_mapping != 0;
  3158. }
  3159. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  3160. {
  3161. #ifdef CONFIG_XFRM
  3162. return skb->sp;
  3163. #else
  3164. return NULL;
  3165. #endif
  3166. }
  3167. /* Keeps track of mac header offset relative to skb->head.
  3168. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  3169. * For non-tunnel skb it points to skb_mac_header() and for
  3170. * tunnel skb it points to outer mac header.
  3171. * Keeps track of level of encapsulation of network headers.
  3172. */
  3173. struct skb_gso_cb {
  3174. union {
  3175. int mac_offset;
  3176. int data_offset;
  3177. };
  3178. int encap_level;
  3179. __wsum csum;
  3180. __u16 csum_start;
  3181. };
  3182. #define SKB_SGO_CB_OFFSET 32
  3183. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
  3184. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  3185. {
  3186. return (skb_mac_header(inner_skb) - inner_skb->head) -
  3187. SKB_GSO_CB(inner_skb)->mac_offset;
  3188. }
  3189. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  3190. {
  3191. int new_headroom, headroom;
  3192. int ret;
  3193. headroom = skb_headroom(skb);
  3194. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  3195. if (ret)
  3196. return ret;
  3197. new_headroom = skb_headroom(skb);
  3198. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  3199. return 0;
  3200. }
  3201. static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
  3202. {
  3203. /* Do not update partial checksums if remote checksum is enabled. */
  3204. if (skb->remcsum_offload)
  3205. return;
  3206. SKB_GSO_CB(skb)->csum = res;
  3207. SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
  3208. }
  3209. /* Compute the checksum for a gso segment. First compute the checksum value
  3210. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  3211. * then add in skb->csum (checksum from csum_start to end of packet).
  3212. * skb->csum and csum_start are then updated to reflect the checksum of the
  3213. * resultant packet starting from the transport header-- the resultant checksum
  3214. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  3215. * header.
  3216. */
  3217. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  3218. {
  3219. unsigned char *csum_start = skb_transport_header(skb);
  3220. int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
  3221. __wsum partial = SKB_GSO_CB(skb)->csum;
  3222. SKB_GSO_CB(skb)->csum = res;
  3223. SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
  3224. return csum_fold(csum_partial(csum_start, plen, partial));
  3225. }
  3226. static inline bool skb_is_gso(const struct sk_buff *skb)
  3227. {
  3228. return skb_shinfo(skb)->gso_size;
  3229. }
  3230. /* Note: Should be called only if skb_is_gso(skb) is true */
  3231. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  3232. {
  3233. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  3234. }
  3235. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  3236. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  3237. {
  3238. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  3239. * wanted then gso_type will be set. */
  3240. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  3241. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  3242. unlikely(shinfo->gso_type == 0)) {
  3243. __skb_warn_lro_forwarding(skb);
  3244. return true;
  3245. }
  3246. return false;
  3247. }
  3248. static inline void skb_forward_csum(struct sk_buff *skb)
  3249. {
  3250. /* Unfortunately we don't support this one. Any brave souls? */
  3251. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3252. skb->ip_summed = CHECKSUM_NONE;
  3253. }
  3254. /**
  3255. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3256. * @skb: skb to check
  3257. *
  3258. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3259. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3260. * use this helper, to document places where we make this assertion.
  3261. */
  3262. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3263. {
  3264. #ifdef DEBUG
  3265. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3266. #endif
  3267. }
  3268. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3269. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3270. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3271. unsigned int transport_len,
  3272. __sum16(*skb_chkf)(struct sk_buff *skb));
  3273. /**
  3274. * skb_head_is_locked - Determine if the skb->head is locked down
  3275. * @skb: skb to check
  3276. *
  3277. * The head on skbs build around a head frag can be removed if they are
  3278. * not cloned. This function returns true if the skb head is locked down
  3279. * due to either being allocated via kmalloc, or by being a clone with
  3280. * multiple references to the head.
  3281. */
  3282. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3283. {
  3284. return !skb->head_frag || skb_cloned(skb);
  3285. }
  3286. /**
  3287. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3288. *
  3289. * @skb: GSO skb
  3290. *
  3291. * skb_gso_network_seglen is used to determine the real size of the
  3292. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3293. *
  3294. * The MAC/L2 header is not accounted for.
  3295. */
  3296. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3297. {
  3298. unsigned int hdr_len = skb_transport_header(skb) -
  3299. skb_network_header(skb);
  3300. return hdr_len + skb_gso_transport_seglen(skb);
  3301. }
  3302. /* Local Checksum Offload.
  3303. * Compute outer checksum based on the assumption that the
  3304. * inner checksum will be offloaded later.
  3305. * See Documentation/networking/checksum-offloads.txt for
  3306. * explanation of how this works.
  3307. * Fill in outer checksum adjustment (e.g. with sum of outer
  3308. * pseudo-header) before calling.
  3309. * Also ensure that inner checksum is in linear data area.
  3310. */
  3311. static inline __wsum lco_csum(struct sk_buff *skb)
  3312. {
  3313. unsigned char *csum_start = skb_checksum_start(skb);
  3314. unsigned char *l4_hdr = skb_transport_header(skb);
  3315. __wsum partial;
  3316. /* Start with complement of inner checksum adjustment */
  3317. partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
  3318. skb->csum_offset));
  3319. /* Add in checksum of our headers (incl. outer checksum
  3320. * adjustment filled in by caller) and return result.
  3321. */
  3322. return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
  3323. }
  3324. #endif /* __KERNEL__ */
  3325. #endif /* _LINUX_SKBUFF_H */