skbuff.h 111 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472347334743475347634773478347934803481348234833484348534863487348834893490349134923493349434953496349734983499350035013502350335043505350635073508350935103511351235133514351535163517351835193520352135223523352435253526352735283529353035313532353335343535353635373538353935403541354235433544354535463547354835493550355135523553355435553556355735583559356035613562356335643565356635673568356935703571357235733574357535763577357835793580358135823583358435853586358735883589359035913592359335943595359635973598359936003601360236033604360536063607360836093610361136123613361436153616361736183619362036213622362336243625362636273628362936303631363236333634363536363637363836393640364136423643364436453646364736483649365036513652365336543655365636573658365936603661366236633664366536663667366836693670367136723673367436753676367736783679368036813682368336843685368636873688368936903691369236933694369536963697369836993700370137023703370437053706370737083709371037113712371337143715371637173718371937203721372237233724372537263727372837293730373137323733373437353736373737383739374037413742374337443745374637473748374937503751375237533754375537563757375837593760376137623763376437653766376737683769377037713772377337743775377637773778377937803781378237833784378537863787378837893790379137923793379437953796379737983799380038013802380338043805380638073808380938103811381238133814381538163817381838193820382138223823382438253826382738283829383038313832383338343835383638373838383938403841384238433844384538463847384838493850385138523853
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/kernel.h>
  16. #include <linux/kmemcheck.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/bug.h>
  20. #include <linux/cache.h>
  21. #include <linux/rbtree.h>
  22. #include <linux/socket.h>
  23. #include <linux/atomic.h>
  24. #include <asm/types.h>
  25. #include <linux/spinlock.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #include <linux/rcupdate.h>
  30. #include <linux/hrtimer.h>
  31. #include <linux/dma-mapping.h>
  32. #include <linux/netdev_features.h>
  33. #include <linux/sched.h>
  34. #include <linux/sched/clock.h>
  35. #include <net/flow_dissector.h>
  36. #include <linux/splice.h>
  37. #include <linux/in6.h>
  38. #include <linux/if_packet.h>
  39. #include <net/flow.h>
  40. /* The interface for checksum offload between the stack and networking drivers
  41. * is as follows...
  42. *
  43. * A. IP checksum related features
  44. *
  45. * Drivers advertise checksum offload capabilities in the features of a device.
  46. * From the stack's point of view these are capabilities offered by the driver,
  47. * a driver typically only advertises features that it is capable of offloading
  48. * to its device.
  49. *
  50. * The checksum related features are:
  51. *
  52. * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
  53. * IP (one's complement) checksum for any combination
  54. * of protocols or protocol layering. The checksum is
  55. * computed and set in a packet per the CHECKSUM_PARTIAL
  56. * interface (see below).
  57. *
  58. * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
  59. * TCP or UDP packets over IPv4. These are specifically
  60. * unencapsulated packets of the form IPv4|TCP or
  61. * IPv4|UDP where the Protocol field in the IPv4 header
  62. * is TCP or UDP. The IPv4 header may contain IP options
  63. * This feature cannot be set in features for a device
  64. * with NETIF_F_HW_CSUM also set. This feature is being
  65. * DEPRECATED (see below).
  66. *
  67. * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
  68. * TCP or UDP packets over IPv6. These are specifically
  69. * unencapsulated packets of the form IPv6|TCP or
  70. * IPv4|UDP where the Next Header field in the IPv6
  71. * header is either TCP or UDP. IPv6 extension headers
  72. * are not supported with this feature. This feature
  73. * cannot be set in features for a device with
  74. * NETIF_F_HW_CSUM also set. This feature is being
  75. * DEPRECATED (see below).
  76. *
  77. * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
  78. * This flag is used only used to disable the RX checksum
  79. * feature for a device. The stack will accept receive
  80. * checksum indication in packets received on a device
  81. * regardless of whether NETIF_F_RXCSUM is set.
  82. *
  83. * B. Checksumming of received packets by device. Indication of checksum
  84. * verification is in set skb->ip_summed. Possible values are:
  85. *
  86. * CHECKSUM_NONE:
  87. *
  88. * Device did not checksum this packet e.g. due to lack of capabilities.
  89. * The packet contains full (though not verified) checksum in packet but
  90. * not in skb->csum. Thus, skb->csum is undefined in this case.
  91. *
  92. * CHECKSUM_UNNECESSARY:
  93. *
  94. * The hardware you're dealing with doesn't calculate the full checksum
  95. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  96. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  97. * if their checksums are okay. skb->csum is still undefined in this case
  98. * though. A driver or device must never modify the checksum field in the
  99. * packet even if checksum is verified.
  100. *
  101. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  102. * TCP: IPv6 and IPv4.
  103. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  104. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  105. * may perform further validation in this case.
  106. * GRE: only if the checksum is present in the header.
  107. * SCTP: indicates the CRC in SCTP header has been validated.
  108. *
  109. * skb->csum_level indicates the number of consecutive checksums found in
  110. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  111. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  112. * and a device is able to verify the checksums for UDP (possibly zero),
  113. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  114. * two. If the device were only able to verify the UDP checksum and not
  115. * GRE, either because it doesn't support GRE checksum of because GRE
  116. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  117. * not considered in this case).
  118. *
  119. * CHECKSUM_COMPLETE:
  120. *
  121. * This is the most generic way. The device supplied checksum of the _whole_
  122. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  123. * hardware doesn't need to parse L3/L4 headers to implement this.
  124. *
  125. * Note: Even if device supports only some protocols, but is able to produce
  126. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  127. *
  128. * CHECKSUM_PARTIAL:
  129. *
  130. * A checksum is set up to be offloaded to a device as described in the
  131. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  132. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  133. * on the same host, or it may be set in the input path in GRO or remote
  134. * checksum offload. For the purposes of checksum verification, the checksum
  135. * referred to by skb->csum_start + skb->csum_offset and any preceding
  136. * checksums in the packet are considered verified. Any checksums in the
  137. * packet that are after the checksum being offloaded are not considered to
  138. * be verified.
  139. *
  140. * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
  141. * in the skb->ip_summed for a packet. Values are:
  142. *
  143. * CHECKSUM_PARTIAL:
  144. *
  145. * The driver is required to checksum the packet as seen by hard_start_xmit()
  146. * from skb->csum_start up to the end, and to record/write the checksum at
  147. * offset skb->csum_start + skb->csum_offset. A driver may verify that the
  148. * csum_start and csum_offset values are valid values given the length and
  149. * offset of the packet, however they should not attempt to validate that the
  150. * checksum refers to a legitimate transport layer checksum-- it is the
  151. * purview of the stack to validate that csum_start and csum_offset are set
  152. * correctly.
  153. *
  154. * When the stack requests checksum offload for a packet, the driver MUST
  155. * ensure that the checksum is set correctly. A driver can either offload the
  156. * checksum calculation to the device, or call skb_checksum_help (in the case
  157. * that the device does not support offload for a particular checksum).
  158. *
  159. * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
  160. * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
  161. * checksum offload capability. If a device has limited checksum capabilities
  162. * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
  163. * described above) a helper function can be called to resolve
  164. * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
  165. * function takes a spec argument that describes the protocol layer that is
  166. * supported for checksum offload and can be called for each packet. If a
  167. * packet does not match the specification for offload, skb_checksum_help
  168. * is called to resolve the checksum.
  169. *
  170. * CHECKSUM_NONE:
  171. *
  172. * The skb was already checksummed by the protocol, or a checksum is not
  173. * required.
  174. *
  175. * CHECKSUM_UNNECESSARY:
  176. *
  177. * This has the same meaning on as CHECKSUM_NONE for checksum offload on
  178. * output.
  179. *
  180. * CHECKSUM_COMPLETE:
  181. * Not used in checksum output. If a driver observes a packet with this value
  182. * set in skbuff, if should treat as CHECKSUM_NONE being set.
  183. *
  184. * D. Non-IP checksum (CRC) offloads
  185. *
  186. * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
  187. * offloading the SCTP CRC in a packet. To perform this offload the stack
  188. * will set set csum_start and csum_offset accordingly, set ip_summed to
  189. * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
  190. * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
  191. * A driver that supports both IP checksum offload and SCTP CRC32c offload
  192. * must verify which offload is configured for a packet by testing the
  193. * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
  194. * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
  195. *
  196. * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
  197. * offloading the FCOE CRC in a packet. To perform this offload the stack
  198. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  199. * accordingly. Note the there is no indication in the skbuff that the
  200. * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
  201. * both IP checksum offload and FCOE CRC offload must verify which offload
  202. * is configured for a packet presumably by inspecting packet headers.
  203. *
  204. * E. Checksumming on output with GSO.
  205. *
  206. * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
  207. * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
  208. * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
  209. * part of the GSO operation is implied. If a checksum is being offloaded
  210. * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
  211. * are set to refer to the outermost checksum being offload (two offloaded
  212. * checksums are possible with UDP encapsulation).
  213. */
  214. /* Don't change this without changing skb_csum_unnecessary! */
  215. #define CHECKSUM_NONE 0
  216. #define CHECKSUM_UNNECESSARY 1
  217. #define CHECKSUM_COMPLETE 2
  218. #define CHECKSUM_PARTIAL 3
  219. /* Maximum value in skb->csum_level */
  220. #define SKB_MAX_CSUM_LEVEL 3
  221. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  222. #define SKB_WITH_OVERHEAD(X) \
  223. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  224. #define SKB_MAX_ORDER(X, ORDER) \
  225. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  226. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  227. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  228. /* return minimum truesize of one skb containing X bytes of data */
  229. #define SKB_TRUESIZE(X) ((X) + \
  230. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  231. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  232. struct net_device;
  233. struct scatterlist;
  234. struct pipe_inode_info;
  235. struct iov_iter;
  236. struct napi_struct;
  237. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  238. struct nf_conntrack {
  239. atomic_t use;
  240. };
  241. #endif
  242. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  243. struct nf_bridge_info {
  244. atomic_t use;
  245. enum {
  246. BRNF_PROTO_UNCHANGED,
  247. BRNF_PROTO_8021Q,
  248. BRNF_PROTO_PPPOE
  249. } orig_proto:8;
  250. u8 pkt_otherhost:1;
  251. u8 in_prerouting:1;
  252. u8 bridged_dnat:1;
  253. __u16 frag_max_size;
  254. struct net_device *physindev;
  255. /* always valid & non-NULL from FORWARD on, for physdev match */
  256. struct net_device *physoutdev;
  257. union {
  258. /* prerouting: detect dnat in orig/reply direction */
  259. __be32 ipv4_daddr;
  260. struct in6_addr ipv6_daddr;
  261. /* after prerouting + nat detected: store original source
  262. * mac since neigh resolution overwrites it, only used while
  263. * skb is out in neigh layer.
  264. */
  265. char neigh_header[8];
  266. };
  267. };
  268. #endif
  269. struct sk_buff_head {
  270. /* These two members must be first. */
  271. struct sk_buff *next;
  272. struct sk_buff *prev;
  273. __u32 qlen;
  274. spinlock_t lock;
  275. };
  276. struct sk_buff;
  277. /* To allow 64K frame to be packed as single skb without frag_list we
  278. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  279. * buffers which do not start on a page boundary.
  280. *
  281. * Since GRO uses frags we allocate at least 16 regardless of page
  282. * size.
  283. */
  284. #if (65536/PAGE_SIZE + 1) < 16
  285. #define MAX_SKB_FRAGS 16UL
  286. #else
  287. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  288. #endif
  289. extern int sysctl_max_skb_frags;
  290. /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
  291. * segment using its current segmentation instead.
  292. */
  293. #define GSO_BY_FRAGS 0xFFFF
  294. typedef struct skb_frag_struct skb_frag_t;
  295. struct skb_frag_struct {
  296. struct {
  297. struct page *p;
  298. } page;
  299. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  300. __u32 page_offset;
  301. __u32 size;
  302. #else
  303. __u16 page_offset;
  304. __u16 size;
  305. #endif
  306. };
  307. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  308. {
  309. return frag->size;
  310. }
  311. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  312. {
  313. frag->size = size;
  314. }
  315. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  316. {
  317. frag->size += delta;
  318. }
  319. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  320. {
  321. frag->size -= delta;
  322. }
  323. #define HAVE_HW_TIME_STAMP
  324. /**
  325. * struct skb_shared_hwtstamps - hardware time stamps
  326. * @hwtstamp: hardware time stamp transformed into duration
  327. * since arbitrary point in time
  328. *
  329. * Software time stamps generated by ktime_get_real() are stored in
  330. * skb->tstamp.
  331. *
  332. * hwtstamps can only be compared against other hwtstamps from
  333. * the same device.
  334. *
  335. * This structure is attached to packets as part of the
  336. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  337. */
  338. struct skb_shared_hwtstamps {
  339. ktime_t hwtstamp;
  340. };
  341. /* Definitions for tx_flags in struct skb_shared_info */
  342. enum {
  343. /* generate hardware time stamp */
  344. SKBTX_HW_TSTAMP = 1 << 0,
  345. /* generate software time stamp when queueing packet to NIC */
  346. SKBTX_SW_TSTAMP = 1 << 1,
  347. /* device driver is going to provide hardware time stamp */
  348. SKBTX_IN_PROGRESS = 1 << 2,
  349. /* device driver supports TX zero-copy buffers */
  350. SKBTX_DEV_ZEROCOPY = 1 << 3,
  351. /* generate wifi status information (where possible) */
  352. SKBTX_WIFI_STATUS = 1 << 4,
  353. /* This indicates at least one fragment might be overwritten
  354. * (as in vmsplice(), sendfile() ...)
  355. * If we need to compute a TX checksum, we'll need to copy
  356. * all frags to avoid possible bad checksum
  357. */
  358. SKBTX_SHARED_FRAG = 1 << 5,
  359. /* generate software time stamp when entering packet scheduling */
  360. SKBTX_SCHED_TSTAMP = 1 << 6,
  361. };
  362. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  363. SKBTX_SCHED_TSTAMP)
  364. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  365. /*
  366. * The callback notifies userspace to release buffers when skb DMA is done in
  367. * lower device, the skb last reference should be 0 when calling this.
  368. * The zerocopy_success argument is true if zero copy transmit occurred,
  369. * false on data copy or out of memory error caused by data copy attempt.
  370. * The ctx field is used to track device context.
  371. * The desc field is used to track userspace buffer index.
  372. */
  373. struct ubuf_info {
  374. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  375. void *ctx;
  376. unsigned long desc;
  377. };
  378. /* This data is invariant across clones and lives at
  379. * the end of the header data, ie. at skb->end.
  380. */
  381. struct skb_shared_info {
  382. unsigned short _unused;
  383. unsigned char nr_frags;
  384. __u8 tx_flags;
  385. unsigned short gso_size;
  386. /* Warning: this field is not always filled in (UFO)! */
  387. unsigned short gso_segs;
  388. struct sk_buff *frag_list;
  389. struct skb_shared_hwtstamps hwtstamps;
  390. unsigned int gso_type;
  391. u32 tskey;
  392. __be32 ip6_frag_id;
  393. /*
  394. * Warning : all fields before dataref are cleared in __alloc_skb()
  395. */
  396. atomic_t dataref;
  397. /* Intermediate layers must ensure that destructor_arg
  398. * remains valid until skb destructor */
  399. void * destructor_arg;
  400. /* must be last field, see pskb_expand_head() */
  401. skb_frag_t frags[MAX_SKB_FRAGS];
  402. };
  403. /* We divide dataref into two halves. The higher 16 bits hold references
  404. * to the payload part of skb->data. The lower 16 bits hold references to
  405. * the entire skb->data. A clone of a headerless skb holds the length of
  406. * the header in skb->hdr_len.
  407. *
  408. * All users must obey the rule that the skb->data reference count must be
  409. * greater than or equal to the payload reference count.
  410. *
  411. * Holding a reference to the payload part means that the user does not
  412. * care about modifications to the header part of skb->data.
  413. */
  414. #define SKB_DATAREF_SHIFT 16
  415. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  416. enum {
  417. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  418. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  419. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  420. };
  421. enum {
  422. SKB_GSO_TCPV4 = 1 << 0,
  423. SKB_GSO_UDP = 1 << 1,
  424. /* This indicates the skb is from an untrusted source. */
  425. SKB_GSO_DODGY = 1 << 2,
  426. /* This indicates the tcp segment has CWR set. */
  427. SKB_GSO_TCP_ECN = 1 << 3,
  428. SKB_GSO_TCP_FIXEDID = 1 << 4,
  429. SKB_GSO_TCPV6 = 1 << 5,
  430. SKB_GSO_FCOE = 1 << 6,
  431. SKB_GSO_GRE = 1 << 7,
  432. SKB_GSO_GRE_CSUM = 1 << 8,
  433. SKB_GSO_IPXIP4 = 1 << 9,
  434. SKB_GSO_IPXIP6 = 1 << 10,
  435. SKB_GSO_UDP_TUNNEL = 1 << 11,
  436. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
  437. SKB_GSO_PARTIAL = 1 << 13,
  438. SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
  439. SKB_GSO_SCTP = 1 << 15,
  440. SKB_GSO_ESP = 1 << 16,
  441. };
  442. #if BITS_PER_LONG > 32
  443. #define NET_SKBUFF_DATA_USES_OFFSET 1
  444. #endif
  445. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  446. typedef unsigned int sk_buff_data_t;
  447. #else
  448. typedef unsigned char *sk_buff_data_t;
  449. #endif
  450. /**
  451. * struct sk_buff - socket buffer
  452. * @next: Next buffer in list
  453. * @prev: Previous buffer in list
  454. * @tstamp: Time we arrived/left
  455. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  456. * @sk: Socket we are owned by
  457. * @dev: Device we arrived on/are leaving by
  458. * @cb: Control buffer. Free for use by every layer. Put private vars here
  459. * @_skb_refdst: destination entry (with norefcount bit)
  460. * @sp: the security path, used for xfrm
  461. * @len: Length of actual data
  462. * @data_len: Data length
  463. * @mac_len: Length of link layer header
  464. * @hdr_len: writable header length of cloned skb
  465. * @csum: Checksum (must include start/offset pair)
  466. * @csum_start: Offset from skb->head where checksumming should start
  467. * @csum_offset: Offset from csum_start where checksum should be stored
  468. * @priority: Packet queueing priority
  469. * @ignore_df: allow local fragmentation
  470. * @cloned: Head may be cloned (check refcnt to be sure)
  471. * @ip_summed: Driver fed us an IP checksum
  472. * @nohdr: Payload reference only, must not modify header
  473. * @pkt_type: Packet class
  474. * @fclone: skbuff clone status
  475. * @ipvs_property: skbuff is owned by ipvs
  476. * @tc_skip_classify: do not classify packet. set by IFB device
  477. * @tc_at_ingress: used within tc_classify to distinguish in/egress
  478. * @tc_redirected: packet was redirected by a tc action
  479. * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
  480. * @peeked: this packet has been seen already, so stats have been
  481. * done for it, don't do them again
  482. * @nf_trace: netfilter packet trace flag
  483. * @protocol: Packet protocol from driver
  484. * @destructor: Destruct function
  485. * @_nfct: Associated connection, if any (with nfctinfo bits)
  486. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  487. * @skb_iif: ifindex of device we arrived on
  488. * @tc_index: Traffic control index
  489. * @hash: the packet hash
  490. * @queue_mapping: Queue mapping for multiqueue devices
  491. * @xmit_more: More SKBs are pending for this queue
  492. * @ndisc_nodetype: router type (from link layer)
  493. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  494. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  495. * ports.
  496. * @sw_hash: indicates hash was computed in software stack
  497. * @wifi_acked_valid: wifi_acked was set
  498. * @wifi_acked: whether frame was acked on wifi or not
  499. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  500. * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
  501. * @dst_pending_confirm: need to confirm neighbour
  502. * @napi_id: id of the NAPI struct this skb came from
  503. * @secmark: security marking
  504. * @mark: Generic packet mark
  505. * @vlan_proto: vlan encapsulation protocol
  506. * @vlan_tci: vlan tag control information
  507. * @inner_protocol: Protocol (encapsulation)
  508. * @inner_transport_header: Inner transport layer header (encapsulation)
  509. * @inner_network_header: Network layer header (encapsulation)
  510. * @inner_mac_header: Link layer header (encapsulation)
  511. * @transport_header: Transport layer header
  512. * @network_header: Network layer header
  513. * @mac_header: Link layer header
  514. * @tail: Tail pointer
  515. * @end: End pointer
  516. * @head: Head of buffer
  517. * @data: Data head pointer
  518. * @truesize: Buffer size
  519. * @users: User count - see {datagram,tcp}.c
  520. */
  521. struct sk_buff {
  522. union {
  523. struct {
  524. /* These two members must be first. */
  525. struct sk_buff *next;
  526. struct sk_buff *prev;
  527. union {
  528. ktime_t tstamp;
  529. u64 skb_mstamp;
  530. };
  531. };
  532. struct rb_node rbnode; /* used in netem & tcp stack */
  533. };
  534. struct sock *sk;
  535. union {
  536. struct net_device *dev;
  537. /* Some protocols might use this space to store information,
  538. * while device pointer would be NULL.
  539. * UDP receive path is one user.
  540. */
  541. unsigned long dev_scratch;
  542. };
  543. /*
  544. * This is the control buffer. It is free to use for every
  545. * layer. Please put your private variables there. If you
  546. * want to keep them across layers you have to do a skb_clone()
  547. * first. This is owned by whoever has the skb queued ATM.
  548. */
  549. char cb[48] __aligned(8);
  550. unsigned long _skb_refdst;
  551. void (*destructor)(struct sk_buff *skb);
  552. #ifdef CONFIG_XFRM
  553. struct sec_path *sp;
  554. #endif
  555. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  556. unsigned long _nfct;
  557. #endif
  558. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  559. struct nf_bridge_info *nf_bridge;
  560. #endif
  561. unsigned int len,
  562. data_len;
  563. __u16 mac_len,
  564. hdr_len;
  565. /* Following fields are _not_ copied in __copy_skb_header()
  566. * Note that queue_mapping is here mostly to fill a hole.
  567. */
  568. kmemcheck_bitfield_begin(flags1);
  569. __u16 queue_mapping;
  570. /* if you move cloned around you also must adapt those constants */
  571. #ifdef __BIG_ENDIAN_BITFIELD
  572. #define CLONED_MASK (1 << 7)
  573. #else
  574. #define CLONED_MASK 1
  575. #endif
  576. #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
  577. __u8 __cloned_offset[0];
  578. __u8 cloned:1,
  579. nohdr:1,
  580. fclone:2,
  581. peeked:1,
  582. head_frag:1,
  583. xmit_more:1,
  584. __unused:1; /* one bit hole */
  585. kmemcheck_bitfield_end(flags1);
  586. /* fields enclosed in headers_start/headers_end are copied
  587. * using a single memcpy() in __copy_skb_header()
  588. */
  589. /* private: */
  590. __u32 headers_start[0];
  591. /* public: */
  592. /* if you move pkt_type around you also must adapt those constants */
  593. #ifdef __BIG_ENDIAN_BITFIELD
  594. #define PKT_TYPE_MAX (7 << 5)
  595. #else
  596. #define PKT_TYPE_MAX 7
  597. #endif
  598. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  599. __u8 __pkt_type_offset[0];
  600. __u8 pkt_type:3;
  601. __u8 pfmemalloc:1;
  602. __u8 ignore_df:1;
  603. __u8 nf_trace:1;
  604. __u8 ip_summed:2;
  605. __u8 ooo_okay:1;
  606. __u8 l4_hash:1;
  607. __u8 sw_hash:1;
  608. __u8 wifi_acked_valid:1;
  609. __u8 wifi_acked:1;
  610. __u8 no_fcs:1;
  611. /* Indicates the inner headers are valid in the skbuff. */
  612. __u8 encapsulation:1;
  613. __u8 encap_hdr_csum:1;
  614. __u8 csum_valid:1;
  615. __u8 csum_complete_sw:1;
  616. __u8 csum_level:2;
  617. __u8 csum_not_inet:1;
  618. __u8 dst_pending_confirm:1;
  619. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  620. __u8 ndisc_nodetype:2;
  621. #endif
  622. __u8 ipvs_property:1;
  623. __u8 inner_protocol_type:1;
  624. __u8 remcsum_offload:1;
  625. #ifdef CONFIG_NET_SWITCHDEV
  626. __u8 offload_fwd_mark:1;
  627. #endif
  628. #ifdef CONFIG_NET_CLS_ACT
  629. __u8 tc_skip_classify:1;
  630. __u8 tc_at_ingress:1;
  631. __u8 tc_redirected:1;
  632. __u8 tc_from_ingress:1;
  633. #endif
  634. #ifdef CONFIG_NET_SCHED
  635. __u16 tc_index; /* traffic control index */
  636. #endif
  637. union {
  638. __wsum csum;
  639. struct {
  640. __u16 csum_start;
  641. __u16 csum_offset;
  642. };
  643. };
  644. __u32 priority;
  645. int skb_iif;
  646. __u32 hash;
  647. __be16 vlan_proto;
  648. __u16 vlan_tci;
  649. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  650. union {
  651. unsigned int napi_id;
  652. unsigned int sender_cpu;
  653. };
  654. #endif
  655. #ifdef CONFIG_NETWORK_SECMARK
  656. __u32 secmark;
  657. #endif
  658. union {
  659. __u32 mark;
  660. __u32 reserved_tailroom;
  661. };
  662. union {
  663. __be16 inner_protocol;
  664. __u8 inner_ipproto;
  665. };
  666. __u16 inner_transport_header;
  667. __u16 inner_network_header;
  668. __u16 inner_mac_header;
  669. __be16 protocol;
  670. __u16 transport_header;
  671. __u16 network_header;
  672. __u16 mac_header;
  673. /* private: */
  674. __u32 headers_end[0];
  675. /* public: */
  676. /* These elements must be at the end, see alloc_skb() for details. */
  677. sk_buff_data_t tail;
  678. sk_buff_data_t end;
  679. unsigned char *head,
  680. *data;
  681. unsigned int truesize;
  682. atomic_t users;
  683. };
  684. #ifdef __KERNEL__
  685. /*
  686. * Handling routines are only of interest to the kernel
  687. */
  688. #include <linux/slab.h>
  689. #define SKB_ALLOC_FCLONE 0x01
  690. #define SKB_ALLOC_RX 0x02
  691. #define SKB_ALLOC_NAPI 0x04
  692. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  693. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  694. {
  695. return unlikely(skb->pfmemalloc);
  696. }
  697. /*
  698. * skb might have a dst pointer attached, refcounted or not.
  699. * _skb_refdst low order bit is set if refcount was _not_ taken
  700. */
  701. #define SKB_DST_NOREF 1UL
  702. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  703. #define SKB_NFCT_PTRMASK ~(7UL)
  704. /**
  705. * skb_dst - returns skb dst_entry
  706. * @skb: buffer
  707. *
  708. * Returns skb dst_entry, regardless of reference taken or not.
  709. */
  710. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  711. {
  712. /* If refdst was not refcounted, check we still are in a
  713. * rcu_read_lock section
  714. */
  715. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  716. !rcu_read_lock_held() &&
  717. !rcu_read_lock_bh_held());
  718. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  719. }
  720. /**
  721. * skb_dst_set - sets skb dst
  722. * @skb: buffer
  723. * @dst: dst entry
  724. *
  725. * Sets skb dst, assuming a reference was taken on dst and should
  726. * be released by skb_dst_drop()
  727. */
  728. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  729. {
  730. skb->_skb_refdst = (unsigned long)dst;
  731. }
  732. /**
  733. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  734. * @skb: buffer
  735. * @dst: dst entry
  736. *
  737. * Sets skb dst, assuming a reference was not taken on dst.
  738. * If dst entry is cached, we do not take reference and dst_release
  739. * will be avoided by refdst_drop. If dst entry is not cached, we take
  740. * reference, so that last dst_release can destroy the dst immediately.
  741. */
  742. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  743. {
  744. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  745. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  746. }
  747. /**
  748. * skb_dst_is_noref - Test if skb dst isn't refcounted
  749. * @skb: buffer
  750. */
  751. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  752. {
  753. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  754. }
  755. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  756. {
  757. return (struct rtable *)skb_dst(skb);
  758. }
  759. /* For mangling skb->pkt_type from user space side from applications
  760. * such as nft, tc, etc, we only allow a conservative subset of
  761. * possible pkt_types to be set.
  762. */
  763. static inline bool skb_pkt_type_ok(u32 ptype)
  764. {
  765. return ptype <= PACKET_OTHERHOST;
  766. }
  767. void kfree_skb(struct sk_buff *skb);
  768. void kfree_skb_list(struct sk_buff *segs);
  769. void skb_tx_error(struct sk_buff *skb);
  770. void consume_skb(struct sk_buff *skb);
  771. void __kfree_skb(struct sk_buff *skb);
  772. extern struct kmem_cache *skbuff_head_cache;
  773. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  774. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  775. bool *fragstolen, int *delta_truesize);
  776. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  777. int node);
  778. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  779. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  780. static inline struct sk_buff *alloc_skb(unsigned int size,
  781. gfp_t priority)
  782. {
  783. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  784. }
  785. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  786. unsigned long data_len,
  787. int max_page_order,
  788. int *errcode,
  789. gfp_t gfp_mask);
  790. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  791. struct sk_buff_fclones {
  792. struct sk_buff skb1;
  793. struct sk_buff skb2;
  794. atomic_t fclone_ref;
  795. };
  796. /**
  797. * skb_fclone_busy - check if fclone is busy
  798. * @sk: socket
  799. * @skb: buffer
  800. *
  801. * Returns true if skb is a fast clone, and its clone is not freed.
  802. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  803. * so we also check that this didnt happen.
  804. */
  805. static inline bool skb_fclone_busy(const struct sock *sk,
  806. const struct sk_buff *skb)
  807. {
  808. const struct sk_buff_fclones *fclones;
  809. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  810. return skb->fclone == SKB_FCLONE_ORIG &&
  811. atomic_read(&fclones->fclone_ref) > 1 &&
  812. fclones->skb2.sk == sk;
  813. }
  814. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  815. gfp_t priority)
  816. {
  817. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  818. }
  819. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  820. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  821. {
  822. return __alloc_skb_head(priority, -1);
  823. }
  824. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  825. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  826. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  827. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  828. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  829. gfp_t gfp_mask, bool fclone);
  830. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  831. gfp_t gfp_mask)
  832. {
  833. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  834. }
  835. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  836. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  837. unsigned int headroom);
  838. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  839. int newtailroom, gfp_t priority);
  840. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  841. int offset, int len);
  842. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  843. int len);
  844. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  845. int skb_pad(struct sk_buff *skb, int pad);
  846. #define dev_kfree_skb(a) consume_skb(a)
  847. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  848. int getfrag(void *from, char *to, int offset,
  849. int len, int odd, struct sk_buff *skb),
  850. void *from, int length);
  851. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  852. int offset, size_t size);
  853. struct skb_seq_state {
  854. __u32 lower_offset;
  855. __u32 upper_offset;
  856. __u32 frag_idx;
  857. __u32 stepped_offset;
  858. struct sk_buff *root_skb;
  859. struct sk_buff *cur_skb;
  860. __u8 *frag_data;
  861. };
  862. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  863. unsigned int to, struct skb_seq_state *st);
  864. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  865. struct skb_seq_state *st);
  866. void skb_abort_seq_read(struct skb_seq_state *st);
  867. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  868. unsigned int to, struct ts_config *config);
  869. /*
  870. * Packet hash types specify the type of hash in skb_set_hash.
  871. *
  872. * Hash types refer to the protocol layer addresses which are used to
  873. * construct a packet's hash. The hashes are used to differentiate or identify
  874. * flows of the protocol layer for the hash type. Hash types are either
  875. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  876. *
  877. * Properties of hashes:
  878. *
  879. * 1) Two packets in different flows have different hash values
  880. * 2) Two packets in the same flow should have the same hash value
  881. *
  882. * A hash at a higher layer is considered to be more specific. A driver should
  883. * set the most specific hash possible.
  884. *
  885. * A driver cannot indicate a more specific hash than the layer at which a hash
  886. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  887. *
  888. * A driver may indicate a hash level which is less specific than the
  889. * actual layer the hash was computed on. For instance, a hash computed
  890. * at L4 may be considered an L3 hash. This should only be done if the
  891. * driver can't unambiguously determine that the HW computed the hash at
  892. * the higher layer. Note that the "should" in the second property above
  893. * permits this.
  894. */
  895. enum pkt_hash_types {
  896. PKT_HASH_TYPE_NONE, /* Undefined type */
  897. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  898. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  899. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  900. };
  901. static inline void skb_clear_hash(struct sk_buff *skb)
  902. {
  903. skb->hash = 0;
  904. skb->sw_hash = 0;
  905. skb->l4_hash = 0;
  906. }
  907. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  908. {
  909. if (!skb->l4_hash)
  910. skb_clear_hash(skb);
  911. }
  912. static inline void
  913. __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
  914. {
  915. skb->l4_hash = is_l4;
  916. skb->sw_hash = is_sw;
  917. skb->hash = hash;
  918. }
  919. static inline void
  920. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  921. {
  922. /* Used by drivers to set hash from HW */
  923. __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
  924. }
  925. static inline void
  926. __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
  927. {
  928. __skb_set_hash(skb, hash, true, is_l4);
  929. }
  930. void __skb_get_hash(struct sk_buff *skb);
  931. u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
  932. u32 skb_get_poff(const struct sk_buff *skb);
  933. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  934. const struct flow_keys *keys, int hlen);
  935. __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
  936. void *data, int hlen_proto);
  937. static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
  938. int thoff, u8 ip_proto)
  939. {
  940. return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
  941. }
  942. void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  943. const struct flow_dissector_key *key,
  944. unsigned int key_count);
  945. bool __skb_flow_dissect(const struct sk_buff *skb,
  946. struct flow_dissector *flow_dissector,
  947. void *target_container,
  948. void *data, __be16 proto, int nhoff, int hlen,
  949. unsigned int flags);
  950. static inline bool skb_flow_dissect(const struct sk_buff *skb,
  951. struct flow_dissector *flow_dissector,
  952. void *target_container, unsigned int flags)
  953. {
  954. return __skb_flow_dissect(skb, flow_dissector, target_container,
  955. NULL, 0, 0, 0, flags);
  956. }
  957. static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
  958. struct flow_keys *flow,
  959. unsigned int flags)
  960. {
  961. memset(flow, 0, sizeof(*flow));
  962. return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
  963. NULL, 0, 0, 0, flags);
  964. }
  965. static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
  966. void *data, __be16 proto,
  967. int nhoff, int hlen,
  968. unsigned int flags)
  969. {
  970. memset(flow, 0, sizeof(*flow));
  971. return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
  972. data, proto, nhoff, hlen, flags);
  973. }
  974. static inline __u32 skb_get_hash(struct sk_buff *skb)
  975. {
  976. if (!skb->l4_hash && !skb->sw_hash)
  977. __skb_get_hash(skb);
  978. return skb->hash;
  979. }
  980. __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
  981. static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
  982. {
  983. if (!skb->l4_hash && !skb->sw_hash) {
  984. struct flow_keys keys;
  985. __u32 hash = __get_hash_from_flowi6(fl6, &keys);
  986. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  987. }
  988. return skb->hash;
  989. }
  990. __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
  991. static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
  992. {
  993. if (!skb->l4_hash && !skb->sw_hash) {
  994. struct flow_keys keys;
  995. __u32 hash = __get_hash_from_flowi4(fl4, &keys);
  996. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  997. }
  998. return skb->hash;
  999. }
  1000. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  1001. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  1002. {
  1003. return skb->hash;
  1004. }
  1005. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  1006. {
  1007. to->hash = from->hash;
  1008. to->sw_hash = from->sw_hash;
  1009. to->l4_hash = from->l4_hash;
  1010. };
  1011. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1012. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1013. {
  1014. return skb->head + skb->end;
  1015. }
  1016. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1017. {
  1018. return skb->end;
  1019. }
  1020. #else
  1021. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1022. {
  1023. return skb->end;
  1024. }
  1025. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1026. {
  1027. return skb->end - skb->head;
  1028. }
  1029. #endif
  1030. /* Internal */
  1031. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  1032. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  1033. {
  1034. return &skb_shinfo(skb)->hwtstamps;
  1035. }
  1036. /**
  1037. * skb_queue_empty - check if a queue is empty
  1038. * @list: queue head
  1039. *
  1040. * Returns true if the queue is empty, false otherwise.
  1041. */
  1042. static inline int skb_queue_empty(const struct sk_buff_head *list)
  1043. {
  1044. return list->next == (const struct sk_buff *) list;
  1045. }
  1046. /**
  1047. * skb_queue_is_last - check if skb is the last entry in the queue
  1048. * @list: queue head
  1049. * @skb: buffer
  1050. *
  1051. * Returns true if @skb is the last buffer on the list.
  1052. */
  1053. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  1054. const struct sk_buff *skb)
  1055. {
  1056. return skb->next == (const struct sk_buff *) list;
  1057. }
  1058. /**
  1059. * skb_queue_is_first - check if skb is the first entry in the queue
  1060. * @list: queue head
  1061. * @skb: buffer
  1062. *
  1063. * Returns true if @skb is the first buffer on the list.
  1064. */
  1065. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  1066. const struct sk_buff *skb)
  1067. {
  1068. return skb->prev == (const struct sk_buff *) list;
  1069. }
  1070. /**
  1071. * skb_queue_next - return the next packet in the queue
  1072. * @list: queue head
  1073. * @skb: current buffer
  1074. *
  1075. * Return the next packet in @list after @skb. It is only valid to
  1076. * call this if skb_queue_is_last() evaluates to false.
  1077. */
  1078. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  1079. const struct sk_buff *skb)
  1080. {
  1081. /* This BUG_ON may seem severe, but if we just return then we
  1082. * are going to dereference garbage.
  1083. */
  1084. BUG_ON(skb_queue_is_last(list, skb));
  1085. return skb->next;
  1086. }
  1087. /**
  1088. * skb_queue_prev - return the prev packet in the queue
  1089. * @list: queue head
  1090. * @skb: current buffer
  1091. *
  1092. * Return the prev packet in @list before @skb. It is only valid to
  1093. * call this if skb_queue_is_first() evaluates to false.
  1094. */
  1095. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  1096. const struct sk_buff *skb)
  1097. {
  1098. /* This BUG_ON may seem severe, but if we just return then we
  1099. * are going to dereference garbage.
  1100. */
  1101. BUG_ON(skb_queue_is_first(list, skb));
  1102. return skb->prev;
  1103. }
  1104. /**
  1105. * skb_get - reference buffer
  1106. * @skb: buffer to reference
  1107. *
  1108. * Makes another reference to a socket buffer and returns a pointer
  1109. * to the buffer.
  1110. */
  1111. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  1112. {
  1113. atomic_inc(&skb->users);
  1114. return skb;
  1115. }
  1116. /*
  1117. * If users == 1, we are the only owner and are can avoid redundant
  1118. * atomic change.
  1119. */
  1120. /**
  1121. * skb_cloned - is the buffer a clone
  1122. * @skb: buffer to check
  1123. *
  1124. * Returns true if the buffer was generated with skb_clone() and is
  1125. * one of multiple shared copies of the buffer. Cloned buffers are
  1126. * shared data so must not be written to under normal circumstances.
  1127. */
  1128. static inline int skb_cloned(const struct sk_buff *skb)
  1129. {
  1130. return skb->cloned &&
  1131. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  1132. }
  1133. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  1134. {
  1135. might_sleep_if(gfpflags_allow_blocking(pri));
  1136. if (skb_cloned(skb))
  1137. return pskb_expand_head(skb, 0, 0, pri);
  1138. return 0;
  1139. }
  1140. /**
  1141. * skb_header_cloned - is the header a clone
  1142. * @skb: buffer to check
  1143. *
  1144. * Returns true if modifying the header part of the buffer requires
  1145. * the data to be copied.
  1146. */
  1147. static inline int skb_header_cloned(const struct sk_buff *skb)
  1148. {
  1149. int dataref;
  1150. if (!skb->cloned)
  1151. return 0;
  1152. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  1153. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  1154. return dataref != 1;
  1155. }
  1156. static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
  1157. {
  1158. might_sleep_if(gfpflags_allow_blocking(pri));
  1159. if (skb_header_cloned(skb))
  1160. return pskb_expand_head(skb, 0, 0, pri);
  1161. return 0;
  1162. }
  1163. /**
  1164. * skb_header_release - release reference to header
  1165. * @skb: buffer to operate on
  1166. *
  1167. * Drop a reference to the header part of the buffer. This is done
  1168. * by acquiring a payload reference. You must not read from the header
  1169. * part of skb->data after this.
  1170. * Note : Check if you can use __skb_header_release() instead.
  1171. */
  1172. static inline void skb_header_release(struct sk_buff *skb)
  1173. {
  1174. BUG_ON(skb->nohdr);
  1175. skb->nohdr = 1;
  1176. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1177. }
  1178. /**
  1179. * __skb_header_release - release reference to header
  1180. * @skb: buffer to operate on
  1181. *
  1182. * Variant of skb_header_release() assuming skb is private to caller.
  1183. * We can avoid one atomic operation.
  1184. */
  1185. static inline void __skb_header_release(struct sk_buff *skb)
  1186. {
  1187. skb->nohdr = 1;
  1188. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1189. }
  1190. /**
  1191. * skb_shared - is the buffer shared
  1192. * @skb: buffer to check
  1193. *
  1194. * Returns true if more than one person has a reference to this
  1195. * buffer.
  1196. */
  1197. static inline int skb_shared(const struct sk_buff *skb)
  1198. {
  1199. return atomic_read(&skb->users) != 1;
  1200. }
  1201. /**
  1202. * skb_share_check - check if buffer is shared and if so clone it
  1203. * @skb: buffer to check
  1204. * @pri: priority for memory allocation
  1205. *
  1206. * If the buffer is shared the buffer is cloned and the old copy
  1207. * drops a reference. A new clone with a single reference is returned.
  1208. * If the buffer is not shared the original buffer is returned. When
  1209. * being called from interrupt status or with spinlocks held pri must
  1210. * be GFP_ATOMIC.
  1211. *
  1212. * NULL is returned on a memory allocation failure.
  1213. */
  1214. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1215. {
  1216. might_sleep_if(gfpflags_allow_blocking(pri));
  1217. if (skb_shared(skb)) {
  1218. struct sk_buff *nskb = skb_clone(skb, pri);
  1219. if (likely(nskb))
  1220. consume_skb(skb);
  1221. else
  1222. kfree_skb(skb);
  1223. skb = nskb;
  1224. }
  1225. return skb;
  1226. }
  1227. /*
  1228. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1229. * packets to handle cases where we have a local reader and forward
  1230. * and a couple of other messy ones. The normal one is tcpdumping
  1231. * a packet thats being forwarded.
  1232. */
  1233. /**
  1234. * skb_unshare - make a copy of a shared buffer
  1235. * @skb: buffer to check
  1236. * @pri: priority for memory allocation
  1237. *
  1238. * If the socket buffer is a clone then this function creates a new
  1239. * copy of the data, drops a reference count on the old copy and returns
  1240. * the new copy with the reference count at 1. If the buffer is not a clone
  1241. * the original buffer is returned. When called with a spinlock held or
  1242. * from interrupt state @pri must be %GFP_ATOMIC
  1243. *
  1244. * %NULL is returned on a memory allocation failure.
  1245. */
  1246. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1247. gfp_t pri)
  1248. {
  1249. might_sleep_if(gfpflags_allow_blocking(pri));
  1250. if (skb_cloned(skb)) {
  1251. struct sk_buff *nskb = skb_copy(skb, pri);
  1252. /* Free our shared copy */
  1253. if (likely(nskb))
  1254. consume_skb(skb);
  1255. else
  1256. kfree_skb(skb);
  1257. skb = nskb;
  1258. }
  1259. return skb;
  1260. }
  1261. /**
  1262. * skb_peek - peek at the head of an &sk_buff_head
  1263. * @list_: list to peek at
  1264. *
  1265. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1266. * be careful with this one. A peek leaves the buffer on the
  1267. * list and someone else may run off with it. You must hold
  1268. * the appropriate locks or have a private queue to do this.
  1269. *
  1270. * Returns %NULL for an empty list or a pointer to the head element.
  1271. * The reference count is not incremented and the reference is therefore
  1272. * volatile. Use with caution.
  1273. */
  1274. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1275. {
  1276. struct sk_buff *skb = list_->next;
  1277. if (skb == (struct sk_buff *)list_)
  1278. skb = NULL;
  1279. return skb;
  1280. }
  1281. /**
  1282. * skb_peek_next - peek skb following the given one from a queue
  1283. * @skb: skb to start from
  1284. * @list_: list to peek at
  1285. *
  1286. * Returns %NULL when the end of the list is met or a pointer to the
  1287. * next element. The reference count is not incremented and the
  1288. * reference is therefore volatile. Use with caution.
  1289. */
  1290. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1291. const struct sk_buff_head *list_)
  1292. {
  1293. struct sk_buff *next = skb->next;
  1294. if (next == (struct sk_buff *)list_)
  1295. next = NULL;
  1296. return next;
  1297. }
  1298. /**
  1299. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1300. * @list_: list to peek at
  1301. *
  1302. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1303. * be careful with this one. A peek leaves the buffer on the
  1304. * list and someone else may run off with it. You must hold
  1305. * the appropriate locks or have a private queue to do this.
  1306. *
  1307. * Returns %NULL for an empty list or a pointer to the tail element.
  1308. * The reference count is not incremented and the reference is therefore
  1309. * volatile. Use with caution.
  1310. */
  1311. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1312. {
  1313. struct sk_buff *skb = list_->prev;
  1314. if (skb == (struct sk_buff *)list_)
  1315. skb = NULL;
  1316. return skb;
  1317. }
  1318. /**
  1319. * skb_queue_len - get queue length
  1320. * @list_: list to measure
  1321. *
  1322. * Return the length of an &sk_buff queue.
  1323. */
  1324. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1325. {
  1326. return list_->qlen;
  1327. }
  1328. /**
  1329. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1330. * @list: queue to initialize
  1331. *
  1332. * This initializes only the list and queue length aspects of
  1333. * an sk_buff_head object. This allows to initialize the list
  1334. * aspects of an sk_buff_head without reinitializing things like
  1335. * the spinlock. It can also be used for on-stack sk_buff_head
  1336. * objects where the spinlock is known to not be used.
  1337. */
  1338. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1339. {
  1340. list->prev = list->next = (struct sk_buff *)list;
  1341. list->qlen = 0;
  1342. }
  1343. /*
  1344. * This function creates a split out lock class for each invocation;
  1345. * this is needed for now since a whole lot of users of the skb-queue
  1346. * infrastructure in drivers have different locking usage (in hardirq)
  1347. * than the networking core (in softirq only). In the long run either the
  1348. * network layer or drivers should need annotation to consolidate the
  1349. * main types of usage into 3 classes.
  1350. */
  1351. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1352. {
  1353. spin_lock_init(&list->lock);
  1354. __skb_queue_head_init(list);
  1355. }
  1356. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1357. struct lock_class_key *class)
  1358. {
  1359. skb_queue_head_init(list);
  1360. lockdep_set_class(&list->lock, class);
  1361. }
  1362. /*
  1363. * Insert an sk_buff on a list.
  1364. *
  1365. * The "__skb_xxxx()" functions are the non-atomic ones that
  1366. * can only be called with interrupts disabled.
  1367. */
  1368. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1369. struct sk_buff_head *list);
  1370. static inline void __skb_insert(struct sk_buff *newsk,
  1371. struct sk_buff *prev, struct sk_buff *next,
  1372. struct sk_buff_head *list)
  1373. {
  1374. newsk->next = next;
  1375. newsk->prev = prev;
  1376. next->prev = prev->next = newsk;
  1377. list->qlen++;
  1378. }
  1379. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1380. struct sk_buff *prev,
  1381. struct sk_buff *next)
  1382. {
  1383. struct sk_buff *first = list->next;
  1384. struct sk_buff *last = list->prev;
  1385. first->prev = prev;
  1386. prev->next = first;
  1387. last->next = next;
  1388. next->prev = last;
  1389. }
  1390. /**
  1391. * skb_queue_splice - join two skb lists, this is designed for stacks
  1392. * @list: the new list to add
  1393. * @head: the place to add it in the first list
  1394. */
  1395. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1396. struct sk_buff_head *head)
  1397. {
  1398. if (!skb_queue_empty(list)) {
  1399. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1400. head->qlen += list->qlen;
  1401. }
  1402. }
  1403. /**
  1404. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1405. * @list: the new list to add
  1406. * @head: the place to add it in the first list
  1407. *
  1408. * The list at @list is reinitialised
  1409. */
  1410. static inline void skb_queue_splice_init(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. __skb_queue_head_init(list);
  1417. }
  1418. }
  1419. /**
  1420. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1421. * @list: the new list to add
  1422. * @head: the place to add it in the first list
  1423. */
  1424. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1425. struct sk_buff_head *head)
  1426. {
  1427. if (!skb_queue_empty(list)) {
  1428. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1429. head->qlen += list->qlen;
  1430. }
  1431. }
  1432. /**
  1433. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1434. * @list: the new list to add
  1435. * @head: the place to add it in the first list
  1436. *
  1437. * Each of the lists is a queue.
  1438. * The list at @list is reinitialised
  1439. */
  1440. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1441. struct sk_buff_head *head)
  1442. {
  1443. if (!skb_queue_empty(list)) {
  1444. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1445. head->qlen += list->qlen;
  1446. __skb_queue_head_init(list);
  1447. }
  1448. }
  1449. /**
  1450. * __skb_queue_after - queue a buffer at the list head
  1451. * @list: list to use
  1452. * @prev: place after this buffer
  1453. * @newsk: buffer to queue
  1454. *
  1455. * Queue a buffer int the middle of a list. This function takes no locks
  1456. * and you must therefore hold required locks before calling it.
  1457. *
  1458. * A buffer cannot be placed on two lists at the same time.
  1459. */
  1460. static inline void __skb_queue_after(struct sk_buff_head *list,
  1461. struct sk_buff *prev,
  1462. struct sk_buff *newsk)
  1463. {
  1464. __skb_insert(newsk, prev, prev->next, list);
  1465. }
  1466. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1467. struct sk_buff_head *list);
  1468. static inline void __skb_queue_before(struct sk_buff_head *list,
  1469. struct sk_buff *next,
  1470. struct sk_buff *newsk)
  1471. {
  1472. __skb_insert(newsk, next->prev, next, list);
  1473. }
  1474. /**
  1475. * __skb_queue_head - queue a buffer at the list head
  1476. * @list: list to use
  1477. * @newsk: buffer to queue
  1478. *
  1479. * Queue a buffer at the start of a list. This function takes no locks
  1480. * and you must therefore hold required locks before calling it.
  1481. *
  1482. * A buffer cannot be placed on two lists at the same time.
  1483. */
  1484. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1485. static inline void __skb_queue_head(struct sk_buff_head *list,
  1486. struct sk_buff *newsk)
  1487. {
  1488. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1489. }
  1490. /**
  1491. * __skb_queue_tail - queue a buffer at the list tail
  1492. * @list: list to use
  1493. * @newsk: buffer to queue
  1494. *
  1495. * Queue a buffer at the end of a list. This function takes no locks
  1496. * and you must therefore hold required locks before calling it.
  1497. *
  1498. * A buffer cannot be placed on two lists at the same time.
  1499. */
  1500. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1501. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1502. struct sk_buff *newsk)
  1503. {
  1504. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1505. }
  1506. /*
  1507. * remove sk_buff from list. _Must_ be called atomically, and with
  1508. * the list known..
  1509. */
  1510. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1511. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1512. {
  1513. struct sk_buff *next, *prev;
  1514. list->qlen--;
  1515. next = skb->next;
  1516. prev = skb->prev;
  1517. skb->next = skb->prev = NULL;
  1518. next->prev = prev;
  1519. prev->next = next;
  1520. }
  1521. /**
  1522. * __skb_dequeue - remove from the head of the queue
  1523. * @list: list to dequeue from
  1524. *
  1525. * Remove the head of the list. This function does not take any locks
  1526. * so must be used with appropriate locks held only. The head item is
  1527. * returned or %NULL if the list is empty.
  1528. */
  1529. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1530. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1531. {
  1532. struct sk_buff *skb = skb_peek(list);
  1533. if (skb)
  1534. __skb_unlink(skb, list);
  1535. return skb;
  1536. }
  1537. /**
  1538. * __skb_dequeue_tail - remove from the tail of the queue
  1539. * @list: list to dequeue from
  1540. *
  1541. * Remove the tail of the list. This function does not take any locks
  1542. * so must be used with appropriate locks held only. The tail item is
  1543. * returned or %NULL if the list is empty.
  1544. */
  1545. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1546. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1547. {
  1548. struct sk_buff *skb = skb_peek_tail(list);
  1549. if (skb)
  1550. __skb_unlink(skb, list);
  1551. return skb;
  1552. }
  1553. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1554. {
  1555. return skb->data_len;
  1556. }
  1557. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1558. {
  1559. return skb->len - skb->data_len;
  1560. }
  1561. static inline unsigned int skb_pagelen(const struct sk_buff *skb)
  1562. {
  1563. unsigned int i, len = 0;
  1564. for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
  1565. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1566. return len + skb_headlen(skb);
  1567. }
  1568. /**
  1569. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1570. * @skb: buffer containing fragment to be initialised
  1571. * @i: paged fragment index to initialise
  1572. * @page: the page to use for this fragment
  1573. * @off: the offset to the data with @page
  1574. * @size: the length of the data
  1575. *
  1576. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1577. * offset @off within @page.
  1578. *
  1579. * Does not take any additional reference on the fragment.
  1580. */
  1581. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1582. struct page *page, int off, int size)
  1583. {
  1584. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1585. /*
  1586. * Propagate page pfmemalloc to the skb if we can. The problem is
  1587. * that not all callers have unique ownership of the page but rely
  1588. * on page_is_pfmemalloc doing the right thing(tm).
  1589. */
  1590. frag->page.p = page;
  1591. frag->page_offset = off;
  1592. skb_frag_size_set(frag, size);
  1593. page = compound_head(page);
  1594. if (page_is_pfmemalloc(page))
  1595. skb->pfmemalloc = true;
  1596. }
  1597. /**
  1598. * skb_fill_page_desc - initialise a paged fragment in an skb
  1599. * @skb: buffer containing fragment to be initialised
  1600. * @i: paged fragment index to initialise
  1601. * @page: the page to use for this fragment
  1602. * @off: the offset to the data with @page
  1603. * @size: the length of the data
  1604. *
  1605. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1606. * @skb to point to @size bytes at offset @off within @page. In
  1607. * addition updates @skb such that @i is the last fragment.
  1608. *
  1609. * Does not take any additional reference on the fragment.
  1610. */
  1611. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1612. struct page *page, int off, int size)
  1613. {
  1614. __skb_fill_page_desc(skb, i, page, off, size);
  1615. skb_shinfo(skb)->nr_frags = i + 1;
  1616. }
  1617. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1618. int size, unsigned int truesize);
  1619. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1620. unsigned int truesize);
  1621. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1622. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1623. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1624. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1625. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1626. {
  1627. return skb->head + skb->tail;
  1628. }
  1629. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1630. {
  1631. skb->tail = skb->data - skb->head;
  1632. }
  1633. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1634. {
  1635. skb_reset_tail_pointer(skb);
  1636. skb->tail += offset;
  1637. }
  1638. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1639. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1640. {
  1641. return skb->tail;
  1642. }
  1643. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1644. {
  1645. skb->tail = skb->data;
  1646. }
  1647. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1648. {
  1649. skb->tail = skb->data + offset;
  1650. }
  1651. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1652. /*
  1653. * Add data to an sk_buff
  1654. */
  1655. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1656. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1657. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1658. {
  1659. unsigned char *tmp = skb_tail_pointer(skb);
  1660. SKB_LINEAR_ASSERT(skb);
  1661. skb->tail += len;
  1662. skb->len += len;
  1663. return tmp;
  1664. }
  1665. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1666. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1667. {
  1668. skb->data -= len;
  1669. skb->len += len;
  1670. return skb->data;
  1671. }
  1672. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1673. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1674. {
  1675. skb->len -= len;
  1676. BUG_ON(skb->len < skb->data_len);
  1677. return skb->data += len;
  1678. }
  1679. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1680. {
  1681. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1682. }
  1683. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1684. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1685. {
  1686. if (len > skb_headlen(skb) &&
  1687. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1688. return NULL;
  1689. skb->len -= len;
  1690. return skb->data += len;
  1691. }
  1692. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1693. {
  1694. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1695. }
  1696. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1697. {
  1698. if (likely(len <= skb_headlen(skb)))
  1699. return 1;
  1700. if (unlikely(len > skb->len))
  1701. return 0;
  1702. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1703. }
  1704. void skb_condense(struct sk_buff *skb);
  1705. /**
  1706. * skb_headroom - bytes at buffer head
  1707. * @skb: buffer to check
  1708. *
  1709. * Return the number of bytes of free space at the head of an &sk_buff.
  1710. */
  1711. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1712. {
  1713. return skb->data - skb->head;
  1714. }
  1715. /**
  1716. * skb_tailroom - bytes at buffer end
  1717. * @skb: buffer to check
  1718. *
  1719. * Return the number of bytes of free space at the tail of an sk_buff
  1720. */
  1721. static inline int skb_tailroom(const struct sk_buff *skb)
  1722. {
  1723. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1724. }
  1725. /**
  1726. * skb_availroom - bytes at buffer end
  1727. * @skb: buffer to check
  1728. *
  1729. * Return the number of bytes of free space at the tail of an sk_buff
  1730. * allocated by sk_stream_alloc()
  1731. */
  1732. static inline int skb_availroom(const struct sk_buff *skb)
  1733. {
  1734. if (skb_is_nonlinear(skb))
  1735. return 0;
  1736. return skb->end - skb->tail - skb->reserved_tailroom;
  1737. }
  1738. /**
  1739. * skb_reserve - adjust headroom
  1740. * @skb: buffer to alter
  1741. * @len: bytes to move
  1742. *
  1743. * Increase the headroom of an empty &sk_buff by reducing the tail
  1744. * room. This is only allowed for an empty buffer.
  1745. */
  1746. static inline void skb_reserve(struct sk_buff *skb, int len)
  1747. {
  1748. skb->data += len;
  1749. skb->tail += len;
  1750. }
  1751. /**
  1752. * skb_tailroom_reserve - adjust reserved_tailroom
  1753. * @skb: buffer to alter
  1754. * @mtu: maximum amount of headlen permitted
  1755. * @needed_tailroom: minimum amount of reserved_tailroom
  1756. *
  1757. * Set reserved_tailroom so that headlen can be as large as possible but
  1758. * not larger than mtu and tailroom cannot be smaller than
  1759. * needed_tailroom.
  1760. * The required headroom should already have been reserved before using
  1761. * this function.
  1762. */
  1763. static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
  1764. unsigned int needed_tailroom)
  1765. {
  1766. SKB_LINEAR_ASSERT(skb);
  1767. if (mtu < skb_tailroom(skb) - needed_tailroom)
  1768. /* use at most mtu */
  1769. skb->reserved_tailroom = skb_tailroom(skb) - mtu;
  1770. else
  1771. /* use up to all available space */
  1772. skb->reserved_tailroom = needed_tailroom;
  1773. }
  1774. #define ENCAP_TYPE_ETHER 0
  1775. #define ENCAP_TYPE_IPPROTO 1
  1776. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1777. __be16 protocol)
  1778. {
  1779. skb->inner_protocol = protocol;
  1780. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1781. }
  1782. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1783. __u8 ipproto)
  1784. {
  1785. skb->inner_ipproto = ipproto;
  1786. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1787. }
  1788. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1789. {
  1790. skb->inner_mac_header = skb->mac_header;
  1791. skb->inner_network_header = skb->network_header;
  1792. skb->inner_transport_header = skb->transport_header;
  1793. }
  1794. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1795. {
  1796. skb->mac_len = skb->network_header - skb->mac_header;
  1797. }
  1798. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1799. *skb)
  1800. {
  1801. return skb->head + skb->inner_transport_header;
  1802. }
  1803. static inline int skb_inner_transport_offset(const struct sk_buff *skb)
  1804. {
  1805. return skb_inner_transport_header(skb) - skb->data;
  1806. }
  1807. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1808. {
  1809. skb->inner_transport_header = skb->data - skb->head;
  1810. }
  1811. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1812. const int offset)
  1813. {
  1814. skb_reset_inner_transport_header(skb);
  1815. skb->inner_transport_header += offset;
  1816. }
  1817. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1818. {
  1819. return skb->head + skb->inner_network_header;
  1820. }
  1821. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1822. {
  1823. skb->inner_network_header = skb->data - skb->head;
  1824. }
  1825. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1826. const int offset)
  1827. {
  1828. skb_reset_inner_network_header(skb);
  1829. skb->inner_network_header += offset;
  1830. }
  1831. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1832. {
  1833. return skb->head + skb->inner_mac_header;
  1834. }
  1835. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1836. {
  1837. skb->inner_mac_header = skb->data - skb->head;
  1838. }
  1839. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1840. const int offset)
  1841. {
  1842. skb_reset_inner_mac_header(skb);
  1843. skb->inner_mac_header += offset;
  1844. }
  1845. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1846. {
  1847. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1848. }
  1849. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1850. {
  1851. return skb->head + skb->transport_header;
  1852. }
  1853. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1854. {
  1855. skb->transport_header = skb->data - skb->head;
  1856. }
  1857. static inline void skb_set_transport_header(struct sk_buff *skb,
  1858. const int offset)
  1859. {
  1860. skb_reset_transport_header(skb);
  1861. skb->transport_header += offset;
  1862. }
  1863. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1864. {
  1865. return skb->head + skb->network_header;
  1866. }
  1867. static inline void skb_reset_network_header(struct sk_buff *skb)
  1868. {
  1869. skb->network_header = skb->data - skb->head;
  1870. }
  1871. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1872. {
  1873. skb_reset_network_header(skb);
  1874. skb->network_header += offset;
  1875. }
  1876. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1877. {
  1878. return skb->head + skb->mac_header;
  1879. }
  1880. static inline int skb_mac_offset(const struct sk_buff *skb)
  1881. {
  1882. return skb_mac_header(skb) - skb->data;
  1883. }
  1884. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1885. {
  1886. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1887. }
  1888. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1889. {
  1890. skb->mac_header = skb->data - skb->head;
  1891. }
  1892. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1893. {
  1894. skb_reset_mac_header(skb);
  1895. skb->mac_header += offset;
  1896. }
  1897. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1898. {
  1899. skb->mac_header = skb->network_header;
  1900. }
  1901. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1902. const int offset_hint)
  1903. {
  1904. struct flow_keys keys;
  1905. if (skb_transport_header_was_set(skb))
  1906. return;
  1907. else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
  1908. skb_set_transport_header(skb, keys.control.thoff);
  1909. else
  1910. skb_set_transport_header(skb, offset_hint);
  1911. }
  1912. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1913. {
  1914. if (skb_mac_header_was_set(skb)) {
  1915. const unsigned char *old_mac = skb_mac_header(skb);
  1916. skb_set_mac_header(skb, -skb->mac_len);
  1917. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1918. }
  1919. }
  1920. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1921. {
  1922. return skb->csum_start - skb_headroom(skb);
  1923. }
  1924. static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
  1925. {
  1926. return skb->head + skb->csum_start;
  1927. }
  1928. static inline int skb_transport_offset(const struct sk_buff *skb)
  1929. {
  1930. return skb_transport_header(skb) - skb->data;
  1931. }
  1932. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1933. {
  1934. return skb->transport_header - skb->network_header;
  1935. }
  1936. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1937. {
  1938. return skb->inner_transport_header - skb->inner_network_header;
  1939. }
  1940. static inline int skb_network_offset(const struct sk_buff *skb)
  1941. {
  1942. return skb_network_header(skb) - skb->data;
  1943. }
  1944. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1945. {
  1946. return skb_inner_network_header(skb) - skb->data;
  1947. }
  1948. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1949. {
  1950. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1951. }
  1952. /*
  1953. * CPUs often take a performance hit when accessing unaligned memory
  1954. * locations. The actual performance hit varies, it can be small if the
  1955. * hardware handles it or large if we have to take an exception and fix it
  1956. * in software.
  1957. *
  1958. * Since an ethernet header is 14 bytes network drivers often end up with
  1959. * the IP header at an unaligned offset. The IP header can be aligned by
  1960. * shifting the start of the packet by 2 bytes. Drivers should do this
  1961. * with:
  1962. *
  1963. * skb_reserve(skb, NET_IP_ALIGN);
  1964. *
  1965. * The downside to this alignment of the IP header is that the DMA is now
  1966. * unaligned. On some architectures the cost of an unaligned DMA is high
  1967. * and this cost outweighs the gains made by aligning the IP header.
  1968. *
  1969. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1970. * to be overridden.
  1971. */
  1972. #ifndef NET_IP_ALIGN
  1973. #define NET_IP_ALIGN 2
  1974. #endif
  1975. /*
  1976. * The networking layer reserves some headroom in skb data (via
  1977. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1978. * the header has to grow. In the default case, if the header has to grow
  1979. * 32 bytes or less we avoid the reallocation.
  1980. *
  1981. * Unfortunately this headroom changes the DMA alignment of the resulting
  1982. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1983. * on some architectures. An architecture can override this value,
  1984. * perhaps setting it to a cacheline in size (since that will maintain
  1985. * cacheline alignment of the DMA). It must be a power of 2.
  1986. *
  1987. * Various parts of the networking layer expect at least 32 bytes of
  1988. * headroom, you should not reduce this.
  1989. *
  1990. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1991. * to reduce average number of cache lines per packet.
  1992. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1993. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1994. */
  1995. #ifndef NET_SKB_PAD
  1996. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1997. #endif
  1998. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1999. static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
  2000. {
  2001. if (unlikely(skb_is_nonlinear(skb))) {
  2002. WARN_ON(1);
  2003. return;
  2004. }
  2005. skb->len = len;
  2006. skb_set_tail_pointer(skb, len);
  2007. }
  2008. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  2009. {
  2010. __skb_set_length(skb, len);
  2011. }
  2012. void skb_trim(struct sk_buff *skb, unsigned int len);
  2013. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  2014. {
  2015. if (skb->data_len)
  2016. return ___pskb_trim(skb, len);
  2017. __skb_trim(skb, len);
  2018. return 0;
  2019. }
  2020. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  2021. {
  2022. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  2023. }
  2024. /**
  2025. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  2026. * @skb: buffer to alter
  2027. * @len: new length
  2028. *
  2029. * This is identical to pskb_trim except that the caller knows that
  2030. * the skb is not cloned so we should never get an error due to out-
  2031. * of-memory.
  2032. */
  2033. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  2034. {
  2035. int err = pskb_trim(skb, len);
  2036. BUG_ON(err);
  2037. }
  2038. static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
  2039. {
  2040. unsigned int diff = len - skb->len;
  2041. if (skb_tailroom(skb) < diff) {
  2042. int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
  2043. GFP_ATOMIC);
  2044. if (ret)
  2045. return ret;
  2046. }
  2047. __skb_set_length(skb, len);
  2048. return 0;
  2049. }
  2050. /**
  2051. * skb_orphan - orphan a buffer
  2052. * @skb: buffer to orphan
  2053. *
  2054. * If a buffer currently has an owner then we call the owner's
  2055. * destructor function and make the @skb unowned. The buffer continues
  2056. * to exist but is no longer charged to its former owner.
  2057. */
  2058. static inline void skb_orphan(struct sk_buff *skb)
  2059. {
  2060. if (skb->destructor) {
  2061. skb->destructor(skb);
  2062. skb->destructor = NULL;
  2063. skb->sk = NULL;
  2064. } else {
  2065. BUG_ON(skb->sk);
  2066. }
  2067. }
  2068. /**
  2069. * skb_orphan_frags - orphan the frags contained in a buffer
  2070. * @skb: buffer to orphan frags from
  2071. * @gfp_mask: allocation mask for replacement pages
  2072. *
  2073. * For each frag in the SKB which needs a destructor (i.e. has an
  2074. * owner) create a copy of that frag and release the original
  2075. * page by calling the destructor.
  2076. */
  2077. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  2078. {
  2079. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  2080. return 0;
  2081. return skb_copy_ubufs(skb, gfp_mask);
  2082. }
  2083. /**
  2084. * __skb_queue_purge - empty a list
  2085. * @list: list to empty
  2086. *
  2087. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  2088. * the list and one reference dropped. This function does not take the
  2089. * list lock and the caller must hold the relevant locks to use it.
  2090. */
  2091. void skb_queue_purge(struct sk_buff_head *list);
  2092. static inline void __skb_queue_purge(struct sk_buff_head *list)
  2093. {
  2094. struct sk_buff *skb;
  2095. while ((skb = __skb_dequeue(list)) != NULL)
  2096. kfree_skb(skb);
  2097. }
  2098. void skb_rbtree_purge(struct rb_root *root);
  2099. void *netdev_alloc_frag(unsigned int fragsz);
  2100. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  2101. gfp_t gfp_mask);
  2102. /**
  2103. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  2104. * @dev: network device to receive on
  2105. * @length: length to allocate
  2106. *
  2107. * Allocate a new &sk_buff and assign it a usage count of one. The
  2108. * buffer has unspecified headroom built in. Users should allocate
  2109. * the headroom they think they need without accounting for the
  2110. * built in space. The built in space is used for optimisations.
  2111. *
  2112. * %NULL is returned if there is no free memory. Although this function
  2113. * allocates memory it can be called from an interrupt.
  2114. */
  2115. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  2116. unsigned int length)
  2117. {
  2118. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  2119. }
  2120. /* legacy helper around __netdev_alloc_skb() */
  2121. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  2122. gfp_t gfp_mask)
  2123. {
  2124. return __netdev_alloc_skb(NULL, length, gfp_mask);
  2125. }
  2126. /* legacy helper around netdev_alloc_skb() */
  2127. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  2128. {
  2129. return netdev_alloc_skb(NULL, length);
  2130. }
  2131. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  2132. unsigned int length, gfp_t gfp)
  2133. {
  2134. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  2135. if (NET_IP_ALIGN && skb)
  2136. skb_reserve(skb, NET_IP_ALIGN);
  2137. return skb;
  2138. }
  2139. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  2140. unsigned int length)
  2141. {
  2142. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  2143. }
  2144. static inline void skb_free_frag(void *addr)
  2145. {
  2146. page_frag_free(addr);
  2147. }
  2148. void *napi_alloc_frag(unsigned int fragsz);
  2149. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  2150. unsigned int length, gfp_t gfp_mask);
  2151. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  2152. unsigned int length)
  2153. {
  2154. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  2155. }
  2156. void napi_consume_skb(struct sk_buff *skb, int budget);
  2157. void __kfree_skb_flush(void);
  2158. void __kfree_skb_defer(struct sk_buff *skb);
  2159. /**
  2160. * __dev_alloc_pages - allocate page for network Rx
  2161. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2162. * @order: size of the allocation
  2163. *
  2164. * Allocate a new page.
  2165. *
  2166. * %NULL is returned if there is no free memory.
  2167. */
  2168. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  2169. unsigned int order)
  2170. {
  2171. /* This piece of code contains several assumptions.
  2172. * 1. This is for device Rx, therefor a cold page is preferred.
  2173. * 2. The expectation is the user wants a compound page.
  2174. * 3. If requesting a order 0 page it will not be compound
  2175. * due to the check to see if order has a value in prep_new_page
  2176. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  2177. * code in gfp_to_alloc_flags that should be enforcing this.
  2178. */
  2179. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  2180. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  2181. }
  2182. static inline struct page *dev_alloc_pages(unsigned int order)
  2183. {
  2184. return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
  2185. }
  2186. /**
  2187. * __dev_alloc_page - allocate a page for network Rx
  2188. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2189. *
  2190. * Allocate a new page.
  2191. *
  2192. * %NULL is returned if there is no free memory.
  2193. */
  2194. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  2195. {
  2196. return __dev_alloc_pages(gfp_mask, 0);
  2197. }
  2198. static inline struct page *dev_alloc_page(void)
  2199. {
  2200. return dev_alloc_pages(0);
  2201. }
  2202. /**
  2203. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  2204. * @page: The page that was allocated from skb_alloc_page
  2205. * @skb: The skb that may need pfmemalloc set
  2206. */
  2207. static inline void skb_propagate_pfmemalloc(struct page *page,
  2208. struct sk_buff *skb)
  2209. {
  2210. if (page_is_pfmemalloc(page))
  2211. skb->pfmemalloc = true;
  2212. }
  2213. /**
  2214. * skb_frag_page - retrieve the page referred to by a paged fragment
  2215. * @frag: the paged fragment
  2216. *
  2217. * Returns the &struct page associated with @frag.
  2218. */
  2219. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2220. {
  2221. return frag->page.p;
  2222. }
  2223. /**
  2224. * __skb_frag_ref - take an addition reference on a paged fragment.
  2225. * @frag: the paged fragment
  2226. *
  2227. * Takes an additional reference on the paged fragment @frag.
  2228. */
  2229. static inline void __skb_frag_ref(skb_frag_t *frag)
  2230. {
  2231. get_page(skb_frag_page(frag));
  2232. }
  2233. /**
  2234. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2235. * @skb: the buffer
  2236. * @f: the fragment offset.
  2237. *
  2238. * Takes an additional reference on the @f'th paged fragment of @skb.
  2239. */
  2240. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2241. {
  2242. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2243. }
  2244. /**
  2245. * __skb_frag_unref - release a reference on a paged fragment.
  2246. * @frag: the paged fragment
  2247. *
  2248. * Releases a reference on the paged fragment @frag.
  2249. */
  2250. static inline void __skb_frag_unref(skb_frag_t *frag)
  2251. {
  2252. put_page(skb_frag_page(frag));
  2253. }
  2254. /**
  2255. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2256. * @skb: the buffer
  2257. * @f: the fragment offset
  2258. *
  2259. * Releases a reference on the @f'th paged fragment of @skb.
  2260. */
  2261. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2262. {
  2263. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2264. }
  2265. /**
  2266. * skb_frag_address - gets the address of the data contained in a paged fragment
  2267. * @frag: the paged fragment buffer
  2268. *
  2269. * Returns the address of the data within @frag. The page must already
  2270. * be mapped.
  2271. */
  2272. static inline void *skb_frag_address(const skb_frag_t *frag)
  2273. {
  2274. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2275. }
  2276. /**
  2277. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2278. * @frag: the paged fragment buffer
  2279. *
  2280. * Returns the address of the data within @frag. Checks that the page
  2281. * is mapped and returns %NULL otherwise.
  2282. */
  2283. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2284. {
  2285. void *ptr = page_address(skb_frag_page(frag));
  2286. if (unlikely(!ptr))
  2287. return NULL;
  2288. return ptr + frag->page_offset;
  2289. }
  2290. /**
  2291. * __skb_frag_set_page - sets the page contained in a paged fragment
  2292. * @frag: the paged fragment
  2293. * @page: the page to set
  2294. *
  2295. * Sets the fragment @frag to contain @page.
  2296. */
  2297. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2298. {
  2299. frag->page.p = page;
  2300. }
  2301. /**
  2302. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2303. * @skb: the buffer
  2304. * @f: the fragment offset
  2305. * @page: the page to set
  2306. *
  2307. * Sets the @f'th fragment of @skb to contain @page.
  2308. */
  2309. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2310. struct page *page)
  2311. {
  2312. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2313. }
  2314. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2315. /**
  2316. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2317. * @dev: the device to map the fragment to
  2318. * @frag: the paged fragment to map
  2319. * @offset: the offset within the fragment (starting at the
  2320. * fragment's own offset)
  2321. * @size: the number of bytes to map
  2322. * @dir: the direction of the mapping (%PCI_DMA_*)
  2323. *
  2324. * Maps the page associated with @frag to @device.
  2325. */
  2326. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2327. const skb_frag_t *frag,
  2328. size_t offset, size_t size,
  2329. enum dma_data_direction dir)
  2330. {
  2331. return dma_map_page(dev, skb_frag_page(frag),
  2332. frag->page_offset + offset, size, dir);
  2333. }
  2334. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2335. gfp_t gfp_mask)
  2336. {
  2337. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2338. }
  2339. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2340. gfp_t gfp_mask)
  2341. {
  2342. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2343. }
  2344. /**
  2345. * skb_clone_writable - is the header of a clone writable
  2346. * @skb: buffer to check
  2347. * @len: length up to which to write
  2348. *
  2349. * Returns true if modifying the header part of the cloned buffer
  2350. * does not requires the data to be copied.
  2351. */
  2352. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2353. {
  2354. return !skb_header_cloned(skb) &&
  2355. skb_headroom(skb) + len <= skb->hdr_len;
  2356. }
  2357. static inline int skb_try_make_writable(struct sk_buff *skb,
  2358. unsigned int write_len)
  2359. {
  2360. return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
  2361. pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
  2362. }
  2363. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2364. int cloned)
  2365. {
  2366. int delta = 0;
  2367. if (headroom > skb_headroom(skb))
  2368. delta = headroom - skb_headroom(skb);
  2369. if (delta || cloned)
  2370. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2371. GFP_ATOMIC);
  2372. return 0;
  2373. }
  2374. /**
  2375. * skb_cow - copy header of skb when it is required
  2376. * @skb: buffer to cow
  2377. * @headroom: needed headroom
  2378. *
  2379. * If the skb passed lacks sufficient headroom or its data part
  2380. * is shared, data is reallocated. If reallocation fails, an error
  2381. * is returned and original skb is not changed.
  2382. *
  2383. * The result is skb with writable area skb->head...skb->tail
  2384. * and at least @headroom of space at head.
  2385. */
  2386. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2387. {
  2388. return __skb_cow(skb, headroom, skb_cloned(skb));
  2389. }
  2390. /**
  2391. * skb_cow_head - skb_cow but only making the head writable
  2392. * @skb: buffer to cow
  2393. * @headroom: needed headroom
  2394. *
  2395. * This function is identical to skb_cow except that we replace the
  2396. * skb_cloned check by skb_header_cloned. It should be used when
  2397. * you only need to push on some header and do not need to modify
  2398. * the data.
  2399. */
  2400. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2401. {
  2402. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2403. }
  2404. /**
  2405. * skb_padto - pad an skbuff up to a minimal size
  2406. * @skb: buffer to pad
  2407. * @len: minimal length
  2408. *
  2409. * Pads up a buffer to ensure the trailing bytes exist and are
  2410. * blanked. If the buffer already contains sufficient data it
  2411. * is untouched. Otherwise it is extended. Returns zero on
  2412. * success. The skb is freed on error.
  2413. */
  2414. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2415. {
  2416. unsigned int size = skb->len;
  2417. if (likely(size >= len))
  2418. return 0;
  2419. return skb_pad(skb, len - size);
  2420. }
  2421. /**
  2422. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2423. * @skb: buffer to pad
  2424. * @len: minimal length
  2425. *
  2426. * Pads up a buffer to ensure the trailing bytes exist and are
  2427. * blanked. If the buffer already contains sufficient data it
  2428. * is untouched. Otherwise it is extended. Returns zero on
  2429. * success. The skb is freed on error.
  2430. */
  2431. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2432. {
  2433. unsigned int size = skb->len;
  2434. if (unlikely(size < len)) {
  2435. len -= size;
  2436. if (skb_pad(skb, len))
  2437. return -ENOMEM;
  2438. __skb_put(skb, len);
  2439. }
  2440. return 0;
  2441. }
  2442. static inline int skb_add_data(struct sk_buff *skb,
  2443. struct iov_iter *from, int copy)
  2444. {
  2445. const int off = skb->len;
  2446. if (skb->ip_summed == CHECKSUM_NONE) {
  2447. __wsum csum = 0;
  2448. if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
  2449. &csum, from)) {
  2450. skb->csum = csum_block_add(skb->csum, csum, off);
  2451. return 0;
  2452. }
  2453. } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
  2454. return 0;
  2455. __skb_trim(skb, off);
  2456. return -EFAULT;
  2457. }
  2458. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2459. const struct page *page, int off)
  2460. {
  2461. if (i) {
  2462. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2463. return page == skb_frag_page(frag) &&
  2464. off == frag->page_offset + skb_frag_size(frag);
  2465. }
  2466. return false;
  2467. }
  2468. static inline int __skb_linearize(struct sk_buff *skb)
  2469. {
  2470. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2471. }
  2472. /**
  2473. * skb_linearize - convert paged skb to linear one
  2474. * @skb: buffer to linarize
  2475. *
  2476. * If there is no free memory -ENOMEM is returned, otherwise zero
  2477. * is returned and the old skb data released.
  2478. */
  2479. static inline int skb_linearize(struct sk_buff *skb)
  2480. {
  2481. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2482. }
  2483. /**
  2484. * skb_has_shared_frag - can any frag be overwritten
  2485. * @skb: buffer to test
  2486. *
  2487. * Return true if the skb has at least one frag that might be modified
  2488. * by an external entity (as in vmsplice()/sendfile())
  2489. */
  2490. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2491. {
  2492. return skb_is_nonlinear(skb) &&
  2493. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2494. }
  2495. /**
  2496. * skb_linearize_cow - make sure skb is linear and writable
  2497. * @skb: buffer to process
  2498. *
  2499. * If there is no free memory -ENOMEM is returned, otherwise zero
  2500. * is returned and the old skb data released.
  2501. */
  2502. static inline int skb_linearize_cow(struct sk_buff *skb)
  2503. {
  2504. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2505. __skb_linearize(skb) : 0;
  2506. }
  2507. static __always_inline void
  2508. __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2509. unsigned int off)
  2510. {
  2511. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2512. skb->csum = csum_block_sub(skb->csum,
  2513. csum_partial(start, len, 0), off);
  2514. else if (skb->ip_summed == CHECKSUM_PARTIAL &&
  2515. skb_checksum_start_offset(skb) < 0)
  2516. skb->ip_summed = CHECKSUM_NONE;
  2517. }
  2518. /**
  2519. * skb_postpull_rcsum - update checksum for received skb after pull
  2520. * @skb: buffer to update
  2521. * @start: start of data before pull
  2522. * @len: length of data pulled
  2523. *
  2524. * After doing a pull on a received packet, you need to call this to
  2525. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2526. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2527. */
  2528. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2529. const void *start, unsigned int len)
  2530. {
  2531. __skb_postpull_rcsum(skb, start, len, 0);
  2532. }
  2533. static __always_inline void
  2534. __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2535. unsigned int off)
  2536. {
  2537. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2538. skb->csum = csum_block_add(skb->csum,
  2539. csum_partial(start, len, 0), off);
  2540. }
  2541. /**
  2542. * skb_postpush_rcsum - update checksum for received skb after push
  2543. * @skb: buffer to update
  2544. * @start: start of data after push
  2545. * @len: length of data pushed
  2546. *
  2547. * After doing a push on a received packet, you need to call this to
  2548. * update the CHECKSUM_COMPLETE checksum.
  2549. */
  2550. static inline void skb_postpush_rcsum(struct sk_buff *skb,
  2551. const void *start, unsigned int len)
  2552. {
  2553. __skb_postpush_rcsum(skb, start, len, 0);
  2554. }
  2555. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2556. /**
  2557. * skb_push_rcsum - push skb and update receive checksum
  2558. * @skb: buffer to update
  2559. * @len: length of data pulled
  2560. *
  2561. * This function performs an skb_push on the packet and updates
  2562. * the CHECKSUM_COMPLETE checksum. It should be used on
  2563. * receive path processing instead of skb_push unless you know
  2564. * that the checksum difference is zero (e.g., a valid IP header)
  2565. * or you are setting ip_summed to CHECKSUM_NONE.
  2566. */
  2567. static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
  2568. unsigned int len)
  2569. {
  2570. skb_push(skb, len);
  2571. skb_postpush_rcsum(skb, skb->data, len);
  2572. return skb->data;
  2573. }
  2574. /**
  2575. * pskb_trim_rcsum - trim received skb and update checksum
  2576. * @skb: buffer to trim
  2577. * @len: new length
  2578. *
  2579. * This is exactly the same as pskb_trim except that it ensures the
  2580. * checksum of received packets are still valid after the operation.
  2581. */
  2582. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2583. {
  2584. if (likely(len >= skb->len))
  2585. return 0;
  2586. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2587. skb->ip_summed = CHECKSUM_NONE;
  2588. return __pskb_trim(skb, len);
  2589. }
  2590. static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2591. {
  2592. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2593. skb->ip_summed = CHECKSUM_NONE;
  2594. __skb_trim(skb, len);
  2595. return 0;
  2596. }
  2597. static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
  2598. {
  2599. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2600. skb->ip_summed = CHECKSUM_NONE;
  2601. return __skb_grow(skb, len);
  2602. }
  2603. #define skb_queue_walk(queue, skb) \
  2604. for (skb = (queue)->next; \
  2605. skb != (struct sk_buff *)(queue); \
  2606. skb = skb->next)
  2607. #define skb_queue_walk_safe(queue, skb, tmp) \
  2608. for (skb = (queue)->next, tmp = skb->next; \
  2609. skb != (struct sk_buff *)(queue); \
  2610. skb = tmp, tmp = skb->next)
  2611. #define skb_queue_walk_from(queue, skb) \
  2612. for (; skb != (struct sk_buff *)(queue); \
  2613. skb = skb->next)
  2614. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2615. for (tmp = skb->next; \
  2616. skb != (struct sk_buff *)(queue); \
  2617. skb = tmp, tmp = skb->next)
  2618. #define skb_queue_reverse_walk(queue, skb) \
  2619. for (skb = (queue)->prev; \
  2620. skb != (struct sk_buff *)(queue); \
  2621. skb = skb->prev)
  2622. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2623. for (skb = (queue)->prev, tmp = skb->prev; \
  2624. skb != (struct sk_buff *)(queue); \
  2625. skb = tmp, tmp = skb->prev)
  2626. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2627. for (tmp = skb->prev; \
  2628. skb != (struct sk_buff *)(queue); \
  2629. skb = tmp, tmp = skb->prev)
  2630. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2631. {
  2632. return skb_shinfo(skb)->frag_list != NULL;
  2633. }
  2634. static inline void skb_frag_list_init(struct sk_buff *skb)
  2635. {
  2636. skb_shinfo(skb)->frag_list = NULL;
  2637. }
  2638. #define skb_walk_frags(skb, iter) \
  2639. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2640. int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
  2641. const struct sk_buff *skb);
  2642. struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
  2643. struct sk_buff_head *queue,
  2644. unsigned int flags,
  2645. void (*destructor)(struct sock *sk,
  2646. struct sk_buff *skb),
  2647. int *peeked, int *off, int *err,
  2648. struct sk_buff **last);
  2649. struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
  2650. void (*destructor)(struct sock *sk,
  2651. struct sk_buff *skb),
  2652. int *peeked, int *off, int *err,
  2653. struct sk_buff **last);
  2654. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2655. void (*destructor)(struct sock *sk,
  2656. struct sk_buff *skb),
  2657. int *peeked, int *off, int *err);
  2658. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2659. int *err);
  2660. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2661. struct poll_table_struct *wait);
  2662. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2663. struct iov_iter *to, int size);
  2664. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2665. struct msghdr *msg, int size)
  2666. {
  2667. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2668. }
  2669. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2670. struct msghdr *msg);
  2671. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2672. struct iov_iter *from, int len);
  2673. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2674. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2675. void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
  2676. static inline void skb_free_datagram_locked(struct sock *sk,
  2677. struct sk_buff *skb)
  2678. {
  2679. __skb_free_datagram_locked(sk, skb, 0);
  2680. }
  2681. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2682. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2683. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2684. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2685. int len, __wsum csum);
  2686. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2687. struct pipe_inode_info *pipe, unsigned int len,
  2688. unsigned int flags);
  2689. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2690. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2691. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2692. int len, int hlen);
  2693. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2694. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2695. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2696. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2697. bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
  2698. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2699. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2700. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2701. int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
  2702. int skb_vlan_pop(struct sk_buff *skb);
  2703. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2704. struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
  2705. gfp_t gfp);
  2706. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2707. {
  2708. return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
  2709. }
  2710. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2711. {
  2712. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2713. }
  2714. struct skb_checksum_ops {
  2715. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2716. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2717. };
  2718. extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
  2719. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2720. __wsum csum, const struct skb_checksum_ops *ops);
  2721. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2722. __wsum csum);
  2723. static inline void * __must_check
  2724. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2725. int len, void *data, int hlen, void *buffer)
  2726. {
  2727. if (hlen - offset >= len)
  2728. return data + offset;
  2729. if (!skb ||
  2730. skb_copy_bits(skb, offset, buffer, len) < 0)
  2731. return NULL;
  2732. return buffer;
  2733. }
  2734. static inline void * __must_check
  2735. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2736. {
  2737. return __skb_header_pointer(skb, offset, len, skb->data,
  2738. skb_headlen(skb), buffer);
  2739. }
  2740. /**
  2741. * skb_needs_linearize - check if we need to linearize a given skb
  2742. * depending on the given device features.
  2743. * @skb: socket buffer to check
  2744. * @features: net device features
  2745. *
  2746. * Returns true if either:
  2747. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2748. * 2. skb is fragmented and the device does not support SG.
  2749. */
  2750. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2751. netdev_features_t features)
  2752. {
  2753. return skb_is_nonlinear(skb) &&
  2754. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2755. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2756. }
  2757. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2758. void *to,
  2759. const unsigned int len)
  2760. {
  2761. memcpy(to, skb->data, len);
  2762. }
  2763. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2764. const int offset, void *to,
  2765. const unsigned int len)
  2766. {
  2767. memcpy(to, skb->data + offset, len);
  2768. }
  2769. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2770. const void *from,
  2771. const unsigned int len)
  2772. {
  2773. memcpy(skb->data, from, len);
  2774. }
  2775. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2776. const int offset,
  2777. const void *from,
  2778. const unsigned int len)
  2779. {
  2780. memcpy(skb->data + offset, from, len);
  2781. }
  2782. void skb_init(void);
  2783. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2784. {
  2785. return skb->tstamp;
  2786. }
  2787. /**
  2788. * skb_get_timestamp - get timestamp from a skb
  2789. * @skb: skb to get stamp from
  2790. * @stamp: pointer to struct timeval to store stamp in
  2791. *
  2792. * Timestamps are stored in the skb as offsets to a base timestamp.
  2793. * This function converts the offset back to a struct timeval and stores
  2794. * it in stamp.
  2795. */
  2796. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2797. struct timeval *stamp)
  2798. {
  2799. *stamp = ktime_to_timeval(skb->tstamp);
  2800. }
  2801. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2802. struct timespec *stamp)
  2803. {
  2804. *stamp = ktime_to_timespec(skb->tstamp);
  2805. }
  2806. static inline void __net_timestamp(struct sk_buff *skb)
  2807. {
  2808. skb->tstamp = ktime_get_real();
  2809. }
  2810. static inline ktime_t net_timedelta(ktime_t t)
  2811. {
  2812. return ktime_sub(ktime_get_real(), t);
  2813. }
  2814. static inline ktime_t net_invalid_timestamp(void)
  2815. {
  2816. return 0;
  2817. }
  2818. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2819. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2820. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2821. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2822. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2823. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2824. {
  2825. }
  2826. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2827. {
  2828. return false;
  2829. }
  2830. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2831. /**
  2832. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2833. *
  2834. * PHY drivers may accept clones of transmitted packets for
  2835. * timestamping via their phy_driver.txtstamp method. These drivers
  2836. * must call this function to return the skb back to the stack with a
  2837. * timestamp.
  2838. *
  2839. * @skb: clone of the the original outgoing packet
  2840. * @hwtstamps: hardware time stamps
  2841. *
  2842. */
  2843. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2844. struct skb_shared_hwtstamps *hwtstamps);
  2845. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2846. struct skb_shared_hwtstamps *hwtstamps,
  2847. struct sock *sk, int tstype);
  2848. /**
  2849. * skb_tstamp_tx - queue clone of skb with send time stamps
  2850. * @orig_skb: the original outgoing packet
  2851. * @hwtstamps: hardware time stamps, may be NULL if not available
  2852. *
  2853. * If the skb has a socket associated, then this function clones the
  2854. * skb (thus sharing the actual data and optional structures), stores
  2855. * the optional hardware time stamping information (if non NULL) or
  2856. * generates a software time stamp (otherwise), then queues the clone
  2857. * to the error queue of the socket. Errors are silently ignored.
  2858. */
  2859. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2860. struct skb_shared_hwtstamps *hwtstamps);
  2861. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2862. {
  2863. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2864. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2865. skb_tstamp_tx(skb, NULL);
  2866. }
  2867. /**
  2868. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2869. *
  2870. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2871. * function immediately before giving the sk_buff to the MAC hardware.
  2872. *
  2873. * Specifically, one should make absolutely sure that this function is
  2874. * called before TX completion of this packet can trigger. Otherwise
  2875. * the packet could potentially already be freed.
  2876. *
  2877. * @skb: A socket buffer.
  2878. */
  2879. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2880. {
  2881. skb_clone_tx_timestamp(skb);
  2882. sw_tx_timestamp(skb);
  2883. }
  2884. /**
  2885. * skb_complete_wifi_ack - deliver skb with wifi status
  2886. *
  2887. * @skb: the original outgoing packet
  2888. * @acked: ack status
  2889. *
  2890. */
  2891. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2892. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2893. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2894. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2895. {
  2896. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2897. skb->csum_valid ||
  2898. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2899. skb_checksum_start_offset(skb) >= 0));
  2900. }
  2901. /**
  2902. * skb_checksum_complete - Calculate checksum of an entire packet
  2903. * @skb: packet to process
  2904. *
  2905. * This function calculates the checksum over the entire packet plus
  2906. * the value of skb->csum. The latter can be used to supply the
  2907. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2908. * checksum.
  2909. *
  2910. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2911. * this function can be used to verify that checksum on received
  2912. * packets. In that case the function should return zero if the
  2913. * checksum is correct. In particular, this function will return zero
  2914. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2915. * hardware has already verified the correctness of the checksum.
  2916. */
  2917. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2918. {
  2919. return skb_csum_unnecessary(skb) ?
  2920. 0 : __skb_checksum_complete(skb);
  2921. }
  2922. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2923. {
  2924. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2925. if (skb->csum_level == 0)
  2926. skb->ip_summed = CHECKSUM_NONE;
  2927. else
  2928. skb->csum_level--;
  2929. }
  2930. }
  2931. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2932. {
  2933. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2934. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2935. skb->csum_level++;
  2936. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2937. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2938. skb->csum_level = 0;
  2939. }
  2940. }
  2941. /* Check if we need to perform checksum complete validation.
  2942. *
  2943. * Returns true if checksum complete is needed, false otherwise
  2944. * (either checksum is unnecessary or zero checksum is allowed).
  2945. */
  2946. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2947. bool zero_okay,
  2948. __sum16 check)
  2949. {
  2950. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2951. skb->csum_valid = 1;
  2952. __skb_decr_checksum_unnecessary(skb);
  2953. return false;
  2954. }
  2955. return true;
  2956. }
  2957. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2958. * in checksum_init.
  2959. */
  2960. #define CHECKSUM_BREAK 76
  2961. /* Unset checksum-complete
  2962. *
  2963. * Unset checksum complete can be done when packet is being modified
  2964. * (uncompressed for instance) and checksum-complete value is
  2965. * invalidated.
  2966. */
  2967. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2968. {
  2969. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2970. skb->ip_summed = CHECKSUM_NONE;
  2971. }
  2972. /* Validate (init) checksum based on checksum complete.
  2973. *
  2974. * Return values:
  2975. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2976. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2977. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2978. * non-zero: value of invalid checksum
  2979. *
  2980. */
  2981. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2982. bool complete,
  2983. __wsum psum)
  2984. {
  2985. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2986. if (!csum_fold(csum_add(psum, skb->csum))) {
  2987. skb->csum_valid = 1;
  2988. return 0;
  2989. }
  2990. }
  2991. skb->csum = psum;
  2992. if (complete || skb->len <= CHECKSUM_BREAK) {
  2993. __sum16 csum;
  2994. csum = __skb_checksum_complete(skb);
  2995. skb->csum_valid = !csum;
  2996. return csum;
  2997. }
  2998. return 0;
  2999. }
  3000. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  3001. {
  3002. return 0;
  3003. }
  3004. /* Perform checksum validate (init). Note that this is a macro since we only
  3005. * want to calculate the pseudo header which is an input function if necessary.
  3006. * First we try to validate without any computation (checksum unnecessary) and
  3007. * then calculate based on checksum complete calling the function to compute
  3008. * pseudo header.
  3009. *
  3010. * Return values:
  3011. * 0: checksum is validated or try to in skb_checksum_complete
  3012. * non-zero: value of invalid checksum
  3013. */
  3014. #define __skb_checksum_validate(skb, proto, complete, \
  3015. zero_okay, check, compute_pseudo) \
  3016. ({ \
  3017. __sum16 __ret = 0; \
  3018. skb->csum_valid = 0; \
  3019. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  3020. __ret = __skb_checksum_validate_complete(skb, \
  3021. complete, compute_pseudo(skb, proto)); \
  3022. __ret; \
  3023. })
  3024. #define skb_checksum_init(skb, proto, compute_pseudo) \
  3025. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  3026. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  3027. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  3028. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  3029. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  3030. #define skb_checksum_validate_zero_check(skb, proto, check, \
  3031. compute_pseudo) \
  3032. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  3033. #define skb_checksum_simple_validate(skb) \
  3034. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  3035. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  3036. {
  3037. return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
  3038. }
  3039. static inline void __skb_checksum_convert(struct sk_buff *skb,
  3040. __sum16 check, __wsum pseudo)
  3041. {
  3042. skb->csum = ~pseudo;
  3043. skb->ip_summed = CHECKSUM_COMPLETE;
  3044. }
  3045. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  3046. do { \
  3047. if (__skb_checksum_convert_check(skb)) \
  3048. __skb_checksum_convert(skb, check, \
  3049. compute_pseudo(skb, proto)); \
  3050. } while (0)
  3051. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  3052. u16 start, u16 offset)
  3053. {
  3054. skb->ip_summed = CHECKSUM_PARTIAL;
  3055. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  3056. skb->csum_offset = offset - start;
  3057. }
  3058. /* Update skbuf and packet to reflect the remote checksum offload operation.
  3059. * When called, ptr indicates the starting point for skb->csum when
  3060. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  3061. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  3062. */
  3063. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  3064. int start, int offset, bool nopartial)
  3065. {
  3066. __wsum delta;
  3067. if (!nopartial) {
  3068. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  3069. return;
  3070. }
  3071. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  3072. __skb_checksum_complete(skb);
  3073. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  3074. }
  3075. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  3076. /* Adjust skb->csum since we changed the packet */
  3077. skb->csum = csum_add(skb->csum, delta);
  3078. }
  3079. static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
  3080. {
  3081. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  3082. return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
  3083. #else
  3084. return NULL;
  3085. #endif
  3086. }
  3087. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3088. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  3089. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  3090. {
  3091. if (nfct && atomic_dec_and_test(&nfct->use))
  3092. nf_conntrack_destroy(nfct);
  3093. }
  3094. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  3095. {
  3096. if (nfct)
  3097. atomic_inc(&nfct->use);
  3098. }
  3099. #endif
  3100. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3101. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  3102. {
  3103. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  3104. kfree(nf_bridge);
  3105. }
  3106. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  3107. {
  3108. if (nf_bridge)
  3109. atomic_inc(&nf_bridge->use);
  3110. }
  3111. #endif /* CONFIG_BRIDGE_NETFILTER */
  3112. static inline void nf_reset(struct sk_buff *skb)
  3113. {
  3114. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3115. nf_conntrack_put(skb_nfct(skb));
  3116. skb->_nfct = 0;
  3117. #endif
  3118. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3119. nf_bridge_put(skb->nf_bridge);
  3120. skb->nf_bridge = NULL;
  3121. #endif
  3122. }
  3123. static inline void nf_reset_trace(struct sk_buff *skb)
  3124. {
  3125. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3126. skb->nf_trace = 0;
  3127. #endif
  3128. }
  3129. /* Note: This doesn't put any conntrack and bridge info in dst. */
  3130. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  3131. bool copy)
  3132. {
  3133. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3134. dst->_nfct = src->_nfct;
  3135. nf_conntrack_get(skb_nfct(src));
  3136. #endif
  3137. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3138. dst->nf_bridge = src->nf_bridge;
  3139. nf_bridge_get(src->nf_bridge);
  3140. #endif
  3141. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3142. if (copy)
  3143. dst->nf_trace = src->nf_trace;
  3144. #endif
  3145. }
  3146. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  3147. {
  3148. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3149. nf_conntrack_put(skb_nfct(dst));
  3150. #endif
  3151. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3152. nf_bridge_put(dst->nf_bridge);
  3153. #endif
  3154. __nf_copy(dst, src, true);
  3155. }
  3156. #ifdef CONFIG_NETWORK_SECMARK
  3157. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3158. {
  3159. to->secmark = from->secmark;
  3160. }
  3161. static inline void skb_init_secmark(struct sk_buff *skb)
  3162. {
  3163. skb->secmark = 0;
  3164. }
  3165. #else
  3166. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3167. { }
  3168. static inline void skb_init_secmark(struct sk_buff *skb)
  3169. { }
  3170. #endif
  3171. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  3172. {
  3173. return !skb->destructor &&
  3174. #if IS_ENABLED(CONFIG_XFRM)
  3175. !skb->sp &&
  3176. #endif
  3177. !skb_nfct(skb) &&
  3178. !skb->_skb_refdst &&
  3179. !skb_has_frag_list(skb);
  3180. }
  3181. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  3182. {
  3183. skb->queue_mapping = queue_mapping;
  3184. }
  3185. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  3186. {
  3187. return skb->queue_mapping;
  3188. }
  3189. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  3190. {
  3191. to->queue_mapping = from->queue_mapping;
  3192. }
  3193. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  3194. {
  3195. skb->queue_mapping = rx_queue + 1;
  3196. }
  3197. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  3198. {
  3199. return skb->queue_mapping - 1;
  3200. }
  3201. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  3202. {
  3203. return skb->queue_mapping != 0;
  3204. }
  3205. static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
  3206. {
  3207. skb->dst_pending_confirm = val;
  3208. }
  3209. static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
  3210. {
  3211. return skb->dst_pending_confirm != 0;
  3212. }
  3213. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  3214. {
  3215. #ifdef CONFIG_XFRM
  3216. return skb->sp;
  3217. #else
  3218. return NULL;
  3219. #endif
  3220. }
  3221. /* Keeps track of mac header offset relative to skb->head.
  3222. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  3223. * For non-tunnel skb it points to skb_mac_header() and for
  3224. * tunnel skb it points to outer mac header.
  3225. * Keeps track of level of encapsulation of network headers.
  3226. */
  3227. struct skb_gso_cb {
  3228. union {
  3229. int mac_offset;
  3230. int data_offset;
  3231. };
  3232. int encap_level;
  3233. __wsum csum;
  3234. __u16 csum_start;
  3235. };
  3236. #define SKB_SGO_CB_OFFSET 32
  3237. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
  3238. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  3239. {
  3240. return (skb_mac_header(inner_skb) - inner_skb->head) -
  3241. SKB_GSO_CB(inner_skb)->mac_offset;
  3242. }
  3243. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  3244. {
  3245. int new_headroom, headroom;
  3246. int ret;
  3247. headroom = skb_headroom(skb);
  3248. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  3249. if (ret)
  3250. return ret;
  3251. new_headroom = skb_headroom(skb);
  3252. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  3253. return 0;
  3254. }
  3255. static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
  3256. {
  3257. /* Do not update partial checksums if remote checksum is enabled. */
  3258. if (skb->remcsum_offload)
  3259. return;
  3260. SKB_GSO_CB(skb)->csum = res;
  3261. SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
  3262. }
  3263. /* Compute the checksum for a gso segment. First compute the checksum value
  3264. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  3265. * then add in skb->csum (checksum from csum_start to end of packet).
  3266. * skb->csum and csum_start are then updated to reflect the checksum of the
  3267. * resultant packet starting from the transport header-- the resultant checksum
  3268. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  3269. * header.
  3270. */
  3271. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  3272. {
  3273. unsigned char *csum_start = skb_transport_header(skb);
  3274. int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
  3275. __wsum partial = SKB_GSO_CB(skb)->csum;
  3276. SKB_GSO_CB(skb)->csum = res;
  3277. SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
  3278. return csum_fold(csum_partial(csum_start, plen, partial));
  3279. }
  3280. static inline bool skb_is_gso(const struct sk_buff *skb)
  3281. {
  3282. return skb_shinfo(skb)->gso_size;
  3283. }
  3284. /* Note: Should be called only if skb_is_gso(skb) is true */
  3285. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  3286. {
  3287. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  3288. }
  3289. static inline void skb_gso_reset(struct sk_buff *skb)
  3290. {
  3291. skb_shinfo(skb)->gso_size = 0;
  3292. skb_shinfo(skb)->gso_segs = 0;
  3293. skb_shinfo(skb)->gso_type = 0;
  3294. }
  3295. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  3296. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  3297. {
  3298. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  3299. * wanted then gso_type will be set. */
  3300. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  3301. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  3302. unlikely(shinfo->gso_type == 0)) {
  3303. __skb_warn_lro_forwarding(skb);
  3304. return true;
  3305. }
  3306. return false;
  3307. }
  3308. static inline void skb_forward_csum(struct sk_buff *skb)
  3309. {
  3310. /* Unfortunately we don't support this one. Any brave souls? */
  3311. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3312. skb->ip_summed = CHECKSUM_NONE;
  3313. }
  3314. /**
  3315. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3316. * @skb: skb to check
  3317. *
  3318. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3319. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3320. * use this helper, to document places where we make this assertion.
  3321. */
  3322. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3323. {
  3324. #ifdef DEBUG
  3325. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3326. #endif
  3327. }
  3328. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3329. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3330. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3331. unsigned int transport_len,
  3332. __sum16(*skb_chkf)(struct sk_buff *skb));
  3333. /**
  3334. * skb_head_is_locked - Determine if the skb->head is locked down
  3335. * @skb: skb to check
  3336. *
  3337. * The head on skbs build around a head frag can be removed if they are
  3338. * not cloned. This function returns true if the skb head is locked down
  3339. * due to either being allocated via kmalloc, or by being a clone with
  3340. * multiple references to the head.
  3341. */
  3342. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3343. {
  3344. return !skb->head_frag || skb_cloned(skb);
  3345. }
  3346. /**
  3347. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3348. *
  3349. * @skb: GSO skb
  3350. *
  3351. * skb_gso_network_seglen is used to determine the real size of the
  3352. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3353. *
  3354. * The MAC/L2 header is not accounted for.
  3355. */
  3356. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3357. {
  3358. unsigned int hdr_len = skb_transport_header(skb) -
  3359. skb_network_header(skb);
  3360. return hdr_len + skb_gso_transport_seglen(skb);
  3361. }
  3362. /* Local Checksum Offload.
  3363. * Compute outer checksum based on the assumption that the
  3364. * inner checksum will be offloaded later.
  3365. * See Documentation/networking/checksum-offloads.txt for
  3366. * explanation of how this works.
  3367. * Fill in outer checksum adjustment (e.g. with sum of outer
  3368. * pseudo-header) before calling.
  3369. * Also ensure that inner checksum is in linear data area.
  3370. */
  3371. static inline __wsum lco_csum(struct sk_buff *skb)
  3372. {
  3373. unsigned char *csum_start = skb_checksum_start(skb);
  3374. unsigned char *l4_hdr = skb_transport_header(skb);
  3375. __wsum partial;
  3376. /* Start with complement of inner checksum adjustment */
  3377. partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
  3378. skb->csum_offset));
  3379. /* Add in checksum of our headers (incl. outer checksum
  3380. * adjustment filled in by caller) and return result.
  3381. */
  3382. return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
  3383. }
  3384. #endif /* __KERNEL__ */
  3385. #endif /* _LINUX_SKBUFF_H */