skbuff.h 110 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472347334743475347634773478347934803481348234833484348534863487348834893490349134923493349434953496349734983499350035013502350335043505350635073508350935103511351235133514351535163517351835193520352135223523352435253526352735283529353035313532353335343535353635373538353935403541354235433544354535463547354835493550355135523553355435553556355735583559356035613562356335643565356635673568356935703571357235733574357535763577357835793580358135823583358435853586358735883589359035913592359335943595359635973598359936003601360236033604360536063607360836093610361136123613361436153616361736183619362036213622362336243625362636273628362936303631363236333634363536363637363836393640364136423643364436453646364736483649365036513652365336543655365636573658365936603661366236633664366536663667366836693670367136723673367436753676367736783679368036813682368336843685368636873688368936903691369236933694369536963697369836993700370137023703370437053706370737083709371037113712371337143715371637173718371937203721372237233724372537263727372837293730373137323733373437353736373737383739374037413742374337443745374637473748374937503751375237533754375537563757375837593760376137623763376437653766376737683769377037713772377337743775377637773778377937803781378237833784378537863787378837893790379137923793379437953796379737983799380038013802380338043805380638073808380938103811381238133814381538163817381838193820382138223823382438253826382738283829383038313832383338343835383638373838383938403841384238433844384538463847384838493850
  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.
  162. * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
  163. * on network device checksumming capabilities: if a packet does not match
  164. * them, skb_checksum_help or skb_crc32c_help (depending on the value of
  165. * csum_not_inet, see item D.) is called to resolve the checksum.
  166. *
  167. * CHECKSUM_NONE:
  168. *
  169. * The skb was already checksummed by the protocol, or a checksum is not
  170. * required.
  171. *
  172. * CHECKSUM_UNNECESSARY:
  173. *
  174. * This has the same meaning on as CHECKSUM_NONE for checksum offload on
  175. * output.
  176. *
  177. * CHECKSUM_COMPLETE:
  178. * Not used in checksum output. If a driver observes a packet with this value
  179. * set in skbuff, if should treat as CHECKSUM_NONE being set.
  180. *
  181. * D. Non-IP checksum (CRC) offloads
  182. *
  183. * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
  184. * offloading the SCTP CRC in a packet. To perform this offload the stack
  185. * will set set csum_start and csum_offset accordingly, set ip_summed to
  186. * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
  187. * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
  188. * A driver that supports both IP checksum offload and SCTP CRC32c offload
  189. * must verify which offload is configured for a packet by testing the
  190. * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
  191. * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
  192. *
  193. * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
  194. * offloading the FCOE CRC in a packet. To perform this offload the stack
  195. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  196. * accordingly. Note the there is no indication in the skbuff that the
  197. * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
  198. * both IP checksum offload and FCOE CRC offload must verify which offload
  199. * is configured for a packet presumably by inspecting packet headers.
  200. *
  201. * E. Checksumming on output with GSO.
  202. *
  203. * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
  204. * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
  205. * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
  206. * part of the GSO operation is implied. If a checksum is being offloaded
  207. * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
  208. * are set to refer to the outermost checksum being offload (two offloaded
  209. * checksums are possible with UDP encapsulation).
  210. */
  211. /* Don't change this without changing skb_csum_unnecessary! */
  212. #define CHECKSUM_NONE 0
  213. #define CHECKSUM_UNNECESSARY 1
  214. #define CHECKSUM_COMPLETE 2
  215. #define CHECKSUM_PARTIAL 3
  216. /* Maximum value in skb->csum_level */
  217. #define SKB_MAX_CSUM_LEVEL 3
  218. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  219. #define SKB_WITH_OVERHEAD(X) \
  220. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  221. #define SKB_MAX_ORDER(X, ORDER) \
  222. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  223. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  224. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  225. /* return minimum truesize of one skb containing X bytes of data */
  226. #define SKB_TRUESIZE(X) ((X) + \
  227. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  228. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  229. struct net_device;
  230. struct scatterlist;
  231. struct pipe_inode_info;
  232. struct iov_iter;
  233. struct napi_struct;
  234. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  235. struct nf_conntrack {
  236. atomic_t use;
  237. };
  238. #endif
  239. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  240. struct nf_bridge_info {
  241. atomic_t use;
  242. enum {
  243. BRNF_PROTO_UNCHANGED,
  244. BRNF_PROTO_8021Q,
  245. BRNF_PROTO_PPPOE
  246. } orig_proto:8;
  247. u8 pkt_otherhost:1;
  248. u8 in_prerouting:1;
  249. u8 bridged_dnat:1;
  250. __u16 frag_max_size;
  251. struct net_device *physindev;
  252. /* always valid & non-NULL from FORWARD on, for physdev match */
  253. struct net_device *physoutdev;
  254. union {
  255. /* prerouting: detect dnat in orig/reply direction */
  256. __be32 ipv4_daddr;
  257. struct in6_addr ipv6_daddr;
  258. /* after prerouting + nat detected: store original source
  259. * mac since neigh resolution overwrites it, only used while
  260. * skb is out in neigh layer.
  261. */
  262. char neigh_header[8];
  263. };
  264. };
  265. #endif
  266. struct sk_buff_head {
  267. /* These two members must be first. */
  268. struct sk_buff *next;
  269. struct sk_buff *prev;
  270. __u32 qlen;
  271. spinlock_t lock;
  272. };
  273. struct sk_buff;
  274. /* To allow 64K frame to be packed as single skb without frag_list we
  275. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  276. * buffers which do not start on a page boundary.
  277. *
  278. * Since GRO uses frags we allocate at least 16 regardless of page
  279. * size.
  280. */
  281. #if (65536/PAGE_SIZE + 1) < 16
  282. #define MAX_SKB_FRAGS 16UL
  283. #else
  284. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  285. #endif
  286. extern int sysctl_max_skb_frags;
  287. /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
  288. * segment using its current segmentation instead.
  289. */
  290. #define GSO_BY_FRAGS 0xFFFF
  291. typedef struct skb_frag_struct skb_frag_t;
  292. struct skb_frag_struct {
  293. struct {
  294. struct page *p;
  295. } page;
  296. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  297. __u32 page_offset;
  298. __u32 size;
  299. #else
  300. __u16 page_offset;
  301. __u16 size;
  302. #endif
  303. };
  304. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  305. {
  306. return frag->size;
  307. }
  308. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  309. {
  310. frag->size = size;
  311. }
  312. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  313. {
  314. frag->size += delta;
  315. }
  316. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  317. {
  318. frag->size -= delta;
  319. }
  320. #define HAVE_HW_TIME_STAMP
  321. /**
  322. * struct skb_shared_hwtstamps - hardware time stamps
  323. * @hwtstamp: hardware time stamp transformed into duration
  324. * since arbitrary point in time
  325. *
  326. * Software time stamps generated by ktime_get_real() are stored in
  327. * skb->tstamp.
  328. *
  329. * hwtstamps can only be compared against other hwtstamps from
  330. * the same device.
  331. *
  332. * This structure is attached to packets as part of the
  333. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  334. */
  335. struct skb_shared_hwtstamps {
  336. ktime_t hwtstamp;
  337. };
  338. /* Definitions for tx_flags in struct skb_shared_info */
  339. enum {
  340. /* generate hardware time stamp */
  341. SKBTX_HW_TSTAMP = 1 << 0,
  342. /* generate software time stamp when queueing packet to NIC */
  343. SKBTX_SW_TSTAMP = 1 << 1,
  344. /* device driver is going to provide hardware time stamp */
  345. SKBTX_IN_PROGRESS = 1 << 2,
  346. /* device driver supports TX zero-copy buffers */
  347. SKBTX_DEV_ZEROCOPY = 1 << 3,
  348. /* generate wifi status information (where possible) */
  349. SKBTX_WIFI_STATUS = 1 << 4,
  350. /* This indicates at least one fragment might be overwritten
  351. * (as in vmsplice(), sendfile() ...)
  352. * If we need to compute a TX checksum, we'll need to copy
  353. * all frags to avoid possible bad checksum
  354. */
  355. SKBTX_SHARED_FRAG = 1 << 5,
  356. /* generate software time stamp when entering packet scheduling */
  357. SKBTX_SCHED_TSTAMP = 1 << 6,
  358. };
  359. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  360. SKBTX_SCHED_TSTAMP)
  361. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  362. /*
  363. * The callback notifies userspace to release buffers when skb DMA is done in
  364. * lower device, the skb last reference should be 0 when calling this.
  365. * The zerocopy_success argument is true if zero copy transmit occurred,
  366. * false on data copy or out of memory error caused by data copy attempt.
  367. * The ctx field is used to track device context.
  368. * The desc field is used to track userspace buffer index.
  369. */
  370. struct ubuf_info {
  371. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  372. void *ctx;
  373. unsigned long desc;
  374. };
  375. /* This data is invariant across clones and lives at
  376. * the end of the header data, ie. at skb->end.
  377. */
  378. struct skb_shared_info {
  379. unsigned short _unused;
  380. unsigned char nr_frags;
  381. __u8 tx_flags;
  382. unsigned short gso_size;
  383. /* Warning: this field is not always filled in (UFO)! */
  384. unsigned short gso_segs;
  385. struct sk_buff *frag_list;
  386. struct skb_shared_hwtstamps hwtstamps;
  387. unsigned int gso_type;
  388. u32 tskey;
  389. __be32 ip6_frag_id;
  390. /*
  391. * Warning : all fields before dataref are cleared in __alloc_skb()
  392. */
  393. atomic_t dataref;
  394. /* Intermediate layers must ensure that destructor_arg
  395. * remains valid until skb destructor */
  396. void * destructor_arg;
  397. /* must be last field, see pskb_expand_head() */
  398. skb_frag_t frags[MAX_SKB_FRAGS];
  399. };
  400. /* We divide dataref into two halves. The higher 16 bits hold references
  401. * to the payload part of skb->data. The lower 16 bits hold references to
  402. * the entire skb->data. A clone of a headerless skb holds the length of
  403. * the header in skb->hdr_len.
  404. *
  405. * All users must obey the rule that the skb->data reference count must be
  406. * greater than or equal to the payload reference count.
  407. *
  408. * Holding a reference to the payload part means that the user does not
  409. * care about modifications to the header part of skb->data.
  410. */
  411. #define SKB_DATAREF_SHIFT 16
  412. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  413. enum {
  414. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  415. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  416. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  417. };
  418. enum {
  419. SKB_GSO_TCPV4 = 1 << 0,
  420. SKB_GSO_UDP = 1 << 1,
  421. /* This indicates the skb is from an untrusted source. */
  422. SKB_GSO_DODGY = 1 << 2,
  423. /* This indicates the tcp segment has CWR set. */
  424. SKB_GSO_TCP_ECN = 1 << 3,
  425. SKB_GSO_TCP_FIXEDID = 1 << 4,
  426. SKB_GSO_TCPV6 = 1 << 5,
  427. SKB_GSO_FCOE = 1 << 6,
  428. SKB_GSO_GRE = 1 << 7,
  429. SKB_GSO_GRE_CSUM = 1 << 8,
  430. SKB_GSO_IPXIP4 = 1 << 9,
  431. SKB_GSO_IPXIP6 = 1 << 10,
  432. SKB_GSO_UDP_TUNNEL = 1 << 11,
  433. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
  434. SKB_GSO_PARTIAL = 1 << 13,
  435. SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
  436. SKB_GSO_SCTP = 1 << 15,
  437. SKB_GSO_ESP = 1 << 16,
  438. };
  439. #if BITS_PER_LONG > 32
  440. #define NET_SKBUFF_DATA_USES_OFFSET 1
  441. #endif
  442. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  443. typedef unsigned int sk_buff_data_t;
  444. #else
  445. typedef unsigned char *sk_buff_data_t;
  446. #endif
  447. /**
  448. * struct sk_buff - socket buffer
  449. * @next: Next buffer in list
  450. * @prev: Previous buffer in list
  451. * @tstamp: Time we arrived/left
  452. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  453. * @sk: Socket we are owned by
  454. * @dev: Device we arrived on/are leaving by
  455. * @cb: Control buffer. Free for use by every layer. Put private vars here
  456. * @_skb_refdst: destination entry (with norefcount bit)
  457. * @sp: the security path, used for xfrm
  458. * @len: Length of actual data
  459. * @data_len: Data length
  460. * @mac_len: Length of link layer header
  461. * @hdr_len: writable header length of cloned skb
  462. * @csum: Checksum (must include start/offset pair)
  463. * @csum_start: Offset from skb->head where checksumming should start
  464. * @csum_offset: Offset from csum_start where checksum should be stored
  465. * @priority: Packet queueing priority
  466. * @ignore_df: allow local fragmentation
  467. * @cloned: Head may be cloned (check refcnt to be sure)
  468. * @ip_summed: Driver fed us an IP checksum
  469. * @nohdr: Payload reference only, must not modify header
  470. * @pkt_type: Packet class
  471. * @fclone: skbuff clone status
  472. * @ipvs_property: skbuff is owned by ipvs
  473. * @tc_skip_classify: do not classify packet. set by IFB device
  474. * @tc_at_ingress: used within tc_classify to distinguish in/egress
  475. * @tc_redirected: packet was redirected by a tc action
  476. * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
  477. * @peeked: this packet has been seen already, so stats have been
  478. * done for it, don't do them again
  479. * @nf_trace: netfilter packet trace flag
  480. * @protocol: Packet protocol from driver
  481. * @destructor: Destruct function
  482. * @_nfct: Associated connection, if any (with nfctinfo bits)
  483. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  484. * @skb_iif: ifindex of device we arrived on
  485. * @tc_index: Traffic control index
  486. * @hash: the packet hash
  487. * @queue_mapping: Queue mapping for multiqueue devices
  488. * @xmit_more: More SKBs are pending for this queue
  489. * @ndisc_nodetype: router type (from link layer)
  490. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  491. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  492. * ports.
  493. * @sw_hash: indicates hash was computed in software stack
  494. * @wifi_acked_valid: wifi_acked was set
  495. * @wifi_acked: whether frame was acked on wifi or not
  496. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  497. * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
  498. * @dst_pending_confirm: need to confirm neighbour
  499. * @napi_id: id of the NAPI struct this skb came from
  500. * @secmark: security marking
  501. * @mark: Generic packet mark
  502. * @vlan_proto: vlan encapsulation protocol
  503. * @vlan_tci: vlan tag control information
  504. * @inner_protocol: Protocol (encapsulation)
  505. * @inner_transport_header: Inner transport layer header (encapsulation)
  506. * @inner_network_header: Network layer header (encapsulation)
  507. * @inner_mac_header: Link layer header (encapsulation)
  508. * @transport_header: Transport layer header
  509. * @network_header: Network layer header
  510. * @mac_header: Link layer header
  511. * @tail: Tail pointer
  512. * @end: End pointer
  513. * @head: Head of buffer
  514. * @data: Data head pointer
  515. * @truesize: Buffer size
  516. * @users: User count - see {datagram,tcp}.c
  517. */
  518. struct sk_buff {
  519. union {
  520. struct {
  521. /* These two members must be first. */
  522. struct sk_buff *next;
  523. struct sk_buff *prev;
  524. union {
  525. ktime_t tstamp;
  526. u64 skb_mstamp;
  527. };
  528. };
  529. struct rb_node rbnode; /* used in netem & tcp stack */
  530. };
  531. struct sock *sk;
  532. union {
  533. struct net_device *dev;
  534. /* Some protocols might use this space to store information,
  535. * while device pointer would be NULL.
  536. * UDP receive path is one user.
  537. */
  538. unsigned long dev_scratch;
  539. };
  540. /*
  541. * This is the control buffer. It is free to use for every
  542. * layer. Please put your private variables there. If you
  543. * want to keep them across layers you have to do a skb_clone()
  544. * first. This is owned by whoever has the skb queued ATM.
  545. */
  546. char cb[48] __aligned(8);
  547. unsigned long _skb_refdst;
  548. void (*destructor)(struct sk_buff *skb);
  549. #ifdef CONFIG_XFRM
  550. struct sec_path *sp;
  551. #endif
  552. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  553. unsigned long _nfct;
  554. #endif
  555. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  556. struct nf_bridge_info *nf_bridge;
  557. #endif
  558. unsigned int len,
  559. data_len;
  560. __u16 mac_len,
  561. hdr_len;
  562. /* Following fields are _not_ copied in __copy_skb_header()
  563. * Note that queue_mapping is here mostly to fill a hole.
  564. */
  565. kmemcheck_bitfield_begin(flags1);
  566. __u16 queue_mapping;
  567. /* if you move cloned around you also must adapt those constants */
  568. #ifdef __BIG_ENDIAN_BITFIELD
  569. #define CLONED_MASK (1 << 7)
  570. #else
  571. #define CLONED_MASK 1
  572. #endif
  573. #define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
  574. __u8 __cloned_offset[0];
  575. __u8 cloned:1,
  576. nohdr:1,
  577. fclone:2,
  578. peeked:1,
  579. head_frag:1,
  580. xmit_more:1,
  581. __unused:1; /* one bit hole */
  582. kmemcheck_bitfield_end(flags1);
  583. /* fields enclosed in headers_start/headers_end are copied
  584. * using a single memcpy() in __copy_skb_header()
  585. */
  586. /* private: */
  587. __u32 headers_start[0];
  588. /* public: */
  589. /* if you move pkt_type around you also must adapt those constants */
  590. #ifdef __BIG_ENDIAN_BITFIELD
  591. #define PKT_TYPE_MAX (7 << 5)
  592. #else
  593. #define PKT_TYPE_MAX 7
  594. #endif
  595. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  596. __u8 __pkt_type_offset[0];
  597. __u8 pkt_type:3;
  598. __u8 pfmemalloc:1;
  599. __u8 ignore_df:1;
  600. __u8 nf_trace:1;
  601. __u8 ip_summed:2;
  602. __u8 ooo_okay:1;
  603. __u8 l4_hash:1;
  604. __u8 sw_hash:1;
  605. __u8 wifi_acked_valid:1;
  606. __u8 wifi_acked:1;
  607. __u8 no_fcs:1;
  608. /* Indicates the inner headers are valid in the skbuff. */
  609. __u8 encapsulation:1;
  610. __u8 encap_hdr_csum:1;
  611. __u8 csum_valid:1;
  612. __u8 csum_complete_sw:1;
  613. __u8 csum_level:2;
  614. __u8 csum_not_inet:1;
  615. __u8 dst_pending_confirm:1;
  616. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  617. __u8 ndisc_nodetype:2;
  618. #endif
  619. __u8 ipvs_property:1;
  620. __u8 inner_protocol_type:1;
  621. __u8 remcsum_offload:1;
  622. #ifdef CONFIG_NET_SWITCHDEV
  623. __u8 offload_fwd_mark:1;
  624. #endif
  625. #ifdef CONFIG_NET_CLS_ACT
  626. __u8 tc_skip_classify:1;
  627. __u8 tc_at_ingress:1;
  628. __u8 tc_redirected:1;
  629. __u8 tc_from_ingress:1;
  630. #endif
  631. #ifdef CONFIG_NET_SCHED
  632. __u16 tc_index; /* traffic control index */
  633. #endif
  634. union {
  635. __wsum csum;
  636. struct {
  637. __u16 csum_start;
  638. __u16 csum_offset;
  639. };
  640. };
  641. __u32 priority;
  642. int skb_iif;
  643. __u32 hash;
  644. __be16 vlan_proto;
  645. __u16 vlan_tci;
  646. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  647. union {
  648. unsigned int napi_id;
  649. unsigned int sender_cpu;
  650. };
  651. #endif
  652. #ifdef CONFIG_NETWORK_SECMARK
  653. __u32 secmark;
  654. #endif
  655. union {
  656. __u32 mark;
  657. __u32 reserved_tailroom;
  658. };
  659. union {
  660. __be16 inner_protocol;
  661. __u8 inner_ipproto;
  662. };
  663. __u16 inner_transport_header;
  664. __u16 inner_network_header;
  665. __u16 inner_mac_header;
  666. __be16 protocol;
  667. __u16 transport_header;
  668. __u16 network_header;
  669. __u16 mac_header;
  670. /* private: */
  671. __u32 headers_end[0];
  672. /* public: */
  673. /* These elements must be at the end, see alloc_skb() for details. */
  674. sk_buff_data_t tail;
  675. sk_buff_data_t end;
  676. unsigned char *head,
  677. *data;
  678. unsigned int truesize;
  679. atomic_t users;
  680. };
  681. #ifdef __KERNEL__
  682. /*
  683. * Handling routines are only of interest to the kernel
  684. */
  685. #include <linux/slab.h>
  686. #define SKB_ALLOC_FCLONE 0x01
  687. #define SKB_ALLOC_RX 0x02
  688. #define SKB_ALLOC_NAPI 0x04
  689. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  690. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  691. {
  692. return unlikely(skb->pfmemalloc);
  693. }
  694. /*
  695. * skb might have a dst pointer attached, refcounted or not.
  696. * _skb_refdst low order bit is set if refcount was _not_ taken
  697. */
  698. #define SKB_DST_NOREF 1UL
  699. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  700. #define SKB_NFCT_PTRMASK ~(7UL)
  701. /**
  702. * skb_dst - returns skb dst_entry
  703. * @skb: buffer
  704. *
  705. * Returns skb dst_entry, regardless of reference taken or not.
  706. */
  707. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  708. {
  709. /* If refdst was not refcounted, check we still are in a
  710. * rcu_read_lock section
  711. */
  712. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  713. !rcu_read_lock_held() &&
  714. !rcu_read_lock_bh_held());
  715. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  716. }
  717. /**
  718. * skb_dst_set - sets skb dst
  719. * @skb: buffer
  720. * @dst: dst entry
  721. *
  722. * Sets skb dst, assuming a reference was taken on dst and should
  723. * be released by skb_dst_drop()
  724. */
  725. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  726. {
  727. skb->_skb_refdst = (unsigned long)dst;
  728. }
  729. /**
  730. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  731. * @skb: buffer
  732. * @dst: dst entry
  733. *
  734. * Sets skb dst, assuming a reference was not taken on dst.
  735. * If dst entry is cached, we do not take reference and dst_release
  736. * will be avoided by refdst_drop. If dst entry is not cached, we take
  737. * reference, so that last dst_release can destroy the dst immediately.
  738. */
  739. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  740. {
  741. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  742. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  743. }
  744. /**
  745. * skb_dst_is_noref - Test if skb dst isn't refcounted
  746. * @skb: buffer
  747. */
  748. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  749. {
  750. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  751. }
  752. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  753. {
  754. return (struct rtable *)skb_dst(skb);
  755. }
  756. /* For mangling skb->pkt_type from user space side from applications
  757. * such as nft, tc, etc, we only allow a conservative subset of
  758. * possible pkt_types to be set.
  759. */
  760. static inline bool skb_pkt_type_ok(u32 ptype)
  761. {
  762. return ptype <= PACKET_OTHERHOST;
  763. }
  764. void kfree_skb(struct sk_buff *skb);
  765. void kfree_skb_list(struct sk_buff *segs);
  766. void skb_tx_error(struct sk_buff *skb);
  767. void consume_skb(struct sk_buff *skb);
  768. void __kfree_skb(struct sk_buff *skb);
  769. extern struct kmem_cache *skbuff_head_cache;
  770. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  771. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  772. bool *fragstolen, int *delta_truesize);
  773. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  774. int node);
  775. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  776. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  777. static inline struct sk_buff *alloc_skb(unsigned int size,
  778. gfp_t priority)
  779. {
  780. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  781. }
  782. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  783. unsigned long data_len,
  784. int max_page_order,
  785. int *errcode,
  786. gfp_t gfp_mask);
  787. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  788. struct sk_buff_fclones {
  789. struct sk_buff skb1;
  790. struct sk_buff skb2;
  791. atomic_t fclone_ref;
  792. };
  793. /**
  794. * skb_fclone_busy - check if fclone is busy
  795. * @sk: socket
  796. * @skb: buffer
  797. *
  798. * Returns true if skb is a fast clone, and its clone is not freed.
  799. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  800. * so we also check that this didnt happen.
  801. */
  802. static inline bool skb_fclone_busy(const struct sock *sk,
  803. const struct sk_buff *skb)
  804. {
  805. const struct sk_buff_fclones *fclones;
  806. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  807. return skb->fclone == SKB_FCLONE_ORIG &&
  808. atomic_read(&fclones->fclone_ref) > 1 &&
  809. fclones->skb2.sk == sk;
  810. }
  811. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  812. gfp_t priority)
  813. {
  814. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  815. }
  816. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  817. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  818. {
  819. return __alloc_skb_head(priority, -1);
  820. }
  821. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  822. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  823. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  824. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  825. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  826. gfp_t gfp_mask, bool fclone);
  827. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  828. gfp_t gfp_mask)
  829. {
  830. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  831. }
  832. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  833. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  834. unsigned int headroom);
  835. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  836. int newtailroom, gfp_t priority);
  837. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  838. int offset, int len);
  839. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  840. int len);
  841. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  842. int skb_pad(struct sk_buff *skb, int pad);
  843. #define dev_kfree_skb(a) consume_skb(a)
  844. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  845. int getfrag(void *from, char *to, int offset,
  846. int len, int odd, struct sk_buff *skb),
  847. void *from, int length);
  848. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  849. int offset, size_t size);
  850. struct skb_seq_state {
  851. __u32 lower_offset;
  852. __u32 upper_offset;
  853. __u32 frag_idx;
  854. __u32 stepped_offset;
  855. struct sk_buff *root_skb;
  856. struct sk_buff *cur_skb;
  857. __u8 *frag_data;
  858. };
  859. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  860. unsigned int to, struct skb_seq_state *st);
  861. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  862. struct skb_seq_state *st);
  863. void skb_abort_seq_read(struct skb_seq_state *st);
  864. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  865. unsigned int to, struct ts_config *config);
  866. /*
  867. * Packet hash types specify the type of hash in skb_set_hash.
  868. *
  869. * Hash types refer to the protocol layer addresses which are used to
  870. * construct a packet's hash. The hashes are used to differentiate or identify
  871. * flows of the protocol layer for the hash type. Hash types are either
  872. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  873. *
  874. * Properties of hashes:
  875. *
  876. * 1) Two packets in different flows have different hash values
  877. * 2) Two packets in the same flow should have the same hash value
  878. *
  879. * A hash at a higher layer is considered to be more specific. A driver should
  880. * set the most specific hash possible.
  881. *
  882. * A driver cannot indicate a more specific hash than the layer at which a hash
  883. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  884. *
  885. * A driver may indicate a hash level which is less specific than the
  886. * actual layer the hash was computed on. For instance, a hash computed
  887. * at L4 may be considered an L3 hash. This should only be done if the
  888. * driver can't unambiguously determine that the HW computed the hash at
  889. * the higher layer. Note that the "should" in the second property above
  890. * permits this.
  891. */
  892. enum pkt_hash_types {
  893. PKT_HASH_TYPE_NONE, /* Undefined type */
  894. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  895. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  896. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  897. };
  898. static inline void skb_clear_hash(struct sk_buff *skb)
  899. {
  900. skb->hash = 0;
  901. skb->sw_hash = 0;
  902. skb->l4_hash = 0;
  903. }
  904. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  905. {
  906. if (!skb->l4_hash)
  907. skb_clear_hash(skb);
  908. }
  909. static inline void
  910. __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
  911. {
  912. skb->l4_hash = is_l4;
  913. skb->sw_hash = is_sw;
  914. skb->hash = hash;
  915. }
  916. static inline void
  917. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  918. {
  919. /* Used by drivers to set hash from HW */
  920. __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
  921. }
  922. static inline void
  923. __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
  924. {
  925. __skb_set_hash(skb, hash, true, is_l4);
  926. }
  927. void __skb_get_hash(struct sk_buff *skb);
  928. u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
  929. u32 skb_get_poff(const struct sk_buff *skb);
  930. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  931. const struct flow_keys *keys, int hlen);
  932. __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
  933. void *data, int hlen_proto);
  934. static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
  935. int thoff, u8 ip_proto)
  936. {
  937. return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
  938. }
  939. void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  940. const struct flow_dissector_key *key,
  941. unsigned int key_count);
  942. bool __skb_flow_dissect(const struct sk_buff *skb,
  943. struct flow_dissector *flow_dissector,
  944. void *target_container,
  945. void *data, __be16 proto, int nhoff, int hlen,
  946. unsigned int flags);
  947. static inline bool skb_flow_dissect(const struct sk_buff *skb,
  948. struct flow_dissector *flow_dissector,
  949. void *target_container, unsigned int flags)
  950. {
  951. return __skb_flow_dissect(skb, flow_dissector, target_container,
  952. NULL, 0, 0, 0, flags);
  953. }
  954. static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
  955. struct flow_keys *flow,
  956. unsigned int flags)
  957. {
  958. memset(flow, 0, sizeof(*flow));
  959. return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
  960. NULL, 0, 0, 0, flags);
  961. }
  962. static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
  963. void *data, __be16 proto,
  964. int nhoff, int hlen,
  965. unsigned int flags)
  966. {
  967. memset(flow, 0, sizeof(*flow));
  968. return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
  969. data, proto, nhoff, hlen, flags);
  970. }
  971. static inline __u32 skb_get_hash(struct sk_buff *skb)
  972. {
  973. if (!skb->l4_hash && !skb->sw_hash)
  974. __skb_get_hash(skb);
  975. return skb->hash;
  976. }
  977. __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
  978. static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
  979. {
  980. if (!skb->l4_hash && !skb->sw_hash) {
  981. struct flow_keys keys;
  982. __u32 hash = __get_hash_from_flowi6(fl6, &keys);
  983. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  984. }
  985. return skb->hash;
  986. }
  987. __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
  988. static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
  989. {
  990. if (!skb->l4_hash && !skb->sw_hash) {
  991. struct flow_keys keys;
  992. __u32 hash = __get_hash_from_flowi4(fl4, &keys);
  993. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  994. }
  995. return skb->hash;
  996. }
  997. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  998. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  999. {
  1000. return skb->hash;
  1001. }
  1002. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  1003. {
  1004. to->hash = from->hash;
  1005. to->sw_hash = from->sw_hash;
  1006. to->l4_hash = from->l4_hash;
  1007. };
  1008. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1009. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1010. {
  1011. return skb->head + skb->end;
  1012. }
  1013. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1014. {
  1015. return skb->end;
  1016. }
  1017. #else
  1018. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1019. {
  1020. return skb->end;
  1021. }
  1022. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1023. {
  1024. return skb->end - skb->head;
  1025. }
  1026. #endif
  1027. /* Internal */
  1028. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  1029. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  1030. {
  1031. return &skb_shinfo(skb)->hwtstamps;
  1032. }
  1033. /**
  1034. * skb_queue_empty - check if a queue is empty
  1035. * @list: queue head
  1036. *
  1037. * Returns true if the queue is empty, false otherwise.
  1038. */
  1039. static inline int skb_queue_empty(const struct sk_buff_head *list)
  1040. {
  1041. return list->next == (const struct sk_buff *) list;
  1042. }
  1043. /**
  1044. * skb_queue_is_last - check if skb is the last entry in the queue
  1045. * @list: queue head
  1046. * @skb: buffer
  1047. *
  1048. * Returns true if @skb is the last buffer on the list.
  1049. */
  1050. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  1051. const struct sk_buff *skb)
  1052. {
  1053. return skb->next == (const struct sk_buff *) list;
  1054. }
  1055. /**
  1056. * skb_queue_is_first - check if skb is the first entry in the queue
  1057. * @list: queue head
  1058. * @skb: buffer
  1059. *
  1060. * Returns true if @skb is the first buffer on the list.
  1061. */
  1062. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  1063. const struct sk_buff *skb)
  1064. {
  1065. return skb->prev == (const struct sk_buff *) list;
  1066. }
  1067. /**
  1068. * skb_queue_next - return the next packet in the queue
  1069. * @list: queue head
  1070. * @skb: current buffer
  1071. *
  1072. * Return the next packet in @list after @skb. It is only valid to
  1073. * call this if skb_queue_is_last() evaluates to false.
  1074. */
  1075. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  1076. const struct sk_buff *skb)
  1077. {
  1078. /* This BUG_ON may seem severe, but if we just return then we
  1079. * are going to dereference garbage.
  1080. */
  1081. BUG_ON(skb_queue_is_last(list, skb));
  1082. return skb->next;
  1083. }
  1084. /**
  1085. * skb_queue_prev - return the prev packet in the queue
  1086. * @list: queue head
  1087. * @skb: current buffer
  1088. *
  1089. * Return the prev packet in @list before @skb. It is only valid to
  1090. * call this if skb_queue_is_first() evaluates to false.
  1091. */
  1092. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  1093. const struct sk_buff *skb)
  1094. {
  1095. /* This BUG_ON may seem severe, but if we just return then we
  1096. * are going to dereference garbage.
  1097. */
  1098. BUG_ON(skb_queue_is_first(list, skb));
  1099. return skb->prev;
  1100. }
  1101. /**
  1102. * skb_get - reference buffer
  1103. * @skb: buffer to reference
  1104. *
  1105. * Makes another reference to a socket buffer and returns a pointer
  1106. * to the buffer.
  1107. */
  1108. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  1109. {
  1110. atomic_inc(&skb->users);
  1111. return skb;
  1112. }
  1113. /*
  1114. * If users == 1, we are the only owner and are can avoid redundant
  1115. * atomic change.
  1116. */
  1117. /**
  1118. * skb_cloned - is the buffer a clone
  1119. * @skb: buffer to check
  1120. *
  1121. * Returns true if the buffer was generated with skb_clone() and is
  1122. * one of multiple shared copies of the buffer. Cloned buffers are
  1123. * shared data so must not be written to under normal circumstances.
  1124. */
  1125. static inline int skb_cloned(const struct sk_buff *skb)
  1126. {
  1127. return skb->cloned &&
  1128. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  1129. }
  1130. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  1131. {
  1132. might_sleep_if(gfpflags_allow_blocking(pri));
  1133. if (skb_cloned(skb))
  1134. return pskb_expand_head(skb, 0, 0, pri);
  1135. return 0;
  1136. }
  1137. /**
  1138. * skb_header_cloned - is the header a clone
  1139. * @skb: buffer to check
  1140. *
  1141. * Returns true if modifying the header part of the buffer requires
  1142. * the data to be copied.
  1143. */
  1144. static inline int skb_header_cloned(const struct sk_buff *skb)
  1145. {
  1146. int dataref;
  1147. if (!skb->cloned)
  1148. return 0;
  1149. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  1150. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  1151. return dataref != 1;
  1152. }
  1153. static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
  1154. {
  1155. might_sleep_if(gfpflags_allow_blocking(pri));
  1156. if (skb_header_cloned(skb))
  1157. return pskb_expand_head(skb, 0, 0, pri);
  1158. return 0;
  1159. }
  1160. /**
  1161. * skb_header_release - release reference to header
  1162. * @skb: buffer to operate on
  1163. *
  1164. * Drop a reference to the header part of the buffer. This is done
  1165. * by acquiring a payload reference. You must not read from the header
  1166. * part of skb->data after this.
  1167. * Note : Check if you can use __skb_header_release() instead.
  1168. */
  1169. static inline void skb_header_release(struct sk_buff *skb)
  1170. {
  1171. BUG_ON(skb->nohdr);
  1172. skb->nohdr = 1;
  1173. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1174. }
  1175. /**
  1176. * __skb_header_release - release reference to header
  1177. * @skb: buffer to operate on
  1178. *
  1179. * Variant of skb_header_release() assuming skb is private to caller.
  1180. * We can avoid one atomic operation.
  1181. */
  1182. static inline void __skb_header_release(struct sk_buff *skb)
  1183. {
  1184. skb->nohdr = 1;
  1185. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1186. }
  1187. /**
  1188. * skb_shared - is the buffer shared
  1189. * @skb: buffer to check
  1190. *
  1191. * Returns true if more than one person has a reference to this
  1192. * buffer.
  1193. */
  1194. static inline int skb_shared(const struct sk_buff *skb)
  1195. {
  1196. return atomic_read(&skb->users) != 1;
  1197. }
  1198. /**
  1199. * skb_share_check - check if buffer is shared and if so clone it
  1200. * @skb: buffer to check
  1201. * @pri: priority for memory allocation
  1202. *
  1203. * If the buffer is shared the buffer is cloned and the old copy
  1204. * drops a reference. A new clone with a single reference is returned.
  1205. * If the buffer is not shared the original buffer is returned. When
  1206. * being called from interrupt status or with spinlocks held pri must
  1207. * be GFP_ATOMIC.
  1208. *
  1209. * NULL is returned on a memory allocation failure.
  1210. */
  1211. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1212. {
  1213. might_sleep_if(gfpflags_allow_blocking(pri));
  1214. if (skb_shared(skb)) {
  1215. struct sk_buff *nskb = skb_clone(skb, pri);
  1216. if (likely(nskb))
  1217. consume_skb(skb);
  1218. else
  1219. kfree_skb(skb);
  1220. skb = nskb;
  1221. }
  1222. return skb;
  1223. }
  1224. /*
  1225. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1226. * packets to handle cases where we have a local reader and forward
  1227. * and a couple of other messy ones. The normal one is tcpdumping
  1228. * a packet thats being forwarded.
  1229. */
  1230. /**
  1231. * skb_unshare - make a copy of a shared buffer
  1232. * @skb: buffer to check
  1233. * @pri: priority for memory allocation
  1234. *
  1235. * If the socket buffer is a clone then this function creates a new
  1236. * copy of the data, drops a reference count on the old copy and returns
  1237. * the new copy with the reference count at 1. If the buffer is not a clone
  1238. * the original buffer is returned. When called with a spinlock held or
  1239. * from interrupt state @pri must be %GFP_ATOMIC
  1240. *
  1241. * %NULL is returned on a memory allocation failure.
  1242. */
  1243. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1244. gfp_t pri)
  1245. {
  1246. might_sleep_if(gfpflags_allow_blocking(pri));
  1247. if (skb_cloned(skb)) {
  1248. struct sk_buff *nskb = skb_copy(skb, pri);
  1249. /* Free our shared copy */
  1250. if (likely(nskb))
  1251. consume_skb(skb);
  1252. else
  1253. kfree_skb(skb);
  1254. skb = nskb;
  1255. }
  1256. return skb;
  1257. }
  1258. /**
  1259. * skb_peek - peek at the head of an &sk_buff_head
  1260. * @list_: list to peek at
  1261. *
  1262. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1263. * be careful with this one. A peek leaves the buffer on the
  1264. * list and someone else may run off with it. You must hold
  1265. * the appropriate locks or have a private queue to do this.
  1266. *
  1267. * Returns %NULL for an empty list or a pointer to the head element.
  1268. * The reference count is not incremented and the reference is therefore
  1269. * volatile. Use with caution.
  1270. */
  1271. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1272. {
  1273. struct sk_buff *skb = list_->next;
  1274. if (skb == (struct sk_buff *)list_)
  1275. skb = NULL;
  1276. return skb;
  1277. }
  1278. /**
  1279. * skb_peek_next - peek skb following the given one from a queue
  1280. * @skb: skb to start from
  1281. * @list_: list to peek at
  1282. *
  1283. * Returns %NULL when the end of the list is met or a pointer to the
  1284. * next element. The reference count is not incremented and the
  1285. * reference is therefore volatile. Use with caution.
  1286. */
  1287. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1288. const struct sk_buff_head *list_)
  1289. {
  1290. struct sk_buff *next = skb->next;
  1291. if (next == (struct sk_buff *)list_)
  1292. next = NULL;
  1293. return next;
  1294. }
  1295. /**
  1296. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1297. * @list_: list to peek at
  1298. *
  1299. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1300. * be careful with this one. A peek leaves the buffer on the
  1301. * list and someone else may run off with it. You must hold
  1302. * the appropriate locks or have a private queue to do this.
  1303. *
  1304. * Returns %NULL for an empty list or a pointer to the tail element.
  1305. * The reference count is not incremented and the reference is therefore
  1306. * volatile. Use with caution.
  1307. */
  1308. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1309. {
  1310. struct sk_buff *skb = list_->prev;
  1311. if (skb == (struct sk_buff *)list_)
  1312. skb = NULL;
  1313. return skb;
  1314. }
  1315. /**
  1316. * skb_queue_len - get queue length
  1317. * @list_: list to measure
  1318. *
  1319. * Return the length of an &sk_buff queue.
  1320. */
  1321. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1322. {
  1323. return list_->qlen;
  1324. }
  1325. /**
  1326. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1327. * @list: queue to initialize
  1328. *
  1329. * This initializes only the list and queue length aspects of
  1330. * an sk_buff_head object. This allows to initialize the list
  1331. * aspects of an sk_buff_head without reinitializing things like
  1332. * the spinlock. It can also be used for on-stack sk_buff_head
  1333. * objects where the spinlock is known to not be used.
  1334. */
  1335. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1336. {
  1337. list->prev = list->next = (struct sk_buff *)list;
  1338. list->qlen = 0;
  1339. }
  1340. /*
  1341. * This function creates a split out lock class for each invocation;
  1342. * this is needed for now since a whole lot of users of the skb-queue
  1343. * infrastructure in drivers have different locking usage (in hardirq)
  1344. * than the networking core (in softirq only). In the long run either the
  1345. * network layer or drivers should need annotation to consolidate the
  1346. * main types of usage into 3 classes.
  1347. */
  1348. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1349. {
  1350. spin_lock_init(&list->lock);
  1351. __skb_queue_head_init(list);
  1352. }
  1353. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1354. struct lock_class_key *class)
  1355. {
  1356. skb_queue_head_init(list);
  1357. lockdep_set_class(&list->lock, class);
  1358. }
  1359. /*
  1360. * Insert an sk_buff on a list.
  1361. *
  1362. * The "__skb_xxxx()" functions are the non-atomic ones that
  1363. * can only be called with interrupts disabled.
  1364. */
  1365. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1366. struct sk_buff_head *list);
  1367. static inline void __skb_insert(struct sk_buff *newsk,
  1368. struct sk_buff *prev, struct sk_buff *next,
  1369. struct sk_buff_head *list)
  1370. {
  1371. newsk->next = next;
  1372. newsk->prev = prev;
  1373. next->prev = prev->next = newsk;
  1374. list->qlen++;
  1375. }
  1376. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1377. struct sk_buff *prev,
  1378. struct sk_buff *next)
  1379. {
  1380. struct sk_buff *first = list->next;
  1381. struct sk_buff *last = list->prev;
  1382. first->prev = prev;
  1383. prev->next = first;
  1384. last->next = next;
  1385. next->prev = last;
  1386. }
  1387. /**
  1388. * skb_queue_splice - join two skb lists, this is designed for stacks
  1389. * @list: the new list to add
  1390. * @head: the place to add it in the first list
  1391. */
  1392. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1393. struct sk_buff_head *head)
  1394. {
  1395. if (!skb_queue_empty(list)) {
  1396. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1397. head->qlen += list->qlen;
  1398. }
  1399. }
  1400. /**
  1401. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1402. * @list: the new list to add
  1403. * @head: the place to add it in the first list
  1404. *
  1405. * The list at @list is reinitialised
  1406. */
  1407. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1408. struct sk_buff_head *head)
  1409. {
  1410. if (!skb_queue_empty(list)) {
  1411. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1412. head->qlen += list->qlen;
  1413. __skb_queue_head_init(list);
  1414. }
  1415. }
  1416. /**
  1417. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1418. * @list: the new list to add
  1419. * @head: the place to add it in the first list
  1420. */
  1421. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1422. struct sk_buff_head *head)
  1423. {
  1424. if (!skb_queue_empty(list)) {
  1425. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1426. head->qlen += list->qlen;
  1427. }
  1428. }
  1429. /**
  1430. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1431. * @list: the new list to add
  1432. * @head: the place to add it in the first list
  1433. *
  1434. * Each of the lists is a queue.
  1435. * The list at @list is reinitialised
  1436. */
  1437. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1438. struct sk_buff_head *head)
  1439. {
  1440. if (!skb_queue_empty(list)) {
  1441. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1442. head->qlen += list->qlen;
  1443. __skb_queue_head_init(list);
  1444. }
  1445. }
  1446. /**
  1447. * __skb_queue_after - queue a buffer at the list head
  1448. * @list: list to use
  1449. * @prev: place after this buffer
  1450. * @newsk: buffer to queue
  1451. *
  1452. * Queue a buffer int the middle of a list. This function takes no locks
  1453. * and you must therefore hold required locks before calling it.
  1454. *
  1455. * A buffer cannot be placed on two lists at the same time.
  1456. */
  1457. static inline void __skb_queue_after(struct sk_buff_head *list,
  1458. struct sk_buff *prev,
  1459. struct sk_buff *newsk)
  1460. {
  1461. __skb_insert(newsk, prev, prev->next, list);
  1462. }
  1463. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1464. struct sk_buff_head *list);
  1465. static inline void __skb_queue_before(struct sk_buff_head *list,
  1466. struct sk_buff *next,
  1467. struct sk_buff *newsk)
  1468. {
  1469. __skb_insert(newsk, next->prev, next, list);
  1470. }
  1471. /**
  1472. * __skb_queue_head - queue a buffer at the list head
  1473. * @list: list to use
  1474. * @newsk: buffer to queue
  1475. *
  1476. * Queue a buffer at the start of a list. This function takes no locks
  1477. * and you must therefore hold required locks before calling it.
  1478. *
  1479. * A buffer cannot be placed on two lists at the same time.
  1480. */
  1481. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1482. static inline void __skb_queue_head(struct sk_buff_head *list,
  1483. struct sk_buff *newsk)
  1484. {
  1485. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1486. }
  1487. /**
  1488. * __skb_queue_tail - queue a buffer at the list tail
  1489. * @list: list to use
  1490. * @newsk: buffer to queue
  1491. *
  1492. * Queue a buffer at the end of a list. This function takes no locks
  1493. * and you must therefore hold required locks before calling it.
  1494. *
  1495. * A buffer cannot be placed on two lists at the same time.
  1496. */
  1497. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1498. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1499. struct sk_buff *newsk)
  1500. {
  1501. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1502. }
  1503. /*
  1504. * remove sk_buff from list. _Must_ be called atomically, and with
  1505. * the list known..
  1506. */
  1507. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1508. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1509. {
  1510. struct sk_buff *next, *prev;
  1511. list->qlen--;
  1512. next = skb->next;
  1513. prev = skb->prev;
  1514. skb->next = skb->prev = NULL;
  1515. next->prev = prev;
  1516. prev->next = next;
  1517. }
  1518. /**
  1519. * __skb_dequeue - remove from the head of the queue
  1520. * @list: list to dequeue from
  1521. *
  1522. * Remove the head of the list. This function does not take any locks
  1523. * so must be used with appropriate locks held only. The head item is
  1524. * returned or %NULL if the list is empty.
  1525. */
  1526. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1527. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1528. {
  1529. struct sk_buff *skb = skb_peek(list);
  1530. if (skb)
  1531. __skb_unlink(skb, list);
  1532. return skb;
  1533. }
  1534. /**
  1535. * __skb_dequeue_tail - remove from the tail of the queue
  1536. * @list: list to dequeue from
  1537. *
  1538. * Remove the tail of the list. This function does not take any locks
  1539. * so must be used with appropriate locks held only. The tail item is
  1540. * returned or %NULL if the list is empty.
  1541. */
  1542. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1543. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1544. {
  1545. struct sk_buff *skb = skb_peek_tail(list);
  1546. if (skb)
  1547. __skb_unlink(skb, list);
  1548. return skb;
  1549. }
  1550. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1551. {
  1552. return skb->data_len;
  1553. }
  1554. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1555. {
  1556. return skb->len - skb->data_len;
  1557. }
  1558. static inline unsigned int skb_pagelen(const struct sk_buff *skb)
  1559. {
  1560. unsigned int i, len = 0;
  1561. for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
  1562. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1563. return len + skb_headlen(skb);
  1564. }
  1565. /**
  1566. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1567. * @skb: buffer containing fragment to be initialised
  1568. * @i: paged fragment index to initialise
  1569. * @page: the page to use for this fragment
  1570. * @off: the offset to the data with @page
  1571. * @size: the length of the data
  1572. *
  1573. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1574. * offset @off within @page.
  1575. *
  1576. * Does not take any additional reference on the fragment.
  1577. */
  1578. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1579. struct page *page, int off, int size)
  1580. {
  1581. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1582. /*
  1583. * Propagate page pfmemalloc to the skb if we can. The problem is
  1584. * that not all callers have unique ownership of the page but rely
  1585. * on page_is_pfmemalloc doing the right thing(tm).
  1586. */
  1587. frag->page.p = page;
  1588. frag->page_offset = off;
  1589. skb_frag_size_set(frag, size);
  1590. page = compound_head(page);
  1591. if (page_is_pfmemalloc(page))
  1592. skb->pfmemalloc = true;
  1593. }
  1594. /**
  1595. * skb_fill_page_desc - initialise a paged fragment in an skb
  1596. * @skb: buffer containing fragment to be initialised
  1597. * @i: paged fragment index to initialise
  1598. * @page: the page to use for this fragment
  1599. * @off: the offset to the data with @page
  1600. * @size: the length of the data
  1601. *
  1602. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1603. * @skb to point to @size bytes at offset @off within @page. In
  1604. * addition updates @skb such that @i is the last fragment.
  1605. *
  1606. * Does not take any additional reference on the fragment.
  1607. */
  1608. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1609. struct page *page, int off, int size)
  1610. {
  1611. __skb_fill_page_desc(skb, i, page, off, size);
  1612. skb_shinfo(skb)->nr_frags = i + 1;
  1613. }
  1614. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1615. int size, unsigned int truesize);
  1616. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1617. unsigned int truesize);
  1618. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1619. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1620. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1621. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1622. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1623. {
  1624. return skb->head + skb->tail;
  1625. }
  1626. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1627. {
  1628. skb->tail = skb->data - skb->head;
  1629. }
  1630. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1631. {
  1632. skb_reset_tail_pointer(skb);
  1633. skb->tail += offset;
  1634. }
  1635. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1636. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1637. {
  1638. return skb->tail;
  1639. }
  1640. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1641. {
  1642. skb->tail = skb->data;
  1643. }
  1644. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1645. {
  1646. skb->tail = skb->data + offset;
  1647. }
  1648. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1649. /*
  1650. * Add data to an sk_buff
  1651. */
  1652. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1653. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1654. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1655. {
  1656. unsigned char *tmp = skb_tail_pointer(skb);
  1657. SKB_LINEAR_ASSERT(skb);
  1658. skb->tail += len;
  1659. skb->len += len;
  1660. return tmp;
  1661. }
  1662. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1663. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1664. {
  1665. skb->data -= len;
  1666. skb->len += len;
  1667. return skb->data;
  1668. }
  1669. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1670. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1671. {
  1672. skb->len -= len;
  1673. BUG_ON(skb->len < skb->data_len);
  1674. return skb->data += len;
  1675. }
  1676. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1677. {
  1678. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1679. }
  1680. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1681. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1682. {
  1683. if (len > skb_headlen(skb) &&
  1684. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1685. return NULL;
  1686. skb->len -= len;
  1687. return skb->data += len;
  1688. }
  1689. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1690. {
  1691. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1692. }
  1693. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1694. {
  1695. if (likely(len <= skb_headlen(skb)))
  1696. return 1;
  1697. if (unlikely(len > skb->len))
  1698. return 0;
  1699. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1700. }
  1701. void skb_condense(struct sk_buff *skb);
  1702. /**
  1703. * skb_headroom - bytes at buffer head
  1704. * @skb: buffer to check
  1705. *
  1706. * Return the number of bytes of free space at the head of an &sk_buff.
  1707. */
  1708. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1709. {
  1710. return skb->data - skb->head;
  1711. }
  1712. /**
  1713. * skb_tailroom - bytes at buffer end
  1714. * @skb: buffer to check
  1715. *
  1716. * Return the number of bytes of free space at the tail of an sk_buff
  1717. */
  1718. static inline int skb_tailroom(const struct sk_buff *skb)
  1719. {
  1720. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1721. }
  1722. /**
  1723. * skb_availroom - bytes at buffer end
  1724. * @skb: buffer to check
  1725. *
  1726. * Return the number of bytes of free space at the tail of an sk_buff
  1727. * allocated by sk_stream_alloc()
  1728. */
  1729. static inline int skb_availroom(const struct sk_buff *skb)
  1730. {
  1731. if (skb_is_nonlinear(skb))
  1732. return 0;
  1733. return skb->end - skb->tail - skb->reserved_tailroom;
  1734. }
  1735. /**
  1736. * skb_reserve - adjust headroom
  1737. * @skb: buffer to alter
  1738. * @len: bytes to move
  1739. *
  1740. * Increase the headroom of an empty &sk_buff by reducing the tail
  1741. * room. This is only allowed for an empty buffer.
  1742. */
  1743. static inline void skb_reserve(struct sk_buff *skb, int len)
  1744. {
  1745. skb->data += len;
  1746. skb->tail += len;
  1747. }
  1748. /**
  1749. * skb_tailroom_reserve - adjust reserved_tailroom
  1750. * @skb: buffer to alter
  1751. * @mtu: maximum amount of headlen permitted
  1752. * @needed_tailroom: minimum amount of reserved_tailroom
  1753. *
  1754. * Set reserved_tailroom so that headlen can be as large as possible but
  1755. * not larger than mtu and tailroom cannot be smaller than
  1756. * needed_tailroom.
  1757. * The required headroom should already have been reserved before using
  1758. * this function.
  1759. */
  1760. static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
  1761. unsigned int needed_tailroom)
  1762. {
  1763. SKB_LINEAR_ASSERT(skb);
  1764. if (mtu < skb_tailroom(skb) - needed_tailroom)
  1765. /* use at most mtu */
  1766. skb->reserved_tailroom = skb_tailroom(skb) - mtu;
  1767. else
  1768. /* use up to all available space */
  1769. skb->reserved_tailroom = needed_tailroom;
  1770. }
  1771. #define ENCAP_TYPE_ETHER 0
  1772. #define ENCAP_TYPE_IPPROTO 1
  1773. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1774. __be16 protocol)
  1775. {
  1776. skb->inner_protocol = protocol;
  1777. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1778. }
  1779. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1780. __u8 ipproto)
  1781. {
  1782. skb->inner_ipproto = ipproto;
  1783. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1784. }
  1785. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1786. {
  1787. skb->inner_mac_header = skb->mac_header;
  1788. skb->inner_network_header = skb->network_header;
  1789. skb->inner_transport_header = skb->transport_header;
  1790. }
  1791. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1792. {
  1793. skb->mac_len = skb->network_header - skb->mac_header;
  1794. }
  1795. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1796. *skb)
  1797. {
  1798. return skb->head + skb->inner_transport_header;
  1799. }
  1800. static inline int skb_inner_transport_offset(const struct sk_buff *skb)
  1801. {
  1802. return skb_inner_transport_header(skb) - skb->data;
  1803. }
  1804. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1805. {
  1806. skb->inner_transport_header = skb->data - skb->head;
  1807. }
  1808. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1809. const int offset)
  1810. {
  1811. skb_reset_inner_transport_header(skb);
  1812. skb->inner_transport_header += offset;
  1813. }
  1814. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1815. {
  1816. return skb->head + skb->inner_network_header;
  1817. }
  1818. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1819. {
  1820. skb->inner_network_header = skb->data - skb->head;
  1821. }
  1822. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1823. const int offset)
  1824. {
  1825. skb_reset_inner_network_header(skb);
  1826. skb->inner_network_header += offset;
  1827. }
  1828. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1829. {
  1830. return skb->head + skb->inner_mac_header;
  1831. }
  1832. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1833. {
  1834. skb->inner_mac_header = skb->data - skb->head;
  1835. }
  1836. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1837. const int offset)
  1838. {
  1839. skb_reset_inner_mac_header(skb);
  1840. skb->inner_mac_header += offset;
  1841. }
  1842. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1843. {
  1844. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1845. }
  1846. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1847. {
  1848. return skb->head + skb->transport_header;
  1849. }
  1850. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1851. {
  1852. skb->transport_header = skb->data - skb->head;
  1853. }
  1854. static inline void skb_set_transport_header(struct sk_buff *skb,
  1855. const int offset)
  1856. {
  1857. skb_reset_transport_header(skb);
  1858. skb->transport_header += offset;
  1859. }
  1860. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1861. {
  1862. return skb->head + skb->network_header;
  1863. }
  1864. static inline void skb_reset_network_header(struct sk_buff *skb)
  1865. {
  1866. skb->network_header = skb->data - skb->head;
  1867. }
  1868. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1869. {
  1870. skb_reset_network_header(skb);
  1871. skb->network_header += offset;
  1872. }
  1873. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1874. {
  1875. return skb->head + skb->mac_header;
  1876. }
  1877. static inline int skb_mac_offset(const struct sk_buff *skb)
  1878. {
  1879. return skb_mac_header(skb) - skb->data;
  1880. }
  1881. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1882. {
  1883. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1884. }
  1885. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1886. {
  1887. skb->mac_header = skb->data - skb->head;
  1888. }
  1889. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1890. {
  1891. skb_reset_mac_header(skb);
  1892. skb->mac_header += offset;
  1893. }
  1894. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1895. {
  1896. skb->mac_header = skb->network_header;
  1897. }
  1898. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1899. const int offset_hint)
  1900. {
  1901. struct flow_keys keys;
  1902. if (skb_transport_header_was_set(skb))
  1903. return;
  1904. else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
  1905. skb_set_transport_header(skb, keys.control.thoff);
  1906. else
  1907. skb_set_transport_header(skb, offset_hint);
  1908. }
  1909. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1910. {
  1911. if (skb_mac_header_was_set(skb)) {
  1912. const unsigned char *old_mac = skb_mac_header(skb);
  1913. skb_set_mac_header(skb, -skb->mac_len);
  1914. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1915. }
  1916. }
  1917. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1918. {
  1919. return skb->csum_start - skb_headroom(skb);
  1920. }
  1921. static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
  1922. {
  1923. return skb->head + skb->csum_start;
  1924. }
  1925. static inline int skb_transport_offset(const struct sk_buff *skb)
  1926. {
  1927. return skb_transport_header(skb) - skb->data;
  1928. }
  1929. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1930. {
  1931. return skb->transport_header - skb->network_header;
  1932. }
  1933. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1934. {
  1935. return skb->inner_transport_header - skb->inner_network_header;
  1936. }
  1937. static inline int skb_network_offset(const struct sk_buff *skb)
  1938. {
  1939. return skb_network_header(skb) - skb->data;
  1940. }
  1941. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1942. {
  1943. return skb_inner_network_header(skb) - skb->data;
  1944. }
  1945. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1946. {
  1947. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1948. }
  1949. /*
  1950. * CPUs often take a performance hit when accessing unaligned memory
  1951. * locations. The actual performance hit varies, it can be small if the
  1952. * hardware handles it or large if we have to take an exception and fix it
  1953. * in software.
  1954. *
  1955. * Since an ethernet header is 14 bytes network drivers often end up with
  1956. * the IP header at an unaligned offset. The IP header can be aligned by
  1957. * shifting the start of the packet by 2 bytes. Drivers should do this
  1958. * with:
  1959. *
  1960. * skb_reserve(skb, NET_IP_ALIGN);
  1961. *
  1962. * The downside to this alignment of the IP header is that the DMA is now
  1963. * unaligned. On some architectures the cost of an unaligned DMA is high
  1964. * and this cost outweighs the gains made by aligning the IP header.
  1965. *
  1966. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1967. * to be overridden.
  1968. */
  1969. #ifndef NET_IP_ALIGN
  1970. #define NET_IP_ALIGN 2
  1971. #endif
  1972. /*
  1973. * The networking layer reserves some headroom in skb data (via
  1974. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1975. * the header has to grow. In the default case, if the header has to grow
  1976. * 32 bytes or less we avoid the reallocation.
  1977. *
  1978. * Unfortunately this headroom changes the DMA alignment of the resulting
  1979. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1980. * on some architectures. An architecture can override this value,
  1981. * perhaps setting it to a cacheline in size (since that will maintain
  1982. * cacheline alignment of the DMA). It must be a power of 2.
  1983. *
  1984. * Various parts of the networking layer expect at least 32 bytes of
  1985. * headroom, you should not reduce this.
  1986. *
  1987. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1988. * to reduce average number of cache lines per packet.
  1989. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1990. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1991. */
  1992. #ifndef NET_SKB_PAD
  1993. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1994. #endif
  1995. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1996. static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
  1997. {
  1998. if (unlikely(skb_is_nonlinear(skb))) {
  1999. WARN_ON(1);
  2000. return;
  2001. }
  2002. skb->len = len;
  2003. skb_set_tail_pointer(skb, len);
  2004. }
  2005. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  2006. {
  2007. __skb_set_length(skb, len);
  2008. }
  2009. void skb_trim(struct sk_buff *skb, unsigned int len);
  2010. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  2011. {
  2012. if (skb->data_len)
  2013. return ___pskb_trim(skb, len);
  2014. __skb_trim(skb, len);
  2015. return 0;
  2016. }
  2017. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  2018. {
  2019. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  2020. }
  2021. /**
  2022. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  2023. * @skb: buffer to alter
  2024. * @len: new length
  2025. *
  2026. * This is identical to pskb_trim except that the caller knows that
  2027. * the skb is not cloned so we should never get an error due to out-
  2028. * of-memory.
  2029. */
  2030. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  2031. {
  2032. int err = pskb_trim(skb, len);
  2033. BUG_ON(err);
  2034. }
  2035. static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
  2036. {
  2037. unsigned int diff = len - skb->len;
  2038. if (skb_tailroom(skb) < diff) {
  2039. int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
  2040. GFP_ATOMIC);
  2041. if (ret)
  2042. return ret;
  2043. }
  2044. __skb_set_length(skb, len);
  2045. return 0;
  2046. }
  2047. /**
  2048. * skb_orphan - orphan a buffer
  2049. * @skb: buffer to orphan
  2050. *
  2051. * If a buffer currently has an owner then we call the owner's
  2052. * destructor function and make the @skb unowned. The buffer continues
  2053. * to exist but is no longer charged to its former owner.
  2054. */
  2055. static inline void skb_orphan(struct sk_buff *skb)
  2056. {
  2057. if (skb->destructor) {
  2058. skb->destructor(skb);
  2059. skb->destructor = NULL;
  2060. skb->sk = NULL;
  2061. } else {
  2062. BUG_ON(skb->sk);
  2063. }
  2064. }
  2065. /**
  2066. * skb_orphan_frags - orphan the frags contained in a buffer
  2067. * @skb: buffer to orphan frags from
  2068. * @gfp_mask: allocation mask for replacement pages
  2069. *
  2070. * For each frag in the SKB which needs a destructor (i.e. has an
  2071. * owner) create a copy of that frag and release the original
  2072. * page by calling the destructor.
  2073. */
  2074. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  2075. {
  2076. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  2077. return 0;
  2078. return skb_copy_ubufs(skb, gfp_mask);
  2079. }
  2080. /**
  2081. * __skb_queue_purge - empty a list
  2082. * @list: list to empty
  2083. *
  2084. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  2085. * the list and one reference dropped. This function does not take the
  2086. * list lock and the caller must hold the relevant locks to use it.
  2087. */
  2088. void skb_queue_purge(struct sk_buff_head *list);
  2089. static inline void __skb_queue_purge(struct sk_buff_head *list)
  2090. {
  2091. struct sk_buff *skb;
  2092. while ((skb = __skb_dequeue(list)) != NULL)
  2093. kfree_skb(skb);
  2094. }
  2095. void skb_rbtree_purge(struct rb_root *root);
  2096. void *netdev_alloc_frag(unsigned int fragsz);
  2097. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  2098. gfp_t gfp_mask);
  2099. /**
  2100. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  2101. * @dev: network device to receive on
  2102. * @length: length to allocate
  2103. *
  2104. * Allocate a new &sk_buff and assign it a usage count of one. The
  2105. * buffer has unspecified headroom built in. Users should allocate
  2106. * the headroom they think they need without accounting for the
  2107. * built in space. The built in space is used for optimisations.
  2108. *
  2109. * %NULL is returned if there is no free memory. Although this function
  2110. * allocates memory it can be called from an interrupt.
  2111. */
  2112. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  2113. unsigned int length)
  2114. {
  2115. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  2116. }
  2117. /* legacy helper around __netdev_alloc_skb() */
  2118. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  2119. gfp_t gfp_mask)
  2120. {
  2121. return __netdev_alloc_skb(NULL, length, gfp_mask);
  2122. }
  2123. /* legacy helper around netdev_alloc_skb() */
  2124. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  2125. {
  2126. return netdev_alloc_skb(NULL, length);
  2127. }
  2128. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  2129. unsigned int length, gfp_t gfp)
  2130. {
  2131. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  2132. if (NET_IP_ALIGN && skb)
  2133. skb_reserve(skb, NET_IP_ALIGN);
  2134. return skb;
  2135. }
  2136. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  2137. unsigned int length)
  2138. {
  2139. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  2140. }
  2141. static inline void skb_free_frag(void *addr)
  2142. {
  2143. page_frag_free(addr);
  2144. }
  2145. void *napi_alloc_frag(unsigned int fragsz);
  2146. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  2147. unsigned int length, gfp_t gfp_mask);
  2148. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  2149. unsigned int length)
  2150. {
  2151. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  2152. }
  2153. void napi_consume_skb(struct sk_buff *skb, int budget);
  2154. void __kfree_skb_flush(void);
  2155. void __kfree_skb_defer(struct sk_buff *skb);
  2156. /**
  2157. * __dev_alloc_pages - allocate page for network Rx
  2158. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2159. * @order: size of the allocation
  2160. *
  2161. * Allocate a new page.
  2162. *
  2163. * %NULL is returned if there is no free memory.
  2164. */
  2165. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  2166. unsigned int order)
  2167. {
  2168. /* This piece of code contains several assumptions.
  2169. * 1. This is for device Rx, therefor a cold page is preferred.
  2170. * 2. The expectation is the user wants a compound page.
  2171. * 3. If requesting a order 0 page it will not be compound
  2172. * due to the check to see if order has a value in prep_new_page
  2173. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  2174. * code in gfp_to_alloc_flags that should be enforcing this.
  2175. */
  2176. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  2177. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  2178. }
  2179. static inline struct page *dev_alloc_pages(unsigned int order)
  2180. {
  2181. return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
  2182. }
  2183. /**
  2184. * __dev_alloc_page - allocate a page for network Rx
  2185. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2186. *
  2187. * Allocate a new page.
  2188. *
  2189. * %NULL is returned if there is no free memory.
  2190. */
  2191. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  2192. {
  2193. return __dev_alloc_pages(gfp_mask, 0);
  2194. }
  2195. static inline struct page *dev_alloc_page(void)
  2196. {
  2197. return dev_alloc_pages(0);
  2198. }
  2199. /**
  2200. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  2201. * @page: The page that was allocated from skb_alloc_page
  2202. * @skb: The skb that may need pfmemalloc set
  2203. */
  2204. static inline void skb_propagate_pfmemalloc(struct page *page,
  2205. struct sk_buff *skb)
  2206. {
  2207. if (page_is_pfmemalloc(page))
  2208. skb->pfmemalloc = true;
  2209. }
  2210. /**
  2211. * skb_frag_page - retrieve the page referred to by a paged fragment
  2212. * @frag: the paged fragment
  2213. *
  2214. * Returns the &struct page associated with @frag.
  2215. */
  2216. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2217. {
  2218. return frag->page.p;
  2219. }
  2220. /**
  2221. * __skb_frag_ref - take an addition reference on a paged fragment.
  2222. * @frag: the paged fragment
  2223. *
  2224. * Takes an additional reference on the paged fragment @frag.
  2225. */
  2226. static inline void __skb_frag_ref(skb_frag_t *frag)
  2227. {
  2228. get_page(skb_frag_page(frag));
  2229. }
  2230. /**
  2231. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2232. * @skb: the buffer
  2233. * @f: the fragment offset.
  2234. *
  2235. * Takes an additional reference on the @f'th paged fragment of @skb.
  2236. */
  2237. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2238. {
  2239. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2240. }
  2241. /**
  2242. * __skb_frag_unref - release a reference on a paged fragment.
  2243. * @frag: the paged fragment
  2244. *
  2245. * Releases a reference on the paged fragment @frag.
  2246. */
  2247. static inline void __skb_frag_unref(skb_frag_t *frag)
  2248. {
  2249. put_page(skb_frag_page(frag));
  2250. }
  2251. /**
  2252. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2253. * @skb: the buffer
  2254. * @f: the fragment offset
  2255. *
  2256. * Releases a reference on the @f'th paged fragment of @skb.
  2257. */
  2258. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2259. {
  2260. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2261. }
  2262. /**
  2263. * skb_frag_address - gets the address of the data contained in a paged fragment
  2264. * @frag: the paged fragment buffer
  2265. *
  2266. * Returns the address of the data within @frag. The page must already
  2267. * be mapped.
  2268. */
  2269. static inline void *skb_frag_address(const skb_frag_t *frag)
  2270. {
  2271. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2272. }
  2273. /**
  2274. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2275. * @frag: the paged fragment buffer
  2276. *
  2277. * Returns the address of the data within @frag. Checks that the page
  2278. * is mapped and returns %NULL otherwise.
  2279. */
  2280. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2281. {
  2282. void *ptr = page_address(skb_frag_page(frag));
  2283. if (unlikely(!ptr))
  2284. return NULL;
  2285. return ptr + frag->page_offset;
  2286. }
  2287. /**
  2288. * __skb_frag_set_page - sets the page contained in a paged fragment
  2289. * @frag: the paged fragment
  2290. * @page: the page to set
  2291. *
  2292. * Sets the fragment @frag to contain @page.
  2293. */
  2294. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2295. {
  2296. frag->page.p = page;
  2297. }
  2298. /**
  2299. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2300. * @skb: the buffer
  2301. * @f: the fragment offset
  2302. * @page: the page to set
  2303. *
  2304. * Sets the @f'th fragment of @skb to contain @page.
  2305. */
  2306. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2307. struct page *page)
  2308. {
  2309. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2310. }
  2311. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2312. /**
  2313. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2314. * @dev: the device to map the fragment to
  2315. * @frag: the paged fragment to map
  2316. * @offset: the offset within the fragment (starting at the
  2317. * fragment's own offset)
  2318. * @size: the number of bytes to map
  2319. * @dir: the direction of the mapping (%PCI_DMA_*)
  2320. *
  2321. * Maps the page associated with @frag to @device.
  2322. */
  2323. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2324. const skb_frag_t *frag,
  2325. size_t offset, size_t size,
  2326. enum dma_data_direction dir)
  2327. {
  2328. return dma_map_page(dev, skb_frag_page(frag),
  2329. frag->page_offset + offset, size, dir);
  2330. }
  2331. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2332. gfp_t gfp_mask)
  2333. {
  2334. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2335. }
  2336. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2337. gfp_t gfp_mask)
  2338. {
  2339. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2340. }
  2341. /**
  2342. * skb_clone_writable - is the header of a clone writable
  2343. * @skb: buffer to check
  2344. * @len: length up to which to write
  2345. *
  2346. * Returns true if modifying the header part of the cloned buffer
  2347. * does not requires the data to be copied.
  2348. */
  2349. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2350. {
  2351. return !skb_header_cloned(skb) &&
  2352. skb_headroom(skb) + len <= skb->hdr_len;
  2353. }
  2354. static inline int skb_try_make_writable(struct sk_buff *skb,
  2355. unsigned int write_len)
  2356. {
  2357. return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
  2358. pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
  2359. }
  2360. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2361. int cloned)
  2362. {
  2363. int delta = 0;
  2364. if (headroom > skb_headroom(skb))
  2365. delta = headroom - skb_headroom(skb);
  2366. if (delta || cloned)
  2367. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2368. GFP_ATOMIC);
  2369. return 0;
  2370. }
  2371. /**
  2372. * skb_cow - copy header of skb when it is required
  2373. * @skb: buffer to cow
  2374. * @headroom: needed headroom
  2375. *
  2376. * If the skb passed lacks sufficient headroom or its data part
  2377. * is shared, data is reallocated. If reallocation fails, an error
  2378. * is returned and original skb is not changed.
  2379. *
  2380. * The result is skb with writable area skb->head...skb->tail
  2381. * and at least @headroom of space at head.
  2382. */
  2383. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2384. {
  2385. return __skb_cow(skb, headroom, skb_cloned(skb));
  2386. }
  2387. /**
  2388. * skb_cow_head - skb_cow but only making the head writable
  2389. * @skb: buffer to cow
  2390. * @headroom: needed headroom
  2391. *
  2392. * This function is identical to skb_cow except that we replace the
  2393. * skb_cloned check by skb_header_cloned. It should be used when
  2394. * you only need to push on some header and do not need to modify
  2395. * the data.
  2396. */
  2397. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2398. {
  2399. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2400. }
  2401. /**
  2402. * skb_padto - pad an skbuff up to a minimal size
  2403. * @skb: buffer to pad
  2404. * @len: minimal length
  2405. *
  2406. * Pads up a buffer to ensure the trailing bytes exist and are
  2407. * blanked. If the buffer already contains sufficient data it
  2408. * is untouched. Otherwise it is extended. Returns zero on
  2409. * success. The skb is freed on error.
  2410. */
  2411. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2412. {
  2413. unsigned int size = skb->len;
  2414. if (likely(size >= len))
  2415. return 0;
  2416. return skb_pad(skb, len - size);
  2417. }
  2418. /**
  2419. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2420. * @skb: buffer to pad
  2421. * @len: minimal length
  2422. *
  2423. * Pads up a buffer to ensure the trailing bytes exist and are
  2424. * blanked. If the buffer already contains sufficient data it
  2425. * is untouched. Otherwise it is extended. Returns zero on
  2426. * success. The skb is freed on error.
  2427. */
  2428. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2429. {
  2430. unsigned int size = skb->len;
  2431. if (unlikely(size < len)) {
  2432. len -= size;
  2433. if (skb_pad(skb, len))
  2434. return -ENOMEM;
  2435. __skb_put(skb, len);
  2436. }
  2437. return 0;
  2438. }
  2439. static inline int skb_add_data(struct sk_buff *skb,
  2440. struct iov_iter *from, int copy)
  2441. {
  2442. const int off = skb->len;
  2443. if (skb->ip_summed == CHECKSUM_NONE) {
  2444. __wsum csum = 0;
  2445. if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
  2446. &csum, from)) {
  2447. skb->csum = csum_block_add(skb->csum, csum, off);
  2448. return 0;
  2449. }
  2450. } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
  2451. return 0;
  2452. __skb_trim(skb, off);
  2453. return -EFAULT;
  2454. }
  2455. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2456. const struct page *page, int off)
  2457. {
  2458. if (i) {
  2459. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2460. return page == skb_frag_page(frag) &&
  2461. off == frag->page_offset + skb_frag_size(frag);
  2462. }
  2463. return false;
  2464. }
  2465. static inline int __skb_linearize(struct sk_buff *skb)
  2466. {
  2467. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2468. }
  2469. /**
  2470. * skb_linearize - convert paged skb to linear one
  2471. * @skb: buffer to linarize
  2472. *
  2473. * If there is no free memory -ENOMEM is returned, otherwise zero
  2474. * is returned and the old skb data released.
  2475. */
  2476. static inline int skb_linearize(struct sk_buff *skb)
  2477. {
  2478. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2479. }
  2480. /**
  2481. * skb_has_shared_frag - can any frag be overwritten
  2482. * @skb: buffer to test
  2483. *
  2484. * Return true if the skb has at least one frag that might be modified
  2485. * by an external entity (as in vmsplice()/sendfile())
  2486. */
  2487. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2488. {
  2489. return skb_is_nonlinear(skb) &&
  2490. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2491. }
  2492. /**
  2493. * skb_linearize_cow - make sure skb is linear and writable
  2494. * @skb: buffer to process
  2495. *
  2496. * If there is no free memory -ENOMEM is returned, otherwise zero
  2497. * is returned and the old skb data released.
  2498. */
  2499. static inline int skb_linearize_cow(struct sk_buff *skb)
  2500. {
  2501. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2502. __skb_linearize(skb) : 0;
  2503. }
  2504. static __always_inline void
  2505. __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2506. unsigned int off)
  2507. {
  2508. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2509. skb->csum = csum_block_sub(skb->csum,
  2510. csum_partial(start, len, 0), off);
  2511. else if (skb->ip_summed == CHECKSUM_PARTIAL &&
  2512. skb_checksum_start_offset(skb) < 0)
  2513. skb->ip_summed = CHECKSUM_NONE;
  2514. }
  2515. /**
  2516. * skb_postpull_rcsum - update checksum for received skb after pull
  2517. * @skb: buffer to update
  2518. * @start: start of data before pull
  2519. * @len: length of data pulled
  2520. *
  2521. * After doing a pull on a received packet, you need to call this to
  2522. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2523. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2524. */
  2525. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2526. const void *start, unsigned int len)
  2527. {
  2528. __skb_postpull_rcsum(skb, start, len, 0);
  2529. }
  2530. static __always_inline void
  2531. __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
  2532. unsigned int off)
  2533. {
  2534. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2535. skb->csum = csum_block_add(skb->csum,
  2536. csum_partial(start, len, 0), off);
  2537. }
  2538. /**
  2539. * skb_postpush_rcsum - update checksum for received skb after push
  2540. * @skb: buffer to update
  2541. * @start: start of data after push
  2542. * @len: length of data pushed
  2543. *
  2544. * After doing a push on a received packet, you need to call this to
  2545. * update the CHECKSUM_COMPLETE checksum.
  2546. */
  2547. static inline void skb_postpush_rcsum(struct sk_buff *skb,
  2548. const void *start, unsigned int len)
  2549. {
  2550. __skb_postpush_rcsum(skb, start, len, 0);
  2551. }
  2552. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2553. /**
  2554. * skb_push_rcsum - push skb and update receive checksum
  2555. * @skb: buffer to update
  2556. * @len: length of data pulled
  2557. *
  2558. * This function performs an skb_push on the packet and updates
  2559. * the CHECKSUM_COMPLETE checksum. It should be used on
  2560. * receive path processing instead of skb_push unless you know
  2561. * that the checksum difference is zero (e.g., a valid IP header)
  2562. * or you are setting ip_summed to CHECKSUM_NONE.
  2563. */
  2564. static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
  2565. unsigned int len)
  2566. {
  2567. skb_push(skb, len);
  2568. skb_postpush_rcsum(skb, skb->data, len);
  2569. return skb->data;
  2570. }
  2571. /**
  2572. * pskb_trim_rcsum - trim received skb and update checksum
  2573. * @skb: buffer to trim
  2574. * @len: new length
  2575. *
  2576. * This is exactly the same as pskb_trim except that it ensures the
  2577. * checksum of received packets are still valid after the operation.
  2578. */
  2579. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2580. {
  2581. if (likely(len >= skb->len))
  2582. return 0;
  2583. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2584. skb->ip_summed = CHECKSUM_NONE;
  2585. return __pskb_trim(skb, len);
  2586. }
  2587. static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2588. {
  2589. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2590. skb->ip_summed = CHECKSUM_NONE;
  2591. __skb_trim(skb, len);
  2592. return 0;
  2593. }
  2594. static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
  2595. {
  2596. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2597. skb->ip_summed = CHECKSUM_NONE;
  2598. return __skb_grow(skb, len);
  2599. }
  2600. #define skb_queue_walk(queue, skb) \
  2601. for (skb = (queue)->next; \
  2602. skb != (struct sk_buff *)(queue); \
  2603. skb = skb->next)
  2604. #define skb_queue_walk_safe(queue, skb, tmp) \
  2605. for (skb = (queue)->next, tmp = skb->next; \
  2606. skb != (struct sk_buff *)(queue); \
  2607. skb = tmp, tmp = skb->next)
  2608. #define skb_queue_walk_from(queue, skb) \
  2609. for (; skb != (struct sk_buff *)(queue); \
  2610. skb = skb->next)
  2611. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2612. for (tmp = skb->next; \
  2613. skb != (struct sk_buff *)(queue); \
  2614. skb = tmp, tmp = skb->next)
  2615. #define skb_queue_reverse_walk(queue, skb) \
  2616. for (skb = (queue)->prev; \
  2617. skb != (struct sk_buff *)(queue); \
  2618. skb = skb->prev)
  2619. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2620. for (skb = (queue)->prev, tmp = skb->prev; \
  2621. skb != (struct sk_buff *)(queue); \
  2622. skb = tmp, tmp = skb->prev)
  2623. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2624. for (tmp = skb->prev; \
  2625. skb != (struct sk_buff *)(queue); \
  2626. skb = tmp, tmp = skb->prev)
  2627. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2628. {
  2629. return skb_shinfo(skb)->frag_list != NULL;
  2630. }
  2631. static inline void skb_frag_list_init(struct sk_buff *skb)
  2632. {
  2633. skb_shinfo(skb)->frag_list = NULL;
  2634. }
  2635. #define skb_walk_frags(skb, iter) \
  2636. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2637. int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
  2638. const struct sk_buff *skb);
  2639. struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
  2640. struct sk_buff_head *queue,
  2641. unsigned int flags,
  2642. void (*destructor)(struct sock *sk,
  2643. struct sk_buff *skb),
  2644. int *peeked, int *off, int *err,
  2645. struct sk_buff **last);
  2646. struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
  2647. void (*destructor)(struct sock *sk,
  2648. struct sk_buff *skb),
  2649. int *peeked, int *off, int *err,
  2650. struct sk_buff **last);
  2651. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2652. void (*destructor)(struct sock *sk,
  2653. struct sk_buff *skb),
  2654. int *peeked, int *off, int *err);
  2655. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2656. int *err);
  2657. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2658. struct poll_table_struct *wait);
  2659. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2660. struct iov_iter *to, int size);
  2661. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2662. struct msghdr *msg, int size)
  2663. {
  2664. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2665. }
  2666. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2667. struct msghdr *msg);
  2668. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2669. struct iov_iter *from, int len);
  2670. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2671. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2672. void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
  2673. static inline void skb_free_datagram_locked(struct sock *sk,
  2674. struct sk_buff *skb)
  2675. {
  2676. __skb_free_datagram_locked(sk, skb, 0);
  2677. }
  2678. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2679. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2680. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2681. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2682. int len, __wsum csum);
  2683. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2684. struct pipe_inode_info *pipe, unsigned int len,
  2685. unsigned int flags);
  2686. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2687. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2688. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2689. int len, int hlen);
  2690. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2691. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2692. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2693. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2694. bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
  2695. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2696. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2697. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2698. int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
  2699. int skb_vlan_pop(struct sk_buff *skb);
  2700. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2701. struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
  2702. gfp_t gfp);
  2703. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2704. {
  2705. return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
  2706. }
  2707. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2708. {
  2709. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2710. }
  2711. struct skb_checksum_ops {
  2712. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2713. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2714. };
  2715. extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
  2716. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2717. __wsum csum, const struct skb_checksum_ops *ops);
  2718. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2719. __wsum csum);
  2720. static inline void * __must_check
  2721. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2722. int len, void *data, int hlen, void *buffer)
  2723. {
  2724. if (hlen - offset >= len)
  2725. return data + offset;
  2726. if (!skb ||
  2727. skb_copy_bits(skb, offset, buffer, len) < 0)
  2728. return NULL;
  2729. return buffer;
  2730. }
  2731. static inline void * __must_check
  2732. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2733. {
  2734. return __skb_header_pointer(skb, offset, len, skb->data,
  2735. skb_headlen(skb), buffer);
  2736. }
  2737. /**
  2738. * skb_needs_linearize - check if we need to linearize a given skb
  2739. * depending on the given device features.
  2740. * @skb: socket buffer to check
  2741. * @features: net device features
  2742. *
  2743. * Returns true if either:
  2744. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2745. * 2. skb is fragmented and the device does not support SG.
  2746. */
  2747. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2748. netdev_features_t features)
  2749. {
  2750. return skb_is_nonlinear(skb) &&
  2751. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2752. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2753. }
  2754. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2755. void *to,
  2756. const unsigned int len)
  2757. {
  2758. memcpy(to, skb->data, len);
  2759. }
  2760. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2761. const int offset, void *to,
  2762. const unsigned int len)
  2763. {
  2764. memcpy(to, skb->data + offset, len);
  2765. }
  2766. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2767. const void *from,
  2768. const unsigned int len)
  2769. {
  2770. memcpy(skb->data, from, len);
  2771. }
  2772. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2773. const int offset,
  2774. const void *from,
  2775. const unsigned int len)
  2776. {
  2777. memcpy(skb->data + offset, from, len);
  2778. }
  2779. void skb_init(void);
  2780. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2781. {
  2782. return skb->tstamp;
  2783. }
  2784. /**
  2785. * skb_get_timestamp - get timestamp from a skb
  2786. * @skb: skb to get stamp from
  2787. * @stamp: pointer to struct timeval to store stamp in
  2788. *
  2789. * Timestamps are stored in the skb as offsets to a base timestamp.
  2790. * This function converts the offset back to a struct timeval and stores
  2791. * it in stamp.
  2792. */
  2793. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2794. struct timeval *stamp)
  2795. {
  2796. *stamp = ktime_to_timeval(skb->tstamp);
  2797. }
  2798. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2799. struct timespec *stamp)
  2800. {
  2801. *stamp = ktime_to_timespec(skb->tstamp);
  2802. }
  2803. static inline void __net_timestamp(struct sk_buff *skb)
  2804. {
  2805. skb->tstamp = ktime_get_real();
  2806. }
  2807. static inline ktime_t net_timedelta(ktime_t t)
  2808. {
  2809. return ktime_sub(ktime_get_real(), t);
  2810. }
  2811. static inline ktime_t net_invalid_timestamp(void)
  2812. {
  2813. return 0;
  2814. }
  2815. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2816. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2817. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2818. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2819. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2820. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2821. {
  2822. }
  2823. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2824. {
  2825. return false;
  2826. }
  2827. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2828. /**
  2829. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2830. *
  2831. * PHY drivers may accept clones of transmitted packets for
  2832. * timestamping via their phy_driver.txtstamp method. These drivers
  2833. * must call this function to return the skb back to the stack with a
  2834. * timestamp.
  2835. *
  2836. * @skb: clone of the the original outgoing packet
  2837. * @hwtstamps: hardware time stamps
  2838. *
  2839. */
  2840. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2841. struct skb_shared_hwtstamps *hwtstamps);
  2842. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2843. struct skb_shared_hwtstamps *hwtstamps,
  2844. struct sock *sk, int tstype);
  2845. /**
  2846. * skb_tstamp_tx - queue clone of skb with send time stamps
  2847. * @orig_skb: the original outgoing packet
  2848. * @hwtstamps: hardware time stamps, may be NULL if not available
  2849. *
  2850. * If the skb has a socket associated, then this function clones the
  2851. * skb (thus sharing the actual data and optional structures), stores
  2852. * the optional hardware time stamping information (if non NULL) or
  2853. * generates a software time stamp (otherwise), then queues the clone
  2854. * to the error queue of the socket. Errors are silently ignored.
  2855. */
  2856. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2857. struct skb_shared_hwtstamps *hwtstamps);
  2858. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2859. {
  2860. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2861. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2862. skb_tstamp_tx(skb, NULL);
  2863. }
  2864. /**
  2865. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2866. *
  2867. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2868. * function immediately before giving the sk_buff to the MAC hardware.
  2869. *
  2870. * Specifically, one should make absolutely sure that this function is
  2871. * called before TX completion of this packet can trigger. Otherwise
  2872. * the packet could potentially already be freed.
  2873. *
  2874. * @skb: A socket buffer.
  2875. */
  2876. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2877. {
  2878. skb_clone_tx_timestamp(skb);
  2879. sw_tx_timestamp(skb);
  2880. }
  2881. /**
  2882. * skb_complete_wifi_ack - deliver skb with wifi status
  2883. *
  2884. * @skb: the original outgoing packet
  2885. * @acked: ack status
  2886. *
  2887. */
  2888. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2889. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2890. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2891. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2892. {
  2893. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2894. skb->csum_valid ||
  2895. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2896. skb_checksum_start_offset(skb) >= 0));
  2897. }
  2898. /**
  2899. * skb_checksum_complete - Calculate checksum of an entire packet
  2900. * @skb: packet to process
  2901. *
  2902. * This function calculates the checksum over the entire packet plus
  2903. * the value of skb->csum. The latter can be used to supply the
  2904. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2905. * checksum.
  2906. *
  2907. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2908. * this function can be used to verify that checksum on received
  2909. * packets. In that case the function should return zero if the
  2910. * checksum is correct. In particular, this function will return zero
  2911. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2912. * hardware has already verified the correctness of the checksum.
  2913. */
  2914. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2915. {
  2916. return skb_csum_unnecessary(skb) ?
  2917. 0 : __skb_checksum_complete(skb);
  2918. }
  2919. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2920. {
  2921. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2922. if (skb->csum_level == 0)
  2923. skb->ip_summed = CHECKSUM_NONE;
  2924. else
  2925. skb->csum_level--;
  2926. }
  2927. }
  2928. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2929. {
  2930. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2931. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2932. skb->csum_level++;
  2933. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2934. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2935. skb->csum_level = 0;
  2936. }
  2937. }
  2938. /* Check if we need to perform checksum complete validation.
  2939. *
  2940. * Returns true if checksum complete is needed, false otherwise
  2941. * (either checksum is unnecessary or zero checksum is allowed).
  2942. */
  2943. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2944. bool zero_okay,
  2945. __sum16 check)
  2946. {
  2947. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2948. skb->csum_valid = 1;
  2949. __skb_decr_checksum_unnecessary(skb);
  2950. return false;
  2951. }
  2952. return true;
  2953. }
  2954. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2955. * in checksum_init.
  2956. */
  2957. #define CHECKSUM_BREAK 76
  2958. /* Unset checksum-complete
  2959. *
  2960. * Unset checksum complete can be done when packet is being modified
  2961. * (uncompressed for instance) and checksum-complete value is
  2962. * invalidated.
  2963. */
  2964. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2965. {
  2966. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2967. skb->ip_summed = CHECKSUM_NONE;
  2968. }
  2969. /* Validate (init) checksum based on checksum complete.
  2970. *
  2971. * Return values:
  2972. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2973. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2974. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2975. * non-zero: value of invalid checksum
  2976. *
  2977. */
  2978. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2979. bool complete,
  2980. __wsum psum)
  2981. {
  2982. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2983. if (!csum_fold(csum_add(psum, skb->csum))) {
  2984. skb->csum_valid = 1;
  2985. return 0;
  2986. }
  2987. }
  2988. skb->csum = psum;
  2989. if (complete || skb->len <= CHECKSUM_BREAK) {
  2990. __sum16 csum;
  2991. csum = __skb_checksum_complete(skb);
  2992. skb->csum_valid = !csum;
  2993. return csum;
  2994. }
  2995. return 0;
  2996. }
  2997. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2998. {
  2999. return 0;
  3000. }
  3001. /* Perform checksum validate (init). Note that this is a macro since we only
  3002. * want to calculate the pseudo header which is an input function if necessary.
  3003. * First we try to validate without any computation (checksum unnecessary) and
  3004. * then calculate based on checksum complete calling the function to compute
  3005. * pseudo header.
  3006. *
  3007. * Return values:
  3008. * 0: checksum is validated or try to in skb_checksum_complete
  3009. * non-zero: value of invalid checksum
  3010. */
  3011. #define __skb_checksum_validate(skb, proto, complete, \
  3012. zero_okay, check, compute_pseudo) \
  3013. ({ \
  3014. __sum16 __ret = 0; \
  3015. skb->csum_valid = 0; \
  3016. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  3017. __ret = __skb_checksum_validate_complete(skb, \
  3018. complete, compute_pseudo(skb, proto)); \
  3019. __ret; \
  3020. })
  3021. #define skb_checksum_init(skb, proto, compute_pseudo) \
  3022. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  3023. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  3024. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  3025. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  3026. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  3027. #define skb_checksum_validate_zero_check(skb, proto, check, \
  3028. compute_pseudo) \
  3029. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  3030. #define skb_checksum_simple_validate(skb) \
  3031. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  3032. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  3033. {
  3034. return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
  3035. }
  3036. static inline void __skb_checksum_convert(struct sk_buff *skb,
  3037. __sum16 check, __wsum pseudo)
  3038. {
  3039. skb->csum = ~pseudo;
  3040. skb->ip_summed = CHECKSUM_COMPLETE;
  3041. }
  3042. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  3043. do { \
  3044. if (__skb_checksum_convert_check(skb)) \
  3045. __skb_checksum_convert(skb, check, \
  3046. compute_pseudo(skb, proto)); \
  3047. } while (0)
  3048. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  3049. u16 start, u16 offset)
  3050. {
  3051. skb->ip_summed = CHECKSUM_PARTIAL;
  3052. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  3053. skb->csum_offset = offset - start;
  3054. }
  3055. /* Update skbuf and packet to reflect the remote checksum offload operation.
  3056. * When called, ptr indicates the starting point for skb->csum when
  3057. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  3058. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  3059. */
  3060. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  3061. int start, int offset, bool nopartial)
  3062. {
  3063. __wsum delta;
  3064. if (!nopartial) {
  3065. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  3066. return;
  3067. }
  3068. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  3069. __skb_checksum_complete(skb);
  3070. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  3071. }
  3072. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  3073. /* Adjust skb->csum since we changed the packet */
  3074. skb->csum = csum_add(skb->csum, delta);
  3075. }
  3076. static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
  3077. {
  3078. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  3079. return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
  3080. #else
  3081. return NULL;
  3082. #endif
  3083. }
  3084. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3085. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  3086. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  3087. {
  3088. if (nfct && atomic_dec_and_test(&nfct->use))
  3089. nf_conntrack_destroy(nfct);
  3090. }
  3091. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  3092. {
  3093. if (nfct)
  3094. atomic_inc(&nfct->use);
  3095. }
  3096. #endif
  3097. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3098. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  3099. {
  3100. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  3101. kfree(nf_bridge);
  3102. }
  3103. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  3104. {
  3105. if (nf_bridge)
  3106. atomic_inc(&nf_bridge->use);
  3107. }
  3108. #endif /* CONFIG_BRIDGE_NETFILTER */
  3109. static inline void nf_reset(struct sk_buff *skb)
  3110. {
  3111. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3112. nf_conntrack_put(skb_nfct(skb));
  3113. skb->_nfct = 0;
  3114. #endif
  3115. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3116. nf_bridge_put(skb->nf_bridge);
  3117. skb->nf_bridge = NULL;
  3118. #endif
  3119. }
  3120. static inline void nf_reset_trace(struct sk_buff *skb)
  3121. {
  3122. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3123. skb->nf_trace = 0;
  3124. #endif
  3125. }
  3126. /* Note: This doesn't put any conntrack and bridge info in dst. */
  3127. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  3128. bool copy)
  3129. {
  3130. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3131. dst->_nfct = src->_nfct;
  3132. nf_conntrack_get(skb_nfct(src));
  3133. #endif
  3134. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3135. dst->nf_bridge = src->nf_bridge;
  3136. nf_bridge_get(src->nf_bridge);
  3137. #endif
  3138. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3139. if (copy)
  3140. dst->nf_trace = src->nf_trace;
  3141. #endif
  3142. }
  3143. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  3144. {
  3145. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3146. nf_conntrack_put(skb_nfct(dst));
  3147. #endif
  3148. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3149. nf_bridge_put(dst->nf_bridge);
  3150. #endif
  3151. __nf_copy(dst, src, true);
  3152. }
  3153. #ifdef CONFIG_NETWORK_SECMARK
  3154. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3155. {
  3156. to->secmark = from->secmark;
  3157. }
  3158. static inline void skb_init_secmark(struct sk_buff *skb)
  3159. {
  3160. skb->secmark = 0;
  3161. }
  3162. #else
  3163. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3164. { }
  3165. static inline void skb_init_secmark(struct sk_buff *skb)
  3166. { }
  3167. #endif
  3168. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  3169. {
  3170. return !skb->destructor &&
  3171. #if IS_ENABLED(CONFIG_XFRM)
  3172. !skb->sp &&
  3173. #endif
  3174. !skb_nfct(skb) &&
  3175. !skb->_skb_refdst &&
  3176. !skb_has_frag_list(skb);
  3177. }
  3178. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  3179. {
  3180. skb->queue_mapping = queue_mapping;
  3181. }
  3182. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  3183. {
  3184. return skb->queue_mapping;
  3185. }
  3186. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  3187. {
  3188. to->queue_mapping = from->queue_mapping;
  3189. }
  3190. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  3191. {
  3192. skb->queue_mapping = rx_queue + 1;
  3193. }
  3194. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  3195. {
  3196. return skb->queue_mapping - 1;
  3197. }
  3198. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  3199. {
  3200. return skb->queue_mapping != 0;
  3201. }
  3202. static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
  3203. {
  3204. skb->dst_pending_confirm = val;
  3205. }
  3206. static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
  3207. {
  3208. return skb->dst_pending_confirm != 0;
  3209. }
  3210. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  3211. {
  3212. #ifdef CONFIG_XFRM
  3213. return skb->sp;
  3214. #else
  3215. return NULL;
  3216. #endif
  3217. }
  3218. /* Keeps track of mac header offset relative to skb->head.
  3219. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  3220. * For non-tunnel skb it points to skb_mac_header() and for
  3221. * tunnel skb it points to outer mac header.
  3222. * Keeps track of level of encapsulation of network headers.
  3223. */
  3224. struct skb_gso_cb {
  3225. union {
  3226. int mac_offset;
  3227. int data_offset;
  3228. };
  3229. int encap_level;
  3230. __wsum csum;
  3231. __u16 csum_start;
  3232. };
  3233. #define SKB_SGO_CB_OFFSET 32
  3234. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
  3235. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  3236. {
  3237. return (skb_mac_header(inner_skb) - inner_skb->head) -
  3238. SKB_GSO_CB(inner_skb)->mac_offset;
  3239. }
  3240. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  3241. {
  3242. int new_headroom, headroom;
  3243. int ret;
  3244. headroom = skb_headroom(skb);
  3245. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  3246. if (ret)
  3247. return ret;
  3248. new_headroom = skb_headroom(skb);
  3249. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  3250. return 0;
  3251. }
  3252. static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
  3253. {
  3254. /* Do not update partial checksums if remote checksum is enabled. */
  3255. if (skb->remcsum_offload)
  3256. return;
  3257. SKB_GSO_CB(skb)->csum = res;
  3258. SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
  3259. }
  3260. /* Compute the checksum for a gso segment. First compute the checksum value
  3261. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  3262. * then add in skb->csum (checksum from csum_start to end of packet).
  3263. * skb->csum and csum_start are then updated to reflect the checksum of the
  3264. * resultant packet starting from the transport header-- the resultant checksum
  3265. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  3266. * header.
  3267. */
  3268. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  3269. {
  3270. unsigned char *csum_start = skb_transport_header(skb);
  3271. int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
  3272. __wsum partial = SKB_GSO_CB(skb)->csum;
  3273. SKB_GSO_CB(skb)->csum = res;
  3274. SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
  3275. return csum_fold(csum_partial(csum_start, plen, partial));
  3276. }
  3277. static inline bool skb_is_gso(const struct sk_buff *skb)
  3278. {
  3279. return skb_shinfo(skb)->gso_size;
  3280. }
  3281. /* Note: Should be called only if skb_is_gso(skb) is true */
  3282. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  3283. {
  3284. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  3285. }
  3286. static inline void skb_gso_reset(struct sk_buff *skb)
  3287. {
  3288. skb_shinfo(skb)->gso_size = 0;
  3289. skb_shinfo(skb)->gso_segs = 0;
  3290. skb_shinfo(skb)->gso_type = 0;
  3291. }
  3292. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  3293. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  3294. {
  3295. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  3296. * wanted then gso_type will be set. */
  3297. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  3298. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  3299. unlikely(shinfo->gso_type == 0)) {
  3300. __skb_warn_lro_forwarding(skb);
  3301. return true;
  3302. }
  3303. return false;
  3304. }
  3305. static inline void skb_forward_csum(struct sk_buff *skb)
  3306. {
  3307. /* Unfortunately we don't support this one. Any brave souls? */
  3308. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3309. skb->ip_summed = CHECKSUM_NONE;
  3310. }
  3311. /**
  3312. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3313. * @skb: skb to check
  3314. *
  3315. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3316. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3317. * use this helper, to document places where we make this assertion.
  3318. */
  3319. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3320. {
  3321. #ifdef DEBUG
  3322. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3323. #endif
  3324. }
  3325. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3326. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3327. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3328. unsigned int transport_len,
  3329. __sum16(*skb_chkf)(struct sk_buff *skb));
  3330. /**
  3331. * skb_head_is_locked - Determine if the skb->head is locked down
  3332. * @skb: skb to check
  3333. *
  3334. * The head on skbs build around a head frag can be removed if they are
  3335. * not cloned. This function returns true if the skb head is locked down
  3336. * due to either being allocated via kmalloc, or by being a clone with
  3337. * multiple references to the head.
  3338. */
  3339. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3340. {
  3341. return !skb->head_frag || skb_cloned(skb);
  3342. }
  3343. /**
  3344. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3345. *
  3346. * @skb: GSO skb
  3347. *
  3348. * skb_gso_network_seglen is used to determine the real size of the
  3349. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3350. *
  3351. * The MAC/L2 header is not accounted for.
  3352. */
  3353. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3354. {
  3355. unsigned int hdr_len = skb_transport_header(skb) -
  3356. skb_network_header(skb);
  3357. return hdr_len + skb_gso_transport_seglen(skb);
  3358. }
  3359. /* Local Checksum Offload.
  3360. * Compute outer checksum based on the assumption that the
  3361. * inner checksum will be offloaded later.
  3362. * See Documentation/networking/checksum-offloads.txt for
  3363. * explanation of how this works.
  3364. * Fill in outer checksum adjustment (e.g. with sum of outer
  3365. * pseudo-header) before calling.
  3366. * Also ensure that inner checksum is in linear data area.
  3367. */
  3368. static inline __wsum lco_csum(struct sk_buff *skb)
  3369. {
  3370. unsigned char *csum_start = skb_checksum_start(skb);
  3371. unsigned char *l4_hdr = skb_transport_header(skb);
  3372. __wsum partial;
  3373. /* Start with complement of inner checksum adjustment */
  3374. partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
  3375. skb->csum_offset));
  3376. /* Add in checksum of our headers (incl. outer checksum
  3377. * adjustment filled in by caller) and return result.
  3378. */
  3379. return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
  3380. }
  3381. #endif /* __KERNEL__ */
  3382. #endif /* _LINUX_SKBUFF_H */