skbuff.h 106 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472347334743475347634773478347934803481348234833484348534863487348834893490349134923493349434953496349734983499350035013502350335043505350635073508350935103511351235133514351535163517351835193520352135223523352435253526352735283529353035313532353335343535353635373538353935403541354235433544354535463547354835493550355135523553355435553556355735583559356035613562356335643565356635673568356935703571357235733574357535763577357835793580358135823583358435853586358735883589359035913592359335943595359635973598359936003601360236033604360536063607360836093610361136123613361436153616361736183619362036213622362336243625362636273628362936303631363236333634363536363637363836393640364136423643364436453646364736483649365036513652365336543655365636573658365936603661366236633664366536663667366836693670367136723673367436753676367736783679368036813682368336843685368636873688
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/kernel.h>
  16. #include <linux/kmemcheck.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/bug.h>
  20. #include <linux/cache.h>
  21. #include <linux/rbtree.h>
  22. #include <linux/socket.h>
  23. #include <linux/atomic.h>
  24. #include <asm/types.h>
  25. #include <linux/spinlock.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #include <linux/rcupdate.h>
  30. #include <linux/hrtimer.h>
  31. #include <linux/dma-mapping.h>
  32. #include <linux/netdev_features.h>
  33. #include <linux/sched.h>
  34. #include <net/flow_dissector.h>
  35. #include <linux/splice.h>
  36. #include <linux/in6.h>
  37. #include <net/flow.h>
  38. /* The interface for checksum offload between the stack and networking drivers
  39. * is as follows...
  40. *
  41. * A. IP checksum related features
  42. *
  43. * Drivers advertise checksum offload capabilities in the features of a device.
  44. * From the stack's point of view these are capabilities offered by the driver,
  45. * a driver typically only advertises features that it is capable of offloading
  46. * to its device.
  47. *
  48. * The checksum related features are:
  49. *
  50. * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
  51. * IP (one's complement) checksum for any combination
  52. * of protocols or protocol layering. The checksum is
  53. * computed and set in a packet per the CHECKSUM_PARTIAL
  54. * interface (see below).
  55. *
  56. * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
  57. * TCP or UDP packets over IPv4. These are specifically
  58. * unencapsulated packets of the form IPv4|TCP or
  59. * IPv4|UDP where the Protocol field in the IPv4 header
  60. * is TCP or UDP. The IPv4 header may contain IP options
  61. * This feature cannot be set in features for a device
  62. * with NETIF_F_HW_CSUM also set. This feature is being
  63. * DEPRECATED (see below).
  64. *
  65. * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
  66. * TCP or UDP packets over IPv6. These are specifically
  67. * unencapsulated packets of the form IPv6|TCP or
  68. * IPv4|UDP where the Next Header field in the IPv6
  69. * header is either TCP or UDP. IPv6 extension headers
  70. * are not supported with this feature. This feature
  71. * cannot be set in features for a device with
  72. * NETIF_F_HW_CSUM also set. This feature is being
  73. * DEPRECATED (see below).
  74. *
  75. * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
  76. * This flag is used only used to disable the RX checksum
  77. * feature for a device. The stack will accept receive
  78. * checksum indication in packets received on a device
  79. * regardless of whether NETIF_F_RXCSUM is set.
  80. *
  81. * B. Checksumming of received packets by device. Indication of checksum
  82. * verification is in set skb->ip_summed. Possible values are:
  83. *
  84. * CHECKSUM_NONE:
  85. *
  86. * Device did not checksum this packet e.g. due to lack of capabilities.
  87. * The packet contains full (though not verified) checksum in packet but
  88. * not in skb->csum. Thus, skb->csum is undefined in this case.
  89. *
  90. * CHECKSUM_UNNECESSARY:
  91. *
  92. * The hardware you're dealing with doesn't calculate the full checksum
  93. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  94. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  95. * if their checksums are okay. skb->csum is still undefined in this case
  96. * though. A driver or device must never modify the checksum field in the
  97. * packet even if checksum is verified.
  98. *
  99. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  100. * TCP: IPv6 and IPv4.
  101. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  102. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  103. * may perform further validation in this case.
  104. * GRE: only if the checksum is present in the header.
  105. * SCTP: indicates the CRC in SCTP header has been validated.
  106. *
  107. * skb->csum_level indicates the number of consecutive checksums found in
  108. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  109. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  110. * and a device is able to verify the checksums for UDP (possibly zero),
  111. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  112. * two. If the device were only able to verify the UDP checksum and not
  113. * GRE, either because it doesn't support GRE checksum of because GRE
  114. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  115. * not considered in this case).
  116. *
  117. * CHECKSUM_COMPLETE:
  118. *
  119. * This is the most generic way. The device supplied checksum of the _whole_
  120. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  121. * hardware doesn't need to parse L3/L4 headers to implement this.
  122. *
  123. * Note: Even if device supports only some protocols, but is able to produce
  124. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  125. *
  126. * CHECKSUM_PARTIAL:
  127. *
  128. * A checksum is set up to be offloaded to a device as described in the
  129. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  130. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  131. * on the same host, or it may be set in the input path in GRO or remote
  132. * checksum offload. For the purposes of checksum verification, the checksum
  133. * referred to by skb->csum_start + skb->csum_offset and any preceding
  134. * checksums in the packet are considered verified. Any checksums in the
  135. * packet that are after the checksum being offloaded are not considered to
  136. * be verified.
  137. *
  138. * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
  139. * in the skb->ip_summed for a packet. Values are:
  140. *
  141. * CHECKSUM_PARTIAL:
  142. *
  143. * The driver is required to checksum the packet as seen by hard_start_xmit()
  144. * from skb->csum_start up to the end, and to record/write the checksum at
  145. * offset skb->csum_start + skb->csum_offset. A driver may verify that the
  146. * csum_start and csum_offset values are valid values given the length and
  147. * offset of the packet, however they should not attempt to validate that the
  148. * checksum refers to a legitimate transport layer checksum-- it is the
  149. * purview of the stack to validate that csum_start and csum_offset are set
  150. * correctly.
  151. *
  152. * When the stack requests checksum offload for a packet, the driver MUST
  153. * ensure that the checksum is set correctly. A driver can either offload the
  154. * checksum calculation to the device, or call skb_checksum_help (in the case
  155. * that the device does not support offload for a particular checksum).
  156. *
  157. * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
  158. * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
  159. * checksum offload capability. If a device has limited checksum capabilities
  160. * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
  161. * described above) a helper function can be called to resolve
  162. * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
  163. * function takes a spec argument that describes the protocol layer that is
  164. * supported for checksum offload and can be called for each packet. If a
  165. * packet does not match the specification for offload, skb_checksum_help
  166. * is called to resolve the checksum.
  167. *
  168. * CHECKSUM_NONE:
  169. *
  170. * The skb was already checksummed by the protocol, or a checksum is not
  171. * required.
  172. *
  173. * CHECKSUM_UNNECESSARY:
  174. *
  175. * This has the same meaning on as CHECKSUM_NONE for checksum offload on
  176. * output.
  177. *
  178. * CHECKSUM_COMPLETE:
  179. * Not used in checksum output. If a driver observes a packet with this value
  180. * set in skbuff, if should treat as CHECKSUM_NONE being set.
  181. *
  182. * D. Non-IP checksum (CRC) offloads
  183. *
  184. * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
  185. * offloading the SCTP CRC in a packet. To perform this offload the stack
  186. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  187. * accordingly. Note the there is no indication in the skbuff that the
  188. * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
  189. * both IP checksum offload and SCTP CRC offload must verify which offload
  190. * is configured for a packet presumably by inspecting packet headers.
  191. *
  192. * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
  193. * offloading the FCOE CRC in a packet. To perform this offload the stack
  194. * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
  195. * accordingly. Note the there is no indication in the skbuff that the
  196. * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
  197. * both IP checksum offload and FCOE CRC offload must verify which offload
  198. * is configured for a packet presumably by inspecting packet headers.
  199. *
  200. * E. Checksumming on output with GSO.
  201. *
  202. * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
  203. * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
  204. * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
  205. * part of the GSO operation is implied. If a checksum is being offloaded
  206. * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
  207. * are set to refer to the outermost checksum being offload (two offloaded
  208. * checksums are possible with UDP encapsulation).
  209. */
  210. /* Don't change this without changing skb_csum_unnecessary! */
  211. #define CHECKSUM_NONE 0
  212. #define CHECKSUM_UNNECESSARY 1
  213. #define CHECKSUM_COMPLETE 2
  214. #define CHECKSUM_PARTIAL 3
  215. /* Maximum value in skb->csum_level */
  216. #define SKB_MAX_CSUM_LEVEL 3
  217. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  218. #define SKB_WITH_OVERHEAD(X) \
  219. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  220. #define SKB_MAX_ORDER(X, ORDER) \
  221. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  222. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  223. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  224. /* return minimum truesize of one skb containing X bytes of data */
  225. #define SKB_TRUESIZE(X) ((X) + \
  226. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  227. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  228. struct net_device;
  229. struct scatterlist;
  230. struct pipe_inode_info;
  231. struct iov_iter;
  232. struct napi_struct;
  233. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  234. struct nf_conntrack {
  235. atomic_t use;
  236. };
  237. #endif
  238. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  239. struct nf_bridge_info {
  240. atomic_t use;
  241. enum {
  242. BRNF_PROTO_UNCHANGED,
  243. BRNF_PROTO_8021Q,
  244. BRNF_PROTO_PPPOE
  245. } orig_proto:8;
  246. u8 pkt_otherhost:1;
  247. u8 in_prerouting:1;
  248. u8 bridged_dnat:1;
  249. __u16 frag_max_size;
  250. struct net_device *physindev;
  251. /* always valid & non-NULL from FORWARD on, for physdev match */
  252. struct net_device *physoutdev;
  253. union {
  254. /* prerouting: detect dnat in orig/reply direction */
  255. __be32 ipv4_daddr;
  256. struct in6_addr ipv6_daddr;
  257. /* after prerouting + nat detected: store original source
  258. * mac since neigh resolution overwrites it, only used while
  259. * skb is out in neigh layer.
  260. */
  261. char neigh_header[8];
  262. };
  263. };
  264. #endif
  265. struct sk_buff_head {
  266. /* These two members must be first. */
  267. struct sk_buff *next;
  268. struct sk_buff *prev;
  269. __u32 qlen;
  270. spinlock_t lock;
  271. };
  272. struct sk_buff;
  273. /* To allow 64K frame to be packed as single skb without frag_list we
  274. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  275. * buffers which do not start on a page boundary.
  276. *
  277. * Since GRO uses frags we allocate at least 16 regardless of page
  278. * size.
  279. */
  280. #if (65536/PAGE_SIZE + 1) < 16
  281. #define MAX_SKB_FRAGS 16UL
  282. #else
  283. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  284. #endif
  285. extern int sysctl_max_skb_frags;
  286. typedef struct skb_frag_struct skb_frag_t;
  287. struct skb_frag_struct {
  288. struct {
  289. struct page *p;
  290. } page;
  291. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  292. __u32 page_offset;
  293. __u32 size;
  294. #else
  295. __u16 page_offset;
  296. __u16 size;
  297. #endif
  298. };
  299. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  300. {
  301. return frag->size;
  302. }
  303. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  304. {
  305. frag->size = size;
  306. }
  307. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  308. {
  309. frag->size += delta;
  310. }
  311. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  312. {
  313. frag->size -= delta;
  314. }
  315. #define HAVE_HW_TIME_STAMP
  316. /**
  317. * struct skb_shared_hwtstamps - hardware time stamps
  318. * @hwtstamp: hardware time stamp transformed into duration
  319. * since arbitrary point in time
  320. *
  321. * Software time stamps generated by ktime_get_real() are stored in
  322. * skb->tstamp.
  323. *
  324. * hwtstamps can only be compared against other hwtstamps from
  325. * the same device.
  326. *
  327. * This structure is attached to packets as part of the
  328. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  329. */
  330. struct skb_shared_hwtstamps {
  331. ktime_t hwtstamp;
  332. };
  333. /* Definitions for tx_flags in struct skb_shared_info */
  334. enum {
  335. /* generate hardware time stamp */
  336. SKBTX_HW_TSTAMP = 1 << 0,
  337. /* generate software time stamp when queueing packet to NIC */
  338. SKBTX_SW_TSTAMP = 1 << 1,
  339. /* device driver is going to provide hardware time stamp */
  340. SKBTX_IN_PROGRESS = 1 << 2,
  341. /* device driver supports TX zero-copy buffers */
  342. SKBTX_DEV_ZEROCOPY = 1 << 3,
  343. /* generate wifi status information (where possible) */
  344. SKBTX_WIFI_STATUS = 1 << 4,
  345. /* This indicates at least one fragment might be overwritten
  346. * (as in vmsplice(), sendfile() ...)
  347. * If we need to compute a TX checksum, we'll need to copy
  348. * all frags to avoid possible bad checksum
  349. */
  350. SKBTX_SHARED_FRAG = 1 << 5,
  351. /* generate software time stamp when entering packet scheduling */
  352. SKBTX_SCHED_TSTAMP = 1 << 6,
  353. /* generate software timestamp on peer data acknowledgment */
  354. SKBTX_ACK_TSTAMP = 1 << 7,
  355. };
  356. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  357. SKBTX_SCHED_TSTAMP | \
  358. SKBTX_ACK_TSTAMP)
  359. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  360. /*
  361. * The callback notifies userspace to release buffers when skb DMA is done in
  362. * lower device, the skb last reference should be 0 when calling this.
  363. * The zerocopy_success argument is true if zero copy transmit occurred,
  364. * false on data copy or out of memory error caused by data copy attempt.
  365. * The ctx field is used to track device context.
  366. * The desc field is used to track userspace buffer index.
  367. */
  368. struct ubuf_info {
  369. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  370. void *ctx;
  371. unsigned long desc;
  372. };
  373. /* This data is invariant across clones and lives at
  374. * the end of the header data, ie. at skb->end.
  375. */
  376. struct skb_shared_info {
  377. unsigned char nr_frags;
  378. __u8 tx_flags;
  379. unsigned short gso_size;
  380. /* Warning: this field is not always filled in (UFO)! */
  381. unsigned short gso_segs;
  382. unsigned short gso_type;
  383. struct sk_buff *frag_list;
  384. struct skb_shared_hwtstamps hwtstamps;
  385. u32 tskey;
  386. __be32 ip6_frag_id;
  387. /*
  388. * Warning : all fields before dataref are cleared in __alloc_skb()
  389. */
  390. atomic_t dataref;
  391. /* Intermediate layers must ensure that destructor_arg
  392. * remains valid until skb destructor */
  393. void * destructor_arg;
  394. /* must be last field, see pskb_expand_head() */
  395. skb_frag_t frags[MAX_SKB_FRAGS];
  396. };
  397. /* We divide dataref into two halves. The higher 16 bits hold references
  398. * to the payload part of skb->data. The lower 16 bits hold references to
  399. * the entire skb->data. A clone of a headerless skb holds the length of
  400. * the header in skb->hdr_len.
  401. *
  402. * All users must obey the rule that the skb->data reference count must be
  403. * greater than or equal to the payload reference count.
  404. *
  405. * Holding a reference to the payload part means that the user does not
  406. * care about modifications to the header part of skb->data.
  407. */
  408. #define SKB_DATAREF_SHIFT 16
  409. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  410. enum {
  411. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  412. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  413. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  414. };
  415. enum {
  416. SKB_GSO_TCPV4 = 1 << 0,
  417. SKB_GSO_UDP = 1 << 1,
  418. /* This indicates the skb is from an untrusted source. */
  419. SKB_GSO_DODGY = 1 << 2,
  420. /* This indicates the tcp segment has CWR set. */
  421. SKB_GSO_TCP_ECN = 1 << 3,
  422. SKB_GSO_TCPV6 = 1 << 4,
  423. SKB_GSO_FCOE = 1 << 5,
  424. SKB_GSO_GRE = 1 << 6,
  425. SKB_GSO_GRE_CSUM = 1 << 7,
  426. SKB_GSO_IPIP = 1 << 8,
  427. SKB_GSO_SIT = 1 << 9,
  428. SKB_GSO_UDP_TUNNEL = 1 << 10,
  429. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  430. SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
  431. };
  432. #if BITS_PER_LONG > 32
  433. #define NET_SKBUFF_DATA_USES_OFFSET 1
  434. #endif
  435. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  436. typedef unsigned int sk_buff_data_t;
  437. #else
  438. typedef unsigned char *sk_buff_data_t;
  439. #endif
  440. /**
  441. * struct skb_mstamp - multi resolution time stamps
  442. * @stamp_us: timestamp in us resolution
  443. * @stamp_jiffies: timestamp in jiffies
  444. */
  445. struct skb_mstamp {
  446. union {
  447. u64 v64;
  448. struct {
  449. u32 stamp_us;
  450. u32 stamp_jiffies;
  451. };
  452. };
  453. };
  454. /**
  455. * skb_mstamp_get - get current timestamp
  456. * @cl: place to store timestamps
  457. */
  458. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  459. {
  460. u64 val = local_clock();
  461. do_div(val, NSEC_PER_USEC);
  462. cl->stamp_us = (u32)val;
  463. cl->stamp_jiffies = (u32)jiffies;
  464. }
  465. /**
  466. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  467. * @t1: pointer to newest sample
  468. * @t0: pointer to oldest sample
  469. */
  470. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  471. const struct skb_mstamp *t0)
  472. {
  473. s32 delta_us = t1->stamp_us - t0->stamp_us;
  474. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  475. /* If delta_us is negative, this might be because interval is too big,
  476. * or local_clock() drift is too big : fallback using jiffies.
  477. */
  478. if (delta_us <= 0 ||
  479. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  480. delta_us = jiffies_to_usecs(delta_jiffies);
  481. return delta_us;
  482. }
  483. static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
  484. const struct skb_mstamp *t0)
  485. {
  486. s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
  487. if (!diff)
  488. diff = t1->stamp_us - t0->stamp_us;
  489. return diff > 0;
  490. }
  491. /**
  492. * struct sk_buff - socket buffer
  493. * @next: Next buffer in list
  494. * @prev: Previous buffer in list
  495. * @tstamp: Time we arrived/left
  496. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  497. * @sk: Socket we are owned by
  498. * @dev: Device we arrived on/are leaving by
  499. * @cb: Control buffer. Free for use by every layer. Put private vars here
  500. * @_skb_refdst: destination entry (with norefcount bit)
  501. * @sp: the security path, used for xfrm
  502. * @len: Length of actual data
  503. * @data_len: Data length
  504. * @mac_len: Length of link layer header
  505. * @hdr_len: writable header length of cloned skb
  506. * @csum: Checksum (must include start/offset pair)
  507. * @csum_start: Offset from skb->head where checksumming should start
  508. * @csum_offset: Offset from csum_start where checksum should be stored
  509. * @priority: Packet queueing priority
  510. * @ignore_df: allow local fragmentation
  511. * @cloned: Head may be cloned (check refcnt to be sure)
  512. * @ip_summed: Driver fed us an IP checksum
  513. * @nohdr: Payload reference only, must not modify header
  514. * @nfctinfo: Relationship of this skb to the connection
  515. * @pkt_type: Packet class
  516. * @fclone: skbuff clone status
  517. * @ipvs_property: skbuff is owned by ipvs
  518. * @peeked: this packet has been seen already, so stats have been
  519. * done for it, don't do them again
  520. * @nf_trace: netfilter packet trace flag
  521. * @protocol: Packet protocol from driver
  522. * @destructor: Destruct function
  523. * @nfct: Associated connection, if any
  524. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  525. * @skb_iif: ifindex of device we arrived on
  526. * @tc_index: Traffic control index
  527. * @tc_verd: traffic control verdict
  528. * @hash: the packet hash
  529. * @queue_mapping: Queue mapping for multiqueue devices
  530. * @xmit_more: More SKBs are pending for this queue
  531. * @ndisc_nodetype: router type (from link layer)
  532. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  533. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  534. * ports.
  535. * @sw_hash: indicates hash was computed in software stack
  536. * @wifi_acked_valid: wifi_acked was set
  537. * @wifi_acked: whether frame was acked on wifi or not
  538. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  539. * @napi_id: id of the NAPI struct this skb came from
  540. * @secmark: security marking
  541. * @offload_fwd_mark: fwding offload mark
  542. * @mark: Generic packet mark
  543. * @vlan_proto: vlan encapsulation protocol
  544. * @vlan_tci: vlan tag control information
  545. * @inner_protocol: Protocol (encapsulation)
  546. * @inner_transport_header: Inner transport layer header (encapsulation)
  547. * @inner_network_header: Network layer header (encapsulation)
  548. * @inner_mac_header: Link layer header (encapsulation)
  549. * @transport_header: Transport layer header
  550. * @network_header: Network layer header
  551. * @mac_header: Link layer header
  552. * @tail: Tail pointer
  553. * @end: End pointer
  554. * @head: Head of buffer
  555. * @data: Data head pointer
  556. * @truesize: Buffer size
  557. * @users: User count - see {datagram,tcp}.c
  558. */
  559. struct sk_buff {
  560. union {
  561. struct {
  562. /* These two members must be first. */
  563. struct sk_buff *next;
  564. struct sk_buff *prev;
  565. union {
  566. ktime_t tstamp;
  567. struct skb_mstamp skb_mstamp;
  568. };
  569. };
  570. struct rb_node rbnode; /* used in netem & tcp stack */
  571. };
  572. struct sock *sk;
  573. struct net_device *dev;
  574. /*
  575. * This is the control buffer. It is free to use for every
  576. * layer. Please put your private variables there. If you
  577. * want to keep them across layers you have to do a skb_clone()
  578. * first. This is owned by whoever has the skb queued ATM.
  579. */
  580. char cb[48] __aligned(8);
  581. unsigned long _skb_refdst;
  582. void (*destructor)(struct sk_buff *skb);
  583. #ifdef CONFIG_XFRM
  584. struct sec_path *sp;
  585. #endif
  586. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  587. struct nf_conntrack *nfct;
  588. #endif
  589. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  590. struct nf_bridge_info *nf_bridge;
  591. #endif
  592. unsigned int len,
  593. data_len;
  594. __u16 mac_len,
  595. hdr_len;
  596. /* Following fields are _not_ copied in __copy_skb_header()
  597. * Note that queue_mapping is here mostly to fill a hole.
  598. */
  599. kmemcheck_bitfield_begin(flags1);
  600. __u16 queue_mapping;
  601. __u8 cloned:1,
  602. nohdr:1,
  603. fclone:2,
  604. peeked:1,
  605. head_frag:1,
  606. xmit_more:1;
  607. /* one bit hole */
  608. kmemcheck_bitfield_end(flags1);
  609. /* fields enclosed in headers_start/headers_end are copied
  610. * using a single memcpy() in __copy_skb_header()
  611. */
  612. /* private: */
  613. __u32 headers_start[0];
  614. /* public: */
  615. /* if you move pkt_type around you also must adapt those constants */
  616. #ifdef __BIG_ENDIAN_BITFIELD
  617. #define PKT_TYPE_MAX (7 << 5)
  618. #else
  619. #define PKT_TYPE_MAX 7
  620. #endif
  621. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  622. __u8 __pkt_type_offset[0];
  623. __u8 pkt_type:3;
  624. __u8 pfmemalloc:1;
  625. __u8 ignore_df:1;
  626. __u8 nfctinfo:3;
  627. __u8 nf_trace:1;
  628. __u8 ip_summed:2;
  629. __u8 ooo_okay:1;
  630. __u8 l4_hash:1;
  631. __u8 sw_hash:1;
  632. __u8 wifi_acked_valid:1;
  633. __u8 wifi_acked:1;
  634. __u8 no_fcs:1;
  635. /* Indicates the inner headers are valid in the skbuff. */
  636. __u8 encapsulation:1;
  637. __u8 encap_hdr_csum:1;
  638. __u8 csum_valid:1;
  639. __u8 csum_complete_sw:1;
  640. __u8 csum_level:2;
  641. __u8 csum_bad:1;
  642. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  643. __u8 ndisc_nodetype:2;
  644. #endif
  645. __u8 ipvs_property:1;
  646. __u8 inner_protocol_type:1;
  647. __u8 remcsum_offload:1;
  648. /* 3 or 5 bit hole */
  649. #ifdef CONFIG_NET_SCHED
  650. __u16 tc_index; /* traffic control index */
  651. #ifdef CONFIG_NET_CLS_ACT
  652. __u16 tc_verd; /* traffic control verdict */
  653. #endif
  654. #endif
  655. union {
  656. __wsum csum;
  657. struct {
  658. __u16 csum_start;
  659. __u16 csum_offset;
  660. };
  661. };
  662. __u32 priority;
  663. int skb_iif;
  664. __u32 hash;
  665. __be16 vlan_proto;
  666. __u16 vlan_tci;
  667. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  668. union {
  669. unsigned int napi_id;
  670. unsigned int sender_cpu;
  671. };
  672. #endif
  673. union {
  674. #ifdef CONFIG_NETWORK_SECMARK
  675. __u32 secmark;
  676. #endif
  677. #ifdef CONFIG_NET_SWITCHDEV
  678. __u32 offload_fwd_mark;
  679. #endif
  680. };
  681. union {
  682. __u32 mark;
  683. __u32 reserved_tailroom;
  684. };
  685. union {
  686. __be16 inner_protocol;
  687. __u8 inner_ipproto;
  688. };
  689. __u16 inner_transport_header;
  690. __u16 inner_network_header;
  691. __u16 inner_mac_header;
  692. __be16 protocol;
  693. __u16 transport_header;
  694. __u16 network_header;
  695. __u16 mac_header;
  696. /* private: */
  697. __u32 headers_end[0];
  698. /* public: */
  699. /* These elements must be at the end, see alloc_skb() for details. */
  700. sk_buff_data_t tail;
  701. sk_buff_data_t end;
  702. unsigned char *head,
  703. *data;
  704. unsigned int truesize;
  705. atomic_t users;
  706. };
  707. #ifdef __KERNEL__
  708. /*
  709. * Handling routines are only of interest to the kernel
  710. */
  711. #include <linux/slab.h>
  712. #define SKB_ALLOC_FCLONE 0x01
  713. #define SKB_ALLOC_RX 0x02
  714. #define SKB_ALLOC_NAPI 0x04
  715. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  716. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  717. {
  718. return unlikely(skb->pfmemalloc);
  719. }
  720. /*
  721. * skb might have a dst pointer attached, refcounted or not.
  722. * _skb_refdst low order bit is set if refcount was _not_ taken
  723. */
  724. #define SKB_DST_NOREF 1UL
  725. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  726. /**
  727. * skb_dst - returns skb dst_entry
  728. * @skb: buffer
  729. *
  730. * Returns skb dst_entry, regardless of reference taken or not.
  731. */
  732. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  733. {
  734. /* If refdst was not refcounted, check we still are in a
  735. * rcu_read_lock section
  736. */
  737. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  738. !rcu_read_lock_held() &&
  739. !rcu_read_lock_bh_held());
  740. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  741. }
  742. /**
  743. * skb_dst_set - sets skb dst
  744. * @skb: buffer
  745. * @dst: dst entry
  746. *
  747. * Sets skb dst, assuming a reference was taken on dst and should
  748. * be released by skb_dst_drop()
  749. */
  750. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  751. {
  752. skb->_skb_refdst = (unsigned long)dst;
  753. }
  754. /**
  755. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  756. * @skb: buffer
  757. * @dst: dst entry
  758. *
  759. * Sets skb dst, assuming a reference was not taken on dst.
  760. * If dst entry is cached, we do not take reference and dst_release
  761. * will be avoided by refdst_drop. If dst entry is not cached, we take
  762. * reference, so that last dst_release can destroy the dst immediately.
  763. */
  764. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  765. {
  766. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  767. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  768. }
  769. /**
  770. * skb_dst_is_noref - Test if skb dst isn't refcounted
  771. * @skb: buffer
  772. */
  773. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  774. {
  775. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  776. }
  777. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  778. {
  779. return (struct rtable *)skb_dst(skb);
  780. }
  781. void kfree_skb(struct sk_buff *skb);
  782. void kfree_skb_list(struct sk_buff *segs);
  783. void skb_tx_error(struct sk_buff *skb);
  784. void consume_skb(struct sk_buff *skb);
  785. void __kfree_skb(struct sk_buff *skb);
  786. extern struct kmem_cache *skbuff_head_cache;
  787. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  788. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  789. bool *fragstolen, int *delta_truesize);
  790. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  791. int node);
  792. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  793. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  794. static inline struct sk_buff *alloc_skb(unsigned int size,
  795. gfp_t priority)
  796. {
  797. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  798. }
  799. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  800. unsigned long data_len,
  801. int max_page_order,
  802. int *errcode,
  803. gfp_t gfp_mask);
  804. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  805. struct sk_buff_fclones {
  806. struct sk_buff skb1;
  807. struct sk_buff skb2;
  808. atomic_t fclone_ref;
  809. };
  810. /**
  811. * skb_fclone_busy - check if fclone is busy
  812. * @skb: buffer
  813. *
  814. * Returns true if skb is a fast clone, and its clone is not freed.
  815. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  816. * so we also check that this didnt happen.
  817. */
  818. static inline bool skb_fclone_busy(const struct sock *sk,
  819. const struct sk_buff *skb)
  820. {
  821. const struct sk_buff_fclones *fclones;
  822. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  823. return skb->fclone == SKB_FCLONE_ORIG &&
  824. atomic_read(&fclones->fclone_ref) > 1 &&
  825. fclones->skb2.sk == sk;
  826. }
  827. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  828. gfp_t priority)
  829. {
  830. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  831. }
  832. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  833. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  834. {
  835. return __alloc_skb_head(priority, -1);
  836. }
  837. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  838. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  839. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  840. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  841. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  842. gfp_t gfp_mask, bool fclone);
  843. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  844. gfp_t gfp_mask)
  845. {
  846. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  847. }
  848. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  849. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  850. unsigned int headroom);
  851. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  852. int newtailroom, gfp_t priority);
  853. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  854. int offset, int len);
  855. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  856. int len);
  857. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  858. int skb_pad(struct sk_buff *skb, int pad);
  859. #define dev_kfree_skb(a) consume_skb(a)
  860. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  861. int getfrag(void *from, char *to, int offset,
  862. int len, int odd, struct sk_buff *skb),
  863. void *from, int length);
  864. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  865. int offset, size_t size);
  866. struct skb_seq_state {
  867. __u32 lower_offset;
  868. __u32 upper_offset;
  869. __u32 frag_idx;
  870. __u32 stepped_offset;
  871. struct sk_buff *root_skb;
  872. struct sk_buff *cur_skb;
  873. __u8 *frag_data;
  874. };
  875. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  876. unsigned int to, struct skb_seq_state *st);
  877. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  878. struct skb_seq_state *st);
  879. void skb_abort_seq_read(struct skb_seq_state *st);
  880. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  881. unsigned int to, struct ts_config *config);
  882. /*
  883. * Packet hash types specify the type of hash in skb_set_hash.
  884. *
  885. * Hash types refer to the protocol layer addresses which are used to
  886. * construct a packet's hash. The hashes are used to differentiate or identify
  887. * flows of the protocol layer for the hash type. Hash types are either
  888. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  889. *
  890. * Properties of hashes:
  891. *
  892. * 1) Two packets in different flows have different hash values
  893. * 2) Two packets in the same flow should have the same hash value
  894. *
  895. * A hash at a higher layer is considered to be more specific. A driver should
  896. * set the most specific hash possible.
  897. *
  898. * A driver cannot indicate a more specific hash than the layer at which a hash
  899. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  900. *
  901. * A driver may indicate a hash level which is less specific than the
  902. * actual layer the hash was computed on. For instance, a hash computed
  903. * at L4 may be considered an L3 hash. This should only be done if the
  904. * driver can't unambiguously determine that the HW computed the hash at
  905. * the higher layer. Note that the "should" in the second property above
  906. * permits this.
  907. */
  908. enum pkt_hash_types {
  909. PKT_HASH_TYPE_NONE, /* Undefined type */
  910. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  911. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  912. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  913. };
  914. static inline void skb_clear_hash(struct sk_buff *skb)
  915. {
  916. skb->hash = 0;
  917. skb->sw_hash = 0;
  918. skb->l4_hash = 0;
  919. }
  920. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  921. {
  922. if (!skb->l4_hash)
  923. skb_clear_hash(skb);
  924. }
  925. static inline void
  926. __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
  927. {
  928. skb->l4_hash = is_l4;
  929. skb->sw_hash = is_sw;
  930. skb->hash = hash;
  931. }
  932. static inline void
  933. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  934. {
  935. /* Used by drivers to set hash from HW */
  936. __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
  937. }
  938. static inline void
  939. __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
  940. {
  941. __skb_set_hash(skb, hash, true, is_l4);
  942. }
  943. void __skb_get_hash(struct sk_buff *skb);
  944. u32 skb_get_poff(const struct sk_buff *skb);
  945. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  946. const struct flow_keys *keys, int hlen);
  947. __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
  948. void *data, int hlen_proto);
  949. static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
  950. int thoff, u8 ip_proto)
  951. {
  952. return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
  953. }
  954. void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  955. const struct flow_dissector_key *key,
  956. unsigned int key_count);
  957. bool __skb_flow_dissect(const struct sk_buff *skb,
  958. struct flow_dissector *flow_dissector,
  959. void *target_container,
  960. void *data, __be16 proto, int nhoff, int hlen,
  961. unsigned int flags);
  962. static inline bool skb_flow_dissect(const struct sk_buff *skb,
  963. struct flow_dissector *flow_dissector,
  964. void *target_container, unsigned int flags)
  965. {
  966. return __skb_flow_dissect(skb, flow_dissector, target_container,
  967. NULL, 0, 0, 0, flags);
  968. }
  969. static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
  970. struct flow_keys *flow,
  971. unsigned int flags)
  972. {
  973. memset(flow, 0, sizeof(*flow));
  974. return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
  975. NULL, 0, 0, 0, flags);
  976. }
  977. static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
  978. void *data, __be16 proto,
  979. int nhoff, int hlen,
  980. unsigned int flags)
  981. {
  982. memset(flow, 0, sizeof(*flow));
  983. return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
  984. data, proto, nhoff, hlen, flags);
  985. }
  986. static inline __u32 skb_get_hash(struct sk_buff *skb)
  987. {
  988. if (!skb->l4_hash && !skb->sw_hash)
  989. __skb_get_hash(skb);
  990. return skb->hash;
  991. }
  992. __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
  993. static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
  994. {
  995. if (!skb->l4_hash && !skb->sw_hash) {
  996. struct flow_keys keys;
  997. __u32 hash = __get_hash_from_flowi6(fl6, &keys);
  998. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  999. }
  1000. return skb->hash;
  1001. }
  1002. __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
  1003. static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
  1004. {
  1005. if (!skb->l4_hash && !skb->sw_hash) {
  1006. struct flow_keys keys;
  1007. __u32 hash = __get_hash_from_flowi4(fl4, &keys);
  1008. __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
  1009. }
  1010. return skb->hash;
  1011. }
  1012. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  1013. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  1014. {
  1015. return skb->hash;
  1016. }
  1017. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  1018. {
  1019. to->hash = from->hash;
  1020. to->sw_hash = from->sw_hash;
  1021. to->l4_hash = from->l4_hash;
  1022. };
  1023. static inline void skb_sender_cpu_clear(struct sk_buff *skb)
  1024. {
  1025. }
  1026. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1027. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1028. {
  1029. return skb->head + skb->end;
  1030. }
  1031. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1032. {
  1033. return skb->end;
  1034. }
  1035. #else
  1036. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  1037. {
  1038. return skb->end;
  1039. }
  1040. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  1041. {
  1042. return skb->end - skb->head;
  1043. }
  1044. #endif
  1045. /* Internal */
  1046. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  1047. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  1048. {
  1049. return &skb_shinfo(skb)->hwtstamps;
  1050. }
  1051. /**
  1052. * skb_queue_empty - check if a queue is empty
  1053. * @list: queue head
  1054. *
  1055. * Returns true if the queue is empty, false otherwise.
  1056. */
  1057. static inline int skb_queue_empty(const struct sk_buff_head *list)
  1058. {
  1059. return list->next == (const struct sk_buff *) list;
  1060. }
  1061. /**
  1062. * skb_queue_is_last - check if skb is the last entry in the queue
  1063. * @list: queue head
  1064. * @skb: buffer
  1065. *
  1066. * Returns true if @skb is the last buffer on the list.
  1067. */
  1068. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  1069. const struct sk_buff *skb)
  1070. {
  1071. return skb->next == (const struct sk_buff *) list;
  1072. }
  1073. /**
  1074. * skb_queue_is_first - check if skb is the first entry in the queue
  1075. * @list: queue head
  1076. * @skb: buffer
  1077. *
  1078. * Returns true if @skb is the first buffer on the list.
  1079. */
  1080. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  1081. const struct sk_buff *skb)
  1082. {
  1083. return skb->prev == (const struct sk_buff *) list;
  1084. }
  1085. /**
  1086. * skb_queue_next - return the next packet in the queue
  1087. * @list: queue head
  1088. * @skb: current buffer
  1089. *
  1090. * Return the next packet in @list after @skb. It is only valid to
  1091. * call this if skb_queue_is_last() evaluates to false.
  1092. */
  1093. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  1094. const struct sk_buff *skb)
  1095. {
  1096. /* This BUG_ON may seem severe, but if we just return then we
  1097. * are going to dereference garbage.
  1098. */
  1099. BUG_ON(skb_queue_is_last(list, skb));
  1100. return skb->next;
  1101. }
  1102. /**
  1103. * skb_queue_prev - return the prev packet in the queue
  1104. * @list: queue head
  1105. * @skb: current buffer
  1106. *
  1107. * Return the prev packet in @list before @skb. It is only valid to
  1108. * call this if skb_queue_is_first() evaluates to false.
  1109. */
  1110. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  1111. const struct sk_buff *skb)
  1112. {
  1113. /* This BUG_ON may seem severe, but if we just return then we
  1114. * are going to dereference garbage.
  1115. */
  1116. BUG_ON(skb_queue_is_first(list, skb));
  1117. return skb->prev;
  1118. }
  1119. /**
  1120. * skb_get - reference buffer
  1121. * @skb: buffer to reference
  1122. *
  1123. * Makes another reference to a socket buffer and returns a pointer
  1124. * to the buffer.
  1125. */
  1126. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  1127. {
  1128. atomic_inc(&skb->users);
  1129. return skb;
  1130. }
  1131. /*
  1132. * If users == 1, we are the only owner and are can avoid redundant
  1133. * atomic change.
  1134. */
  1135. /**
  1136. * skb_cloned - is the buffer a clone
  1137. * @skb: buffer to check
  1138. *
  1139. * Returns true if the buffer was generated with skb_clone() and is
  1140. * one of multiple shared copies of the buffer. Cloned buffers are
  1141. * shared data so must not be written to under normal circumstances.
  1142. */
  1143. static inline int skb_cloned(const struct sk_buff *skb)
  1144. {
  1145. return skb->cloned &&
  1146. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  1147. }
  1148. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  1149. {
  1150. might_sleep_if(gfpflags_allow_blocking(pri));
  1151. if (skb_cloned(skb))
  1152. return pskb_expand_head(skb, 0, 0, pri);
  1153. return 0;
  1154. }
  1155. /**
  1156. * skb_header_cloned - is the header a clone
  1157. * @skb: buffer to check
  1158. *
  1159. * Returns true if modifying the header part of the buffer requires
  1160. * the data to be copied.
  1161. */
  1162. static inline int skb_header_cloned(const struct sk_buff *skb)
  1163. {
  1164. int dataref;
  1165. if (!skb->cloned)
  1166. return 0;
  1167. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  1168. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  1169. return dataref != 1;
  1170. }
  1171. /**
  1172. * skb_header_release - release reference to header
  1173. * @skb: buffer to operate on
  1174. *
  1175. * Drop a reference to the header part of the buffer. This is done
  1176. * by acquiring a payload reference. You must not read from the header
  1177. * part of skb->data after this.
  1178. * Note : Check if you can use __skb_header_release() instead.
  1179. */
  1180. static inline void skb_header_release(struct sk_buff *skb)
  1181. {
  1182. BUG_ON(skb->nohdr);
  1183. skb->nohdr = 1;
  1184. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1185. }
  1186. /**
  1187. * __skb_header_release - release reference to header
  1188. * @skb: buffer to operate on
  1189. *
  1190. * Variant of skb_header_release() assuming skb is private to caller.
  1191. * We can avoid one atomic operation.
  1192. */
  1193. static inline void __skb_header_release(struct sk_buff *skb)
  1194. {
  1195. skb->nohdr = 1;
  1196. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1197. }
  1198. /**
  1199. * skb_shared - is the buffer shared
  1200. * @skb: buffer to check
  1201. *
  1202. * Returns true if more than one person has a reference to this
  1203. * buffer.
  1204. */
  1205. static inline int skb_shared(const struct sk_buff *skb)
  1206. {
  1207. return atomic_read(&skb->users) != 1;
  1208. }
  1209. /**
  1210. * skb_share_check - check if buffer is shared and if so clone it
  1211. * @skb: buffer to check
  1212. * @pri: priority for memory allocation
  1213. *
  1214. * If the buffer is shared the buffer is cloned and the old copy
  1215. * drops a reference. A new clone with a single reference is returned.
  1216. * If the buffer is not shared the original buffer is returned. When
  1217. * being called from interrupt status or with spinlocks held pri must
  1218. * be GFP_ATOMIC.
  1219. *
  1220. * NULL is returned on a memory allocation failure.
  1221. */
  1222. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1223. {
  1224. might_sleep_if(gfpflags_allow_blocking(pri));
  1225. if (skb_shared(skb)) {
  1226. struct sk_buff *nskb = skb_clone(skb, pri);
  1227. if (likely(nskb))
  1228. consume_skb(skb);
  1229. else
  1230. kfree_skb(skb);
  1231. skb = nskb;
  1232. }
  1233. return skb;
  1234. }
  1235. /*
  1236. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1237. * packets to handle cases where we have a local reader and forward
  1238. * and a couple of other messy ones. The normal one is tcpdumping
  1239. * a packet thats being forwarded.
  1240. */
  1241. /**
  1242. * skb_unshare - make a copy of a shared buffer
  1243. * @skb: buffer to check
  1244. * @pri: priority for memory allocation
  1245. *
  1246. * If the socket buffer is a clone then this function creates a new
  1247. * copy of the data, drops a reference count on the old copy and returns
  1248. * the new copy with the reference count at 1. If the buffer is not a clone
  1249. * the original buffer is returned. When called with a spinlock held or
  1250. * from interrupt state @pri must be %GFP_ATOMIC
  1251. *
  1252. * %NULL is returned on a memory allocation failure.
  1253. */
  1254. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1255. gfp_t pri)
  1256. {
  1257. might_sleep_if(gfpflags_allow_blocking(pri));
  1258. if (skb_cloned(skb)) {
  1259. struct sk_buff *nskb = skb_copy(skb, pri);
  1260. /* Free our shared copy */
  1261. if (likely(nskb))
  1262. consume_skb(skb);
  1263. else
  1264. kfree_skb(skb);
  1265. skb = nskb;
  1266. }
  1267. return skb;
  1268. }
  1269. /**
  1270. * skb_peek - peek at the head of an &sk_buff_head
  1271. * @list_: list to peek at
  1272. *
  1273. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1274. * be careful with this one. A peek leaves the buffer on the
  1275. * list and someone else may run off with it. You must hold
  1276. * the appropriate locks or have a private queue to do this.
  1277. *
  1278. * Returns %NULL for an empty list or a pointer to the head element.
  1279. * The reference count is not incremented and the reference is therefore
  1280. * volatile. Use with caution.
  1281. */
  1282. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1283. {
  1284. struct sk_buff *skb = list_->next;
  1285. if (skb == (struct sk_buff *)list_)
  1286. skb = NULL;
  1287. return skb;
  1288. }
  1289. /**
  1290. * skb_peek_next - peek skb following the given one from a queue
  1291. * @skb: skb to start from
  1292. * @list_: list to peek at
  1293. *
  1294. * Returns %NULL when the end of the list is met or a pointer to the
  1295. * next element. The reference count is not incremented and the
  1296. * reference is therefore volatile. Use with caution.
  1297. */
  1298. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1299. const struct sk_buff_head *list_)
  1300. {
  1301. struct sk_buff *next = skb->next;
  1302. if (next == (struct sk_buff *)list_)
  1303. next = NULL;
  1304. return next;
  1305. }
  1306. /**
  1307. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1308. * @list_: list to peek at
  1309. *
  1310. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1311. * be careful with this one. A peek leaves the buffer on the
  1312. * list and someone else may run off with it. You must hold
  1313. * the appropriate locks or have a private queue to do this.
  1314. *
  1315. * Returns %NULL for an empty list or a pointer to the tail element.
  1316. * The reference count is not incremented and the reference is therefore
  1317. * volatile. Use with caution.
  1318. */
  1319. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1320. {
  1321. struct sk_buff *skb = list_->prev;
  1322. if (skb == (struct sk_buff *)list_)
  1323. skb = NULL;
  1324. return skb;
  1325. }
  1326. /**
  1327. * skb_queue_len - get queue length
  1328. * @list_: list to measure
  1329. *
  1330. * Return the length of an &sk_buff queue.
  1331. */
  1332. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1333. {
  1334. return list_->qlen;
  1335. }
  1336. /**
  1337. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1338. * @list: queue to initialize
  1339. *
  1340. * This initializes only the list and queue length aspects of
  1341. * an sk_buff_head object. This allows to initialize the list
  1342. * aspects of an sk_buff_head without reinitializing things like
  1343. * the spinlock. It can also be used for on-stack sk_buff_head
  1344. * objects where the spinlock is known to not be used.
  1345. */
  1346. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1347. {
  1348. list->prev = list->next = (struct sk_buff *)list;
  1349. list->qlen = 0;
  1350. }
  1351. /*
  1352. * This function creates a split out lock class for each invocation;
  1353. * this is needed for now since a whole lot of users of the skb-queue
  1354. * infrastructure in drivers have different locking usage (in hardirq)
  1355. * than the networking core (in softirq only). In the long run either the
  1356. * network layer or drivers should need annotation to consolidate the
  1357. * main types of usage into 3 classes.
  1358. */
  1359. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1360. {
  1361. spin_lock_init(&list->lock);
  1362. __skb_queue_head_init(list);
  1363. }
  1364. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1365. struct lock_class_key *class)
  1366. {
  1367. skb_queue_head_init(list);
  1368. lockdep_set_class(&list->lock, class);
  1369. }
  1370. /*
  1371. * Insert an sk_buff on a list.
  1372. *
  1373. * The "__skb_xxxx()" functions are the non-atomic ones that
  1374. * can only be called with interrupts disabled.
  1375. */
  1376. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1377. struct sk_buff_head *list);
  1378. static inline void __skb_insert(struct sk_buff *newsk,
  1379. struct sk_buff *prev, struct sk_buff *next,
  1380. struct sk_buff_head *list)
  1381. {
  1382. newsk->next = next;
  1383. newsk->prev = prev;
  1384. next->prev = prev->next = newsk;
  1385. list->qlen++;
  1386. }
  1387. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1388. struct sk_buff *prev,
  1389. struct sk_buff *next)
  1390. {
  1391. struct sk_buff *first = list->next;
  1392. struct sk_buff *last = list->prev;
  1393. first->prev = prev;
  1394. prev->next = first;
  1395. last->next = next;
  1396. next->prev = last;
  1397. }
  1398. /**
  1399. * skb_queue_splice - join two skb lists, this is designed for stacks
  1400. * @list: the new list to add
  1401. * @head: the place to add it in the first list
  1402. */
  1403. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1404. struct sk_buff_head *head)
  1405. {
  1406. if (!skb_queue_empty(list)) {
  1407. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1408. head->qlen += list->qlen;
  1409. }
  1410. }
  1411. /**
  1412. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1413. * @list: the new list to add
  1414. * @head: the place to add it in the first list
  1415. *
  1416. * The list at @list is reinitialised
  1417. */
  1418. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1419. struct sk_buff_head *head)
  1420. {
  1421. if (!skb_queue_empty(list)) {
  1422. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1423. head->qlen += list->qlen;
  1424. __skb_queue_head_init(list);
  1425. }
  1426. }
  1427. /**
  1428. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1429. * @list: the new list to add
  1430. * @head: the place to add it in the first list
  1431. */
  1432. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1433. struct sk_buff_head *head)
  1434. {
  1435. if (!skb_queue_empty(list)) {
  1436. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1437. head->qlen += list->qlen;
  1438. }
  1439. }
  1440. /**
  1441. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1442. * @list: the new list to add
  1443. * @head: the place to add it in the first list
  1444. *
  1445. * Each of the lists is a queue.
  1446. * The list at @list is reinitialised
  1447. */
  1448. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1449. struct sk_buff_head *head)
  1450. {
  1451. if (!skb_queue_empty(list)) {
  1452. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1453. head->qlen += list->qlen;
  1454. __skb_queue_head_init(list);
  1455. }
  1456. }
  1457. /**
  1458. * __skb_queue_after - queue a buffer at the list head
  1459. * @list: list to use
  1460. * @prev: place after this buffer
  1461. * @newsk: buffer to queue
  1462. *
  1463. * Queue a buffer int the middle of a list. This function takes no locks
  1464. * and you must therefore hold required locks before calling it.
  1465. *
  1466. * A buffer cannot be placed on two lists at the same time.
  1467. */
  1468. static inline void __skb_queue_after(struct sk_buff_head *list,
  1469. struct sk_buff *prev,
  1470. struct sk_buff *newsk)
  1471. {
  1472. __skb_insert(newsk, prev, prev->next, list);
  1473. }
  1474. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1475. struct sk_buff_head *list);
  1476. static inline void __skb_queue_before(struct sk_buff_head *list,
  1477. struct sk_buff *next,
  1478. struct sk_buff *newsk)
  1479. {
  1480. __skb_insert(newsk, next->prev, next, list);
  1481. }
  1482. /**
  1483. * __skb_queue_head - queue a buffer at the list head
  1484. * @list: list to use
  1485. * @newsk: buffer to queue
  1486. *
  1487. * Queue a buffer at the start of a list. This function takes no locks
  1488. * and you must therefore hold required locks before calling it.
  1489. *
  1490. * A buffer cannot be placed on two lists at the same time.
  1491. */
  1492. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1493. static inline void __skb_queue_head(struct sk_buff_head *list,
  1494. struct sk_buff *newsk)
  1495. {
  1496. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1497. }
  1498. /**
  1499. * __skb_queue_tail - queue a buffer at the list tail
  1500. * @list: list to use
  1501. * @newsk: buffer to queue
  1502. *
  1503. * Queue a buffer at the end of a list. This function takes no locks
  1504. * and you must therefore hold required locks before calling it.
  1505. *
  1506. * A buffer cannot be placed on two lists at the same time.
  1507. */
  1508. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1509. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1510. struct sk_buff *newsk)
  1511. {
  1512. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1513. }
  1514. /*
  1515. * remove sk_buff from list. _Must_ be called atomically, and with
  1516. * the list known..
  1517. */
  1518. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1519. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1520. {
  1521. struct sk_buff *next, *prev;
  1522. list->qlen--;
  1523. next = skb->next;
  1524. prev = skb->prev;
  1525. skb->next = skb->prev = NULL;
  1526. next->prev = prev;
  1527. prev->next = next;
  1528. }
  1529. /**
  1530. * __skb_dequeue - remove from the head of the queue
  1531. * @list: list to dequeue from
  1532. *
  1533. * Remove the head of the list. This function does not take any locks
  1534. * so must be used with appropriate locks held only. The head item is
  1535. * returned or %NULL if the list is empty.
  1536. */
  1537. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1538. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1539. {
  1540. struct sk_buff *skb = skb_peek(list);
  1541. if (skb)
  1542. __skb_unlink(skb, list);
  1543. return skb;
  1544. }
  1545. /**
  1546. * __skb_dequeue_tail - remove from the tail of the queue
  1547. * @list: list to dequeue from
  1548. *
  1549. * Remove the tail of the list. This function does not take any locks
  1550. * so must be used with appropriate locks held only. The tail item is
  1551. * returned or %NULL if the list is empty.
  1552. */
  1553. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1554. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1555. {
  1556. struct sk_buff *skb = skb_peek_tail(list);
  1557. if (skb)
  1558. __skb_unlink(skb, list);
  1559. return skb;
  1560. }
  1561. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1562. {
  1563. return skb->data_len;
  1564. }
  1565. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1566. {
  1567. return skb->len - skb->data_len;
  1568. }
  1569. static inline int skb_pagelen(const struct sk_buff *skb)
  1570. {
  1571. int i, len = 0;
  1572. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1573. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1574. return len + skb_headlen(skb);
  1575. }
  1576. /**
  1577. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1578. * @skb: buffer containing fragment to be initialised
  1579. * @i: paged fragment index to initialise
  1580. * @page: the page to use for this fragment
  1581. * @off: the offset to the data with @page
  1582. * @size: the length of the data
  1583. *
  1584. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1585. * offset @off within @page.
  1586. *
  1587. * Does not take any additional reference on the fragment.
  1588. */
  1589. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1590. struct page *page, int off, int size)
  1591. {
  1592. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1593. /*
  1594. * Propagate page pfmemalloc to the skb if we can. The problem is
  1595. * that not all callers have unique ownership of the page but rely
  1596. * on page_is_pfmemalloc doing the right thing(tm).
  1597. */
  1598. frag->page.p = page;
  1599. frag->page_offset = off;
  1600. skb_frag_size_set(frag, size);
  1601. page = compound_head(page);
  1602. if (page_is_pfmemalloc(page))
  1603. skb->pfmemalloc = true;
  1604. }
  1605. /**
  1606. * skb_fill_page_desc - initialise a paged fragment in an skb
  1607. * @skb: buffer containing fragment to be initialised
  1608. * @i: paged fragment index to initialise
  1609. * @page: the page to use for this fragment
  1610. * @off: the offset to the data with @page
  1611. * @size: the length of the data
  1612. *
  1613. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1614. * @skb to point to @size bytes at offset @off within @page. In
  1615. * addition updates @skb such that @i is the last fragment.
  1616. *
  1617. * Does not take any additional reference on the fragment.
  1618. */
  1619. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1620. struct page *page, int off, int size)
  1621. {
  1622. __skb_fill_page_desc(skb, i, page, off, size);
  1623. skb_shinfo(skb)->nr_frags = i + 1;
  1624. }
  1625. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1626. int size, unsigned int truesize);
  1627. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1628. unsigned int truesize);
  1629. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1630. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1631. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1632. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1633. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1634. {
  1635. return skb->head + skb->tail;
  1636. }
  1637. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1638. {
  1639. skb->tail = skb->data - skb->head;
  1640. }
  1641. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1642. {
  1643. skb_reset_tail_pointer(skb);
  1644. skb->tail += offset;
  1645. }
  1646. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1647. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1648. {
  1649. return skb->tail;
  1650. }
  1651. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1652. {
  1653. skb->tail = skb->data;
  1654. }
  1655. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1656. {
  1657. skb->tail = skb->data + offset;
  1658. }
  1659. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1660. /*
  1661. * Add data to an sk_buff
  1662. */
  1663. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1664. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1665. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1666. {
  1667. unsigned char *tmp = skb_tail_pointer(skb);
  1668. SKB_LINEAR_ASSERT(skb);
  1669. skb->tail += len;
  1670. skb->len += len;
  1671. return tmp;
  1672. }
  1673. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1674. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1675. {
  1676. skb->data -= len;
  1677. skb->len += len;
  1678. return skb->data;
  1679. }
  1680. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1681. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1682. {
  1683. skb->len -= len;
  1684. BUG_ON(skb->len < skb->data_len);
  1685. return skb->data += len;
  1686. }
  1687. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1688. {
  1689. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1690. }
  1691. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1692. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1693. {
  1694. if (len > skb_headlen(skb) &&
  1695. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1696. return NULL;
  1697. skb->len -= len;
  1698. return skb->data += len;
  1699. }
  1700. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1701. {
  1702. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1703. }
  1704. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1705. {
  1706. if (likely(len <= skb_headlen(skb)))
  1707. return 1;
  1708. if (unlikely(len > skb->len))
  1709. return 0;
  1710. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1711. }
  1712. /**
  1713. * skb_headroom - bytes at buffer head
  1714. * @skb: buffer to check
  1715. *
  1716. * Return the number of bytes of free space at the head of an &sk_buff.
  1717. */
  1718. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1719. {
  1720. return skb->data - skb->head;
  1721. }
  1722. /**
  1723. * skb_tailroom - 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. */
  1728. static inline int skb_tailroom(const struct sk_buff *skb)
  1729. {
  1730. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1731. }
  1732. /**
  1733. * skb_availroom - bytes at buffer end
  1734. * @skb: buffer to check
  1735. *
  1736. * Return the number of bytes of free space at the tail of an sk_buff
  1737. * allocated by sk_stream_alloc()
  1738. */
  1739. static inline int skb_availroom(const struct sk_buff *skb)
  1740. {
  1741. if (skb_is_nonlinear(skb))
  1742. return 0;
  1743. return skb->end - skb->tail - skb->reserved_tailroom;
  1744. }
  1745. /**
  1746. * skb_reserve - adjust headroom
  1747. * @skb: buffer to alter
  1748. * @len: bytes to move
  1749. *
  1750. * Increase the headroom of an empty &sk_buff by reducing the tail
  1751. * room. This is only allowed for an empty buffer.
  1752. */
  1753. static inline void skb_reserve(struct sk_buff *skb, int len)
  1754. {
  1755. skb->data += len;
  1756. skb->tail += len;
  1757. }
  1758. #define ENCAP_TYPE_ETHER 0
  1759. #define ENCAP_TYPE_IPPROTO 1
  1760. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1761. __be16 protocol)
  1762. {
  1763. skb->inner_protocol = protocol;
  1764. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1765. }
  1766. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1767. __u8 ipproto)
  1768. {
  1769. skb->inner_ipproto = ipproto;
  1770. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1771. }
  1772. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1773. {
  1774. skb->inner_mac_header = skb->mac_header;
  1775. skb->inner_network_header = skb->network_header;
  1776. skb->inner_transport_header = skb->transport_header;
  1777. }
  1778. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1779. {
  1780. skb->mac_len = skb->network_header - skb->mac_header;
  1781. }
  1782. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1783. *skb)
  1784. {
  1785. return skb->head + skb->inner_transport_header;
  1786. }
  1787. static inline int skb_inner_transport_offset(const struct sk_buff *skb)
  1788. {
  1789. return skb_inner_transport_header(skb) - skb->data;
  1790. }
  1791. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1792. {
  1793. skb->inner_transport_header = skb->data - skb->head;
  1794. }
  1795. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1796. const int offset)
  1797. {
  1798. skb_reset_inner_transport_header(skb);
  1799. skb->inner_transport_header += offset;
  1800. }
  1801. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1802. {
  1803. return skb->head + skb->inner_network_header;
  1804. }
  1805. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1806. {
  1807. skb->inner_network_header = skb->data - skb->head;
  1808. }
  1809. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1810. const int offset)
  1811. {
  1812. skb_reset_inner_network_header(skb);
  1813. skb->inner_network_header += offset;
  1814. }
  1815. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1816. {
  1817. return skb->head + skb->inner_mac_header;
  1818. }
  1819. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1820. {
  1821. skb->inner_mac_header = skb->data - skb->head;
  1822. }
  1823. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1824. const int offset)
  1825. {
  1826. skb_reset_inner_mac_header(skb);
  1827. skb->inner_mac_header += offset;
  1828. }
  1829. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1830. {
  1831. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1832. }
  1833. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1834. {
  1835. return skb->head + skb->transport_header;
  1836. }
  1837. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1838. {
  1839. skb->transport_header = skb->data - skb->head;
  1840. }
  1841. static inline void skb_set_transport_header(struct sk_buff *skb,
  1842. const int offset)
  1843. {
  1844. skb_reset_transport_header(skb);
  1845. skb->transport_header += offset;
  1846. }
  1847. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1848. {
  1849. return skb->head + skb->network_header;
  1850. }
  1851. static inline void skb_reset_network_header(struct sk_buff *skb)
  1852. {
  1853. skb->network_header = skb->data - skb->head;
  1854. }
  1855. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1856. {
  1857. skb_reset_network_header(skb);
  1858. skb->network_header += offset;
  1859. }
  1860. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1861. {
  1862. return skb->head + skb->mac_header;
  1863. }
  1864. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1865. {
  1866. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1867. }
  1868. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1869. {
  1870. skb->mac_header = skb->data - skb->head;
  1871. }
  1872. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1873. {
  1874. skb_reset_mac_header(skb);
  1875. skb->mac_header += offset;
  1876. }
  1877. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1878. {
  1879. skb->mac_header = skb->network_header;
  1880. }
  1881. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1882. const int offset_hint)
  1883. {
  1884. struct flow_keys keys;
  1885. if (skb_transport_header_was_set(skb))
  1886. return;
  1887. else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
  1888. skb_set_transport_header(skb, keys.control.thoff);
  1889. else
  1890. skb_set_transport_header(skb, offset_hint);
  1891. }
  1892. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1893. {
  1894. if (skb_mac_header_was_set(skb)) {
  1895. const unsigned char *old_mac = skb_mac_header(skb);
  1896. skb_set_mac_header(skb, -skb->mac_len);
  1897. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1898. }
  1899. }
  1900. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1901. {
  1902. return skb->csum_start - skb_headroom(skb);
  1903. }
  1904. static inline int skb_transport_offset(const struct sk_buff *skb)
  1905. {
  1906. return skb_transport_header(skb) - skb->data;
  1907. }
  1908. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1909. {
  1910. return skb->transport_header - skb->network_header;
  1911. }
  1912. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1913. {
  1914. return skb->inner_transport_header - skb->inner_network_header;
  1915. }
  1916. static inline int skb_network_offset(const struct sk_buff *skb)
  1917. {
  1918. return skb_network_header(skb) - skb->data;
  1919. }
  1920. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1921. {
  1922. return skb_inner_network_header(skb) - skb->data;
  1923. }
  1924. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1925. {
  1926. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1927. }
  1928. /*
  1929. * CPUs often take a performance hit when accessing unaligned memory
  1930. * locations. The actual performance hit varies, it can be small if the
  1931. * hardware handles it or large if we have to take an exception and fix it
  1932. * in software.
  1933. *
  1934. * Since an ethernet header is 14 bytes network drivers often end up with
  1935. * the IP header at an unaligned offset. The IP header can be aligned by
  1936. * shifting the start of the packet by 2 bytes. Drivers should do this
  1937. * with:
  1938. *
  1939. * skb_reserve(skb, NET_IP_ALIGN);
  1940. *
  1941. * The downside to this alignment of the IP header is that the DMA is now
  1942. * unaligned. On some architectures the cost of an unaligned DMA is high
  1943. * and this cost outweighs the gains made by aligning the IP header.
  1944. *
  1945. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1946. * to be overridden.
  1947. */
  1948. #ifndef NET_IP_ALIGN
  1949. #define NET_IP_ALIGN 2
  1950. #endif
  1951. /*
  1952. * The networking layer reserves some headroom in skb data (via
  1953. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1954. * the header has to grow. In the default case, if the header has to grow
  1955. * 32 bytes or less we avoid the reallocation.
  1956. *
  1957. * Unfortunately this headroom changes the DMA alignment of the resulting
  1958. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1959. * on some architectures. An architecture can override this value,
  1960. * perhaps setting it to a cacheline in size (since that will maintain
  1961. * cacheline alignment of the DMA). It must be a power of 2.
  1962. *
  1963. * Various parts of the networking layer expect at least 32 bytes of
  1964. * headroom, you should not reduce this.
  1965. *
  1966. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1967. * to reduce average number of cache lines per packet.
  1968. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1969. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1970. */
  1971. #ifndef NET_SKB_PAD
  1972. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1973. #endif
  1974. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1975. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1976. {
  1977. if (unlikely(skb_is_nonlinear(skb))) {
  1978. WARN_ON(1);
  1979. return;
  1980. }
  1981. skb->len = len;
  1982. skb_set_tail_pointer(skb, len);
  1983. }
  1984. void skb_trim(struct sk_buff *skb, unsigned int len);
  1985. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1986. {
  1987. if (skb->data_len)
  1988. return ___pskb_trim(skb, len);
  1989. __skb_trim(skb, len);
  1990. return 0;
  1991. }
  1992. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1993. {
  1994. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1995. }
  1996. /**
  1997. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1998. * @skb: buffer to alter
  1999. * @len: new length
  2000. *
  2001. * This is identical to pskb_trim except that the caller knows that
  2002. * the skb is not cloned so we should never get an error due to out-
  2003. * of-memory.
  2004. */
  2005. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  2006. {
  2007. int err = pskb_trim(skb, len);
  2008. BUG_ON(err);
  2009. }
  2010. /**
  2011. * skb_orphan - orphan a buffer
  2012. * @skb: buffer to orphan
  2013. *
  2014. * If a buffer currently has an owner then we call the owner's
  2015. * destructor function and make the @skb unowned. The buffer continues
  2016. * to exist but is no longer charged to its former owner.
  2017. */
  2018. static inline void skb_orphan(struct sk_buff *skb)
  2019. {
  2020. if (skb->destructor) {
  2021. skb->destructor(skb);
  2022. skb->destructor = NULL;
  2023. skb->sk = NULL;
  2024. } else {
  2025. BUG_ON(skb->sk);
  2026. }
  2027. }
  2028. /**
  2029. * skb_orphan_frags - orphan the frags contained in a buffer
  2030. * @skb: buffer to orphan frags from
  2031. * @gfp_mask: allocation mask for replacement pages
  2032. *
  2033. * For each frag in the SKB which needs a destructor (i.e. has an
  2034. * owner) create a copy of that frag and release the original
  2035. * page by calling the destructor.
  2036. */
  2037. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  2038. {
  2039. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  2040. return 0;
  2041. return skb_copy_ubufs(skb, gfp_mask);
  2042. }
  2043. /**
  2044. * __skb_queue_purge - empty a list
  2045. * @list: list to empty
  2046. *
  2047. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  2048. * the list and one reference dropped. This function does not take the
  2049. * list lock and the caller must hold the relevant locks to use it.
  2050. */
  2051. void skb_queue_purge(struct sk_buff_head *list);
  2052. static inline void __skb_queue_purge(struct sk_buff_head *list)
  2053. {
  2054. struct sk_buff *skb;
  2055. while ((skb = __skb_dequeue(list)) != NULL)
  2056. kfree_skb(skb);
  2057. }
  2058. void *netdev_alloc_frag(unsigned int fragsz);
  2059. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  2060. gfp_t gfp_mask);
  2061. /**
  2062. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  2063. * @dev: network device to receive on
  2064. * @length: length to allocate
  2065. *
  2066. * Allocate a new &sk_buff and assign it a usage count of one. The
  2067. * buffer has unspecified headroom built in. Users should allocate
  2068. * the headroom they think they need without accounting for the
  2069. * built in space. The built in space is used for optimisations.
  2070. *
  2071. * %NULL is returned if there is no free memory. Although this function
  2072. * allocates memory it can be called from an interrupt.
  2073. */
  2074. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  2075. unsigned int length)
  2076. {
  2077. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  2078. }
  2079. /* legacy helper around __netdev_alloc_skb() */
  2080. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  2081. gfp_t gfp_mask)
  2082. {
  2083. return __netdev_alloc_skb(NULL, length, gfp_mask);
  2084. }
  2085. /* legacy helper around netdev_alloc_skb() */
  2086. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  2087. {
  2088. return netdev_alloc_skb(NULL, length);
  2089. }
  2090. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  2091. unsigned int length, gfp_t gfp)
  2092. {
  2093. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  2094. if (NET_IP_ALIGN && skb)
  2095. skb_reserve(skb, NET_IP_ALIGN);
  2096. return skb;
  2097. }
  2098. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  2099. unsigned int length)
  2100. {
  2101. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  2102. }
  2103. static inline void skb_free_frag(void *addr)
  2104. {
  2105. __free_page_frag(addr);
  2106. }
  2107. void *napi_alloc_frag(unsigned int fragsz);
  2108. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  2109. unsigned int length, gfp_t gfp_mask);
  2110. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  2111. unsigned int length)
  2112. {
  2113. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  2114. }
  2115. /**
  2116. * __dev_alloc_pages - allocate page for network Rx
  2117. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2118. * @order: size of the allocation
  2119. *
  2120. * Allocate a new page.
  2121. *
  2122. * %NULL is returned if there is no free memory.
  2123. */
  2124. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  2125. unsigned int order)
  2126. {
  2127. /* This piece of code contains several assumptions.
  2128. * 1. This is for device Rx, therefor a cold page is preferred.
  2129. * 2. The expectation is the user wants a compound page.
  2130. * 3. If requesting a order 0 page it will not be compound
  2131. * due to the check to see if order has a value in prep_new_page
  2132. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  2133. * code in gfp_to_alloc_flags that should be enforcing this.
  2134. */
  2135. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  2136. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  2137. }
  2138. static inline struct page *dev_alloc_pages(unsigned int order)
  2139. {
  2140. return __dev_alloc_pages(GFP_ATOMIC, order);
  2141. }
  2142. /**
  2143. * __dev_alloc_page - allocate a page for network Rx
  2144. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  2145. *
  2146. * Allocate a new page.
  2147. *
  2148. * %NULL is returned if there is no free memory.
  2149. */
  2150. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  2151. {
  2152. return __dev_alloc_pages(gfp_mask, 0);
  2153. }
  2154. static inline struct page *dev_alloc_page(void)
  2155. {
  2156. return __dev_alloc_page(GFP_ATOMIC);
  2157. }
  2158. /**
  2159. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  2160. * @page: The page that was allocated from skb_alloc_page
  2161. * @skb: The skb that may need pfmemalloc set
  2162. */
  2163. static inline void skb_propagate_pfmemalloc(struct page *page,
  2164. struct sk_buff *skb)
  2165. {
  2166. if (page_is_pfmemalloc(page))
  2167. skb->pfmemalloc = true;
  2168. }
  2169. /**
  2170. * skb_frag_page - retrieve the page referred to by a paged fragment
  2171. * @frag: the paged fragment
  2172. *
  2173. * Returns the &struct page associated with @frag.
  2174. */
  2175. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2176. {
  2177. return frag->page.p;
  2178. }
  2179. /**
  2180. * __skb_frag_ref - take an addition reference on a paged fragment.
  2181. * @frag: the paged fragment
  2182. *
  2183. * Takes an additional reference on the paged fragment @frag.
  2184. */
  2185. static inline void __skb_frag_ref(skb_frag_t *frag)
  2186. {
  2187. get_page(skb_frag_page(frag));
  2188. }
  2189. /**
  2190. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2191. * @skb: the buffer
  2192. * @f: the fragment offset.
  2193. *
  2194. * Takes an additional reference on the @f'th paged fragment of @skb.
  2195. */
  2196. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2197. {
  2198. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2199. }
  2200. /**
  2201. * __skb_frag_unref - release a reference on a paged fragment.
  2202. * @frag: the paged fragment
  2203. *
  2204. * Releases a reference on the paged fragment @frag.
  2205. */
  2206. static inline void __skb_frag_unref(skb_frag_t *frag)
  2207. {
  2208. put_page(skb_frag_page(frag));
  2209. }
  2210. /**
  2211. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2212. * @skb: the buffer
  2213. * @f: the fragment offset
  2214. *
  2215. * Releases a reference on the @f'th paged fragment of @skb.
  2216. */
  2217. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2218. {
  2219. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2220. }
  2221. /**
  2222. * skb_frag_address - gets the address of the data contained in a paged fragment
  2223. * @frag: the paged fragment buffer
  2224. *
  2225. * Returns the address of the data within @frag. The page must already
  2226. * be mapped.
  2227. */
  2228. static inline void *skb_frag_address(const skb_frag_t *frag)
  2229. {
  2230. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2231. }
  2232. /**
  2233. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2234. * @frag: the paged fragment buffer
  2235. *
  2236. * Returns the address of the data within @frag. Checks that the page
  2237. * is mapped and returns %NULL otherwise.
  2238. */
  2239. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2240. {
  2241. void *ptr = page_address(skb_frag_page(frag));
  2242. if (unlikely(!ptr))
  2243. return NULL;
  2244. return ptr + frag->page_offset;
  2245. }
  2246. /**
  2247. * __skb_frag_set_page - sets the page contained in a paged fragment
  2248. * @frag: the paged fragment
  2249. * @page: the page to set
  2250. *
  2251. * Sets the fragment @frag to contain @page.
  2252. */
  2253. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2254. {
  2255. frag->page.p = page;
  2256. }
  2257. /**
  2258. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2259. * @skb: the buffer
  2260. * @f: the fragment offset
  2261. * @page: the page to set
  2262. *
  2263. * Sets the @f'th fragment of @skb to contain @page.
  2264. */
  2265. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2266. struct page *page)
  2267. {
  2268. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2269. }
  2270. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2271. /**
  2272. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2273. * @dev: the device to map the fragment to
  2274. * @frag: the paged fragment to map
  2275. * @offset: the offset within the fragment (starting at the
  2276. * fragment's own offset)
  2277. * @size: the number of bytes to map
  2278. * @dir: the direction of the mapping (%PCI_DMA_*)
  2279. *
  2280. * Maps the page associated with @frag to @device.
  2281. */
  2282. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2283. const skb_frag_t *frag,
  2284. size_t offset, size_t size,
  2285. enum dma_data_direction dir)
  2286. {
  2287. return dma_map_page(dev, skb_frag_page(frag),
  2288. frag->page_offset + offset, size, dir);
  2289. }
  2290. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2291. gfp_t gfp_mask)
  2292. {
  2293. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2294. }
  2295. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2296. gfp_t gfp_mask)
  2297. {
  2298. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2299. }
  2300. /**
  2301. * skb_clone_writable - is the header of a clone writable
  2302. * @skb: buffer to check
  2303. * @len: length up to which to write
  2304. *
  2305. * Returns true if modifying the header part of the cloned buffer
  2306. * does not requires the data to be copied.
  2307. */
  2308. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2309. {
  2310. return !skb_header_cloned(skb) &&
  2311. skb_headroom(skb) + len <= skb->hdr_len;
  2312. }
  2313. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2314. int cloned)
  2315. {
  2316. int delta = 0;
  2317. if (headroom > skb_headroom(skb))
  2318. delta = headroom - skb_headroom(skb);
  2319. if (delta || cloned)
  2320. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2321. GFP_ATOMIC);
  2322. return 0;
  2323. }
  2324. /**
  2325. * skb_cow - copy header of skb when it is required
  2326. * @skb: buffer to cow
  2327. * @headroom: needed headroom
  2328. *
  2329. * If the skb passed lacks sufficient headroom or its data part
  2330. * is shared, data is reallocated. If reallocation fails, an error
  2331. * is returned and original skb is not changed.
  2332. *
  2333. * The result is skb with writable area skb->head...skb->tail
  2334. * and at least @headroom of space at head.
  2335. */
  2336. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2337. {
  2338. return __skb_cow(skb, headroom, skb_cloned(skb));
  2339. }
  2340. /**
  2341. * skb_cow_head - skb_cow but only making the head writable
  2342. * @skb: buffer to cow
  2343. * @headroom: needed headroom
  2344. *
  2345. * This function is identical to skb_cow except that we replace the
  2346. * skb_cloned check by skb_header_cloned. It should be used when
  2347. * you only need to push on some header and do not need to modify
  2348. * the data.
  2349. */
  2350. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2351. {
  2352. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2353. }
  2354. /**
  2355. * skb_padto - pad an skbuff up to a minimal size
  2356. * @skb: buffer to pad
  2357. * @len: minimal length
  2358. *
  2359. * Pads up a buffer to ensure the trailing bytes exist and are
  2360. * blanked. If the buffer already contains sufficient data it
  2361. * is untouched. Otherwise it is extended. Returns zero on
  2362. * success. The skb is freed on error.
  2363. */
  2364. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2365. {
  2366. unsigned int size = skb->len;
  2367. if (likely(size >= len))
  2368. return 0;
  2369. return skb_pad(skb, len - size);
  2370. }
  2371. /**
  2372. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2373. * @skb: buffer to pad
  2374. * @len: minimal length
  2375. *
  2376. * Pads up a buffer to ensure the trailing bytes exist and are
  2377. * blanked. If the buffer already contains sufficient data it
  2378. * is untouched. Otherwise it is extended. Returns zero on
  2379. * success. The skb is freed on error.
  2380. */
  2381. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2382. {
  2383. unsigned int size = skb->len;
  2384. if (unlikely(size < len)) {
  2385. len -= size;
  2386. if (skb_pad(skb, len))
  2387. return -ENOMEM;
  2388. __skb_put(skb, len);
  2389. }
  2390. return 0;
  2391. }
  2392. static inline int skb_add_data(struct sk_buff *skb,
  2393. struct iov_iter *from, int copy)
  2394. {
  2395. const int off = skb->len;
  2396. if (skb->ip_summed == CHECKSUM_NONE) {
  2397. __wsum csum = 0;
  2398. if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
  2399. &csum, from) == copy) {
  2400. skb->csum = csum_block_add(skb->csum, csum, off);
  2401. return 0;
  2402. }
  2403. } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
  2404. return 0;
  2405. __skb_trim(skb, off);
  2406. return -EFAULT;
  2407. }
  2408. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2409. const struct page *page, int off)
  2410. {
  2411. if (i) {
  2412. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2413. return page == skb_frag_page(frag) &&
  2414. off == frag->page_offset + skb_frag_size(frag);
  2415. }
  2416. return false;
  2417. }
  2418. static inline int __skb_linearize(struct sk_buff *skb)
  2419. {
  2420. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2421. }
  2422. /**
  2423. * skb_linearize - convert paged skb to linear one
  2424. * @skb: buffer to linarize
  2425. *
  2426. * If there is no free memory -ENOMEM is returned, otherwise zero
  2427. * is returned and the old skb data released.
  2428. */
  2429. static inline int skb_linearize(struct sk_buff *skb)
  2430. {
  2431. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2432. }
  2433. /**
  2434. * skb_has_shared_frag - can any frag be overwritten
  2435. * @skb: buffer to test
  2436. *
  2437. * Return true if the skb has at least one frag that might be modified
  2438. * by an external entity (as in vmsplice()/sendfile())
  2439. */
  2440. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2441. {
  2442. return skb_is_nonlinear(skb) &&
  2443. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2444. }
  2445. /**
  2446. * skb_linearize_cow - make sure skb is linear and writable
  2447. * @skb: buffer to process
  2448. *
  2449. * If there is no free memory -ENOMEM is returned, otherwise zero
  2450. * is returned and the old skb data released.
  2451. */
  2452. static inline int skb_linearize_cow(struct sk_buff *skb)
  2453. {
  2454. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2455. __skb_linearize(skb) : 0;
  2456. }
  2457. /**
  2458. * skb_postpull_rcsum - update checksum for received skb after pull
  2459. * @skb: buffer to update
  2460. * @start: start of data before pull
  2461. * @len: length of data pulled
  2462. *
  2463. * After doing a pull on a received packet, you need to call this to
  2464. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2465. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2466. */
  2467. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2468. const void *start, unsigned int len)
  2469. {
  2470. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2471. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2472. else if (skb->ip_summed == CHECKSUM_PARTIAL &&
  2473. skb_checksum_start_offset(skb) < 0)
  2474. skb->ip_summed = CHECKSUM_NONE;
  2475. }
  2476. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2477. static inline void skb_postpush_rcsum(struct sk_buff *skb,
  2478. const void *start, unsigned int len)
  2479. {
  2480. /* For performing the reverse operation to skb_postpull_rcsum(),
  2481. * we can instead of ...
  2482. *
  2483. * skb->csum = csum_add(skb->csum, csum_partial(start, len, 0));
  2484. *
  2485. * ... just use this equivalent version here to save a few
  2486. * instructions. Feeding csum of 0 in csum_partial() and later
  2487. * on adding skb->csum is equivalent to feed skb->csum in the
  2488. * first place.
  2489. */
  2490. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2491. skb->csum = csum_partial(start, len, skb->csum);
  2492. }
  2493. /**
  2494. * pskb_trim_rcsum - trim received skb and update checksum
  2495. * @skb: buffer to trim
  2496. * @len: new length
  2497. *
  2498. * This is exactly the same as pskb_trim except that it ensures the
  2499. * checksum of received packets are still valid after the operation.
  2500. */
  2501. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2502. {
  2503. if (likely(len >= skb->len))
  2504. return 0;
  2505. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2506. skb->ip_summed = CHECKSUM_NONE;
  2507. return __pskb_trim(skb, len);
  2508. }
  2509. #define skb_queue_walk(queue, skb) \
  2510. for (skb = (queue)->next; \
  2511. skb != (struct sk_buff *)(queue); \
  2512. skb = skb->next)
  2513. #define skb_queue_walk_safe(queue, skb, tmp) \
  2514. for (skb = (queue)->next, tmp = skb->next; \
  2515. skb != (struct sk_buff *)(queue); \
  2516. skb = tmp, tmp = skb->next)
  2517. #define skb_queue_walk_from(queue, skb) \
  2518. for (; skb != (struct sk_buff *)(queue); \
  2519. skb = skb->next)
  2520. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2521. for (tmp = skb->next; \
  2522. skb != (struct sk_buff *)(queue); \
  2523. skb = tmp, tmp = skb->next)
  2524. #define skb_queue_reverse_walk(queue, skb) \
  2525. for (skb = (queue)->prev; \
  2526. skb != (struct sk_buff *)(queue); \
  2527. skb = skb->prev)
  2528. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2529. for (skb = (queue)->prev, tmp = skb->prev; \
  2530. skb != (struct sk_buff *)(queue); \
  2531. skb = tmp, tmp = skb->prev)
  2532. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2533. for (tmp = skb->prev; \
  2534. skb != (struct sk_buff *)(queue); \
  2535. skb = tmp, tmp = skb->prev)
  2536. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2537. {
  2538. return skb_shinfo(skb)->frag_list != NULL;
  2539. }
  2540. static inline void skb_frag_list_init(struct sk_buff *skb)
  2541. {
  2542. skb_shinfo(skb)->frag_list = NULL;
  2543. }
  2544. #define skb_walk_frags(skb, iter) \
  2545. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2546. int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
  2547. const struct sk_buff *skb);
  2548. struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
  2549. int *peeked, int *off, int *err,
  2550. struct sk_buff **last);
  2551. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2552. int *peeked, int *off, int *err);
  2553. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2554. int *err);
  2555. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2556. struct poll_table_struct *wait);
  2557. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2558. struct iov_iter *to, int size);
  2559. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2560. struct msghdr *msg, int size)
  2561. {
  2562. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2563. }
  2564. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2565. struct msghdr *msg);
  2566. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2567. struct iov_iter *from, int len);
  2568. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2569. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2570. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2571. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2572. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2573. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2574. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2575. int len, __wsum csum);
  2576. ssize_t skb_socket_splice(struct sock *sk,
  2577. struct pipe_inode_info *pipe,
  2578. struct splice_pipe_desc *spd);
  2579. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2580. struct pipe_inode_info *pipe, unsigned int len,
  2581. unsigned int flags,
  2582. ssize_t (*splice_cb)(struct sock *,
  2583. struct pipe_inode_info *,
  2584. struct splice_pipe_desc *));
  2585. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2586. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2587. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2588. int len, int hlen);
  2589. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2590. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2591. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2592. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2593. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2594. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2595. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2596. int skb_vlan_pop(struct sk_buff *skb);
  2597. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2598. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2599. {
  2600. return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2601. }
  2602. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2603. {
  2604. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2605. }
  2606. struct skb_checksum_ops {
  2607. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2608. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2609. };
  2610. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2611. __wsum csum, const struct skb_checksum_ops *ops);
  2612. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2613. __wsum csum);
  2614. static inline void * __must_check
  2615. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2616. int len, void *data, int hlen, void *buffer)
  2617. {
  2618. if (hlen - offset >= len)
  2619. return data + offset;
  2620. if (!skb ||
  2621. skb_copy_bits(skb, offset, buffer, len) < 0)
  2622. return NULL;
  2623. return buffer;
  2624. }
  2625. static inline void * __must_check
  2626. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2627. {
  2628. return __skb_header_pointer(skb, offset, len, skb->data,
  2629. skb_headlen(skb), buffer);
  2630. }
  2631. /**
  2632. * skb_needs_linearize - check if we need to linearize a given skb
  2633. * depending on the given device features.
  2634. * @skb: socket buffer to check
  2635. * @features: net device features
  2636. *
  2637. * Returns true if either:
  2638. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2639. * 2. skb is fragmented and the device does not support SG.
  2640. */
  2641. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2642. netdev_features_t features)
  2643. {
  2644. return skb_is_nonlinear(skb) &&
  2645. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2646. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2647. }
  2648. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2649. void *to,
  2650. const unsigned int len)
  2651. {
  2652. memcpy(to, skb->data, len);
  2653. }
  2654. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2655. const int offset, void *to,
  2656. const unsigned int len)
  2657. {
  2658. memcpy(to, skb->data + offset, len);
  2659. }
  2660. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2661. const void *from,
  2662. const unsigned int len)
  2663. {
  2664. memcpy(skb->data, from, len);
  2665. }
  2666. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2667. const int offset,
  2668. const void *from,
  2669. const unsigned int len)
  2670. {
  2671. memcpy(skb->data + offset, from, len);
  2672. }
  2673. void skb_init(void);
  2674. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2675. {
  2676. return skb->tstamp;
  2677. }
  2678. /**
  2679. * skb_get_timestamp - get timestamp from a skb
  2680. * @skb: skb to get stamp from
  2681. * @stamp: pointer to struct timeval to store stamp in
  2682. *
  2683. * Timestamps are stored in the skb as offsets to a base timestamp.
  2684. * This function converts the offset back to a struct timeval and stores
  2685. * it in stamp.
  2686. */
  2687. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2688. struct timeval *stamp)
  2689. {
  2690. *stamp = ktime_to_timeval(skb->tstamp);
  2691. }
  2692. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2693. struct timespec *stamp)
  2694. {
  2695. *stamp = ktime_to_timespec(skb->tstamp);
  2696. }
  2697. static inline void __net_timestamp(struct sk_buff *skb)
  2698. {
  2699. skb->tstamp = ktime_get_real();
  2700. }
  2701. static inline ktime_t net_timedelta(ktime_t t)
  2702. {
  2703. return ktime_sub(ktime_get_real(), t);
  2704. }
  2705. static inline ktime_t net_invalid_timestamp(void)
  2706. {
  2707. return ktime_set(0, 0);
  2708. }
  2709. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2710. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2711. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2712. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2713. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2714. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2715. {
  2716. }
  2717. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2718. {
  2719. return false;
  2720. }
  2721. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2722. /**
  2723. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2724. *
  2725. * PHY drivers may accept clones of transmitted packets for
  2726. * timestamping via their phy_driver.txtstamp method. These drivers
  2727. * must call this function to return the skb back to the stack with a
  2728. * timestamp.
  2729. *
  2730. * @skb: clone of the the original outgoing packet
  2731. * @hwtstamps: hardware time stamps
  2732. *
  2733. */
  2734. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2735. struct skb_shared_hwtstamps *hwtstamps);
  2736. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2737. struct skb_shared_hwtstamps *hwtstamps,
  2738. struct sock *sk, int tstype);
  2739. /**
  2740. * skb_tstamp_tx - queue clone of skb with send time stamps
  2741. * @orig_skb: the original outgoing packet
  2742. * @hwtstamps: hardware time stamps, may be NULL if not available
  2743. *
  2744. * If the skb has a socket associated, then this function clones the
  2745. * skb (thus sharing the actual data and optional structures), stores
  2746. * the optional hardware time stamping information (if non NULL) or
  2747. * generates a software time stamp (otherwise), then queues the clone
  2748. * to the error queue of the socket. Errors are silently ignored.
  2749. */
  2750. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2751. struct skb_shared_hwtstamps *hwtstamps);
  2752. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2753. {
  2754. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2755. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2756. skb_tstamp_tx(skb, NULL);
  2757. }
  2758. /**
  2759. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2760. *
  2761. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2762. * function immediately before giving the sk_buff to the MAC hardware.
  2763. *
  2764. * Specifically, one should make absolutely sure that this function is
  2765. * called before TX completion of this packet can trigger. Otherwise
  2766. * the packet could potentially already be freed.
  2767. *
  2768. * @skb: A socket buffer.
  2769. */
  2770. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2771. {
  2772. skb_clone_tx_timestamp(skb);
  2773. sw_tx_timestamp(skb);
  2774. }
  2775. /**
  2776. * skb_complete_wifi_ack - deliver skb with wifi status
  2777. *
  2778. * @skb: the original outgoing packet
  2779. * @acked: ack status
  2780. *
  2781. */
  2782. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2783. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2784. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2785. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2786. {
  2787. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2788. skb->csum_valid ||
  2789. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2790. skb_checksum_start_offset(skb) >= 0));
  2791. }
  2792. /**
  2793. * skb_checksum_complete - Calculate checksum of an entire packet
  2794. * @skb: packet to process
  2795. *
  2796. * This function calculates the checksum over the entire packet plus
  2797. * the value of skb->csum. The latter can be used to supply the
  2798. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2799. * checksum.
  2800. *
  2801. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2802. * this function can be used to verify that checksum on received
  2803. * packets. In that case the function should return zero if the
  2804. * checksum is correct. In particular, this function will return zero
  2805. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2806. * hardware has already verified the correctness of the checksum.
  2807. */
  2808. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2809. {
  2810. return skb_csum_unnecessary(skb) ?
  2811. 0 : __skb_checksum_complete(skb);
  2812. }
  2813. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2814. {
  2815. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2816. if (skb->csum_level == 0)
  2817. skb->ip_summed = CHECKSUM_NONE;
  2818. else
  2819. skb->csum_level--;
  2820. }
  2821. }
  2822. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2823. {
  2824. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2825. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2826. skb->csum_level++;
  2827. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2828. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2829. skb->csum_level = 0;
  2830. }
  2831. }
  2832. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2833. {
  2834. /* Mark current checksum as bad (typically called from GRO
  2835. * path). In the case that ip_summed is CHECKSUM_NONE
  2836. * this must be the first checksum encountered in the packet.
  2837. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2838. * checksum after the last one validated. For UDP, a zero
  2839. * checksum can not be marked as bad.
  2840. */
  2841. if (skb->ip_summed == CHECKSUM_NONE ||
  2842. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2843. skb->csum_bad = 1;
  2844. }
  2845. /* Check if we need to perform checksum complete validation.
  2846. *
  2847. * Returns true if checksum complete is needed, false otherwise
  2848. * (either checksum is unnecessary or zero checksum is allowed).
  2849. */
  2850. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2851. bool zero_okay,
  2852. __sum16 check)
  2853. {
  2854. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2855. skb->csum_valid = 1;
  2856. __skb_decr_checksum_unnecessary(skb);
  2857. return false;
  2858. }
  2859. return true;
  2860. }
  2861. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2862. * in checksum_init.
  2863. */
  2864. #define CHECKSUM_BREAK 76
  2865. /* Unset checksum-complete
  2866. *
  2867. * Unset checksum complete can be done when packet is being modified
  2868. * (uncompressed for instance) and checksum-complete value is
  2869. * invalidated.
  2870. */
  2871. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2872. {
  2873. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2874. skb->ip_summed = CHECKSUM_NONE;
  2875. }
  2876. /* Validate (init) checksum based on checksum complete.
  2877. *
  2878. * Return values:
  2879. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2880. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2881. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2882. * non-zero: value of invalid checksum
  2883. *
  2884. */
  2885. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2886. bool complete,
  2887. __wsum psum)
  2888. {
  2889. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2890. if (!csum_fold(csum_add(psum, skb->csum))) {
  2891. skb->csum_valid = 1;
  2892. return 0;
  2893. }
  2894. } else if (skb->csum_bad) {
  2895. /* ip_summed == CHECKSUM_NONE in this case */
  2896. return (__force __sum16)1;
  2897. }
  2898. skb->csum = psum;
  2899. if (complete || skb->len <= CHECKSUM_BREAK) {
  2900. __sum16 csum;
  2901. csum = __skb_checksum_complete(skb);
  2902. skb->csum_valid = !csum;
  2903. return csum;
  2904. }
  2905. return 0;
  2906. }
  2907. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2908. {
  2909. return 0;
  2910. }
  2911. /* Perform checksum validate (init). Note that this is a macro since we only
  2912. * want to calculate the pseudo header which is an input function if necessary.
  2913. * First we try to validate without any computation (checksum unnecessary) and
  2914. * then calculate based on checksum complete calling the function to compute
  2915. * pseudo header.
  2916. *
  2917. * Return values:
  2918. * 0: checksum is validated or try to in skb_checksum_complete
  2919. * non-zero: value of invalid checksum
  2920. */
  2921. #define __skb_checksum_validate(skb, proto, complete, \
  2922. zero_okay, check, compute_pseudo) \
  2923. ({ \
  2924. __sum16 __ret = 0; \
  2925. skb->csum_valid = 0; \
  2926. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2927. __ret = __skb_checksum_validate_complete(skb, \
  2928. complete, compute_pseudo(skb, proto)); \
  2929. __ret; \
  2930. })
  2931. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2932. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2933. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2934. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2935. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2936. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2937. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2938. compute_pseudo) \
  2939. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  2940. #define skb_checksum_simple_validate(skb) \
  2941. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2942. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2943. {
  2944. return (skb->ip_summed == CHECKSUM_NONE &&
  2945. skb->csum_valid && !skb->csum_bad);
  2946. }
  2947. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2948. __sum16 check, __wsum pseudo)
  2949. {
  2950. skb->csum = ~pseudo;
  2951. skb->ip_summed = CHECKSUM_COMPLETE;
  2952. }
  2953. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  2954. do { \
  2955. if (__skb_checksum_convert_check(skb)) \
  2956. __skb_checksum_convert(skb, check, \
  2957. compute_pseudo(skb, proto)); \
  2958. } while (0)
  2959. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  2960. u16 start, u16 offset)
  2961. {
  2962. skb->ip_summed = CHECKSUM_PARTIAL;
  2963. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  2964. skb->csum_offset = offset - start;
  2965. }
  2966. /* Update skbuf and packet to reflect the remote checksum offload operation.
  2967. * When called, ptr indicates the starting point for skb->csum when
  2968. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  2969. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  2970. */
  2971. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  2972. int start, int offset, bool nopartial)
  2973. {
  2974. __wsum delta;
  2975. if (!nopartial) {
  2976. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  2977. return;
  2978. }
  2979. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  2980. __skb_checksum_complete(skb);
  2981. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  2982. }
  2983. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  2984. /* Adjust skb->csum since we changed the packet */
  2985. skb->csum = csum_add(skb->csum, delta);
  2986. }
  2987. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2988. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2989. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2990. {
  2991. if (nfct && atomic_dec_and_test(&nfct->use))
  2992. nf_conntrack_destroy(nfct);
  2993. }
  2994. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2995. {
  2996. if (nfct)
  2997. atomic_inc(&nfct->use);
  2998. }
  2999. #endif
  3000. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3001. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  3002. {
  3003. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  3004. kfree(nf_bridge);
  3005. }
  3006. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  3007. {
  3008. if (nf_bridge)
  3009. atomic_inc(&nf_bridge->use);
  3010. }
  3011. #endif /* CONFIG_BRIDGE_NETFILTER */
  3012. static inline void nf_reset(struct sk_buff *skb)
  3013. {
  3014. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3015. nf_conntrack_put(skb->nfct);
  3016. skb->nfct = NULL;
  3017. #endif
  3018. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3019. nf_bridge_put(skb->nf_bridge);
  3020. skb->nf_bridge = NULL;
  3021. #endif
  3022. }
  3023. static inline void nf_reset_trace(struct sk_buff *skb)
  3024. {
  3025. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3026. skb->nf_trace = 0;
  3027. #endif
  3028. }
  3029. /* Note: This doesn't put any conntrack and bridge info in dst. */
  3030. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  3031. bool copy)
  3032. {
  3033. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3034. dst->nfct = src->nfct;
  3035. nf_conntrack_get(src->nfct);
  3036. if (copy)
  3037. dst->nfctinfo = src->nfctinfo;
  3038. #endif
  3039. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3040. dst->nf_bridge = src->nf_bridge;
  3041. nf_bridge_get(src->nf_bridge);
  3042. #endif
  3043. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  3044. if (copy)
  3045. dst->nf_trace = src->nf_trace;
  3046. #endif
  3047. }
  3048. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  3049. {
  3050. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  3051. nf_conntrack_put(dst->nfct);
  3052. #endif
  3053. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  3054. nf_bridge_put(dst->nf_bridge);
  3055. #endif
  3056. __nf_copy(dst, src, true);
  3057. }
  3058. #ifdef CONFIG_NETWORK_SECMARK
  3059. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3060. {
  3061. to->secmark = from->secmark;
  3062. }
  3063. static inline void skb_init_secmark(struct sk_buff *skb)
  3064. {
  3065. skb->secmark = 0;
  3066. }
  3067. #else
  3068. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  3069. { }
  3070. static inline void skb_init_secmark(struct sk_buff *skb)
  3071. { }
  3072. #endif
  3073. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  3074. {
  3075. return !skb->destructor &&
  3076. #if IS_ENABLED(CONFIG_XFRM)
  3077. !skb->sp &&
  3078. #endif
  3079. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  3080. !skb->nfct &&
  3081. #endif
  3082. !skb->_skb_refdst &&
  3083. !skb_has_frag_list(skb);
  3084. }
  3085. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  3086. {
  3087. skb->queue_mapping = queue_mapping;
  3088. }
  3089. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  3090. {
  3091. return skb->queue_mapping;
  3092. }
  3093. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  3094. {
  3095. to->queue_mapping = from->queue_mapping;
  3096. }
  3097. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  3098. {
  3099. skb->queue_mapping = rx_queue + 1;
  3100. }
  3101. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  3102. {
  3103. return skb->queue_mapping - 1;
  3104. }
  3105. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  3106. {
  3107. return skb->queue_mapping != 0;
  3108. }
  3109. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  3110. {
  3111. #ifdef CONFIG_XFRM
  3112. return skb->sp;
  3113. #else
  3114. return NULL;
  3115. #endif
  3116. }
  3117. /* Keeps track of mac header offset relative to skb->head.
  3118. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  3119. * For non-tunnel skb it points to skb_mac_header() and for
  3120. * tunnel skb it points to outer mac header.
  3121. * Keeps track of level of encapsulation of network headers.
  3122. */
  3123. struct skb_gso_cb {
  3124. int mac_offset;
  3125. int encap_level;
  3126. __u16 csum_start;
  3127. };
  3128. #define SKB_SGO_CB_OFFSET 32
  3129. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
  3130. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  3131. {
  3132. return (skb_mac_header(inner_skb) - inner_skb->head) -
  3133. SKB_GSO_CB(inner_skb)->mac_offset;
  3134. }
  3135. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  3136. {
  3137. int new_headroom, headroom;
  3138. int ret;
  3139. headroom = skb_headroom(skb);
  3140. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  3141. if (ret)
  3142. return ret;
  3143. new_headroom = skb_headroom(skb);
  3144. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  3145. return 0;
  3146. }
  3147. /* Compute the checksum for a gso segment. First compute the checksum value
  3148. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  3149. * then add in skb->csum (checksum from csum_start to end of packet).
  3150. * skb->csum and csum_start are then updated to reflect the checksum of the
  3151. * resultant packet starting from the transport header-- the resultant checksum
  3152. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  3153. * header.
  3154. */
  3155. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  3156. {
  3157. int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
  3158. skb_transport_offset(skb);
  3159. __wsum partial;
  3160. partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
  3161. skb->csum = res;
  3162. SKB_GSO_CB(skb)->csum_start -= plen;
  3163. return csum_fold(partial);
  3164. }
  3165. static inline bool skb_is_gso(const struct sk_buff *skb)
  3166. {
  3167. return skb_shinfo(skb)->gso_size;
  3168. }
  3169. /* Note: Should be called only if skb_is_gso(skb) is true */
  3170. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  3171. {
  3172. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  3173. }
  3174. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  3175. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  3176. {
  3177. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  3178. * wanted then gso_type will be set. */
  3179. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  3180. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  3181. unlikely(shinfo->gso_type == 0)) {
  3182. __skb_warn_lro_forwarding(skb);
  3183. return true;
  3184. }
  3185. return false;
  3186. }
  3187. static inline void skb_forward_csum(struct sk_buff *skb)
  3188. {
  3189. /* Unfortunately we don't support this one. Any brave souls? */
  3190. if (skb->ip_summed == CHECKSUM_COMPLETE)
  3191. skb->ip_summed = CHECKSUM_NONE;
  3192. }
  3193. /**
  3194. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3195. * @skb: skb to check
  3196. *
  3197. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3198. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3199. * use this helper, to document places where we make this assertion.
  3200. */
  3201. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3202. {
  3203. #ifdef DEBUG
  3204. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3205. #endif
  3206. }
  3207. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3208. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3209. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3210. unsigned int transport_len,
  3211. __sum16(*skb_chkf)(struct sk_buff *skb));
  3212. /**
  3213. * skb_head_is_locked - Determine if the skb->head is locked down
  3214. * @skb: skb to check
  3215. *
  3216. * The head on skbs build around a head frag can be removed if they are
  3217. * not cloned. This function returns true if the skb head is locked down
  3218. * due to either being allocated via kmalloc, or by being a clone with
  3219. * multiple references to the head.
  3220. */
  3221. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3222. {
  3223. return !skb->head_frag || skb_cloned(skb);
  3224. }
  3225. /**
  3226. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3227. *
  3228. * @skb: GSO skb
  3229. *
  3230. * skb_gso_network_seglen is used to determine the real size of the
  3231. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3232. *
  3233. * The MAC/L2 header is not accounted for.
  3234. */
  3235. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3236. {
  3237. unsigned int hdr_len = skb_transport_header(skb) -
  3238. skb_network_header(skb);
  3239. return hdr_len + skb_gso_transport_seglen(skb);
  3240. }
  3241. #endif /* __KERNEL__ */
  3242. #endif /* _LINUX_SKBUFF_H */