skbuff.h 97 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472
  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. /* A. Checksumming of received packets by device.
  38. *
  39. * CHECKSUM_NONE:
  40. *
  41. * Device failed to checksum this packet e.g. due to lack of capabilities.
  42. * The packet contains full (though not verified) checksum in packet but
  43. * not in skb->csum. Thus, skb->csum is undefined in this case.
  44. *
  45. * CHECKSUM_UNNECESSARY:
  46. *
  47. * The hardware you're dealing with doesn't calculate the full checksum
  48. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  49. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  50. * if their checksums are okay. skb->csum is still undefined in this case
  51. * though. It is a bad option, but, unfortunately, nowadays most vendors do
  52. * this. Apparently with the secret goal to sell you new devices, when you
  53. * will add new protocol to your host, f.e. IPv6 8)
  54. *
  55. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  56. * TCP: IPv6 and IPv4.
  57. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  58. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  59. * may perform further validation in this case.
  60. * GRE: only if the checksum is present in the header.
  61. * SCTP: indicates the CRC in SCTP header has been validated.
  62. *
  63. * skb->csum_level indicates the number of consecutive checksums found in
  64. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  65. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  66. * and a device is able to verify the checksums for UDP (possibly zero),
  67. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  68. * two. If the device were only able to verify the UDP checksum and not
  69. * GRE, either because it doesn't support GRE checksum of because GRE
  70. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  71. * not considered in this case).
  72. *
  73. * CHECKSUM_COMPLETE:
  74. *
  75. * This is the most generic way. The device supplied checksum of the _whole_
  76. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  77. * hardware doesn't need to parse L3/L4 headers to implement this.
  78. *
  79. * Note: Even if device supports only some protocols, but is able to produce
  80. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  81. *
  82. * CHECKSUM_PARTIAL:
  83. *
  84. * A checksum is set up to be offloaded to a device as described in the
  85. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  86. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  87. * on the same host, or it may be set in the input path in GRO or remote
  88. * checksum offload. For the purposes of checksum verification, the checksum
  89. * referred to by skb->csum_start + skb->csum_offset and any preceding
  90. * checksums in the packet are considered verified. Any checksums in the
  91. * packet that are after the checksum being offloaded are not considered to
  92. * be verified.
  93. *
  94. * B. Checksumming on output.
  95. *
  96. * CHECKSUM_NONE:
  97. *
  98. * The skb was already checksummed by the protocol, or a checksum is not
  99. * required.
  100. *
  101. * CHECKSUM_PARTIAL:
  102. *
  103. * The device is required to checksum the packet as seen by hard_start_xmit()
  104. * from skb->csum_start up to the end, and to record/write the checksum at
  105. * offset skb->csum_start + skb->csum_offset.
  106. *
  107. * The device must show its capabilities in dev->features, set up at device
  108. * setup time, e.g. netdev_features.h:
  109. *
  110. * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
  111. * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
  112. * IPv4. Sigh. Vendors like this way for an unknown reason.
  113. * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
  114. * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
  115. * NETIF_F_... - Well, you get the picture.
  116. *
  117. * CHECKSUM_UNNECESSARY:
  118. *
  119. * Normally, the device will do per protocol specific checksumming. Protocol
  120. * implementations that do not want the NIC to perform the checksum
  121. * calculation should use this flag in their outgoing skbs.
  122. *
  123. * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
  124. * offload. Correspondingly, the FCoE protocol driver
  125. * stack should use CHECKSUM_UNNECESSARY.
  126. *
  127. * Any questions? No questions, good. --ANK
  128. */
  129. /* Don't change this without changing skb_csum_unnecessary! */
  130. #define CHECKSUM_NONE 0
  131. #define CHECKSUM_UNNECESSARY 1
  132. #define CHECKSUM_COMPLETE 2
  133. #define CHECKSUM_PARTIAL 3
  134. /* Maximum value in skb->csum_level */
  135. #define SKB_MAX_CSUM_LEVEL 3
  136. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  137. #define SKB_WITH_OVERHEAD(X) \
  138. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  139. #define SKB_MAX_ORDER(X, ORDER) \
  140. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  141. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  142. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  143. /* return minimum truesize of one skb containing X bytes of data */
  144. #define SKB_TRUESIZE(X) ((X) + \
  145. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  146. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  147. struct net_device;
  148. struct scatterlist;
  149. struct pipe_inode_info;
  150. struct iov_iter;
  151. struct napi_struct;
  152. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  153. struct nf_conntrack {
  154. atomic_t use;
  155. };
  156. #endif
  157. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  158. struct nf_bridge_info {
  159. atomic_t use;
  160. enum {
  161. BRNF_PROTO_UNCHANGED,
  162. BRNF_PROTO_8021Q,
  163. BRNF_PROTO_PPPOE
  164. } orig_proto:8;
  165. bool pkt_otherhost;
  166. __u16 frag_max_size;
  167. unsigned int mask;
  168. struct net_device *physindev;
  169. union {
  170. struct net_device *physoutdev;
  171. char neigh_header[8];
  172. };
  173. union {
  174. __be32 ipv4_daddr;
  175. struct in6_addr ipv6_daddr;
  176. };
  177. };
  178. #endif
  179. struct sk_buff_head {
  180. /* These two members must be first. */
  181. struct sk_buff *next;
  182. struct sk_buff *prev;
  183. __u32 qlen;
  184. spinlock_t lock;
  185. };
  186. struct sk_buff;
  187. /* To allow 64K frame to be packed as single skb without frag_list we
  188. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  189. * buffers which do not start on a page boundary.
  190. *
  191. * Since GRO uses frags we allocate at least 16 regardless of page
  192. * size.
  193. */
  194. #if (65536/PAGE_SIZE + 1) < 16
  195. #define MAX_SKB_FRAGS 16UL
  196. #else
  197. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  198. #endif
  199. typedef struct skb_frag_struct skb_frag_t;
  200. struct skb_frag_struct {
  201. struct {
  202. struct page *p;
  203. } page;
  204. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  205. __u32 page_offset;
  206. __u32 size;
  207. #else
  208. __u16 page_offset;
  209. __u16 size;
  210. #endif
  211. };
  212. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  213. {
  214. return frag->size;
  215. }
  216. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  217. {
  218. frag->size = size;
  219. }
  220. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  221. {
  222. frag->size += delta;
  223. }
  224. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  225. {
  226. frag->size -= delta;
  227. }
  228. #define HAVE_HW_TIME_STAMP
  229. /**
  230. * struct skb_shared_hwtstamps - hardware time stamps
  231. * @hwtstamp: hardware time stamp transformed into duration
  232. * since arbitrary point in time
  233. *
  234. * Software time stamps generated by ktime_get_real() are stored in
  235. * skb->tstamp.
  236. *
  237. * hwtstamps can only be compared against other hwtstamps from
  238. * the same device.
  239. *
  240. * This structure is attached to packets as part of the
  241. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  242. */
  243. struct skb_shared_hwtstamps {
  244. ktime_t hwtstamp;
  245. };
  246. /* Definitions for tx_flags in struct skb_shared_info */
  247. enum {
  248. /* generate hardware time stamp */
  249. SKBTX_HW_TSTAMP = 1 << 0,
  250. /* generate software time stamp when queueing packet to NIC */
  251. SKBTX_SW_TSTAMP = 1 << 1,
  252. /* device driver is going to provide hardware time stamp */
  253. SKBTX_IN_PROGRESS = 1 << 2,
  254. /* device driver supports TX zero-copy buffers */
  255. SKBTX_DEV_ZEROCOPY = 1 << 3,
  256. /* generate wifi status information (where possible) */
  257. SKBTX_WIFI_STATUS = 1 << 4,
  258. /* This indicates at least one fragment might be overwritten
  259. * (as in vmsplice(), sendfile() ...)
  260. * If we need to compute a TX checksum, we'll need to copy
  261. * all frags to avoid possible bad checksum
  262. */
  263. SKBTX_SHARED_FRAG = 1 << 5,
  264. /* generate software time stamp when entering packet scheduling */
  265. SKBTX_SCHED_TSTAMP = 1 << 6,
  266. /* generate software timestamp on peer data acknowledgment */
  267. SKBTX_ACK_TSTAMP = 1 << 7,
  268. };
  269. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  270. SKBTX_SCHED_TSTAMP | \
  271. SKBTX_ACK_TSTAMP)
  272. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  273. /*
  274. * The callback notifies userspace to release buffers when skb DMA is done in
  275. * lower device, the skb last reference should be 0 when calling this.
  276. * The zerocopy_success argument is true if zero copy transmit occurred,
  277. * false on data copy or out of memory error caused by data copy attempt.
  278. * The ctx field is used to track device context.
  279. * The desc field is used to track userspace buffer index.
  280. */
  281. struct ubuf_info {
  282. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  283. void *ctx;
  284. unsigned long desc;
  285. };
  286. /* This data is invariant across clones and lives at
  287. * the end of the header data, ie. at skb->end.
  288. */
  289. struct skb_shared_info {
  290. unsigned char nr_frags;
  291. __u8 tx_flags;
  292. unsigned short gso_size;
  293. /* Warning: this field is not always filled in (UFO)! */
  294. unsigned short gso_segs;
  295. unsigned short gso_type;
  296. struct sk_buff *frag_list;
  297. struct skb_shared_hwtstamps hwtstamps;
  298. u32 tskey;
  299. __be32 ip6_frag_id;
  300. /*
  301. * Warning : all fields before dataref are cleared in __alloc_skb()
  302. */
  303. atomic_t dataref;
  304. /* Intermediate layers must ensure that destructor_arg
  305. * remains valid until skb destructor */
  306. void * destructor_arg;
  307. /* must be last field, see pskb_expand_head() */
  308. skb_frag_t frags[MAX_SKB_FRAGS];
  309. };
  310. /* We divide dataref into two halves. The higher 16 bits hold references
  311. * to the payload part of skb->data. The lower 16 bits hold references to
  312. * the entire skb->data. A clone of a headerless skb holds the length of
  313. * the header in skb->hdr_len.
  314. *
  315. * All users must obey the rule that the skb->data reference count must be
  316. * greater than or equal to the payload reference count.
  317. *
  318. * Holding a reference to the payload part means that the user does not
  319. * care about modifications to the header part of skb->data.
  320. */
  321. #define SKB_DATAREF_SHIFT 16
  322. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  323. enum {
  324. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  325. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  326. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  327. };
  328. enum {
  329. SKB_GSO_TCPV4 = 1 << 0,
  330. SKB_GSO_UDP = 1 << 1,
  331. /* This indicates the skb is from an untrusted source. */
  332. SKB_GSO_DODGY = 1 << 2,
  333. /* This indicates the tcp segment has CWR set. */
  334. SKB_GSO_TCP_ECN = 1 << 3,
  335. SKB_GSO_TCPV6 = 1 << 4,
  336. SKB_GSO_FCOE = 1 << 5,
  337. SKB_GSO_GRE = 1 << 6,
  338. SKB_GSO_GRE_CSUM = 1 << 7,
  339. SKB_GSO_IPIP = 1 << 8,
  340. SKB_GSO_SIT = 1 << 9,
  341. SKB_GSO_UDP_TUNNEL = 1 << 10,
  342. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  343. SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
  344. };
  345. #if BITS_PER_LONG > 32
  346. #define NET_SKBUFF_DATA_USES_OFFSET 1
  347. #endif
  348. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  349. typedef unsigned int sk_buff_data_t;
  350. #else
  351. typedef unsigned char *sk_buff_data_t;
  352. #endif
  353. /**
  354. * struct skb_mstamp - multi resolution time stamps
  355. * @stamp_us: timestamp in us resolution
  356. * @stamp_jiffies: timestamp in jiffies
  357. */
  358. struct skb_mstamp {
  359. union {
  360. u64 v64;
  361. struct {
  362. u32 stamp_us;
  363. u32 stamp_jiffies;
  364. };
  365. };
  366. };
  367. /**
  368. * skb_mstamp_get - get current timestamp
  369. * @cl: place to store timestamps
  370. */
  371. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  372. {
  373. u64 val = local_clock();
  374. do_div(val, NSEC_PER_USEC);
  375. cl->stamp_us = (u32)val;
  376. cl->stamp_jiffies = (u32)jiffies;
  377. }
  378. /**
  379. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  380. * @t1: pointer to newest sample
  381. * @t0: pointer to oldest sample
  382. */
  383. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  384. const struct skb_mstamp *t0)
  385. {
  386. s32 delta_us = t1->stamp_us - t0->stamp_us;
  387. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  388. /* If delta_us is negative, this might be because interval is too big,
  389. * or local_clock() drift is too big : fallback using jiffies.
  390. */
  391. if (delta_us <= 0 ||
  392. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  393. delta_us = jiffies_to_usecs(delta_jiffies);
  394. return delta_us;
  395. }
  396. /**
  397. * struct sk_buff - socket buffer
  398. * @next: Next buffer in list
  399. * @prev: Previous buffer in list
  400. * @tstamp: Time we arrived/left
  401. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  402. * @sk: Socket we are owned by
  403. * @dev: Device we arrived on/are leaving by
  404. * @cb: Control buffer. Free for use by every layer. Put private vars here
  405. * @_skb_refdst: destination entry (with norefcount bit)
  406. * @sp: the security path, used for xfrm
  407. * @len: Length of actual data
  408. * @data_len: Data length
  409. * @mac_len: Length of link layer header
  410. * @hdr_len: writable header length of cloned skb
  411. * @csum: Checksum (must include start/offset pair)
  412. * @csum_start: Offset from skb->head where checksumming should start
  413. * @csum_offset: Offset from csum_start where checksum should be stored
  414. * @priority: Packet queueing priority
  415. * @ignore_df: allow local fragmentation
  416. * @cloned: Head may be cloned (check refcnt to be sure)
  417. * @ip_summed: Driver fed us an IP checksum
  418. * @nohdr: Payload reference only, must not modify header
  419. * @nfctinfo: Relationship of this skb to the connection
  420. * @pkt_type: Packet class
  421. * @fclone: skbuff clone status
  422. * @ipvs_property: skbuff is owned by ipvs
  423. * @peeked: this packet has been seen already, so stats have been
  424. * done for it, don't do them again
  425. * @nf_trace: netfilter packet trace flag
  426. * @protocol: Packet protocol from driver
  427. * @destructor: Destruct function
  428. * @nfct: Associated connection, if any
  429. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  430. * @skb_iif: ifindex of device we arrived on
  431. * @tc_index: Traffic control index
  432. * @tc_verd: traffic control verdict
  433. * @hash: the packet hash
  434. * @queue_mapping: Queue mapping for multiqueue devices
  435. * @xmit_more: More SKBs are pending for this queue
  436. * @ndisc_nodetype: router type (from link layer)
  437. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  438. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  439. * ports.
  440. * @sw_hash: indicates hash was computed in software stack
  441. * @wifi_acked_valid: wifi_acked was set
  442. * @wifi_acked: whether frame was acked on wifi or not
  443. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  444. * @napi_id: id of the NAPI struct this skb came from
  445. * @secmark: security marking
  446. * @mark: Generic packet mark
  447. * @vlan_proto: vlan encapsulation protocol
  448. * @vlan_tci: vlan tag control information
  449. * @inner_protocol: Protocol (encapsulation)
  450. * @inner_transport_header: Inner transport layer header (encapsulation)
  451. * @inner_network_header: Network layer header (encapsulation)
  452. * @inner_mac_header: Link layer header (encapsulation)
  453. * @transport_header: Transport layer header
  454. * @network_header: Network layer header
  455. * @mac_header: Link layer header
  456. * @tail: Tail pointer
  457. * @end: End pointer
  458. * @head: Head of buffer
  459. * @data: Data head pointer
  460. * @truesize: Buffer size
  461. * @users: User count - see {datagram,tcp}.c
  462. */
  463. struct sk_buff {
  464. union {
  465. struct {
  466. /* These two members must be first. */
  467. struct sk_buff *next;
  468. struct sk_buff *prev;
  469. union {
  470. ktime_t tstamp;
  471. struct skb_mstamp skb_mstamp;
  472. };
  473. };
  474. struct rb_node rbnode; /* used in netem & tcp stack */
  475. };
  476. struct sock *sk;
  477. struct net_device *dev;
  478. /*
  479. * This is the control buffer. It is free to use for every
  480. * layer. Please put your private variables there. If you
  481. * want to keep them across layers you have to do a skb_clone()
  482. * first. This is owned by whoever has the skb queued ATM.
  483. */
  484. char cb[48] __aligned(8);
  485. unsigned long _skb_refdst;
  486. void (*destructor)(struct sk_buff *skb);
  487. #ifdef CONFIG_XFRM
  488. struct sec_path *sp;
  489. #endif
  490. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  491. struct nf_conntrack *nfct;
  492. #endif
  493. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  494. struct nf_bridge_info *nf_bridge;
  495. #endif
  496. unsigned int len,
  497. data_len;
  498. __u16 mac_len,
  499. hdr_len;
  500. /* Following fields are _not_ copied in __copy_skb_header()
  501. * Note that queue_mapping is here mostly to fill a hole.
  502. */
  503. kmemcheck_bitfield_begin(flags1);
  504. __u16 queue_mapping;
  505. __u8 cloned:1,
  506. nohdr:1,
  507. fclone:2,
  508. peeked:1,
  509. head_frag:1,
  510. xmit_more:1;
  511. /* one bit hole */
  512. kmemcheck_bitfield_end(flags1);
  513. /* fields enclosed in headers_start/headers_end are copied
  514. * using a single memcpy() in __copy_skb_header()
  515. */
  516. /* private: */
  517. __u32 headers_start[0];
  518. /* public: */
  519. /* if you move pkt_type around you also must adapt those constants */
  520. #ifdef __BIG_ENDIAN_BITFIELD
  521. #define PKT_TYPE_MAX (7 << 5)
  522. #else
  523. #define PKT_TYPE_MAX 7
  524. #endif
  525. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  526. __u8 __pkt_type_offset[0];
  527. __u8 pkt_type:3;
  528. __u8 pfmemalloc:1;
  529. __u8 ignore_df:1;
  530. __u8 nfctinfo:3;
  531. __u8 nf_trace:1;
  532. __u8 ip_summed:2;
  533. __u8 ooo_okay:1;
  534. __u8 l4_hash:1;
  535. __u8 sw_hash:1;
  536. __u8 wifi_acked_valid:1;
  537. __u8 wifi_acked:1;
  538. __u8 no_fcs:1;
  539. /* Indicates the inner headers are valid in the skbuff. */
  540. __u8 encapsulation:1;
  541. __u8 encap_hdr_csum:1;
  542. __u8 csum_valid:1;
  543. __u8 csum_complete_sw:1;
  544. __u8 csum_level:2;
  545. __u8 csum_bad:1;
  546. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  547. __u8 ndisc_nodetype:2;
  548. #endif
  549. __u8 ipvs_property:1;
  550. __u8 inner_protocol_type:1;
  551. __u8 remcsum_offload:1;
  552. /* 3 or 5 bit hole */
  553. #ifdef CONFIG_NET_SCHED
  554. __u16 tc_index; /* traffic control index */
  555. #ifdef CONFIG_NET_CLS_ACT
  556. __u16 tc_verd; /* traffic control verdict */
  557. #endif
  558. #endif
  559. union {
  560. __wsum csum;
  561. struct {
  562. __u16 csum_start;
  563. __u16 csum_offset;
  564. };
  565. };
  566. __u32 priority;
  567. int skb_iif;
  568. __u32 hash;
  569. __be16 vlan_proto;
  570. __u16 vlan_tci;
  571. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  572. union {
  573. unsigned int napi_id;
  574. unsigned int sender_cpu;
  575. };
  576. #endif
  577. #ifdef CONFIG_NETWORK_SECMARK
  578. __u32 secmark;
  579. #endif
  580. union {
  581. __u32 mark;
  582. __u32 reserved_tailroom;
  583. };
  584. union {
  585. __be16 inner_protocol;
  586. __u8 inner_ipproto;
  587. };
  588. __u16 inner_transport_header;
  589. __u16 inner_network_header;
  590. __u16 inner_mac_header;
  591. __be16 protocol;
  592. __u16 transport_header;
  593. __u16 network_header;
  594. __u16 mac_header;
  595. /* private: */
  596. __u32 headers_end[0];
  597. /* public: */
  598. /* These elements must be at the end, see alloc_skb() for details. */
  599. sk_buff_data_t tail;
  600. sk_buff_data_t end;
  601. unsigned char *head,
  602. *data;
  603. unsigned int truesize;
  604. atomic_t users;
  605. };
  606. #ifdef __KERNEL__
  607. /*
  608. * Handling routines are only of interest to the kernel
  609. */
  610. #include <linux/slab.h>
  611. #define SKB_ALLOC_FCLONE 0x01
  612. #define SKB_ALLOC_RX 0x02
  613. #define SKB_ALLOC_NAPI 0x04
  614. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  615. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  616. {
  617. return unlikely(skb->pfmemalloc);
  618. }
  619. /*
  620. * skb might have a dst pointer attached, refcounted or not.
  621. * _skb_refdst low order bit is set if refcount was _not_ taken
  622. */
  623. #define SKB_DST_NOREF 1UL
  624. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  625. /**
  626. * skb_dst - returns skb dst_entry
  627. * @skb: buffer
  628. *
  629. * Returns skb dst_entry, regardless of reference taken or not.
  630. */
  631. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  632. {
  633. /* If refdst was not refcounted, check we still are in a
  634. * rcu_read_lock section
  635. */
  636. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  637. !rcu_read_lock_held() &&
  638. !rcu_read_lock_bh_held());
  639. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  640. }
  641. /**
  642. * skb_dst_set - sets skb dst
  643. * @skb: buffer
  644. * @dst: dst entry
  645. *
  646. * Sets skb dst, assuming a reference was taken on dst and should
  647. * be released by skb_dst_drop()
  648. */
  649. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  650. {
  651. skb->_skb_refdst = (unsigned long)dst;
  652. }
  653. /**
  654. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  655. * @skb: buffer
  656. * @dst: dst entry
  657. *
  658. * Sets skb dst, assuming a reference was not taken on dst.
  659. * If dst entry is cached, we do not take reference and dst_release
  660. * will be avoided by refdst_drop. If dst entry is not cached, we take
  661. * reference, so that last dst_release can destroy the dst immediately.
  662. */
  663. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  664. {
  665. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  666. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  667. }
  668. /**
  669. * skb_dst_is_noref - Test if skb dst isn't refcounted
  670. * @skb: buffer
  671. */
  672. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  673. {
  674. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  675. }
  676. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  677. {
  678. return (struct rtable *)skb_dst(skb);
  679. }
  680. void kfree_skb(struct sk_buff *skb);
  681. void kfree_skb_list(struct sk_buff *segs);
  682. void skb_tx_error(struct sk_buff *skb);
  683. void consume_skb(struct sk_buff *skb);
  684. void __kfree_skb(struct sk_buff *skb);
  685. extern struct kmem_cache *skbuff_head_cache;
  686. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  687. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  688. bool *fragstolen, int *delta_truesize);
  689. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  690. int node);
  691. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  692. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  693. static inline struct sk_buff *alloc_skb(unsigned int size,
  694. gfp_t priority)
  695. {
  696. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  697. }
  698. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  699. unsigned long data_len,
  700. int max_page_order,
  701. int *errcode,
  702. gfp_t gfp_mask);
  703. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  704. struct sk_buff_fclones {
  705. struct sk_buff skb1;
  706. struct sk_buff skb2;
  707. atomic_t fclone_ref;
  708. };
  709. /**
  710. * skb_fclone_busy - check if fclone is busy
  711. * @skb: buffer
  712. *
  713. * Returns true is skb is a fast clone, and its clone is not freed.
  714. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  715. * so we also check that this didnt happen.
  716. */
  717. static inline bool skb_fclone_busy(const struct sock *sk,
  718. const struct sk_buff *skb)
  719. {
  720. const struct sk_buff_fclones *fclones;
  721. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  722. return skb->fclone == SKB_FCLONE_ORIG &&
  723. atomic_read(&fclones->fclone_ref) > 1 &&
  724. fclones->skb2.sk == sk;
  725. }
  726. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  727. gfp_t priority)
  728. {
  729. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  730. }
  731. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  732. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  733. {
  734. return __alloc_skb_head(priority, -1);
  735. }
  736. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  737. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  738. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  739. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  740. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  741. gfp_t gfp_mask, bool fclone);
  742. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  743. gfp_t gfp_mask)
  744. {
  745. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  746. }
  747. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  748. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  749. unsigned int headroom);
  750. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  751. int newtailroom, gfp_t priority);
  752. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  753. int offset, int len);
  754. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  755. int len);
  756. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  757. int skb_pad(struct sk_buff *skb, int pad);
  758. #define dev_kfree_skb(a) consume_skb(a)
  759. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  760. int getfrag(void *from, char *to, int offset,
  761. int len, int odd, struct sk_buff *skb),
  762. void *from, int length);
  763. int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
  764. int offset, size_t size);
  765. struct skb_seq_state {
  766. __u32 lower_offset;
  767. __u32 upper_offset;
  768. __u32 frag_idx;
  769. __u32 stepped_offset;
  770. struct sk_buff *root_skb;
  771. struct sk_buff *cur_skb;
  772. __u8 *frag_data;
  773. };
  774. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  775. unsigned int to, struct skb_seq_state *st);
  776. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  777. struct skb_seq_state *st);
  778. void skb_abort_seq_read(struct skb_seq_state *st);
  779. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  780. unsigned int to, struct ts_config *config);
  781. /*
  782. * Packet hash types specify the type of hash in skb_set_hash.
  783. *
  784. * Hash types refer to the protocol layer addresses which are used to
  785. * construct a packet's hash. The hashes are used to differentiate or identify
  786. * flows of the protocol layer for the hash type. Hash types are either
  787. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  788. *
  789. * Properties of hashes:
  790. *
  791. * 1) Two packets in different flows have different hash values
  792. * 2) Two packets in the same flow should have the same hash value
  793. *
  794. * A hash at a higher layer is considered to be more specific. A driver should
  795. * set the most specific hash possible.
  796. *
  797. * A driver cannot indicate a more specific hash than the layer at which a hash
  798. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  799. *
  800. * A driver may indicate a hash level which is less specific than the
  801. * actual layer the hash was computed on. For instance, a hash computed
  802. * at L4 may be considered an L3 hash. This should only be done if the
  803. * driver can't unambiguously determine that the HW computed the hash at
  804. * the higher layer. Note that the "should" in the second property above
  805. * permits this.
  806. */
  807. enum pkt_hash_types {
  808. PKT_HASH_TYPE_NONE, /* Undefined type */
  809. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  810. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  811. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  812. };
  813. static inline void
  814. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  815. {
  816. skb->l4_hash = (type == PKT_HASH_TYPE_L4);
  817. skb->sw_hash = 0;
  818. skb->hash = hash;
  819. }
  820. static inline __u32 skb_get_hash(struct sk_buff *skb)
  821. {
  822. if (!skb->l4_hash && !skb->sw_hash)
  823. __skb_get_hash(skb);
  824. return skb->hash;
  825. }
  826. __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
  827. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  828. {
  829. return skb->hash;
  830. }
  831. static inline void skb_clear_hash(struct sk_buff *skb)
  832. {
  833. skb->hash = 0;
  834. skb->sw_hash = 0;
  835. skb->l4_hash = 0;
  836. }
  837. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  838. {
  839. if (!skb->l4_hash)
  840. skb_clear_hash(skb);
  841. }
  842. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  843. {
  844. to->hash = from->hash;
  845. to->sw_hash = from->sw_hash;
  846. to->l4_hash = from->l4_hash;
  847. };
  848. static inline void skb_sender_cpu_clear(struct sk_buff *skb)
  849. {
  850. #ifdef CONFIG_XPS
  851. skb->sender_cpu = 0;
  852. #endif
  853. }
  854. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  855. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  856. {
  857. return skb->head + skb->end;
  858. }
  859. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  860. {
  861. return skb->end;
  862. }
  863. #else
  864. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  865. {
  866. return skb->end;
  867. }
  868. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  869. {
  870. return skb->end - skb->head;
  871. }
  872. #endif
  873. /* Internal */
  874. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  875. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  876. {
  877. return &skb_shinfo(skb)->hwtstamps;
  878. }
  879. /**
  880. * skb_queue_empty - check if a queue is empty
  881. * @list: queue head
  882. *
  883. * Returns true if the queue is empty, false otherwise.
  884. */
  885. static inline int skb_queue_empty(const struct sk_buff_head *list)
  886. {
  887. return list->next == (const struct sk_buff *) list;
  888. }
  889. /**
  890. * skb_queue_is_last - check if skb is the last entry in the queue
  891. * @list: queue head
  892. * @skb: buffer
  893. *
  894. * Returns true if @skb is the last buffer on the list.
  895. */
  896. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  897. const struct sk_buff *skb)
  898. {
  899. return skb->next == (const struct sk_buff *) list;
  900. }
  901. /**
  902. * skb_queue_is_first - check if skb is the first entry in the queue
  903. * @list: queue head
  904. * @skb: buffer
  905. *
  906. * Returns true if @skb is the first buffer on the list.
  907. */
  908. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  909. const struct sk_buff *skb)
  910. {
  911. return skb->prev == (const struct sk_buff *) list;
  912. }
  913. /**
  914. * skb_queue_next - return the next packet in the queue
  915. * @list: queue head
  916. * @skb: current buffer
  917. *
  918. * Return the next packet in @list after @skb. It is only valid to
  919. * call this if skb_queue_is_last() evaluates to false.
  920. */
  921. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  922. const struct sk_buff *skb)
  923. {
  924. /* This BUG_ON may seem severe, but if we just return then we
  925. * are going to dereference garbage.
  926. */
  927. BUG_ON(skb_queue_is_last(list, skb));
  928. return skb->next;
  929. }
  930. /**
  931. * skb_queue_prev - return the prev packet in the queue
  932. * @list: queue head
  933. * @skb: current buffer
  934. *
  935. * Return the prev packet in @list before @skb. It is only valid to
  936. * call this if skb_queue_is_first() evaluates to false.
  937. */
  938. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  939. const struct sk_buff *skb)
  940. {
  941. /* This BUG_ON may seem severe, but if we just return then we
  942. * are going to dereference garbage.
  943. */
  944. BUG_ON(skb_queue_is_first(list, skb));
  945. return skb->prev;
  946. }
  947. /**
  948. * skb_get - reference buffer
  949. * @skb: buffer to reference
  950. *
  951. * Makes another reference to a socket buffer and returns a pointer
  952. * to the buffer.
  953. */
  954. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  955. {
  956. atomic_inc(&skb->users);
  957. return skb;
  958. }
  959. /*
  960. * If users == 1, we are the only owner and are can avoid redundant
  961. * atomic change.
  962. */
  963. /**
  964. * skb_cloned - is the buffer a clone
  965. * @skb: buffer to check
  966. *
  967. * Returns true if the buffer was generated with skb_clone() and is
  968. * one of multiple shared copies of the buffer. Cloned buffers are
  969. * shared data so must not be written to under normal circumstances.
  970. */
  971. static inline int skb_cloned(const struct sk_buff *skb)
  972. {
  973. return skb->cloned &&
  974. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  975. }
  976. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  977. {
  978. might_sleep_if(pri & __GFP_WAIT);
  979. if (skb_cloned(skb))
  980. return pskb_expand_head(skb, 0, 0, pri);
  981. return 0;
  982. }
  983. /**
  984. * skb_header_cloned - is the header a clone
  985. * @skb: buffer to check
  986. *
  987. * Returns true if modifying the header part of the buffer requires
  988. * the data to be copied.
  989. */
  990. static inline int skb_header_cloned(const struct sk_buff *skb)
  991. {
  992. int dataref;
  993. if (!skb->cloned)
  994. return 0;
  995. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  996. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  997. return dataref != 1;
  998. }
  999. /**
  1000. * skb_header_release - release reference to header
  1001. * @skb: buffer to operate on
  1002. *
  1003. * Drop a reference to the header part of the buffer. This is done
  1004. * by acquiring a payload reference. You must not read from the header
  1005. * part of skb->data after this.
  1006. * Note : Check if you can use __skb_header_release() instead.
  1007. */
  1008. static inline void skb_header_release(struct sk_buff *skb)
  1009. {
  1010. BUG_ON(skb->nohdr);
  1011. skb->nohdr = 1;
  1012. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1013. }
  1014. /**
  1015. * __skb_header_release - release reference to header
  1016. * @skb: buffer to operate on
  1017. *
  1018. * Variant of skb_header_release() assuming skb is private to caller.
  1019. * We can avoid one atomic operation.
  1020. */
  1021. static inline void __skb_header_release(struct sk_buff *skb)
  1022. {
  1023. skb->nohdr = 1;
  1024. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1025. }
  1026. /**
  1027. * skb_shared - is the buffer shared
  1028. * @skb: buffer to check
  1029. *
  1030. * Returns true if more than one person has a reference to this
  1031. * buffer.
  1032. */
  1033. static inline int skb_shared(const struct sk_buff *skb)
  1034. {
  1035. return atomic_read(&skb->users) != 1;
  1036. }
  1037. /**
  1038. * skb_share_check - check if buffer is shared and if so clone it
  1039. * @skb: buffer to check
  1040. * @pri: priority for memory allocation
  1041. *
  1042. * If the buffer is shared the buffer is cloned and the old copy
  1043. * drops a reference. A new clone with a single reference is returned.
  1044. * If the buffer is not shared the original buffer is returned. When
  1045. * being called from interrupt status or with spinlocks held pri must
  1046. * be GFP_ATOMIC.
  1047. *
  1048. * NULL is returned on a memory allocation failure.
  1049. */
  1050. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1051. {
  1052. might_sleep_if(pri & __GFP_WAIT);
  1053. if (skb_shared(skb)) {
  1054. struct sk_buff *nskb = skb_clone(skb, pri);
  1055. if (likely(nskb))
  1056. consume_skb(skb);
  1057. else
  1058. kfree_skb(skb);
  1059. skb = nskb;
  1060. }
  1061. return skb;
  1062. }
  1063. /*
  1064. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1065. * packets to handle cases where we have a local reader and forward
  1066. * and a couple of other messy ones. The normal one is tcpdumping
  1067. * a packet thats being forwarded.
  1068. */
  1069. /**
  1070. * skb_unshare - make a copy of a shared buffer
  1071. * @skb: buffer to check
  1072. * @pri: priority for memory allocation
  1073. *
  1074. * If the socket buffer is a clone then this function creates a new
  1075. * copy of the data, drops a reference count on the old copy and returns
  1076. * the new copy with the reference count at 1. If the buffer is not a clone
  1077. * the original buffer is returned. When called with a spinlock held or
  1078. * from interrupt state @pri must be %GFP_ATOMIC
  1079. *
  1080. * %NULL is returned on a memory allocation failure.
  1081. */
  1082. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1083. gfp_t pri)
  1084. {
  1085. might_sleep_if(pri & __GFP_WAIT);
  1086. if (skb_cloned(skb)) {
  1087. struct sk_buff *nskb = skb_copy(skb, pri);
  1088. /* Free our shared copy */
  1089. if (likely(nskb))
  1090. consume_skb(skb);
  1091. else
  1092. kfree_skb(skb);
  1093. skb = nskb;
  1094. }
  1095. return skb;
  1096. }
  1097. /**
  1098. * skb_peek - peek at the head of an &sk_buff_head
  1099. * @list_: list to peek at
  1100. *
  1101. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1102. * be careful with this one. A peek leaves the buffer on the
  1103. * list and someone else may run off with it. You must hold
  1104. * the appropriate locks or have a private queue to do this.
  1105. *
  1106. * Returns %NULL for an empty list or a pointer to the head element.
  1107. * The reference count is not incremented and the reference is therefore
  1108. * volatile. Use with caution.
  1109. */
  1110. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1111. {
  1112. struct sk_buff *skb = list_->next;
  1113. if (skb == (struct sk_buff *)list_)
  1114. skb = NULL;
  1115. return skb;
  1116. }
  1117. /**
  1118. * skb_peek_next - peek skb following the given one from a queue
  1119. * @skb: skb to start from
  1120. * @list_: list to peek at
  1121. *
  1122. * Returns %NULL when the end of the list is met or a pointer to the
  1123. * next element. The reference count is not incremented and the
  1124. * reference is therefore volatile. Use with caution.
  1125. */
  1126. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1127. const struct sk_buff_head *list_)
  1128. {
  1129. struct sk_buff *next = skb->next;
  1130. if (next == (struct sk_buff *)list_)
  1131. next = NULL;
  1132. return next;
  1133. }
  1134. /**
  1135. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1136. * @list_: list to peek at
  1137. *
  1138. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1139. * be careful with this one. A peek leaves the buffer on the
  1140. * list and someone else may run off with it. You must hold
  1141. * the appropriate locks or have a private queue to do this.
  1142. *
  1143. * Returns %NULL for an empty list or a pointer to the tail element.
  1144. * The reference count is not incremented and the reference is therefore
  1145. * volatile. Use with caution.
  1146. */
  1147. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1148. {
  1149. struct sk_buff *skb = list_->prev;
  1150. if (skb == (struct sk_buff *)list_)
  1151. skb = NULL;
  1152. return skb;
  1153. }
  1154. /**
  1155. * skb_queue_len - get queue length
  1156. * @list_: list to measure
  1157. *
  1158. * Return the length of an &sk_buff queue.
  1159. */
  1160. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1161. {
  1162. return list_->qlen;
  1163. }
  1164. /**
  1165. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1166. * @list: queue to initialize
  1167. *
  1168. * This initializes only the list and queue length aspects of
  1169. * an sk_buff_head object. This allows to initialize the list
  1170. * aspects of an sk_buff_head without reinitializing things like
  1171. * the spinlock. It can also be used for on-stack sk_buff_head
  1172. * objects where the spinlock is known to not be used.
  1173. */
  1174. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1175. {
  1176. list->prev = list->next = (struct sk_buff *)list;
  1177. list->qlen = 0;
  1178. }
  1179. /*
  1180. * This function creates a split out lock class for each invocation;
  1181. * this is needed for now since a whole lot of users of the skb-queue
  1182. * infrastructure in drivers have different locking usage (in hardirq)
  1183. * than the networking core (in softirq only). In the long run either the
  1184. * network layer or drivers should need annotation to consolidate the
  1185. * main types of usage into 3 classes.
  1186. */
  1187. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1188. {
  1189. spin_lock_init(&list->lock);
  1190. __skb_queue_head_init(list);
  1191. }
  1192. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1193. struct lock_class_key *class)
  1194. {
  1195. skb_queue_head_init(list);
  1196. lockdep_set_class(&list->lock, class);
  1197. }
  1198. /*
  1199. * Insert an sk_buff on a list.
  1200. *
  1201. * The "__skb_xxxx()" functions are the non-atomic ones that
  1202. * can only be called with interrupts disabled.
  1203. */
  1204. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1205. struct sk_buff_head *list);
  1206. static inline void __skb_insert(struct sk_buff *newsk,
  1207. struct sk_buff *prev, struct sk_buff *next,
  1208. struct sk_buff_head *list)
  1209. {
  1210. newsk->next = next;
  1211. newsk->prev = prev;
  1212. next->prev = prev->next = newsk;
  1213. list->qlen++;
  1214. }
  1215. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1216. struct sk_buff *prev,
  1217. struct sk_buff *next)
  1218. {
  1219. struct sk_buff *first = list->next;
  1220. struct sk_buff *last = list->prev;
  1221. first->prev = prev;
  1222. prev->next = first;
  1223. last->next = next;
  1224. next->prev = last;
  1225. }
  1226. /**
  1227. * skb_queue_splice - join two skb lists, this is designed for stacks
  1228. * @list: the new list to add
  1229. * @head: the place to add it in the first list
  1230. */
  1231. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1232. struct sk_buff_head *head)
  1233. {
  1234. if (!skb_queue_empty(list)) {
  1235. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1236. head->qlen += list->qlen;
  1237. }
  1238. }
  1239. /**
  1240. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1241. * @list: the new list to add
  1242. * @head: the place to add it in the first list
  1243. *
  1244. * The list at @list is reinitialised
  1245. */
  1246. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1247. struct sk_buff_head *head)
  1248. {
  1249. if (!skb_queue_empty(list)) {
  1250. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1251. head->qlen += list->qlen;
  1252. __skb_queue_head_init(list);
  1253. }
  1254. }
  1255. /**
  1256. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1257. * @list: the new list to add
  1258. * @head: the place to add it in the first list
  1259. */
  1260. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1261. struct sk_buff_head *head)
  1262. {
  1263. if (!skb_queue_empty(list)) {
  1264. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1265. head->qlen += list->qlen;
  1266. }
  1267. }
  1268. /**
  1269. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1270. * @list: the new list to add
  1271. * @head: the place to add it in the first list
  1272. *
  1273. * Each of the lists is a queue.
  1274. * The list at @list is reinitialised
  1275. */
  1276. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1277. struct sk_buff_head *head)
  1278. {
  1279. if (!skb_queue_empty(list)) {
  1280. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1281. head->qlen += list->qlen;
  1282. __skb_queue_head_init(list);
  1283. }
  1284. }
  1285. /**
  1286. * __skb_queue_after - queue a buffer at the list head
  1287. * @list: list to use
  1288. * @prev: place after this buffer
  1289. * @newsk: buffer to queue
  1290. *
  1291. * Queue a buffer int the middle of a list. This function takes no locks
  1292. * and you must therefore hold required locks before calling it.
  1293. *
  1294. * A buffer cannot be placed on two lists at the same time.
  1295. */
  1296. static inline void __skb_queue_after(struct sk_buff_head *list,
  1297. struct sk_buff *prev,
  1298. struct sk_buff *newsk)
  1299. {
  1300. __skb_insert(newsk, prev, prev->next, list);
  1301. }
  1302. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1303. struct sk_buff_head *list);
  1304. static inline void __skb_queue_before(struct sk_buff_head *list,
  1305. struct sk_buff *next,
  1306. struct sk_buff *newsk)
  1307. {
  1308. __skb_insert(newsk, next->prev, next, list);
  1309. }
  1310. /**
  1311. * __skb_queue_head - queue a buffer at the list head
  1312. * @list: list to use
  1313. * @newsk: buffer to queue
  1314. *
  1315. * Queue a buffer at the start of a list. This function takes no locks
  1316. * and you must therefore hold required locks before calling it.
  1317. *
  1318. * A buffer cannot be placed on two lists at the same time.
  1319. */
  1320. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1321. static inline void __skb_queue_head(struct sk_buff_head *list,
  1322. struct sk_buff *newsk)
  1323. {
  1324. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1325. }
  1326. /**
  1327. * __skb_queue_tail - queue a buffer at the list tail
  1328. * @list: list to use
  1329. * @newsk: buffer to queue
  1330. *
  1331. * Queue a buffer at the end of a list. This function takes no locks
  1332. * and you must therefore hold required locks before calling it.
  1333. *
  1334. * A buffer cannot be placed on two lists at the same time.
  1335. */
  1336. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1337. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1338. struct sk_buff *newsk)
  1339. {
  1340. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1341. }
  1342. /*
  1343. * remove sk_buff from list. _Must_ be called atomically, and with
  1344. * the list known..
  1345. */
  1346. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1347. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1348. {
  1349. struct sk_buff *next, *prev;
  1350. list->qlen--;
  1351. next = skb->next;
  1352. prev = skb->prev;
  1353. skb->next = skb->prev = NULL;
  1354. next->prev = prev;
  1355. prev->next = next;
  1356. }
  1357. /**
  1358. * __skb_dequeue - remove from the head of the queue
  1359. * @list: list to dequeue from
  1360. *
  1361. * Remove the head of the list. This function does not take any locks
  1362. * so must be used with appropriate locks held only. The head item is
  1363. * returned or %NULL if the list is empty.
  1364. */
  1365. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1366. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1367. {
  1368. struct sk_buff *skb = skb_peek(list);
  1369. if (skb)
  1370. __skb_unlink(skb, list);
  1371. return skb;
  1372. }
  1373. /**
  1374. * __skb_dequeue_tail - remove from the tail of the queue
  1375. * @list: list to dequeue from
  1376. *
  1377. * Remove the tail of the list. This function does not take any locks
  1378. * so must be used with appropriate locks held only. The tail item is
  1379. * returned or %NULL if the list is empty.
  1380. */
  1381. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1382. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1383. {
  1384. struct sk_buff *skb = skb_peek_tail(list);
  1385. if (skb)
  1386. __skb_unlink(skb, list);
  1387. return skb;
  1388. }
  1389. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1390. {
  1391. return skb->data_len;
  1392. }
  1393. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1394. {
  1395. return skb->len - skb->data_len;
  1396. }
  1397. static inline int skb_pagelen(const struct sk_buff *skb)
  1398. {
  1399. int i, len = 0;
  1400. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1401. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1402. return len + skb_headlen(skb);
  1403. }
  1404. /**
  1405. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1406. * @skb: buffer containing fragment to be initialised
  1407. * @i: paged fragment index to initialise
  1408. * @page: the page to use for this fragment
  1409. * @off: the offset to the data with @page
  1410. * @size: the length of the data
  1411. *
  1412. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1413. * offset @off within @page.
  1414. *
  1415. * Does not take any additional reference on the fragment.
  1416. */
  1417. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1418. struct page *page, int off, int size)
  1419. {
  1420. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1421. /*
  1422. * Propagate page->pfmemalloc to the skb if we can. The problem is
  1423. * that not all callers have unique ownership of the page. If
  1424. * pfmemalloc is set, we check the mapping as a mapping implies
  1425. * page->index is set (index and pfmemalloc share space).
  1426. * If it's a valid mapping, we cannot use page->pfmemalloc but we
  1427. * do not lose pfmemalloc information as the pages would not be
  1428. * allocated using __GFP_MEMALLOC.
  1429. */
  1430. frag->page.p = page;
  1431. frag->page_offset = off;
  1432. skb_frag_size_set(frag, size);
  1433. page = compound_head(page);
  1434. if (page->pfmemalloc && !page->mapping)
  1435. skb->pfmemalloc = true;
  1436. }
  1437. /**
  1438. * skb_fill_page_desc - initialise a paged fragment in an skb
  1439. * @skb: buffer containing fragment to be initialised
  1440. * @i: paged fragment index to initialise
  1441. * @page: the page to use for this fragment
  1442. * @off: the offset to the data with @page
  1443. * @size: the length of the data
  1444. *
  1445. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1446. * @skb to point to @size bytes at offset @off within @page. In
  1447. * addition updates @skb such that @i is the last fragment.
  1448. *
  1449. * Does not take any additional reference on the fragment.
  1450. */
  1451. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1452. struct page *page, int off, int size)
  1453. {
  1454. __skb_fill_page_desc(skb, i, page, off, size);
  1455. skb_shinfo(skb)->nr_frags = i + 1;
  1456. }
  1457. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1458. int size, unsigned int truesize);
  1459. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1460. unsigned int truesize);
  1461. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1462. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1463. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1464. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1465. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1466. {
  1467. return skb->head + skb->tail;
  1468. }
  1469. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1470. {
  1471. skb->tail = skb->data - skb->head;
  1472. }
  1473. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1474. {
  1475. skb_reset_tail_pointer(skb);
  1476. skb->tail += offset;
  1477. }
  1478. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1479. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1480. {
  1481. return skb->tail;
  1482. }
  1483. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1484. {
  1485. skb->tail = skb->data;
  1486. }
  1487. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1488. {
  1489. skb->tail = skb->data + offset;
  1490. }
  1491. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1492. /*
  1493. * Add data to an sk_buff
  1494. */
  1495. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1496. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1497. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1498. {
  1499. unsigned char *tmp = skb_tail_pointer(skb);
  1500. SKB_LINEAR_ASSERT(skb);
  1501. skb->tail += len;
  1502. skb->len += len;
  1503. return tmp;
  1504. }
  1505. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1506. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1507. {
  1508. skb->data -= len;
  1509. skb->len += len;
  1510. return skb->data;
  1511. }
  1512. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1513. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1514. {
  1515. skb->len -= len;
  1516. BUG_ON(skb->len < skb->data_len);
  1517. return skb->data += len;
  1518. }
  1519. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1520. {
  1521. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1522. }
  1523. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1524. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1525. {
  1526. if (len > skb_headlen(skb) &&
  1527. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1528. return NULL;
  1529. skb->len -= len;
  1530. return skb->data += len;
  1531. }
  1532. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1533. {
  1534. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1535. }
  1536. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1537. {
  1538. if (likely(len <= skb_headlen(skb)))
  1539. return 1;
  1540. if (unlikely(len > skb->len))
  1541. return 0;
  1542. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1543. }
  1544. /**
  1545. * skb_headroom - bytes at buffer head
  1546. * @skb: buffer to check
  1547. *
  1548. * Return the number of bytes of free space at the head of an &sk_buff.
  1549. */
  1550. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1551. {
  1552. return skb->data - skb->head;
  1553. }
  1554. /**
  1555. * skb_tailroom - bytes at buffer end
  1556. * @skb: buffer to check
  1557. *
  1558. * Return the number of bytes of free space at the tail of an sk_buff
  1559. */
  1560. static inline int skb_tailroom(const struct sk_buff *skb)
  1561. {
  1562. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1563. }
  1564. /**
  1565. * skb_availroom - bytes at buffer end
  1566. * @skb: buffer to check
  1567. *
  1568. * Return the number of bytes of free space at the tail of an sk_buff
  1569. * allocated by sk_stream_alloc()
  1570. */
  1571. static inline int skb_availroom(const struct sk_buff *skb)
  1572. {
  1573. if (skb_is_nonlinear(skb))
  1574. return 0;
  1575. return skb->end - skb->tail - skb->reserved_tailroom;
  1576. }
  1577. /**
  1578. * skb_reserve - adjust headroom
  1579. * @skb: buffer to alter
  1580. * @len: bytes to move
  1581. *
  1582. * Increase the headroom of an empty &sk_buff by reducing the tail
  1583. * room. This is only allowed for an empty buffer.
  1584. */
  1585. static inline void skb_reserve(struct sk_buff *skb, int len)
  1586. {
  1587. skb->data += len;
  1588. skb->tail += len;
  1589. }
  1590. #define ENCAP_TYPE_ETHER 0
  1591. #define ENCAP_TYPE_IPPROTO 1
  1592. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1593. __be16 protocol)
  1594. {
  1595. skb->inner_protocol = protocol;
  1596. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1597. }
  1598. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1599. __u8 ipproto)
  1600. {
  1601. skb->inner_ipproto = ipproto;
  1602. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1603. }
  1604. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1605. {
  1606. skb->inner_mac_header = skb->mac_header;
  1607. skb->inner_network_header = skb->network_header;
  1608. skb->inner_transport_header = skb->transport_header;
  1609. }
  1610. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1611. {
  1612. skb->mac_len = skb->network_header - skb->mac_header;
  1613. }
  1614. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1615. *skb)
  1616. {
  1617. return skb->head + skb->inner_transport_header;
  1618. }
  1619. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1620. {
  1621. skb->inner_transport_header = skb->data - skb->head;
  1622. }
  1623. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1624. const int offset)
  1625. {
  1626. skb_reset_inner_transport_header(skb);
  1627. skb->inner_transport_header += offset;
  1628. }
  1629. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1630. {
  1631. return skb->head + skb->inner_network_header;
  1632. }
  1633. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1634. {
  1635. skb->inner_network_header = skb->data - skb->head;
  1636. }
  1637. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1638. const int offset)
  1639. {
  1640. skb_reset_inner_network_header(skb);
  1641. skb->inner_network_header += offset;
  1642. }
  1643. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1644. {
  1645. return skb->head + skb->inner_mac_header;
  1646. }
  1647. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1648. {
  1649. skb->inner_mac_header = skb->data - skb->head;
  1650. }
  1651. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1652. const int offset)
  1653. {
  1654. skb_reset_inner_mac_header(skb);
  1655. skb->inner_mac_header += offset;
  1656. }
  1657. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1658. {
  1659. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1660. }
  1661. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1662. {
  1663. return skb->head + skb->transport_header;
  1664. }
  1665. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1666. {
  1667. skb->transport_header = skb->data - skb->head;
  1668. }
  1669. static inline void skb_set_transport_header(struct sk_buff *skb,
  1670. const int offset)
  1671. {
  1672. skb_reset_transport_header(skb);
  1673. skb->transport_header += offset;
  1674. }
  1675. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1676. {
  1677. return skb->head + skb->network_header;
  1678. }
  1679. static inline void skb_reset_network_header(struct sk_buff *skb)
  1680. {
  1681. skb->network_header = skb->data - skb->head;
  1682. }
  1683. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1684. {
  1685. skb_reset_network_header(skb);
  1686. skb->network_header += offset;
  1687. }
  1688. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1689. {
  1690. return skb->head + skb->mac_header;
  1691. }
  1692. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1693. {
  1694. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1695. }
  1696. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1697. {
  1698. skb->mac_header = skb->data - skb->head;
  1699. }
  1700. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1701. {
  1702. skb_reset_mac_header(skb);
  1703. skb->mac_header += offset;
  1704. }
  1705. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1706. {
  1707. skb->mac_header = skb->network_header;
  1708. }
  1709. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1710. const int offset_hint)
  1711. {
  1712. struct flow_keys keys;
  1713. if (skb_transport_header_was_set(skb))
  1714. return;
  1715. else if (skb_flow_dissect_flow_keys(skb, &keys))
  1716. skb_set_transport_header(skb, keys.control.thoff);
  1717. else
  1718. skb_set_transport_header(skb, offset_hint);
  1719. }
  1720. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1721. {
  1722. if (skb_mac_header_was_set(skb)) {
  1723. const unsigned char *old_mac = skb_mac_header(skb);
  1724. skb_set_mac_header(skb, -skb->mac_len);
  1725. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1726. }
  1727. }
  1728. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1729. {
  1730. return skb->csum_start - skb_headroom(skb);
  1731. }
  1732. static inline int skb_transport_offset(const struct sk_buff *skb)
  1733. {
  1734. return skb_transport_header(skb) - skb->data;
  1735. }
  1736. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1737. {
  1738. return skb->transport_header - skb->network_header;
  1739. }
  1740. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1741. {
  1742. return skb->inner_transport_header - skb->inner_network_header;
  1743. }
  1744. static inline int skb_network_offset(const struct sk_buff *skb)
  1745. {
  1746. return skb_network_header(skb) - skb->data;
  1747. }
  1748. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1749. {
  1750. return skb_inner_network_header(skb) - skb->data;
  1751. }
  1752. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1753. {
  1754. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1755. }
  1756. /*
  1757. * CPUs often take a performance hit when accessing unaligned memory
  1758. * locations. The actual performance hit varies, it can be small if the
  1759. * hardware handles it or large if we have to take an exception and fix it
  1760. * in software.
  1761. *
  1762. * Since an ethernet header is 14 bytes network drivers often end up with
  1763. * the IP header at an unaligned offset. The IP header can be aligned by
  1764. * shifting the start of the packet by 2 bytes. Drivers should do this
  1765. * with:
  1766. *
  1767. * skb_reserve(skb, NET_IP_ALIGN);
  1768. *
  1769. * The downside to this alignment of the IP header is that the DMA is now
  1770. * unaligned. On some architectures the cost of an unaligned DMA is high
  1771. * and this cost outweighs the gains made by aligning the IP header.
  1772. *
  1773. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1774. * to be overridden.
  1775. */
  1776. #ifndef NET_IP_ALIGN
  1777. #define NET_IP_ALIGN 2
  1778. #endif
  1779. /*
  1780. * The networking layer reserves some headroom in skb data (via
  1781. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1782. * the header has to grow. In the default case, if the header has to grow
  1783. * 32 bytes or less we avoid the reallocation.
  1784. *
  1785. * Unfortunately this headroom changes the DMA alignment of the resulting
  1786. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1787. * on some architectures. An architecture can override this value,
  1788. * perhaps setting it to a cacheline in size (since that will maintain
  1789. * cacheline alignment of the DMA). It must be a power of 2.
  1790. *
  1791. * Various parts of the networking layer expect at least 32 bytes of
  1792. * headroom, you should not reduce this.
  1793. *
  1794. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1795. * to reduce average number of cache lines per packet.
  1796. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1797. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1798. */
  1799. #ifndef NET_SKB_PAD
  1800. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1801. #endif
  1802. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1803. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1804. {
  1805. if (unlikely(skb_is_nonlinear(skb))) {
  1806. WARN_ON(1);
  1807. return;
  1808. }
  1809. skb->len = len;
  1810. skb_set_tail_pointer(skb, len);
  1811. }
  1812. void skb_trim(struct sk_buff *skb, unsigned int len);
  1813. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1814. {
  1815. if (skb->data_len)
  1816. return ___pskb_trim(skb, len);
  1817. __skb_trim(skb, len);
  1818. return 0;
  1819. }
  1820. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1821. {
  1822. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1823. }
  1824. /**
  1825. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1826. * @skb: buffer to alter
  1827. * @len: new length
  1828. *
  1829. * This is identical to pskb_trim except that the caller knows that
  1830. * the skb is not cloned so we should never get an error due to out-
  1831. * of-memory.
  1832. */
  1833. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  1834. {
  1835. int err = pskb_trim(skb, len);
  1836. BUG_ON(err);
  1837. }
  1838. /**
  1839. * skb_orphan - orphan a buffer
  1840. * @skb: buffer to orphan
  1841. *
  1842. * If a buffer currently has an owner then we call the owner's
  1843. * destructor function and make the @skb unowned. The buffer continues
  1844. * to exist but is no longer charged to its former owner.
  1845. */
  1846. static inline void skb_orphan(struct sk_buff *skb)
  1847. {
  1848. if (skb->destructor) {
  1849. skb->destructor(skb);
  1850. skb->destructor = NULL;
  1851. skb->sk = NULL;
  1852. } else {
  1853. BUG_ON(skb->sk);
  1854. }
  1855. }
  1856. /**
  1857. * skb_orphan_frags - orphan the frags contained in a buffer
  1858. * @skb: buffer to orphan frags from
  1859. * @gfp_mask: allocation mask for replacement pages
  1860. *
  1861. * For each frag in the SKB which needs a destructor (i.e. has an
  1862. * owner) create a copy of that frag and release the original
  1863. * page by calling the destructor.
  1864. */
  1865. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  1866. {
  1867. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  1868. return 0;
  1869. return skb_copy_ubufs(skb, gfp_mask);
  1870. }
  1871. /**
  1872. * __skb_queue_purge - empty a list
  1873. * @list: list to empty
  1874. *
  1875. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  1876. * the list and one reference dropped. This function does not take the
  1877. * list lock and the caller must hold the relevant locks to use it.
  1878. */
  1879. void skb_queue_purge(struct sk_buff_head *list);
  1880. static inline void __skb_queue_purge(struct sk_buff_head *list)
  1881. {
  1882. struct sk_buff *skb;
  1883. while ((skb = __skb_dequeue(list)) != NULL)
  1884. kfree_skb(skb);
  1885. }
  1886. void *netdev_alloc_frag(unsigned int fragsz);
  1887. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  1888. gfp_t gfp_mask);
  1889. /**
  1890. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  1891. * @dev: network device to receive on
  1892. * @length: length to allocate
  1893. *
  1894. * Allocate a new &sk_buff and assign it a usage count of one. The
  1895. * buffer has unspecified headroom built in. Users should allocate
  1896. * the headroom they think they need without accounting for the
  1897. * built in space. The built in space is used for optimisations.
  1898. *
  1899. * %NULL is returned if there is no free memory. Although this function
  1900. * allocates memory it can be called from an interrupt.
  1901. */
  1902. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  1903. unsigned int length)
  1904. {
  1905. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  1906. }
  1907. /* legacy helper around __netdev_alloc_skb() */
  1908. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  1909. gfp_t gfp_mask)
  1910. {
  1911. return __netdev_alloc_skb(NULL, length, gfp_mask);
  1912. }
  1913. /* legacy helper around netdev_alloc_skb() */
  1914. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  1915. {
  1916. return netdev_alloc_skb(NULL, length);
  1917. }
  1918. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  1919. unsigned int length, gfp_t gfp)
  1920. {
  1921. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  1922. if (NET_IP_ALIGN && skb)
  1923. skb_reserve(skb, NET_IP_ALIGN);
  1924. return skb;
  1925. }
  1926. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  1927. unsigned int length)
  1928. {
  1929. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  1930. }
  1931. static inline void skb_free_frag(void *addr)
  1932. {
  1933. __free_page_frag(addr);
  1934. }
  1935. void *napi_alloc_frag(unsigned int fragsz);
  1936. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  1937. unsigned int length, gfp_t gfp_mask);
  1938. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  1939. unsigned int length)
  1940. {
  1941. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  1942. }
  1943. /**
  1944. * __dev_alloc_pages - allocate page for network Rx
  1945. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1946. * @order: size of the allocation
  1947. *
  1948. * Allocate a new page.
  1949. *
  1950. * %NULL is returned if there is no free memory.
  1951. */
  1952. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  1953. unsigned int order)
  1954. {
  1955. /* This piece of code contains several assumptions.
  1956. * 1. This is for device Rx, therefor a cold page is preferred.
  1957. * 2. The expectation is the user wants a compound page.
  1958. * 3. If requesting a order 0 page it will not be compound
  1959. * due to the check to see if order has a value in prep_new_page
  1960. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  1961. * code in gfp_to_alloc_flags that should be enforcing this.
  1962. */
  1963. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  1964. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  1965. }
  1966. static inline struct page *dev_alloc_pages(unsigned int order)
  1967. {
  1968. return __dev_alloc_pages(GFP_ATOMIC, order);
  1969. }
  1970. /**
  1971. * __dev_alloc_page - allocate a page for network Rx
  1972. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1973. *
  1974. * Allocate a new page.
  1975. *
  1976. * %NULL is returned if there is no free memory.
  1977. */
  1978. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  1979. {
  1980. return __dev_alloc_pages(gfp_mask, 0);
  1981. }
  1982. static inline struct page *dev_alloc_page(void)
  1983. {
  1984. return __dev_alloc_page(GFP_ATOMIC);
  1985. }
  1986. /**
  1987. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  1988. * @page: The page that was allocated from skb_alloc_page
  1989. * @skb: The skb that may need pfmemalloc set
  1990. */
  1991. static inline void skb_propagate_pfmemalloc(struct page *page,
  1992. struct sk_buff *skb)
  1993. {
  1994. if (page && page->pfmemalloc)
  1995. skb->pfmemalloc = true;
  1996. }
  1997. /**
  1998. * skb_frag_page - retrieve the page referred to by a paged fragment
  1999. * @frag: the paged fragment
  2000. *
  2001. * Returns the &struct page associated with @frag.
  2002. */
  2003. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  2004. {
  2005. return frag->page.p;
  2006. }
  2007. /**
  2008. * __skb_frag_ref - take an addition reference on a paged fragment.
  2009. * @frag: the paged fragment
  2010. *
  2011. * Takes an additional reference on the paged fragment @frag.
  2012. */
  2013. static inline void __skb_frag_ref(skb_frag_t *frag)
  2014. {
  2015. get_page(skb_frag_page(frag));
  2016. }
  2017. /**
  2018. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2019. * @skb: the buffer
  2020. * @f: the fragment offset.
  2021. *
  2022. * Takes an additional reference on the @f'th paged fragment of @skb.
  2023. */
  2024. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2025. {
  2026. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2027. }
  2028. /**
  2029. * __skb_frag_unref - release a reference on a paged fragment.
  2030. * @frag: the paged fragment
  2031. *
  2032. * Releases a reference on the paged fragment @frag.
  2033. */
  2034. static inline void __skb_frag_unref(skb_frag_t *frag)
  2035. {
  2036. put_page(skb_frag_page(frag));
  2037. }
  2038. /**
  2039. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2040. * @skb: the buffer
  2041. * @f: the fragment offset
  2042. *
  2043. * Releases a reference on the @f'th paged fragment of @skb.
  2044. */
  2045. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2046. {
  2047. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2048. }
  2049. /**
  2050. * skb_frag_address - gets the address of the data contained in a paged fragment
  2051. * @frag: the paged fragment buffer
  2052. *
  2053. * Returns the address of the data within @frag. The page must already
  2054. * be mapped.
  2055. */
  2056. static inline void *skb_frag_address(const skb_frag_t *frag)
  2057. {
  2058. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2059. }
  2060. /**
  2061. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2062. * @frag: the paged fragment buffer
  2063. *
  2064. * Returns the address of the data within @frag. Checks that the page
  2065. * is mapped and returns %NULL otherwise.
  2066. */
  2067. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2068. {
  2069. void *ptr = page_address(skb_frag_page(frag));
  2070. if (unlikely(!ptr))
  2071. return NULL;
  2072. return ptr + frag->page_offset;
  2073. }
  2074. /**
  2075. * __skb_frag_set_page - sets the page contained in a paged fragment
  2076. * @frag: the paged fragment
  2077. * @page: the page to set
  2078. *
  2079. * Sets the fragment @frag to contain @page.
  2080. */
  2081. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2082. {
  2083. frag->page.p = page;
  2084. }
  2085. /**
  2086. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2087. * @skb: the buffer
  2088. * @f: the fragment offset
  2089. * @page: the page to set
  2090. *
  2091. * Sets the @f'th fragment of @skb to contain @page.
  2092. */
  2093. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2094. struct page *page)
  2095. {
  2096. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2097. }
  2098. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2099. /**
  2100. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2101. * @dev: the device to map the fragment to
  2102. * @frag: the paged fragment to map
  2103. * @offset: the offset within the fragment (starting at the
  2104. * fragment's own offset)
  2105. * @size: the number of bytes to map
  2106. * @dir: the direction of the mapping (%PCI_DMA_*)
  2107. *
  2108. * Maps the page associated with @frag to @device.
  2109. */
  2110. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2111. const skb_frag_t *frag,
  2112. size_t offset, size_t size,
  2113. enum dma_data_direction dir)
  2114. {
  2115. return dma_map_page(dev, skb_frag_page(frag),
  2116. frag->page_offset + offset, size, dir);
  2117. }
  2118. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2119. gfp_t gfp_mask)
  2120. {
  2121. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2122. }
  2123. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2124. gfp_t gfp_mask)
  2125. {
  2126. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2127. }
  2128. /**
  2129. * skb_clone_writable - is the header of a clone writable
  2130. * @skb: buffer to check
  2131. * @len: length up to which to write
  2132. *
  2133. * Returns true if modifying the header part of the cloned buffer
  2134. * does not requires the data to be copied.
  2135. */
  2136. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2137. {
  2138. return !skb_header_cloned(skb) &&
  2139. skb_headroom(skb) + len <= skb->hdr_len;
  2140. }
  2141. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2142. int cloned)
  2143. {
  2144. int delta = 0;
  2145. if (headroom > skb_headroom(skb))
  2146. delta = headroom - skb_headroom(skb);
  2147. if (delta || cloned)
  2148. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2149. GFP_ATOMIC);
  2150. return 0;
  2151. }
  2152. /**
  2153. * skb_cow - copy header of skb when it is required
  2154. * @skb: buffer to cow
  2155. * @headroom: needed headroom
  2156. *
  2157. * If the skb passed lacks sufficient headroom or its data part
  2158. * is shared, data is reallocated. If reallocation fails, an error
  2159. * is returned and original skb is not changed.
  2160. *
  2161. * The result is skb with writable area skb->head...skb->tail
  2162. * and at least @headroom of space at head.
  2163. */
  2164. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2165. {
  2166. return __skb_cow(skb, headroom, skb_cloned(skb));
  2167. }
  2168. /**
  2169. * skb_cow_head - skb_cow but only making the head writable
  2170. * @skb: buffer to cow
  2171. * @headroom: needed headroom
  2172. *
  2173. * This function is identical to skb_cow except that we replace the
  2174. * skb_cloned check by skb_header_cloned. It should be used when
  2175. * you only need to push on some header and do not need to modify
  2176. * the data.
  2177. */
  2178. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2179. {
  2180. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2181. }
  2182. /**
  2183. * skb_padto - pad an skbuff up to a minimal size
  2184. * @skb: buffer to pad
  2185. * @len: minimal length
  2186. *
  2187. * Pads up a buffer to ensure the trailing bytes exist and are
  2188. * blanked. If the buffer already contains sufficient data it
  2189. * is untouched. Otherwise it is extended. Returns zero on
  2190. * success. The skb is freed on error.
  2191. */
  2192. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2193. {
  2194. unsigned int size = skb->len;
  2195. if (likely(size >= len))
  2196. return 0;
  2197. return skb_pad(skb, len - size);
  2198. }
  2199. /**
  2200. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2201. * @skb: buffer to pad
  2202. * @len: minimal length
  2203. *
  2204. * Pads up a buffer to ensure the trailing bytes exist and are
  2205. * blanked. If the buffer already contains sufficient data it
  2206. * is untouched. Otherwise it is extended. Returns zero on
  2207. * success. The skb is freed on error.
  2208. */
  2209. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2210. {
  2211. unsigned int size = skb->len;
  2212. if (unlikely(size < len)) {
  2213. len -= size;
  2214. if (skb_pad(skb, len))
  2215. return -ENOMEM;
  2216. __skb_put(skb, len);
  2217. }
  2218. return 0;
  2219. }
  2220. static inline int skb_add_data(struct sk_buff *skb,
  2221. struct iov_iter *from, int copy)
  2222. {
  2223. const int off = skb->len;
  2224. if (skb->ip_summed == CHECKSUM_NONE) {
  2225. __wsum csum = 0;
  2226. if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
  2227. &csum, from) == copy) {
  2228. skb->csum = csum_block_add(skb->csum, csum, off);
  2229. return 0;
  2230. }
  2231. } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
  2232. return 0;
  2233. __skb_trim(skb, off);
  2234. return -EFAULT;
  2235. }
  2236. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2237. const struct page *page, int off)
  2238. {
  2239. if (i) {
  2240. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2241. return page == skb_frag_page(frag) &&
  2242. off == frag->page_offset + skb_frag_size(frag);
  2243. }
  2244. return false;
  2245. }
  2246. static inline int __skb_linearize(struct sk_buff *skb)
  2247. {
  2248. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2249. }
  2250. /**
  2251. * skb_linearize - convert paged skb to linear one
  2252. * @skb: buffer to linarize
  2253. *
  2254. * If there is no free memory -ENOMEM is returned, otherwise zero
  2255. * is returned and the old skb data released.
  2256. */
  2257. static inline int skb_linearize(struct sk_buff *skb)
  2258. {
  2259. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2260. }
  2261. /**
  2262. * skb_has_shared_frag - can any frag be overwritten
  2263. * @skb: buffer to test
  2264. *
  2265. * Return true if the skb has at least one frag that might be modified
  2266. * by an external entity (as in vmsplice()/sendfile())
  2267. */
  2268. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2269. {
  2270. return skb_is_nonlinear(skb) &&
  2271. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2272. }
  2273. /**
  2274. * skb_linearize_cow - make sure skb is linear and writable
  2275. * @skb: buffer to process
  2276. *
  2277. * If there is no free memory -ENOMEM is returned, otherwise zero
  2278. * is returned and the old skb data released.
  2279. */
  2280. static inline int skb_linearize_cow(struct sk_buff *skb)
  2281. {
  2282. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2283. __skb_linearize(skb) : 0;
  2284. }
  2285. /**
  2286. * skb_postpull_rcsum - update checksum for received skb after pull
  2287. * @skb: buffer to update
  2288. * @start: start of data before pull
  2289. * @len: length of data pulled
  2290. *
  2291. * After doing a pull on a received packet, you need to call this to
  2292. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2293. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2294. */
  2295. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2296. const void *start, unsigned int len)
  2297. {
  2298. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2299. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2300. }
  2301. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2302. /**
  2303. * pskb_trim_rcsum - trim received skb and update checksum
  2304. * @skb: buffer to trim
  2305. * @len: new length
  2306. *
  2307. * This is exactly the same as pskb_trim except that it ensures the
  2308. * checksum of received packets are still valid after the operation.
  2309. */
  2310. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2311. {
  2312. if (likely(len >= skb->len))
  2313. return 0;
  2314. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2315. skb->ip_summed = CHECKSUM_NONE;
  2316. return __pskb_trim(skb, len);
  2317. }
  2318. #define skb_queue_walk(queue, skb) \
  2319. for (skb = (queue)->next; \
  2320. skb != (struct sk_buff *)(queue); \
  2321. skb = skb->next)
  2322. #define skb_queue_walk_safe(queue, skb, tmp) \
  2323. for (skb = (queue)->next, tmp = skb->next; \
  2324. skb != (struct sk_buff *)(queue); \
  2325. skb = tmp, tmp = skb->next)
  2326. #define skb_queue_walk_from(queue, skb) \
  2327. for (; skb != (struct sk_buff *)(queue); \
  2328. skb = skb->next)
  2329. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2330. for (tmp = skb->next; \
  2331. skb != (struct sk_buff *)(queue); \
  2332. skb = tmp, tmp = skb->next)
  2333. #define skb_queue_reverse_walk(queue, skb) \
  2334. for (skb = (queue)->prev; \
  2335. skb != (struct sk_buff *)(queue); \
  2336. skb = skb->prev)
  2337. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2338. for (skb = (queue)->prev, tmp = skb->prev; \
  2339. skb != (struct sk_buff *)(queue); \
  2340. skb = tmp, tmp = skb->prev)
  2341. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2342. for (tmp = skb->prev; \
  2343. skb != (struct sk_buff *)(queue); \
  2344. skb = tmp, tmp = skb->prev)
  2345. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2346. {
  2347. return skb_shinfo(skb)->frag_list != NULL;
  2348. }
  2349. static inline void skb_frag_list_init(struct sk_buff *skb)
  2350. {
  2351. skb_shinfo(skb)->frag_list = NULL;
  2352. }
  2353. static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
  2354. {
  2355. frag->next = skb_shinfo(skb)->frag_list;
  2356. skb_shinfo(skb)->frag_list = frag;
  2357. }
  2358. #define skb_walk_frags(skb, iter) \
  2359. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2360. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2361. int *peeked, int *off, int *err);
  2362. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2363. int *err);
  2364. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2365. struct poll_table_struct *wait);
  2366. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2367. struct iov_iter *to, int size);
  2368. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2369. struct msghdr *msg, int size)
  2370. {
  2371. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2372. }
  2373. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2374. struct msghdr *msg);
  2375. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2376. struct iov_iter *from, int len);
  2377. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2378. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2379. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2380. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2381. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2382. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2383. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2384. int len, __wsum csum);
  2385. ssize_t skb_socket_splice(struct sock *sk,
  2386. struct pipe_inode_info *pipe,
  2387. struct splice_pipe_desc *spd);
  2388. int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
  2389. struct pipe_inode_info *pipe, unsigned int len,
  2390. unsigned int flags,
  2391. ssize_t (*splice_cb)(struct sock *,
  2392. struct pipe_inode_info *,
  2393. struct splice_pipe_desc *));
  2394. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2395. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2396. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2397. int len, int hlen);
  2398. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2399. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2400. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2401. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2402. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2403. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2404. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2405. int skb_vlan_pop(struct sk_buff *skb);
  2406. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2407. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2408. {
  2409. return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2410. }
  2411. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2412. {
  2413. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2414. }
  2415. struct skb_checksum_ops {
  2416. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2417. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2418. };
  2419. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2420. __wsum csum, const struct skb_checksum_ops *ops);
  2421. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2422. __wsum csum);
  2423. static inline void * __must_check
  2424. __skb_header_pointer(const struct sk_buff *skb, int offset,
  2425. int len, void *data, int hlen, void *buffer)
  2426. {
  2427. if (hlen - offset >= len)
  2428. return data + offset;
  2429. if (!skb ||
  2430. skb_copy_bits(skb, offset, buffer, len) < 0)
  2431. return NULL;
  2432. return buffer;
  2433. }
  2434. static inline void * __must_check
  2435. skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
  2436. {
  2437. return __skb_header_pointer(skb, offset, len, skb->data,
  2438. skb_headlen(skb), buffer);
  2439. }
  2440. /**
  2441. * skb_needs_linearize - check if we need to linearize a given skb
  2442. * depending on the given device features.
  2443. * @skb: socket buffer to check
  2444. * @features: net device features
  2445. *
  2446. * Returns true if either:
  2447. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2448. * 2. skb is fragmented and the device does not support SG.
  2449. */
  2450. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2451. netdev_features_t features)
  2452. {
  2453. return skb_is_nonlinear(skb) &&
  2454. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2455. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2456. }
  2457. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2458. void *to,
  2459. const unsigned int len)
  2460. {
  2461. memcpy(to, skb->data, len);
  2462. }
  2463. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2464. const int offset, void *to,
  2465. const unsigned int len)
  2466. {
  2467. memcpy(to, skb->data + offset, len);
  2468. }
  2469. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2470. const void *from,
  2471. const unsigned int len)
  2472. {
  2473. memcpy(skb->data, from, len);
  2474. }
  2475. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2476. const int offset,
  2477. const void *from,
  2478. const unsigned int len)
  2479. {
  2480. memcpy(skb->data + offset, from, len);
  2481. }
  2482. void skb_init(void);
  2483. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2484. {
  2485. return skb->tstamp;
  2486. }
  2487. /**
  2488. * skb_get_timestamp - get timestamp from a skb
  2489. * @skb: skb to get stamp from
  2490. * @stamp: pointer to struct timeval to store stamp in
  2491. *
  2492. * Timestamps are stored in the skb as offsets to a base timestamp.
  2493. * This function converts the offset back to a struct timeval and stores
  2494. * it in stamp.
  2495. */
  2496. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2497. struct timeval *stamp)
  2498. {
  2499. *stamp = ktime_to_timeval(skb->tstamp);
  2500. }
  2501. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2502. struct timespec *stamp)
  2503. {
  2504. *stamp = ktime_to_timespec(skb->tstamp);
  2505. }
  2506. static inline void __net_timestamp(struct sk_buff *skb)
  2507. {
  2508. skb->tstamp = ktime_get_real();
  2509. }
  2510. static inline ktime_t net_timedelta(ktime_t t)
  2511. {
  2512. return ktime_sub(ktime_get_real(), t);
  2513. }
  2514. static inline ktime_t net_invalid_timestamp(void)
  2515. {
  2516. return ktime_set(0, 0);
  2517. }
  2518. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2519. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2520. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2521. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2522. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2523. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2524. {
  2525. }
  2526. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2527. {
  2528. return false;
  2529. }
  2530. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2531. /**
  2532. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2533. *
  2534. * PHY drivers may accept clones of transmitted packets for
  2535. * timestamping via their phy_driver.txtstamp method. These drivers
  2536. * must call this function to return the skb back to the stack, with
  2537. * or without a timestamp.
  2538. *
  2539. * @skb: clone of the the original outgoing packet
  2540. * @hwtstamps: hardware time stamps, may be NULL if not available
  2541. *
  2542. */
  2543. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2544. struct skb_shared_hwtstamps *hwtstamps);
  2545. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2546. struct skb_shared_hwtstamps *hwtstamps,
  2547. struct sock *sk, int tstype);
  2548. /**
  2549. * skb_tstamp_tx - queue clone of skb with send time stamps
  2550. * @orig_skb: the original outgoing packet
  2551. * @hwtstamps: hardware time stamps, may be NULL if not available
  2552. *
  2553. * If the skb has a socket associated, then this function clones the
  2554. * skb (thus sharing the actual data and optional structures), stores
  2555. * the optional hardware time stamping information (if non NULL) or
  2556. * generates a software time stamp (otherwise), then queues the clone
  2557. * to the error queue of the socket. Errors are silently ignored.
  2558. */
  2559. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2560. struct skb_shared_hwtstamps *hwtstamps);
  2561. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2562. {
  2563. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2564. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2565. skb_tstamp_tx(skb, NULL);
  2566. }
  2567. /**
  2568. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2569. *
  2570. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2571. * function immediately before giving the sk_buff to the MAC hardware.
  2572. *
  2573. * Specifically, one should make absolutely sure that this function is
  2574. * called before TX completion of this packet can trigger. Otherwise
  2575. * the packet could potentially already be freed.
  2576. *
  2577. * @skb: A socket buffer.
  2578. */
  2579. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2580. {
  2581. skb_clone_tx_timestamp(skb);
  2582. sw_tx_timestamp(skb);
  2583. }
  2584. /**
  2585. * skb_complete_wifi_ack - deliver skb with wifi status
  2586. *
  2587. * @skb: the original outgoing packet
  2588. * @acked: ack status
  2589. *
  2590. */
  2591. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2592. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2593. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2594. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2595. {
  2596. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2597. skb->csum_valid ||
  2598. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2599. skb_checksum_start_offset(skb) >= 0));
  2600. }
  2601. /**
  2602. * skb_checksum_complete - Calculate checksum of an entire packet
  2603. * @skb: packet to process
  2604. *
  2605. * This function calculates the checksum over the entire packet plus
  2606. * the value of skb->csum. The latter can be used to supply the
  2607. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2608. * checksum.
  2609. *
  2610. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2611. * this function can be used to verify that checksum on received
  2612. * packets. In that case the function should return zero if the
  2613. * checksum is correct. In particular, this function will return zero
  2614. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2615. * hardware has already verified the correctness of the checksum.
  2616. */
  2617. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2618. {
  2619. return skb_csum_unnecessary(skb) ?
  2620. 0 : __skb_checksum_complete(skb);
  2621. }
  2622. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2623. {
  2624. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2625. if (skb->csum_level == 0)
  2626. skb->ip_summed = CHECKSUM_NONE;
  2627. else
  2628. skb->csum_level--;
  2629. }
  2630. }
  2631. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2632. {
  2633. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2634. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2635. skb->csum_level++;
  2636. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2637. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2638. skb->csum_level = 0;
  2639. }
  2640. }
  2641. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2642. {
  2643. /* Mark current checksum as bad (typically called from GRO
  2644. * path). In the case that ip_summed is CHECKSUM_NONE
  2645. * this must be the first checksum encountered in the packet.
  2646. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2647. * checksum after the last one validated. For UDP, a zero
  2648. * checksum can not be marked as bad.
  2649. */
  2650. if (skb->ip_summed == CHECKSUM_NONE ||
  2651. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2652. skb->csum_bad = 1;
  2653. }
  2654. /* Check if we need to perform checksum complete validation.
  2655. *
  2656. * Returns true if checksum complete is needed, false otherwise
  2657. * (either checksum is unnecessary or zero checksum is allowed).
  2658. */
  2659. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2660. bool zero_okay,
  2661. __sum16 check)
  2662. {
  2663. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2664. skb->csum_valid = 1;
  2665. __skb_decr_checksum_unnecessary(skb);
  2666. return false;
  2667. }
  2668. return true;
  2669. }
  2670. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2671. * in checksum_init.
  2672. */
  2673. #define CHECKSUM_BREAK 76
  2674. /* Unset checksum-complete
  2675. *
  2676. * Unset checksum complete can be done when packet is being modified
  2677. * (uncompressed for instance) and checksum-complete value is
  2678. * invalidated.
  2679. */
  2680. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2681. {
  2682. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2683. skb->ip_summed = CHECKSUM_NONE;
  2684. }
  2685. /* Validate (init) checksum based on checksum complete.
  2686. *
  2687. * Return values:
  2688. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2689. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2690. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2691. * non-zero: value of invalid checksum
  2692. *
  2693. */
  2694. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2695. bool complete,
  2696. __wsum psum)
  2697. {
  2698. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2699. if (!csum_fold(csum_add(psum, skb->csum))) {
  2700. skb->csum_valid = 1;
  2701. return 0;
  2702. }
  2703. } else if (skb->csum_bad) {
  2704. /* ip_summed == CHECKSUM_NONE in this case */
  2705. return (__force __sum16)1;
  2706. }
  2707. skb->csum = psum;
  2708. if (complete || skb->len <= CHECKSUM_BREAK) {
  2709. __sum16 csum;
  2710. csum = __skb_checksum_complete(skb);
  2711. skb->csum_valid = !csum;
  2712. return csum;
  2713. }
  2714. return 0;
  2715. }
  2716. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2717. {
  2718. return 0;
  2719. }
  2720. /* Perform checksum validate (init). Note that this is a macro since we only
  2721. * want to calculate the pseudo header which is an input function if necessary.
  2722. * First we try to validate without any computation (checksum unnecessary) and
  2723. * then calculate based on checksum complete calling the function to compute
  2724. * pseudo header.
  2725. *
  2726. * Return values:
  2727. * 0: checksum is validated or try to in skb_checksum_complete
  2728. * non-zero: value of invalid checksum
  2729. */
  2730. #define __skb_checksum_validate(skb, proto, complete, \
  2731. zero_okay, check, compute_pseudo) \
  2732. ({ \
  2733. __sum16 __ret = 0; \
  2734. skb->csum_valid = 0; \
  2735. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2736. __ret = __skb_checksum_validate_complete(skb, \
  2737. complete, compute_pseudo(skb, proto)); \
  2738. __ret; \
  2739. })
  2740. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2741. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2742. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2743. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2744. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2745. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2746. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2747. compute_pseudo) \
  2748. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  2749. #define skb_checksum_simple_validate(skb) \
  2750. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2751. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2752. {
  2753. return (skb->ip_summed == CHECKSUM_NONE &&
  2754. skb->csum_valid && !skb->csum_bad);
  2755. }
  2756. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2757. __sum16 check, __wsum pseudo)
  2758. {
  2759. skb->csum = ~pseudo;
  2760. skb->ip_summed = CHECKSUM_COMPLETE;
  2761. }
  2762. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  2763. do { \
  2764. if (__skb_checksum_convert_check(skb)) \
  2765. __skb_checksum_convert(skb, check, \
  2766. compute_pseudo(skb, proto)); \
  2767. } while (0)
  2768. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  2769. u16 start, u16 offset)
  2770. {
  2771. skb->ip_summed = CHECKSUM_PARTIAL;
  2772. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  2773. skb->csum_offset = offset - start;
  2774. }
  2775. /* Update skbuf and packet to reflect the remote checksum offload operation.
  2776. * When called, ptr indicates the starting point for skb->csum when
  2777. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  2778. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  2779. */
  2780. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  2781. int start, int offset, bool nopartial)
  2782. {
  2783. __wsum delta;
  2784. if (!nopartial) {
  2785. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  2786. return;
  2787. }
  2788. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  2789. __skb_checksum_complete(skb);
  2790. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  2791. }
  2792. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  2793. /* Adjust skb->csum since we changed the packet */
  2794. skb->csum = csum_add(skb->csum, delta);
  2795. }
  2796. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2797. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2798. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2799. {
  2800. if (nfct && atomic_dec_and_test(&nfct->use))
  2801. nf_conntrack_destroy(nfct);
  2802. }
  2803. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2804. {
  2805. if (nfct)
  2806. atomic_inc(&nfct->use);
  2807. }
  2808. #endif
  2809. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2810. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  2811. {
  2812. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  2813. kfree(nf_bridge);
  2814. }
  2815. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  2816. {
  2817. if (nf_bridge)
  2818. atomic_inc(&nf_bridge->use);
  2819. }
  2820. #endif /* CONFIG_BRIDGE_NETFILTER */
  2821. static inline void nf_reset(struct sk_buff *skb)
  2822. {
  2823. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2824. nf_conntrack_put(skb->nfct);
  2825. skb->nfct = NULL;
  2826. #endif
  2827. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2828. nf_bridge_put(skb->nf_bridge);
  2829. skb->nf_bridge = NULL;
  2830. #endif
  2831. }
  2832. static inline void nf_reset_trace(struct sk_buff *skb)
  2833. {
  2834. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2835. skb->nf_trace = 0;
  2836. #endif
  2837. }
  2838. /* Note: This doesn't put any conntrack and bridge info in dst. */
  2839. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  2840. bool copy)
  2841. {
  2842. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2843. dst->nfct = src->nfct;
  2844. nf_conntrack_get(src->nfct);
  2845. if (copy)
  2846. dst->nfctinfo = src->nfctinfo;
  2847. #endif
  2848. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2849. dst->nf_bridge = src->nf_bridge;
  2850. nf_bridge_get(src->nf_bridge);
  2851. #endif
  2852. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2853. if (copy)
  2854. dst->nf_trace = src->nf_trace;
  2855. #endif
  2856. }
  2857. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2858. {
  2859. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2860. nf_conntrack_put(dst->nfct);
  2861. #endif
  2862. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2863. nf_bridge_put(dst->nf_bridge);
  2864. #endif
  2865. __nf_copy(dst, src, true);
  2866. }
  2867. #ifdef CONFIG_NETWORK_SECMARK
  2868. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2869. {
  2870. to->secmark = from->secmark;
  2871. }
  2872. static inline void skb_init_secmark(struct sk_buff *skb)
  2873. {
  2874. skb->secmark = 0;
  2875. }
  2876. #else
  2877. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2878. { }
  2879. static inline void skb_init_secmark(struct sk_buff *skb)
  2880. { }
  2881. #endif
  2882. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  2883. {
  2884. return !skb->destructor &&
  2885. #if IS_ENABLED(CONFIG_XFRM)
  2886. !skb->sp &&
  2887. #endif
  2888. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  2889. !skb->nfct &&
  2890. #endif
  2891. !skb->_skb_refdst &&
  2892. !skb_has_frag_list(skb);
  2893. }
  2894. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  2895. {
  2896. skb->queue_mapping = queue_mapping;
  2897. }
  2898. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  2899. {
  2900. return skb->queue_mapping;
  2901. }
  2902. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  2903. {
  2904. to->queue_mapping = from->queue_mapping;
  2905. }
  2906. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  2907. {
  2908. skb->queue_mapping = rx_queue + 1;
  2909. }
  2910. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  2911. {
  2912. return skb->queue_mapping - 1;
  2913. }
  2914. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  2915. {
  2916. return skb->queue_mapping != 0;
  2917. }
  2918. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  2919. {
  2920. #ifdef CONFIG_XFRM
  2921. return skb->sp;
  2922. #else
  2923. return NULL;
  2924. #endif
  2925. }
  2926. /* Keeps track of mac header offset relative to skb->head.
  2927. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  2928. * For non-tunnel skb it points to skb_mac_header() and for
  2929. * tunnel skb it points to outer mac header.
  2930. * Keeps track of level of encapsulation of network headers.
  2931. */
  2932. struct skb_gso_cb {
  2933. int mac_offset;
  2934. int encap_level;
  2935. __u16 csum_start;
  2936. };
  2937. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
  2938. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  2939. {
  2940. return (skb_mac_header(inner_skb) - inner_skb->head) -
  2941. SKB_GSO_CB(inner_skb)->mac_offset;
  2942. }
  2943. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  2944. {
  2945. int new_headroom, headroom;
  2946. int ret;
  2947. headroom = skb_headroom(skb);
  2948. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  2949. if (ret)
  2950. return ret;
  2951. new_headroom = skb_headroom(skb);
  2952. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  2953. return 0;
  2954. }
  2955. /* Compute the checksum for a gso segment. First compute the checksum value
  2956. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  2957. * then add in skb->csum (checksum from csum_start to end of packet).
  2958. * skb->csum and csum_start are then updated to reflect the checksum of the
  2959. * resultant packet starting from the transport header-- the resultant checksum
  2960. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  2961. * header.
  2962. */
  2963. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  2964. {
  2965. int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
  2966. skb_transport_offset(skb);
  2967. __wsum partial;
  2968. partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
  2969. skb->csum = res;
  2970. SKB_GSO_CB(skb)->csum_start -= plen;
  2971. return csum_fold(partial);
  2972. }
  2973. static inline bool skb_is_gso(const struct sk_buff *skb)
  2974. {
  2975. return skb_shinfo(skb)->gso_size;
  2976. }
  2977. /* Note: Should be called only if skb_is_gso(skb) is true */
  2978. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  2979. {
  2980. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  2981. }
  2982. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  2983. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  2984. {
  2985. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  2986. * wanted then gso_type will be set. */
  2987. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  2988. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  2989. unlikely(shinfo->gso_type == 0)) {
  2990. __skb_warn_lro_forwarding(skb);
  2991. return true;
  2992. }
  2993. return false;
  2994. }
  2995. static inline void skb_forward_csum(struct sk_buff *skb)
  2996. {
  2997. /* Unfortunately we don't support this one. Any brave souls? */
  2998. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2999. skb->ip_summed = CHECKSUM_NONE;
  3000. }
  3001. /**
  3002. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  3003. * @skb: skb to check
  3004. *
  3005. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  3006. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  3007. * use this helper, to document places where we make this assertion.
  3008. */
  3009. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  3010. {
  3011. #ifdef DEBUG
  3012. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  3013. #endif
  3014. }
  3015. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3016. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3017. struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
  3018. unsigned int transport_len,
  3019. __sum16(*skb_chkf)(struct sk_buff *skb));
  3020. /**
  3021. * skb_head_is_locked - Determine if the skb->head is locked down
  3022. * @skb: skb to check
  3023. *
  3024. * The head on skbs build around a head frag can be removed if they are
  3025. * not cloned. This function returns true if the skb head is locked down
  3026. * due to either being allocated via kmalloc, or by being a clone with
  3027. * multiple references to the head.
  3028. */
  3029. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3030. {
  3031. return !skb->head_frag || skb_cloned(skb);
  3032. }
  3033. /**
  3034. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3035. *
  3036. * @skb: GSO skb
  3037. *
  3038. * skb_gso_network_seglen is used to determine the real size of the
  3039. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3040. *
  3041. * The MAC/L2 header is not accounted for.
  3042. */
  3043. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3044. {
  3045. unsigned int hdr_len = skb_transport_header(skb) -
  3046. skb_network_header(skb);
  3047. return hdr_len + skb_gso_transport_seglen(skb);
  3048. }
  3049. #endif /* __KERNEL__ */
  3050. #endif /* _LINUX_SKBUFF_H */