skbuff.h 96 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460246124622463246424652466246724682469247024712472247324742475247624772478247924802481248224832484248524862487248824892490249124922493249424952496249724982499250025012502250325042505250625072508250925102511251225132514251525162517251825192520252125222523252425252526252725282529253025312532253325342535253625372538253925402541254225432544254525462547254825492550255125522553255425552556255725582559256025612562256325642565256625672568256925702571257225732574257525762577257825792580258125822583258425852586258725882589259025912592259325942595259625972598259926002601260226032604260526062607260826092610261126122613261426152616261726182619262026212622262326242625262626272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697269826992700270127022703270427052706270727082709271027112712271327142715271627172718271927202721272227232724272527262727272827292730273127322733273427352736273727382739274027412742274327442745274627472748274927502751275227532754275527562757275827592760276127622763276427652766276727682769277027712772277327742775277627772778277927802781278227832784278527862787278827892790279127922793279427952796279727982799280028012802280328042805280628072808280928102811281228132814281528162817281828192820282128222823282428252826282728282829283028312832283328342835283628372838283928402841284228432844284528462847284828492850285128522853285428552856285728582859286028612862286328642865286628672868286928702871287228732874287528762877287828792880288128822883288428852886288728882889289028912892289328942895289628972898289929002901290229032904290529062907290829092910291129122913291429152916291729182919292029212922292329242925292629272928292929302931293229332934293529362937293829392940294129422943294429452946294729482949295029512952295329542955295629572958295929602961296229632964296529662967296829692970297129722973297429752976297729782979298029812982298329842985298629872988298929902991299229932994299529962997299829993000300130023003300430053006300730083009301030113012301330143015301630173018301930203021302230233024302530263027302830293030303130323033303430353036303730383039304030413042304330443045304630473048304930503051305230533054305530563057305830593060306130623063306430653066306730683069307030713072307330743075307630773078307930803081308230833084308530863087308830893090309130923093309430953096309730983099310031013102310331043105310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135313631373138313931403141314231433144314531463147314831493150315131523153315431553156315731583159316031613162316331643165316631673168316931703171317231733174317531763177317831793180318131823183318431853186318731883189319031913192319331943195319631973198319932003201320232033204320532063207320832093210321132123213321432153216321732183219322032213222322332243225322632273228322932303231323232333234323532363237323832393240324132423243324432453246324732483249325032513252325332543255325632573258325932603261326232633264326532663267326832693270327132723273327432753276327732783279328032813282328332843285328632873288328932903291329232933294329532963297329832993300330133023303330433053306330733083309331033113312331333143315331633173318331933203321332233233324332533263327332833293330333133323333333433353336333733383339334033413342334333443345334633473348334933503351335233533354335533563357335833593360336133623363336433653366336733683369337033713372337333743375337633773378337933803381338233833384338533863387338833893390339133923393339433953396339733983399340034013402340334043405340634073408340934103411341234133414341534163417341834193420342134223423342434253426342734283429343034313432
  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_keys.h>
  35. /* A. Checksumming of received packets by device.
  36. *
  37. * CHECKSUM_NONE:
  38. *
  39. * Device failed to checksum this packet e.g. due to lack of capabilities.
  40. * The packet contains full (though not verified) checksum in packet but
  41. * not in skb->csum. Thus, skb->csum is undefined in this case.
  42. *
  43. * CHECKSUM_UNNECESSARY:
  44. *
  45. * The hardware you're dealing with doesn't calculate the full checksum
  46. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  47. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  48. * if their checksums are okay. skb->csum is still undefined in this case
  49. * though. It is a bad option, but, unfortunately, nowadays most vendors do
  50. * this. Apparently with the secret goal to sell you new devices, when you
  51. * will add new protocol to your host, f.e. IPv6 8)
  52. *
  53. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  54. * TCP: IPv6 and IPv4.
  55. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  56. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  57. * may perform further validation in this case.
  58. * GRE: only if the checksum is present in the header.
  59. * SCTP: indicates the CRC in SCTP header has been validated.
  60. *
  61. * skb->csum_level indicates the number of consecutive checksums found in
  62. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  63. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  64. * and a device is able to verify the checksums for UDP (possibly zero),
  65. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  66. * two. If the device were only able to verify the UDP checksum and not
  67. * GRE, either because it doesn't support GRE checksum of because GRE
  68. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  69. * not considered in this case).
  70. *
  71. * CHECKSUM_COMPLETE:
  72. *
  73. * This is the most generic way. The device supplied checksum of the _whole_
  74. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  75. * hardware doesn't need to parse L3/L4 headers to implement this.
  76. *
  77. * Note: Even if device supports only some protocols, but is able to produce
  78. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  79. *
  80. * CHECKSUM_PARTIAL:
  81. *
  82. * A checksum is set up to be offloaded to a device as described in the
  83. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  84. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  85. * on the same host, or it may be set in the input path in GRO or remote
  86. * checksum offload. For the purposes of checksum verification, the checksum
  87. * referred to by skb->csum_start + skb->csum_offset and any preceding
  88. * checksums in the packet are considered verified. Any checksums in the
  89. * packet that are after the checksum being offloaded are not considered to
  90. * be verified.
  91. *
  92. * B. Checksumming on output.
  93. *
  94. * CHECKSUM_NONE:
  95. *
  96. * The skb was already checksummed by the protocol, or a checksum is not
  97. * required.
  98. *
  99. * CHECKSUM_PARTIAL:
  100. *
  101. * The device is required to checksum the packet as seen by hard_start_xmit()
  102. * from skb->csum_start up to the end, and to record/write the checksum at
  103. * offset skb->csum_start + skb->csum_offset.
  104. *
  105. * The device must show its capabilities in dev->features, set up at device
  106. * setup time, e.g. netdev_features.h:
  107. *
  108. * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
  109. * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
  110. * IPv4. Sigh. Vendors like this way for an unknown reason.
  111. * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
  112. * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
  113. * NETIF_F_... - Well, you get the picture.
  114. *
  115. * CHECKSUM_UNNECESSARY:
  116. *
  117. * Normally, the device will do per protocol specific checksumming. Protocol
  118. * implementations that do not want the NIC to perform the checksum
  119. * calculation should use this flag in their outgoing skbs.
  120. *
  121. * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
  122. * offload. Correspondingly, the FCoE protocol driver
  123. * stack should use CHECKSUM_UNNECESSARY.
  124. *
  125. * Any questions? No questions, good. --ANK
  126. */
  127. /* Don't change this without changing skb_csum_unnecessary! */
  128. #define CHECKSUM_NONE 0
  129. #define CHECKSUM_UNNECESSARY 1
  130. #define CHECKSUM_COMPLETE 2
  131. #define CHECKSUM_PARTIAL 3
  132. /* Maximum value in skb->csum_level */
  133. #define SKB_MAX_CSUM_LEVEL 3
  134. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  135. #define SKB_WITH_OVERHEAD(X) \
  136. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  137. #define SKB_MAX_ORDER(X, ORDER) \
  138. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  139. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  140. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  141. /* return minimum truesize of one skb containing X bytes of data */
  142. #define SKB_TRUESIZE(X) ((X) + \
  143. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  144. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  145. struct net_device;
  146. struct scatterlist;
  147. struct pipe_inode_info;
  148. struct iov_iter;
  149. struct napi_struct;
  150. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  151. struct nf_conntrack {
  152. atomic_t use;
  153. };
  154. #endif
  155. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  156. struct nf_bridge_info {
  157. atomic_t use;
  158. unsigned int mask;
  159. struct net_device *physindev;
  160. struct net_device *physoutdev;
  161. unsigned long data[32 / sizeof(unsigned long)];
  162. };
  163. #endif
  164. struct sk_buff_head {
  165. /* These two members must be first. */
  166. struct sk_buff *next;
  167. struct sk_buff *prev;
  168. __u32 qlen;
  169. spinlock_t lock;
  170. };
  171. struct sk_buff;
  172. /* To allow 64K frame to be packed as single skb without frag_list we
  173. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  174. * buffers which do not start on a page boundary.
  175. *
  176. * Since GRO uses frags we allocate at least 16 regardless of page
  177. * size.
  178. */
  179. #if (65536/PAGE_SIZE + 1) < 16
  180. #define MAX_SKB_FRAGS 16UL
  181. #else
  182. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  183. #endif
  184. typedef struct skb_frag_struct skb_frag_t;
  185. struct skb_frag_struct {
  186. struct {
  187. struct page *p;
  188. } page;
  189. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  190. __u32 page_offset;
  191. __u32 size;
  192. #else
  193. __u16 page_offset;
  194. __u16 size;
  195. #endif
  196. };
  197. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  198. {
  199. return frag->size;
  200. }
  201. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  202. {
  203. frag->size = size;
  204. }
  205. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  206. {
  207. frag->size += delta;
  208. }
  209. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  210. {
  211. frag->size -= delta;
  212. }
  213. #define HAVE_HW_TIME_STAMP
  214. /**
  215. * struct skb_shared_hwtstamps - hardware time stamps
  216. * @hwtstamp: hardware time stamp transformed into duration
  217. * since arbitrary point in time
  218. *
  219. * Software time stamps generated by ktime_get_real() are stored in
  220. * skb->tstamp.
  221. *
  222. * hwtstamps can only be compared against other hwtstamps from
  223. * the same device.
  224. *
  225. * This structure is attached to packets as part of the
  226. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  227. */
  228. struct skb_shared_hwtstamps {
  229. ktime_t hwtstamp;
  230. };
  231. /* Definitions for tx_flags in struct skb_shared_info */
  232. enum {
  233. /* generate hardware time stamp */
  234. SKBTX_HW_TSTAMP = 1 << 0,
  235. /* generate software time stamp when queueing packet to NIC */
  236. SKBTX_SW_TSTAMP = 1 << 1,
  237. /* device driver is going to provide hardware time stamp */
  238. SKBTX_IN_PROGRESS = 1 << 2,
  239. /* device driver supports TX zero-copy buffers */
  240. SKBTX_DEV_ZEROCOPY = 1 << 3,
  241. /* generate wifi status information (where possible) */
  242. SKBTX_WIFI_STATUS = 1 << 4,
  243. /* This indicates at least one fragment might be overwritten
  244. * (as in vmsplice(), sendfile() ...)
  245. * If we need to compute a TX checksum, we'll need to copy
  246. * all frags to avoid possible bad checksum
  247. */
  248. SKBTX_SHARED_FRAG = 1 << 5,
  249. /* generate software time stamp when entering packet scheduling */
  250. SKBTX_SCHED_TSTAMP = 1 << 6,
  251. /* generate software timestamp on peer data acknowledgment */
  252. SKBTX_ACK_TSTAMP = 1 << 7,
  253. };
  254. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  255. SKBTX_SCHED_TSTAMP | \
  256. SKBTX_ACK_TSTAMP)
  257. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  258. /*
  259. * The callback notifies userspace to release buffers when skb DMA is done in
  260. * lower device, the skb last reference should be 0 when calling this.
  261. * The zerocopy_success argument is true if zero copy transmit occurred,
  262. * false on data copy or out of memory error caused by data copy attempt.
  263. * The ctx field is used to track device context.
  264. * The desc field is used to track userspace buffer index.
  265. */
  266. struct ubuf_info {
  267. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  268. void *ctx;
  269. unsigned long desc;
  270. };
  271. /* This data is invariant across clones and lives at
  272. * the end of the header data, ie. at skb->end.
  273. */
  274. struct skb_shared_info {
  275. unsigned char nr_frags;
  276. __u8 tx_flags;
  277. unsigned short gso_size;
  278. /* Warning: this field is not always filled in (UFO)! */
  279. unsigned short gso_segs;
  280. unsigned short gso_type;
  281. struct sk_buff *frag_list;
  282. struct skb_shared_hwtstamps hwtstamps;
  283. u32 tskey;
  284. __be32 ip6_frag_id;
  285. /*
  286. * Warning : all fields before dataref are cleared in __alloc_skb()
  287. */
  288. atomic_t dataref;
  289. /* Intermediate layers must ensure that destructor_arg
  290. * remains valid until skb destructor */
  291. void * destructor_arg;
  292. /* must be last field, see pskb_expand_head() */
  293. skb_frag_t frags[MAX_SKB_FRAGS];
  294. };
  295. /* We divide dataref into two halves. The higher 16 bits hold references
  296. * to the payload part of skb->data. The lower 16 bits hold references to
  297. * the entire skb->data. A clone of a headerless skb holds the length of
  298. * the header in skb->hdr_len.
  299. *
  300. * All users must obey the rule that the skb->data reference count must be
  301. * greater than or equal to the payload reference count.
  302. *
  303. * Holding a reference to the payload part means that the user does not
  304. * care about modifications to the header part of skb->data.
  305. */
  306. #define SKB_DATAREF_SHIFT 16
  307. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  308. enum {
  309. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  310. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  311. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  312. };
  313. enum {
  314. SKB_GSO_TCPV4 = 1 << 0,
  315. SKB_GSO_UDP = 1 << 1,
  316. /* This indicates the skb is from an untrusted source. */
  317. SKB_GSO_DODGY = 1 << 2,
  318. /* This indicates the tcp segment has CWR set. */
  319. SKB_GSO_TCP_ECN = 1 << 3,
  320. SKB_GSO_TCPV6 = 1 << 4,
  321. SKB_GSO_FCOE = 1 << 5,
  322. SKB_GSO_GRE = 1 << 6,
  323. SKB_GSO_GRE_CSUM = 1 << 7,
  324. SKB_GSO_IPIP = 1 << 8,
  325. SKB_GSO_SIT = 1 << 9,
  326. SKB_GSO_UDP_TUNNEL = 1 << 10,
  327. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  328. SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
  329. };
  330. #if BITS_PER_LONG > 32
  331. #define NET_SKBUFF_DATA_USES_OFFSET 1
  332. #endif
  333. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  334. typedef unsigned int sk_buff_data_t;
  335. #else
  336. typedef unsigned char *sk_buff_data_t;
  337. #endif
  338. /**
  339. * struct skb_mstamp - multi resolution time stamps
  340. * @stamp_us: timestamp in us resolution
  341. * @stamp_jiffies: timestamp in jiffies
  342. */
  343. struct skb_mstamp {
  344. union {
  345. u64 v64;
  346. struct {
  347. u32 stamp_us;
  348. u32 stamp_jiffies;
  349. };
  350. };
  351. };
  352. /**
  353. * skb_mstamp_get - get current timestamp
  354. * @cl: place to store timestamps
  355. */
  356. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  357. {
  358. u64 val = local_clock();
  359. do_div(val, NSEC_PER_USEC);
  360. cl->stamp_us = (u32)val;
  361. cl->stamp_jiffies = (u32)jiffies;
  362. }
  363. /**
  364. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  365. * @t1: pointer to newest sample
  366. * @t0: pointer to oldest sample
  367. */
  368. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  369. const struct skb_mstamp *t0)
  370. {
  371. s32 delta_us = t1->stamp_us - t0->stamp_us;
  372. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  373. /* If delta_us is negative, this might be because interval is too big,
  374. * or local_clock() drift is too big : fallback using jiffies.
  375. */
  376. if (delta_us <= 0 ||
  377. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  378. delta_us = jiffies_to_usecs(delta_jiffies);
  379. return delta_us;
  380. }
  381. /**
  382. * struct sk_buff - socket buffer
  383. * @next: Next buffer in list
  384. * @prev: Previous buffer in list
  385. * @tstamp: Time we arrived/left
  386. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  387. * @sk: Socket we are owned by
  388. * @dev: Device we arrived on/are leaving by
  389. * @cb: Control buffer. Free for use by every layer. Put private vars here
  390. * @_skb_refdst: destination entry (with norefcount bit)
  391. * @sp: the security path, used for xfrm
  392. * @len: Length of actual data
  393. * @data_len: Data length
  394. * @mac_len: Length of link layer header
  395. * @hdr_len: writable header length of cloned skb
  396. * @csum: Checksum (must include start/offset pair)
  397. * @csum_start: Offset from skb->head where checksumming should start
  398. * @csum_offset: Offset from csum_start where checksum should be stored
  399. * @priority: Packet queueing priority
  400. * @ignore_df: allow local fragmentation
  401. * @cloned: Head may be cloned (check refcnt to be sure)
  402. * @ip_summed: Driver fed us an IP checksum
  403. * @nohdr: Payload reference only, must not modify header
  404. * @nfctinfo: Relationship of this skb to the connection
  405. * @pkt_type: Packet class
  406. * @fclone: skbuff clone status
  407. * @ipvs_property: skbuff is owned by ipvs
  408. * @peeked: this packet has been seen already, so stats have been
  409. * done for it, don't do them again
  410. * @nf_trace: netfilter packet trace flag
  411. * @protocol: Packet protocol from driver
  412. * @destructor: Destruct function
  413. * @nfct: Associated connection, if any
  414. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  415. * @skb_iif: ifindex of device we arrived on
  416. * @tc_index: Traffic control index
  417. * @tc_verd: traffic control verdict
  418. * @hash: the packet hash
  419. * @queue_mapping: Queue mapping for multiqueue devices
  420. * @xmit_more: More SKBs are pending for this queue
  421. * @ndisc_nodetype: router type (from link layer)
  422. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  423. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  424. * ports.
  425. * @sw_hash: indicates hash was computed in software stack
  426. * @wifi_acked_valid: wifi_acked was set
  427. * @wifi_acked: whether frame was acked on wifi or not
  428. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  429. * @napi_id: id of the NAPI struct this skb came from
  430. * @secmark: security marking
  431. * @mark: Generic packet mark
  432. * @dropcount: total number of sk_receive_queue overflows
  433. * @vlan_proto: vlan encapsulation protocol
  434. * @vlan_tci: vlan tag control information
  435. * @inner_protocol: Protocol (encapsulation)
  436. * @inner_transport_header: Inner transport layer header (encapsulation)
  437. * @inner_network_header: Network layer header (encapsulation)
  438. * @inner_mac_header: Link layer header (encapsulation)
  439. * @transport_header: Transport layer header
  440. * @network_header: Network layer header
  441. * @mac_header: Link layer header
  442. * @tail: Tail pointer
  443. * @end: End pointer
  444. * @head: Head of buffer
  445. * @data: Data head pointer
  446. * @truesize: Buffer size
  447. * @users: User count - see {datagram,tcp}.c
  448. */
  449. struct sk_buff {
  450. union {
  451. struct {
  452. /* These two members must be first. */
  453. struct sk_buff *next;
  454. struct sk_buff *prev;
  455. union {
  456. ktime_t tstamp;
  457. struct skb_mstamp skb_mstamp;
  458. };
  459. };
  460. struct rb_node rbnode; /* used in netem & tcp stack */
  461. };
  462. struct sock *sk;
  463. struct net_device *dev;
  464. /*
  465. * This is the control buffer. It is free to use for every
  466. * layer. Please put your private variables there. If you
  467. * want to keep them across layers you have to do a skb_clone()
  468. * first. This is owned by whoever has the skb queued ATM.
  469. */
  470. char cb[48] __aligned(8);
  471. unsigned long _skb_refdst;
  472. void (*destructor)(struct sk_buff *skb);
  473. #ifdef CONFIG_XFRM
  474. struct sec_path *sp;
  475. #endif
  476. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  477. struct nf_conntrack *nfct;
  478. #endif
  479. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  480. struct nf_bridge_info *nf_bridge;
  481. #endif
  482. unsigned int len,
  483. data_len;
  484. __u16 mac_len,
  485. hdr_len;
  486. /* Following fields are _not_ copied in __copy_skb_header()
  487. * Note that queue_mapping is here mostly to fill a hole.
  488. */
  489. kmemcheck_bitfield_begin(flags1);
  490. __u16 queue_mapping;
  491. __u8 cloned:1,
  492. nohdr:1,
  493. fclone:2,
  494. peeked:1,
  495. head_frag:1,
  496. xmit_more:1;
  497. /* one bit hole */
  498. kmemcheck_bitfield_end(flags1);
  499. /* fields enclosed in headers_start/headers_end are copied
  500. * using a single memcpy() in __copy_skb_header()
  501. */
  502. /* private: */
  503. __u32 headers_start[0];
  504. /* public: */
  505. /* if you move pkt_type around you also must adapt those constants */
  506. #ifdef __BIG_ENDIAN_BITFIELD
  507. #define PKT_TYPE_MAX (7 << 5)
  508. #else
  509. #define PKT_TYPE_MAX 7
  510. #endif
  511. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  512. __u8 __pkt_type_offset[0];
  513. __u8 pkt_type:3;
  514. __u8 pfmemalloc:1;
  515. __u8 ignore_df:1;
  516. __u8 nfctinfo:3;
  517. __u8 nf_trace:1;
  518. __u8 ip_summed:2;
  519. __u8 ooo_okay:1;
  520. __u8 l4_hash:1;
  521. __u8 sw_hash:1;
  522. __u8 wifi_acked_valid:1;
  523. __u8 wifi_acked:1;
  524. __u8 no_fcs:1;
  525. /* Indicates the inner headers are valid in the skbuff. */
  526. __u8 encapsulation:1;
  527. __u8 encap_hdr_csum:1;
  528. __u8 csum_valid:1;
  529. __u8 csum_complete_sw:1;
  530. __u8 csum_level:2;
  531. __u8 csum_bad:1;
  532. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  533. __u8 ndisc_nodetype:2;
  534. #endif
  535. __u8 ipvs_property:1;
  536. __u8 inner_protocol_type:1;
  537. __u8 remcsum_offload:1;
  538. /* 3 or 5 bit hole */
  539. #ifdef CONFIG_NET_SCHED
  540. __u16 tc_index; /* traffic control index */
  541. #ifdef CONFIG_NET_CLS_ACT
  542. __u16 tc_verd; /* traffic control verdict */
  543. #endif
  544. #endif
  545. union {
  546. __wsum csum;
  547. struct {
  548. __u16 csum_start;
  549. __u16 csum_offset;
  550. };
  551. };
  552. __u32 priority;
  553. int skb_iif;
  554. __u32 hash;
  555. __be16 vlan_proto;
  556. __u16 vlan_tci;
  557. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  558. union {
  559. unsigned int napi_id;
  560. unsigned int sender_cpu;
  561. };
  562. #endif
  563. #ifdef CONFIG_NETWORK_SECMARK
  564. __u32 secmark;
  565. #endif
  566. union {
  567. __u32 mark;
  568. __u32 dropcount;
  569. __u32 reserved_tailroom;
  570. };
  571. union {
  572. __be16 inner_protocol;
  573. __u8 inner_ipproto;
  574. };
  575. __u16 inner_transport_header;
  576. __u16 inner_network_header;
  577. __u16 inner_mac_header;
  578. __be16 protocol;
  579. __u16 transport_header;
  580. __u16 network_header;
  581. __u16 mac_header;
  582. /* private: */
  583. __u32 headers_end[0];
  584. /* public: */
  585. /* These elements must be at the end, see alloc_skb() for details. */
  586. sk_buff_data_t tail;
  587. sk_buff_data_t end;
  588. unsigned char *head,
  589. *data;
  590. unsigned int truesize;
  591. atomic_t users;
  592. };
  593. #ifdef __KERNEL__
  594. /*
  595. * Handling routines are only of interest to the kernel
  596. */
  597. #include <linux/slab.h>
  598. #define SKB_ALLOC_FCLONE 0x01
  599. #define SKB_ALLOC_RX 0x02
  600. #define SKB_ALLOC_NAPI 0x04
  601. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  602. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  603. {
  604. return unlikely(skb->pfmemalloc);
  605. }
  606. /*
  607. * skb might have a dst pointer attached, refcounted or not.
  608. * _skb_refdst low order bit is set if refcount was _not_ taken
  609. */
  610. #define SKB_DST_NOREF 1UL
  611. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  612. /**
  613. * skb_dst - returns skb dst_entry
  614. * @skb: buffer
  615. *
  616. * Returns skb dst_entry, regardless of reference taken or not.
  617. */
  618. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  619. {
  620. /* If refdst was not refcounted, check we still are in a
  621. * rcu_read_lock section
  622. */
  623. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  624. !rcu_read_lock_held() &&
  625. !rcu_read_lock_bh_held());
  626. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  627. }
  628. /**
  629. * skb_dst_set - sets skb dst
  630. * @skb: buffer
  631. * @dst: dst entry
  632. *
  633. * Sets skb dst, assuming a reference was taken on dst and should
  634. * be released by skb_dst_drop()
  635. */
  636. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  637. {
  638. skb->_skb_refdst = (unsigned long)dst;
  639. }
  640. /**
  641. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  642. * @skb: buffer
  643. * @dst: dst entry
  644. *
  645. * Sets skb dst, assuming a reference was not taken on dst.
  646. * If dst entry is cached, we do not take reference and dst_release
  647. * will be avoided by refdst_drop. If dst entry is not cached, we take
  648. * reference, so that last dst_release can destroy the dst immediately.
  649. */
  650. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  651. {
  652. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  653. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  654. }
  655. /**
  656. * skb_dst_is_noref - Test if skb dst isn't refcounted
  657. * @skb: buffer
  658. */
  659. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  660. {
  661. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  662. }
  663. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  664. {
  665. return (struct rtable *)skb_dst(skb);
  666. }
  667. void kfree_skb(struct sk_buff *skb);
  668. void kfree_skb_list(struct sk_buff *segs);
  669. void skb_tx_error(struct sk_buff *skb);
  670. void consume_skb(struct sk_buff *skb);
  671. void __kfree_skb(struct sk_buff *skb);
  672. extern struct kmem_cache *skbuff_head_cache;
  673. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  674. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  675. bool *fragstolen, int *delta_truesize);
  676. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  677. int node);
  678. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  679. static inline struct sk_buff *alloc_skb(unsigned int size,
  680. gfp_t priority)
  681. {
  682. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  683. }
  684. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  685. unsigned long data_len,
  686. int max_page_order,
  687. int *errcode,
  688. gfp_t gfp_mask);
  689. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  690. struct sk_buff_fclones {
  691. struct sk_buff skb1;
  692. struct sk_buff skb2;
  693. atomic_t fclone_ref;
  694. };
  695. /**
  696. * skb_fclone_busy - check if fclone is busy
  697. * @skb: buffer
  698. *
  699. * Returns true is skb is a fast clone, and its clone is not freed.
  700. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  701. * so we also check that this didnt happen.
  702. */
  703. static inline bool skb_fclone_busy(const struct sock *sk,
  704. const struct sk_buff *skb)
  705. {
  706. const struct sk_buff_fclones *fclones;
  707. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  708. return skb->fclone == SKB_FCLONE_ORIG &&
  709. atomic_read(&fclones->fclone_ref) > 1 &&
  710. fclones->skb2.sk == sk;
  711. }
  712. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  713. gfp_t priority)
  714. {
  715. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  716. }
  717. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  718. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  719. {
  720. return __alloc_skb_head(priority, -1);
  721. }
  722. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  723. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  724. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  725. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  726. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  727. gfp_t gfp_mask, bool fclone);
  728. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  729. gfp_t gfp_mask)
  730. {
  731. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  732. }
  733. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  734. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  735. unsigned int headroom);
  736. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  737. int newtailroom, gfp_t priority);
  738. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  739. int offset, int len);
  740. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  741. int len);
  742. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  743. int skb_pad(struct sk_buff *skb, int pad);
  744. #define dev_kfree_skb(a) consume_skb(a)
  745. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  746. int getfrag(void *from, char *to, int offset,
  747. int len, int odd, struct sk_buff *skb),
  748. void *from, int length);
  749. struct skb_seq_state {
  750. __u32 lower_offset;
  751. __u32 upper_offset;
  752. __u32 frag_idx;
  753. __u32 stepped_offset;
  754. struct sk_buff *root_skb;
  755. struct sk_buff *cur_skb;
  756. __u8 *frag_data;
  757. };
  758. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  759. unsigned int to, struct skb_seq_state *st);
  760. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  761. struct skb_seq_state *st);
  762. void skb_abort_seq_read(struct skb_seq_state *st);
  763. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  764. unsigned int to, struct ts_config *config);
  765. /*
  766. * Packet hash types specify the type of hash in skb_set_hash.
  767. *
  768. * Hash types refer to the protocol layer addresses which are used to
  769. * construct a packet's hash. The hashes are used to differentiate or identify
  770. * flows of the protocol layer for the hash type. Hash types are either
  771. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  772. *
  773. * Properties of hashes:
  774. *
  775. * 1) Two packets in different flows have different hash values
  776. * 2) Two packets in the same flow should have the same hash value
  777. *
  778. * A hash at a higher layer is considered to be more specific. A driver should
  779. * set the most specific hash possible.
  780. *
  781. * A driver cannot indicate a more specific hash than the layer at which a hash
  782. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  783. *
  784. * A driver may indicate a hash level which is less specific than the
  785. * actual layer the hash was computed on. For instance, a hash computed
  786. * at L4 may be considered an L3 hash. This should only be done if the
  787. * driver can't unambiguously determine that the HW computed the hash at
  788. * the higher layer. Note that the "should" in the second property above
  789. * permits this.
  790. */
  791. enum pkt_hash_types {
  792. PKT_HASH_TYPE_NONE, /* Undefined type */
  793. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  794. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  795. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  796. };
  797. static inline void
  798. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  799. {
  800. skb->l4_hash = (type == PKT_HASH_TYPE_L4);
  801. skb->sw_hash = 0;
  802. skb->hash = hash;
  803. }
  804. void __skb_get_hash(struct sk_buff *skb);
  805. static inline __u32 skb_get_hash(struct sk_buff *skb)
  806. {
  807. if (!skb->l4_hash && !skb->sw_hash)
  808. __skb_get_hash(skb);
  809. return skb->hash;
  810. }
  811. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  812. {
  813. return skb->hash;
  814. }
  815. static inline void skb_clear_hash(struct sk_buff *skb)
  816. {
  817. skb->hash = 0;
  818. skb->sw_hash = 0;
  819. skb->l4_hash = 0;
  820. }
  821. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  822. {
  823. if (!skb->l4_hash)
  824. skb_clear_hash(skb);
  825. }
  826. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  827. {
  828. to->hash = from->hash;
  829. to->sw_hash = from->sw_hash;
  830. to->l4_hash = from->l4_hash;
  831. };
  832. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  833. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  834. {
  835. return skb->head + skb->end;
  836. }
  837. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  838. {
  839. return skb->end;
  840. }
  841. #else
  842. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  843. {
  844. return skb->end;
  845. }
  846. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  847. {
  848. return skb->end - skb->head;
  849. }
  850. #endif
  851. /* Internal */
  852. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  853. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  854. {
  855. return &skb_shinfo(skb)->hwtstamps;
  856. }
  857. /**
  858. * skb_queue_empty - check if a queue is empty
  859. * @list: queue head
  860. *
  861. * Returns true if the queue is empty, false otherwise.
  862. */
  863. static inline int skb_queue_empty(const struct sk_buff_head *list)
  864. {
  865. return list->next == (const struct sk_buff *) list;
  866. }
  867. /**
  868. * skb_queue_is_last - check if skb is the last entry in the queue
  869. * @list: queue head
  870. * @skb: buffer
  871. *
  872. * Returns true if @skb is the last buffer on the list.
  873. */
  874. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  875. const struct sk_buff *skb)
  876. {
  877. return skb->next == (const struct sk_buff *) list;
  878. }
  879. /**
  880. * skb_queue_is_first - check if skb is the first entry in the queue
  881. * @list: queue head
  882. * @skb: buffer
  883. *
  884. * Returns true if @skb is the first buffer on the list.
  885. */
  886. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  887. const struct sk_buff *skb)
  888. {
  889. return skb->prev == (const struct sk_buff *) list;
  890. }
  891. /**
  892. * skb_queue_next - return the next packet in the queue
  893. * @list: queue head
  894. * @skb: current buffer
  895. *
  896. * Return the next packet in @list after @skb. It is only valid to
  897. * call this if skb_queue_is_last() evaluates to false.
  898. */
  899. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  900. const struct sk_buff *skb)
  901. {
  902. /* This BUG_ON may seem severe, but if we just return then we
  903. * are going to dereference garbage.
  904. */
  905. BUG_ON(skb_queue_is_last(list, skb));
  906. return skb->next;
  907. }
  908. /**
  909. * skb_queue_prev - return the prev packet in the queue
  910. * @list: queue head
  911. * @skb: current buffer
  912. *
  913. * Return the prev packet in @list before @skb. It is only valid to
  914. * call this if skb_queue_is_first() evaluates to false.
  915. */
  916. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  917. const struct sk_buff *skb)
  918. {
  919. /* This BUG_ON may seem severe, but if we just return then we
  920. * are going to dereference garbage.
  921. */
  922. BUG_ON(skb_queue_is_first(list, skb));
  923. return skb->prev;
  924. }
  925. /**
  926. * skb_get - reference buffer
  927. * @skb: buffer to reference
  928. *
  929. * Makes another reference to a socket buffer and returns a pointer
  930. * to the buffer.
  931. */
  932. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  933. {
  934. atomic_inc(&skb->users);
  935. return skb;
  936. }
  937. /*
  938. * If users == 1, we are the only owner and are can avoid redundant
  939. * atomic change.
  940. */
  941. /**
  942. * skb_cloned - is the buffer a clone
  943. * @skb: buffer to check
  944. *
  945. * Returns true if the buffer was generated with skb_clone() and is
  946. * one of multiple shared copies of the buffer. Cloned buffers are
  947. * shared data so must not be written to under normal circumstances.
  948. */
  949. static inline int skb_cloned(const struct sk_buff *skb)
  950. {
  951. return skb->cloned &&
  952. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  953. }
  954. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  955. {
  956. might_sleep_if(pri & __GFP_WAIT);
  957. if (skb_cloned(skb))
  958. return pskb_expand_head(skb, 0, 0, pri);
  959. return 0;
  960. }
  961. /**
  962. * skb_header_cloned - is the header a clone
  963. * @skb: buffer to check
  964. *
  965. * Returns true if modifying the header part of the buffer requires
  966. * the data to be copied.
  967. */
  968. static inline int skb_header_cloned(const struct sk_buff *skb)
  969. {
  970. int dataref;
  971. if (!skb->cloned)
  972. return 0;
  973. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  974. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  975. return dataref != 1;
  976. }
  977. /**
  978. * skb_header_release - release reference to header
  979. * @skb: buffer to operate on
  980. *
  981. * Drop a reference to the header part of the buffer. This is done
  982. * by acquiring a payload reference. You must not read from the header
  983. * part of skb->data after this.
  984. * Note : Check if you can use __skb_header_release() instead.
  985. */
  986. static inline void skb_header_release(struct sk_buff *skb)
  987. {
  988. BUG_ON(skb->nohdr);
  989. skb->nohdr = 1;
  990. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  991. }
  992. /**
  993. * __skb_header_release - release reference to header
  994. * @skb: buffer to operate on
  995. *
  996. * Variant of skb_header_release() assuming skb is private to caller.
  997. * We can avoid one atomic operation.
  998. */
  999. static inline void __skb_header_release(struct sk_buff *skb)
  1000. {
  1001. skb->nohdr = 1;
  1002. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1003. }
  1004. /**
  1005. * skb_shared - is the buffer shared
  1006. * @skb: buffer to check
  1007. *
  1008. * Returns true if more than one person has a reference to this
  1009. * buffer.
  1010. */
  1011. static inline int skb_shared(const struct sk_buff *skb)
  1012. {
  1013. return atomic_read(&skb->users) != 1;
  1014. }
  1015. /**
  1016. * skb_share_check - check if buffer is shared and if so clone it
  1017. * @skb: buffer to check
  1018. * @pri: priority for memory allocation
  1019. *
  1020. * If the buffer is shared the buffer is cloned and the old copy
  1021. * drops a reference. A new clone with a single reference is returned.
  1022. * If the buffer is not shared the original buffer is returned. When
  1023. * being called from interrupt status or with spinlocks held pri must
  1024. * be GFP_ATOMIC.
  1025. *
  1026. * NULL is returned on a memory allocation failure.
  1027. */
  1028. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1029. {
  1030. might_sleep_if(pri & __GFP_WAIT);
  1031. if (skb_shared(skb)) {
  1032. struct sk_buff *nskb = skb_clone(skb, pri);
  1033. if (likely(nskb))
  1034. consume_skb(skb);
  1035. else
  1036. kfree_skb(skb);
  1037. skb = nskb;
  1038. }
  1039. return skb;
  1040. }
  1041. /*
  1042. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1043. * packets to handle cases where we have a local reader and forward
  1044. * and a couple of other messy ones. The normal one is tcpdumping
  1045. * a packet thats being forwarded.
  1046. */
  1047. /**
  1048. * skb_unshare - make a copy of a shared buffer
  1049. * @skb: buffer to check
  1050. * @pri: priority for memory allocation
  1051. *
  1052. * If the socket buffer is a clone then this function creates a new
  1053. * copy of the data, drops a reference count on the old copy and returns
  1054. * the new copy with the reference count at 1. If the buffer is not a clone
  1055. * the original buffer is returned. When called with a spinlock held or
  1056. * from interrupt state @pri must be %GFP_ATOMIC
  1057. *
  1058. * %NULL is returned on a memory allocation failure.
  1059. */
  1060. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1061. gfp_t pri)
  1062. {
  1063. might_sleep_if(pri & __GFP_WAIT);
  1064. if (skb_cloned(skb)) {
  1065. struct sk_buff *nskb = skb_copy(skb, pri);
  1066. /* Free our shared copy */
  1067. if (likely(nskb))
  1068. consume_skb(skb);
  1069. else
  1070. kfree_skb(skb);
  1071. skb = nskb;
  1072. }
  1073. return skb;
  1074. }
  1075. /**
  1076. * skb_peek - peek at the head of an &sk_buff_head
  1077. * @list_: list to peek at
  1078. *
  1079. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1080. * be careful with this one. A peek leaves the buffer on the
  1081. * list and someone else may run off with it. You must hold
  1082. * the appropriate locks or have a private queue to do this.
  1083. *
  1084. * Returns %NULL for an empty list or a pointer to the head element.
  1085. * The reference count is not incremented and the reference is therefore
  1086. * volatile. Use with caution.
  1087. */
  1088. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1089. {
  1090. struct sk_buff *skb = list_->next;
  1091. if (skb == (struct sk_buff *)list_)
  1092. skb = NULL;
  1093. return skb;
  1094. }
  1095. /**
  1096. * skb_peek_next - peek skb following the given one from a queue
  1097. * @skb: skb to start from
  1098. * @list_: list to peek at
  1099. *
  1100. * Returns %NULL when the end of the list is met or a pointer to the
  1101. * next element. The reference count is not incremented and the
  1102. * reference is therefore volatile. Use with caution.
  1103. */
  1104. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1105. const struct sk_buff_head *list_)
  1106. {
  1107. struct sk_buff *next = skb->next;
  1108. if (next == (struct sk_buff *)list_)
  1109. next = NULL;
  1110. return next;
  1111. }
  1112. /**
  1113. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1114. * @list_: list to peek at
  1115. *
  1116. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1117. * be careful with this one. A peek leaves the buffer on the
  1118. * list and someone else may run off with it. You must hold
  1119. * the appropriate locks or have a private queue to do this.
  1120. *
  1121. * Returns %NULL for an empty list or a pointer to the tail element.
  1122. * The reference count is not incremented and the reference is therefore
  1123. * volatile. Use with caution.
  1124. */
  1125. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1126. {
  1127. struct sk_buff *skb = list_->prev;
  1128. if (skb == (struct sk_buff *)list_)
  1129. skb = NULL;
  1130. return skb;
  1131. }
  1132. /**
  1133. * skb_queue_len - get queue length
  1134. * @list_: list to measure
  1135. *
  1136. * Return the length of an &sk_buff queue.
  1137. */
  1138. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1139. {
  1140. return list_->qlen;
  1141. }
  1142. /**
  1143. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1144. * @list: queue to initialize
  1145. *
  1146. * This initializes only the list and queue length aspects of
  1147. * an sk_buff_head object. This allows to initialize the list
  1148. * aspects of an sk_buff_head without reinitializing things like
  1149. * the spinlock. It can also be used for on-stack sk_buff_head
  1150. * objects where the spinlock is known to not be used.
  1151. */
  1152. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1153. {
  1154. list->prev = list->next = (struct sk_buff *)list;
  1155. list->qlen = 0;
  1156. }
  1157. /*
  1158. * This function creates a split out lock class for each invocation;
  1159. * this is needed for now since a whole lot of users of the skb-queue
  1160. * infrastructure in drivers have different locking usage (in hardirq)
  1161. * than the networking core (in softirq only). In the long run either the
  1162. * network layer or drivers should need annotation to consolidate the
  1163. * main types of usage into 3 classes.
  1164. */
  1165. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1166. {
  1167. spin_lock_init(&list->lock);
  1168. __skb_queue_head_init(list);
  1169. }
  1170. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1171. struct lock_class_key *class)
  1172. {
  1173. skb_queue_head_init(list);
  1174. lockdep_set_class(&list->lock, class);
  1175. }
  1176. /*
  1177. * Insert an sk_buff on a list.
  1178. *
  1179. * The "__skb_xxxx()" functions are the non-atomic ones that
  1180. * can only be called with interrupts disabled.
  1181. */
  1182. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1183. struct sk_buff_head *list);
  1184. static inline void __skb_insert(struct sk_buff *newsk,
  1185. struct sk_buff *prev, struct sk_buff *next,
  1186. struct sk_buff_head *list)
  1187. {
  1188. newsk->next = next;
  1189. newsk->prev = prev;
  1190. next->prev = prev->next = newsk;
  1191. list->qlen++;
  1192. }
  1193. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1194. struct sk_buff *prev,
  1195. struct sk_buff *next)
  1196. {
  1197. struct sk_buff *first = list->next;
  1198. struct sk_buff *last = list->prev;
  1199. first->prev = prev;
  1200. prev->next = first;
  1201. last->next = next;
  1202. next->prev = last;
  1203. }
  1204. /**
  1205. * skb_queue_splice - join two skb lists, this is designed for stacks
  1206. * @list: the new list to add
  1207. * @head: the place to add it in the first list
  1208. */
  1209. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1210. struct sk_buff_head *head)
  1211. {
  1212. if (!skb_queue_empty(list)) {
  1213. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1214. head->qlen += list->qlen;
  1215. }
  1216. }
  1217. /**
  1218. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1219. * @list: the new list to add
  1220. * @head: the place to add it in the first list
  1221. *
  1222. * The list at @list is reinitialised
  1223. */
  1224. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1225. struct sk_buff_head *head)
  1226. {
  1227. if (!skb_queue_empty(list)) {
  1228. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1229. head->qlen += list->qlen;
  1230. __skb_queue_head_init(list);
  1231. }
  1232. }
  1233. /**
  1234. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1235. * @list: the new list to add
  1236. * @head: the place to add it in the first list
  1237. */
  1238. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1239. struct sk_buff_head *head)
  1240. {
  1241. if (!skb_queue_empty(list)) {
  1242. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1243. head->qlen += list->qlen;
  1244. }
  1245. }
  1246. /**
  1247. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1248. * @list: the new list to add
  1249. * @head: the place to add it in the first list
  1250. *
  1251. * Each of the lists is a queue.
  1252. * The list at @list is reinitialised
  1253. */
  1254. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1255. struct sk_buff_head *head)
  1256. {
  1257. if (!skb_queue_empty(list)) {
  1258. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1259. head->qlen += list->qlen;
  1260. __skb_queue_head_init(list);
  1261. }
  1262. }
  1263. /**
  1264. * __skb_queue_after - queue a buffer at the list head
  1265. * @list: list to use
  1266. * @prev: place after this buffer
  1267. * @newsk: buffer to queue
  1268. *
  1269. * Queue a buffer int the middle of a list. This function takes no locks
  1270. * and you must therefore hold required locks before calling it.
  1271. *
  1272. * A buffer cannot be placed on two lists at the same time.
  1273. */
  1274. static inline void __skb_queue_after(struct sk_buff_head *list,
  1275. struct sk_buff *prev,
  1276. struct sk_buff *newsk)
  1277. {
  1278. __skb_insert(newsk, prev, prev->next, list);
  1279. }
  1280. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1281. struct sk_buff_head *list);
  1282. static inline void __skb_queue_before(struct sk_buff_head *list,
  1283. struct sk_buff *next,
  1284. struct sk_buff *newsk)
  1285. {
  1286. __skb_insert(newsk, next->prev, next, list);
  1287. }
  1288. /**
  1289. * __skb_queue_head - queue a buffer at the list head
  1290. * @list: list to use
  1291. * @newsk: buffer to queue
  1292. *
  1293. * Queue a buffer at the start of a list. This function takes no locks
  1294. * and you must therefore hold required locks before calling it.
  1295. *
  1296. * A buffer cannot be placed on two lists at the same time.
  1297. */
  1298. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1299. static inline void __skb_queue_head(struct sk_buff_head *list,
  1300. struct sk_buff *newsk)
  1301. {
  1302. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1303. }
  1304. /**
  1305. * __skb_queue_tail - queue a buffer at the list tail
  1306. * @list: list to use
  1307. * @newsk: buffer to queue
  1308. *
  1309. * Queue a buffer at the end of a list. This function takes no locks
  1310. * and you must therefore hold required locks before calling it.
  1311. *
  1312. * A buffer cannot be placed on two lists at the same time.
  1313. */
  1314. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1315. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1316. struct sk_buff *newsk)
  1317. {
  1318. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1319. }
  1320. /*
  1321. * remove sk_buff from list. _Must_ be called atomically, and with
  1322. * the list known..
  1323. */
  1324. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1325. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1326. {
  1327. struct sk_buff *next, *prev;
  1328. list->qlen--;
  1329. next = skb->next;
  1330. prev = skb->prev;
  1331. skb->next = skb->prev = NULL;
  1332. next->prev = prev;
  1333. prev->next = next;
  1334. }
  1335. /**
  1336. * __skb_dequeue - remove from the head of the queue
  1337. * @list: list to dequeue from
  1338. *
  1339. * Remove the head of the list. This function does not take any locks
  1340. * so must be used with appropriate locks held only. The head item is
  1341. * returned or %NULL if the list is empty.
  1342. */
  1343. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1344. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1345. {
  1346. struct sk_buff *skb = skb_peek(list);
  1347. if (skb)
  1348. __skb_unlink(skb, list);
  1349. return skb;
  1350. }
  1351. /**
  1352. * __skb_dequeue_tail - remove from the tail of the queue
  1353. * @list: list to dequeue from
  1354. *
  1355. * Remove the tail of the list. This function does not take any locks
  1356. * so must be used with appropriate locks held only. The tail item is
  1357. * returned or %NULL if the list is empty.
  1358. */
  1359. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1360. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1361. {
  1362. struct sk_buff *skb = skb_peek_tail(list);
  1363. if (skb)
  1364. __skb_unlink(skb, list);
  1365. return skb;
  1366. }
  1367. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1368. {
  1369. return skb->data_len;
  1370. }
  1371. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1372. {
  1373. return skb->len - skb->data_len;
  1374. }
  1375. static inline int skb_pagelen(const struct sk_buff *skb)
  1376. {
  1377. int i, len = 0;
  1378. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1379. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1380. return len + skb_headlen(skb);
  1381. }
  1382. /**
  1383. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1384. * @skb: buffer containing fragment to be initialised
  1385. * @i: paged fragment index to initialise
  1386. * @page: the page to use for this fragment
  1387. * @off: the offset to the data with @page
  1388. * @size: the length of the data
  1389. *
  1390. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1391. * offset @off within @page.
  1392. *
  1393. * Does not take any additional reference on the fragment.
  1394. */
  1395. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1396. struct page *page, int off, int size)
  1397. {
  1398. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1399. /*
  1400. * Propagate page->pfmemalloc to the skb if we can. The problem is
  1401. * that not all callers have unique ownership of the page. If
  1402. * pfmemalloc is set, we check the mapping as a mapping implies
  1403. * page->index is set (index and pfmemalloc share space).
  1404. * If it's a valid mapping, we cannot use page->pfmemalloc but we
  1405. * do not lose pfmemalloc information as the pages would not be
  1406. * allocated using __GFP_MEMALLOC.
  1407. */
  1408. frag->page.p = page;
  1409. frag->page_offset = off;
  1410. skb_frag_size_set(frag, size);
  1411. page = compound_head(page);
  1412. if (page->pfmemalloc && !page->mapping)
  1413. skb->pfmemalloc = true;
  1414. }
  1415. /**
  1416. * skb_fill_page_desc - initialise a paged fragment in an skb
  1417. * @skb: buffer containing fragment to be initialised
  1418. * @i: paged fragment index to initialise
  1419. * @page: the page to use for this fragment
  1420. * @off: the offset to the data with @page
  1421. * @size: the length of the data
  1422. *
  1423. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1424. * @skb to point to @size bytes at offset @off within @page. In
  1425. * addition updates @skb such that @i is the last fragment.
  1426. *
  1427. * Does not take any additional reference on the fragment.
  1428. */
  1429. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1430. struct page *page, int off, int size)
  1431. {
  1432. __skb_fill_page_desc(skb, i, page, off, size);
  1433. skb_shinfo(skb)->nr_frags = i + 1;
  1434. }
  1435. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1436. int size, unsigned int truesize);
  1437. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1438. unsigned int truesize);
  1439. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1440. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1441. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1442. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1443. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1444. {
  1445. return skb->head + skb->tail;
  1446. }
  1447. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1448. {
  1449. skb->tail = skb->data - skb->head;
  1450. }
  1451. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1452. {
  1453. skb_reset_tail_pointer(skb);
  1454. skb->tail += offset;
  1455. }
  1456. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1457. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1458. {
  1459. return skb->tail;
  1460. }
  1461. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1462. {
  1463. skb->tail = skb->data;
  1464. }
  1465. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1466. {
  1467. skb->tail = skb->data + offset;
  1468. }
  1469. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1470. /*
  1471. * Add data to an sk_buff
  1472. */
  1473. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1474. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1475. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1476. {
  1477. unsigned char *tmp = skb_tail_pointer(skb);
  1478. SKB_LINEAR_ASSERT(skb);
  1479. skb->tail += len;
  1480. skb->len += len;
  1481. return tmp;
  1482. }
  1483. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1484. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1485. {
  1486. skb->data -= len;
  1487. skb->len += len;
  1488. return skb->data;
  1489. }
  1490. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1491. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1492. {
  1493. skb->len -= len;
  1494. BUG_ON(skb->len < skb->data_len);
  1495. return skb->data += len;
  1496. }
  1497. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1498. {
  1499. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1500. }
  1501. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1502. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1503. {
  1504. if (len > skb_headlen(skb) &&
  1505. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1506. return NULL;
  1507. skb->len -= len;
  1508. return skb->data += len;
  1509. }
  1510. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1511. {
  1512. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1513. }
  1514. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1515. {
  1516. if (likely(len <= skb_headlen(skb)))
  1517. return 1;
  1518. if (unlikely(len > skb->len))
  1519. return 0;
  1520. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1521. }
  1522. /**
  1523. * skb_headroom - bytes at buffer head
  1524. * @skb: buffer to check
  1525. *
  1526. * Return the number of bytes of free space at the head of an &sk_buff.
  1527. */
  1528. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1529. {
  1530. return skb->data - skb->head;
  1531. }
  1532. /**
  1533. * skb_tailroom - bytes at buffer end
  1534. * @skb: buffer to check
  1535. *
  1536. * Return the number of bytes of free space at the tail of an sk_buff
  1537. */
  1538. static inline int skb_tailroom(const struct sk_buff *skb)
  1539. {
  1540. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1541. }
  1542. /**
  1543. * skb_availroom - bytes at buffer end
  1544. * @skb: buffer to check
  1545. *
  1546. * Return the number of bytes of free space at the tail of an sk_buff
  1547. * allocated by sk_stream_alloc()
  1548. */
  1549. static inline int skb_availroom(const struct sk_buff *skb)
  1550. {
  1551. if (skb_is_nonlinear(skb))
  1552. return 0;
  1553. return skb->end - skb->tail - skb->reserved_tailroom;
  1554. }
  1555. /**
  1556. * skb_reserve - adjust headroom
  1557. * @skb: buffer to alter
  1558. * @len: bytes to move
  1559. *
  1560. * Increase the headroom of an empty &sk_buff by reducing the tail
  1561. * room. This is only allowed for an empty buffer.
  1562. */
  1563. static inline void skb_reserve(struct sk_buff *skb, int len)
  1564. {
  1565. skb->data += len;
  1566. skb->tail += len;
  1567. }
  1568. #define ENCAP_TYPE_ETHER 0
  1569. #define ENCAP_TYPE_IPPROTO 1
  1570. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1571. __be16 protocol)
  1572. {
  1573. skb->inner_protocol = protocol;
  1574. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1575. }
  1576. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1577. __u8 ipproto)
  1578. {
  1579. skb->inner_ipproto = ipproto;
  1580. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1581. }
  1582. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1583. {
  1584. skb->inner_mac_header = skb->mac_header;
  1585. skb->inner_network_header = skb->network_header;
  1586. skb->inner_transport_header = skb->transport_header;
  1587. }
  1588. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1589. {
  1590. skb->mac_len = skb->network_header - skb->mac_header;
  1591. }
  1592. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1593. *skb)
  1594. {
  1595. return skb->head + skb->inner_transport_header;
  1596. }
  1597. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1598. {
  1599. skb->inner_transport_header = skb->data - skb->head;
  1600. }
  1601. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1602. const int offset)
  1603. {
  1604. skb_reset_inner_transport_header(skb);
  1605. skb->inner_transport_header += offset;
  1606. }
  1607. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1608. {
  1609. return skb->head + skb->inner_network_header;
  1610. }
  1611. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1612. {
  1613. skb->inner_network_header = skb->data - skb->head;
  1614. }
  1615. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1616. const int offset)
  1617. {
  1618. skb_reset_inner_network_header(skb);
  1619. skb->inner_network_header += offset;
  1620. }
  1621. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1622. {
  1623. return skb->head + skb->inner_mac_header;
  1624. }
  1625. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1626. {
  1627. skb->inner_mac_header = skb->data - skb->head;
  1628. }
  1629. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1630. const int offset)
  1631. {
  1632. skb_reset_inner_mac_header(skb);
  1633. skb->inner_mac_header += offset;
  1634. }
  1635. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1636. {
  1637. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1638. }
  1639. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1640. {
  1641. return skb->head + skb->transport_header;
  1642. }
  1643. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1644. {
  1645. skb->transport_header = skb->data - skb->head;
  1646. }
  1647. static inline void skb_set_transport_header(struct sk_buff *skb,
  1648. const int offset)
  1649. {
  1650. skb_reset_transport_header(skb);
  1651. skb->transport_header += offset;
  1652. }
  1653. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1654. {
  1655. return skb->head + skb->network_header;
  1656. }
  1657. static inline void skb_reset_network_header(struct sk_buff *skb)
  1658. {
  1659. skb->network_header = skb->data - skb->head;
  1660. }
  1661. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1662. {
  1663. skb_reset_network_header(skb);
  1664. skb->network_header += offset;
  1665. }
  1666. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1667. {
  1668. return skb->head + skb->mac_header;
  1669. }
  1670. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1671. {
  1672. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1673. }
  1674. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1675. {
  1676. skb->mac_header = skb->data - skb->head;
  1677. }
  1678. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1679. {
  1680. skb_reset_mac_header(skb);
  1681. skb->mac_header += offset;
  1682. }
  1683. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1684. {
  1685. skb->mac_header = skb->network_header;
  1686. }
  1687. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1688. const int offset_hint)
  1689. {
  1690. struct flow_keys keys;
  1691. if (skb_transport_header_was_set(skb))
  1692. return;
  1693. else if (skb_flow_dissect(skb, &keys))
  1694. skb_set_transport_header(skb, keys.thoff);
  1695. else
  1696. skb_set_transport_header(skb, offset_hint);
  1697. }
  1698. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1699. {
  1700. if (skb_mac_header_was_set(skb)) {
  1701. const unsigned char *old_mac = skb_mac_header(skb);
  1702. skb_set_mac_header(skb, -skb->mac_len);
  1703. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1704. }
  1705. }
  1706. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1707. {
  1708. return skb->csum_start - skb_headroom(skb);
  1709. }
  1710. static inline int skb_transport_offset(const struct sk_buff *skb)
  1711. {
  1712. return skb_transport_header(skb) - skb->data;
  1713. }
  1714. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1715. {
  1716. return skb->transport_header - skb->network_header;
  1717. }
  1718. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1719. {
  1720. return skb->inner_transport_header - skb->inner_network_header;
  1721. }
  1722. static inline int skb_network_offset(const struct sk_buff *skb)
  1723. {
  1724. return skb_network_header(skb) - skb->data;
  1725. }
  1726. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1727. {
  1728. return skb_inner_network_header(skb) - skb->data;
  1729. }
  1730. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1731. {
  1732. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1733. }
  1734. /*
  1735. * CPUs often take a performance hit when accessing unaligned memory
  1736. * locations. The actual performance hit varies, it can be small if the
  1737. * hardware handles it or large if we have to take an exception and fix it
  1738. * in software.
  1739. *
  1740. * Since an ethernet header is 14 bytes network drivers often end up with
  1741. * the IP header at an unaligned offset. The IP header can be aligned by
  1742. * shifting the start of the packet by 2 bytes. Drivers should do this
  1743. * with:
  1744. *
  1745. * skb_reserve(skb, NET_IP_ALIGN);
  1746. *
  1747. * The downside to this alignment of the IP header is that the DMA is now
  1748. * unaligned. On some architectures the cost of an unaligned DMA is high
  1749. * and this cost outweighs the gains made by aligning the IP header.
  1750. *
  1751. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1752. * to be overridden.
  1753. */
  1754. #ifndef NET_IP_ALIGN
  1755. #define NET_IP_ALIGN 2
  1756. #endif
  1757. /*
  1758. * The networking layer reserves some headroom in skb data (via
  1759. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1760. * the header has to grow. In the default case, if the header has to grow
  1761. * 32 bytes or less we avoid the reallocation.
  1762. *
  1763. * Unfortunately this headroom changes the DMA alignment of the resulting
  1764. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1765. * on some architectures. An architecture can override this value,
  1766. * perhaps setting it to a cacheline in size (since that will maintain
  1767. * cacheline alignment of the DMA). It must be a power of 2.
  1768. *
  1769. * Various parts of the networking layer expect at least 32 bytes of
  1770. * headroom, you should not reduce this.
  1771. *
  1772. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1773. * to reduce average number of cache lines per packet.
  1774. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1775. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1776. */
  1777. #ifndef NET_SKB_PAD
  1778. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1779. #endif
  1780. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1781. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1782. {
  1783. if (unlikely(skb_is_nonlinear(skb))) {
  1784. WARN_ON(1);
  1785. return;
  1786. }
  1787. skb->len = len;
  1788. skb_set_tail_pointer(skb, len);
  1789. }
  1790. void skb_trim(struct sk_buff *skb, unsigned int len);
  1791. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1792. {
  1793. if (skb->data_len)
  1794. return ___pskb_trim(skb, len);
  1795. __skb_trim(skb, len);
  1796. return 0;
  1797. }
  1798. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1799. {
  1800. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1801. }
  1802. /**
  1803. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1804. * @skb: buffer to alter
  1805. * @len: new length
  1806. *
  1807. * This is identical to pskb_trim except that the caller knows that
  1808. * the skb is not cloned so we should never get an error due to out-
  1809. * of-memory.
  1810. */
  1811. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  1812. {
  1813. int err = pskb_trim(skb, len);
  1814. BUG_ON(err);
  1815. }
  1816. /**
  1817. * skb_orphan - orphan a buffer
  1818. * @skb: buffer to orphan
  1819. *
  1820. * If a buffer currently has an owner then we call the owner's
  1821. * destructor function and make the @skb unowned. The buffer continues
  1822. * to exist but is no longer charged to its former owner.
  1823. */
  1824. static inline void skb_orphan(struct sk_buff *skb)
  1825. {
  1826. if (skb->destructor) {
  1827. skb->destructor(skb);
  1828. skb->destructor = NULL;
  1829. skb->sk = NULL;
  1830. } else {
  1831. BUG_ON(skb->sk);
  1832. }
  1833. }
  1834. /**
  1835. * skb_orphan_frags - orphan the frags contained in a buffer
  1836. * @skb: buffer to orphan frags from
  1837. * @gfp_mask: allocation mask for replacement pages
  1838. *
  1839. * For each frag in the SKB which needs a destructor (i.e. has an
  1840. * owner) create a copy of that frag and release the original
  1841. * page by calling the destructor.
  1842. */
  1843. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  1844. {
  1845. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  1846. return 0;
  1847. return skb_copy_ubufs(skb, gfp_mask);
  1848. }
  1849. /**
  1850. * __skb_queue_purge - empty a list
  1851. * @list: list to empty
  1852. *
  1853. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  1854. * the list and one reference dropped. This function does not take the
  1855. * list lock and the caller must hold the relevant locks to use it.
  1856. */
  1857. void skb_queue_purge(struct sk_buff_head *list);
  1858. static inline void __skb_queue_purge(struct sk_buff_head *list)
  1859. {
  1860. struct sk_buff *skb;
  1861. while ((skb = __skb_dequeue(list)) != NULL)
  1862. kfree_skb(skb);
  1863. }
  1864. #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
  1865. #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
  1866. #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
  1867. void *netdev_alloc_frag(unsigned int fragsz);
  1868. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  1869. gfp_t gfp_mask);
  1870. /**
  1871. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  1872. * @dev: network device to receive on
  1873. * @length: length to allocate
  1874. *
  1875. * Allocate a new &sk_buff and assign it a usage count of one. The
  1876. * buffer has unspecified headroom built in. Users should allocate
  1877. * the headroom they think they need without accounting for the
  1878. * built in space. The built in space is used for optimisations.
  1879. *
  1880. * %NULL is returned if there is no free memory. Although this function
  1881. * allocates memory it can be called from an interrupt.
  1882. */
  1883. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  1884. unsigned int length)
  1885. {
  1886. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  1887. }
  1888. /* legacy helper around __netdev_alloc_skb() */
  1889. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  1890. gfp_t gfp_mask)
  1891. {
  1892. return __netdev_alloc_skb(NULL, length, gfp_mask);
  1893. }
  1894. /* legacy helper around netdev_alloc_skb() */
  1895. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  1896. {
  1897. return netdev_alloc_skb(NULL, length);
  1898. }
  1899. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  1900. unsigned int length, gfp_t gfp)
  1901. {
  1902. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  1903. if (NET_IP_ALIGN && skb)
  1904. skb_reserve(skb, NET_IP_ALIGN);
  1905. return skb;
  1906. }
  1907. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  1908. unsigned int length)
  1909. {
  1910. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  1911. }
  1912. void *napi_alloc_frag(unsigned int fragsz);
  1913. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  1914. unsigned int length, gfp_t gfp_mask);
  1915. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  1916. unsigned int length)
  1917. {
  1918. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  1919. }
  1920. /**
  1921. * __dev_alloc_pages - allocate page for network Rx
  1922. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1923. * @order: size of the allocation
  1924. *
  1925. * Allocate a new page.
  1926. *
  1927. * %NULL is returned if there is no free memory.
  1928. */
  1929. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  1930. unsigned int order)
  1931. {
  1932. /* This piece of code contains several assumptions.
  1933. * 1. This is for device Rx, therefor a cold page is preferred.
  1934. * 2. The expectation is the user wants a compound page.
  1935. * 3. If requesting a order 0 page it will not be compound
  1936. * due to the check to see if order has a value in prep_new_page
  1937. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  1938. * code in gfp_to_alloc_flags that should be enforcing this.
  1939. */
  1940. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  1941. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  1942. }
  1943. static inline struct page *dev_alloc_pages(unsigned int order)
  1944. {
  1945. return __dev_alloc_pages(GFP_ATOMIC, order);
  1946. }
  1947. /**
  1948. * __dev_alloc_page - allocate a page for network Rx
  1949. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1950. *
  1951. * Allocate a new page.
  1952. *
  1953. * %NULL is returned if there is no free memory.
  1954. */
  1955. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  1956. {
  1957. return __dev_alloc_pages(gfp_mask, 0);
  1958. }
  1959. static inline struct page *dev_alloc_page(void)
  1960. {
  1961. return __dev_alloc_page(GFP_ATOMIC);
  1962. }
  1963. /**
  1964. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  1965. * @page: The page that was allocated from skb_alloc_page
  1966. * @skb: The skb that may need pfmemalloc set
  1967. */
  1968. static inline void skb_propagate_pfmemalloc(struct page *page,
  1969. struct sk_buff *skb)
  1970. {
  1971. if (page && page->pfmemalloc)
  1972. skb->pfmemalloc = true;
  1973. }
  1974. /**
  1975. * skb_frag_page - retrieve the page referred to by a paged fragment
  1976. * @frag: the paged fragment
  1977. *
  1978. * Returns the &struct page associated with @frag.
  1979. */
  1980. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  1981. {
  1982. return frag->page.p;
  1983. }
  1984. /**
  1985. * __skb_frag_ref - take an addition reference on a paged fragment.
  1986. * @frag: the paged fragment
  1987. *
  1988. * Takes an additional reference on the paged fragment @frag.
  1989. */
  1990. static inline void __skb_frag_ref(skb_frag_t *frag)
  1991. {
  1992. get_page(skb_frag_page(frag));
  1993. }
  1994. /**
  1995. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  1996. * @skb: the buffer
  1997. * @f: the fragment offset.
  1998. *
  1999. * Takes an additional reference on the @f'th paged fragment of @skb.
  2000. */
  2001. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2002. {
  2003. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2004. }
  2005. /**
  2006. * __skb_frag_unref - release a reference on a paged fragment.
  2007. * @frag: the paged fragment
  2008. *
  2009. * Releases a reference on the paged fragment @frag.
  2010. */
  2011. static inline void __skb_frag_unref(skb_frag_t *frag)
  2012. {
  2013. put_page(skb_frag_page(frag));
  2014. }
  2015. /**
  2016. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2017. * @skb: the buffer
  2018. * @f: the fragment offset
  2019. *
  2020. * Releases a reference on the @f'th paged fragment of @skb.
  2021. */
  2022. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2023. {
  2024. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2025. }
  2026. /**
  2027. * skb_frag_address - gets the address of the data contained in a paged fragment
  2028. * @frag: the paged fragment buffer
  2029. *
  2030. * Returns the address of the data within @frag. The page must already
  2031. * be mapped.
  2032. */
  2033. static inline void *skb_frag_address(const skb_frag_t *frag)
  2034. {
  2035. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2036. }
  2037. /**
  2038. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2039. * @frag: the paged fragment buffer
  2040. *
  2041. * Returns the address of the data within @frag. Checks that the page
  2042. * is mapped and returns %NULL otherwise.
  2043. */
  2044. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2045. {
  2046. void *ptr = page_address(skb_frag_page(frag));
  2047. if (unlikely(!ptr))
  2048. return NULL;
  2049. return ptr + frag->page_offset;
  2050. }
  2051. /**
  2052. * __skb_frag_set_page - sets the page contained in a paged fragment
  2053. * @frag: the paged fragment
  2054. * @page: the page to set
  2055. *
  2056. * Sets the fragment @frag to contain @page.
  2057. */
  2058. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2059. {
  2060. frag->page.p = page;
  2061. }
  2062. /**
  2063. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2064. * @skb: the buffer
  2065. * @f: the fragment offset
  2066. * @page: the page to set
  2067. *
  2068. * Sets the @f'th fragment of @skb to contain @page.
  2069. */
  2070. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2071. struct page *page)
  2072. {
  2073. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2074. }
  2075. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2076. /**
  2077. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2078. * @dev: the device to map the fragment to
  2079. * @frag: the paged fragment to map
  2080. * @offset: the offset within the fragment (starting at the
  2081. * fragment's own offset)
  2082. * @size: the number of bytes to map
  2083. * @dir: the direction of the mapping (%PCI_DMA_*)
  2084. *
  2085. * Maps the page associated with @frag to @device.
  2086. */
  2087. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2088. const skb_frag_t *frag,
  2089. size_t offset, size_t size,
  2090. enum dma_data_direction dir)
  2091. {
  2092. return dma_map_page(dev, skb_frag_page(frag),
  2093. frag->page_offset + offset, size, dir);
  2094. }
  2095. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2096. gfp_t gfp_mask)
  2097. {
  2098. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2099. }
  2100. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2101. gfp_t gfp_mask)
  2102. {
  2103. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2104. }
  2105. /**
  2106. * skb_clone_writable - is the header of a clone writable
  2107. * @skb: buffer to check
  2108. * @len: length up to which to write
  2109. *
  2110. * Returns true if modifying the header part of the cloned buffer
  2111. * does not requires the data to be copied.
  2112. */
  2113. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2114. {
  2115. return !skb_header_cloned(skb) &&
  2116. skb_headroom(skb) + len <= skb->hdr_len;
  2117. }
  2118. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2119. int cloned)
  2120. {
  2121. int delta = 0;
  2122. if (headroom > skb_headroom(skb))
  2123. delta = headroom - skb_headroom(skb);
  2124. if (delta || cloned)
  2125. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2126. GFP_ATOMIC);
  2127. return 0;
  2128. }
  2129. /**
  2130. * skb_cow - copy header of skb when it is required
  2131. * @skb: buffer to cow
  2132. * @headroom: needed headroom
  2133. *
  2134. * If the skb passed lacks sufficient headroom or its data part
  2135. * is shared, data is reallocated. If reallocation fails, an error
  2136. * is returned and original skb is not changed.
  2137. *
  2138. * The result is skb with writable area skb->head...skb->tail
  2139. * and at least @headroom of space at head.
  2140. */
  2141. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2142. {
  2143. return __skb_cow(skb, headroom, skb_cloned(skb));
  2144. }
  2145. /**
  2146. * skb_cow_head - skb_cow but only making the head writable
  2147. * @skb: buffer to cow
  2148. * @headroom: needed headroom
  2149. *
  2150. * This function is identical to skb_cow except that we replace the
  2151. * skb_cloned check by skb_header_cloned. It should be used when
  2152. * you only need to push on some header and do not need to modify
  2153. * the data.
  2154. */
  2155. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2156. {
  2157. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2158. }
  2159. /**
  2160. * skb_padto - pad an skbuff up to a minimal size
  2161. * @skb: buffer to pad
  2162. * @len: minimal length
  2163. *
  2164. * Pads up a buffer to ensure the trailing bytes exist and are
  2165. * blanked. If the buffer already contains sufficient data it
  2166. * is untouched. Otherwise it is extended. Returns zero on
  2167. * success. The skb is freed on error.
  2168. */
  2169. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2170. {
  2171. unsigned int size = skb->len;
  2172. if (likely(size >= len))
  2173. return 0;
  2174. return skb_pad(skb, len - size);
  2175. }
  2176. /**
  2177. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2178. * @skb: buffer to pad
  2179. * @len: minimal length
  2180. *
  2181. * Pads up a buffer to ensure the trailing bytes exist and are
  2182. * blanked. If the buffer already contains sufficient data it
  2183. * is untouched. Otherwise it is extended. Returns zero on
  2184. * success. The skb is freed on error.
  2185. */
  2186. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2187. {
  2188. unsigned int size = skb->len;
  2189. if (unlikely(size < len)) {
  2190. len -= size;
  2191. if (skb_pad(skb, len))
  2192. return -ENOMEM;
  2193. __skb_put(skb, len);
  2194. }
  2195. return 0;
  2196. }
  2197. static inline int skb_add_data(struct sk_buff *skb,
  2198. struct iov_iter *from, int copy)
  2199. {
  2200. const int off = skb->len;
  2201. if (skb->ip_summed == CHECKSUM_NONE) {
  2202. __wsum csum = 0;
  2203. if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
  2204. &csum, from) == copy) {
  2205. skb->csum = csum_block_add(skb->csum, csum, off);
  2206. return 0;
  2207. }
  2208. } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
  2209. return 0;
  2210. __skb_trim(skb, off);
  2211. return -EFAULT;
  2212. }
  2213. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2214. const struct page *page, int off)
  2215. {
  2216. if (i) {
  2217. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2218. return page == skb_frag_page(frag) &&
  2219. off == frag->page_offset + skb_frag_size(frag);
  2220. }
  2221. return false;
  2222. }
  2223. static inline int __skb_linearize(struct sk_buff *skb)
  2224. {
  2225. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2226. }
  2227. /**
  2228. * skb_linearize - convert paged skb to linear one
  2229. * @skb: buffer to linarize
  2230. *
  2231. * If there is no free memory -ENOMEM is returned, otherwise zero
  2232. * is returned and the old skb data released.
  2233. */
  2234. static inline int skb_linearize(struct sk_buff *skb)
  2235. {
  2236. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2237. }
  2238. /**
  2239. * skb_has_shared_frag - can any frag be overwritten
  2240. * @skb: buffer to test
  2241. *
  2242. * Return true if the skb has at least one frag that might be modified
  2243. * by an external entity (as in vmsplice()/sendfile())
  2244. */
  2245. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2246. {
  2247. return skb_is_nonlinear(skb) &&
  2248. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2249. }
  2250. /**
  2251. * skb_linearize_cow - make sure skb is linear and writable
  2252. * @skb: buffer to process
  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_cow(struct sk_buff *skb)
  2258. {
  2259. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2260. __skb_linearize(skb) : 0;
  2261. }
  2262. /**
  2263. * skb_postpull_rcsum - update checksum for received skb after pull
  2264. * @skb: buffer to update
  2265. * @start: start of data before pull
  2266. * @len: length of data pulled
  2267. *
  2268. * After doing a pull on a received packet, you need to call this to
  2269. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2270. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2271. */
  2272. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2273. const void *start, unsigned int len)
  2274. {
  2275. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2276. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2277. }
  2278. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2279. /**
  2280. * pskb_trim_rcsum - trim received skb and update checksum
  2281. * @skb: buffer to trim
  2282. * @len: new length
  2283. *
  2284. * This is exactly the same as pskb_trim except that it ensures the
  2285. * checksum of received packets are still valid after the operation.
  2286. */
  2287. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2288. {
  2289. if (likely(len >= skb->len))
  2290. return 0;
  2291. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2292. skb->ip_summed = CHECKSUM_NONE;
  2293. return __pskb_trim(skb, len);
  2294. }
  2295. #define skb_queue_walk(queue, skb) \
  2296. for (skb = (queue)->next; \
  2297. skb != (struct sk_buff *)(queue); \
  2298. skb = skb->next)
  2299. #define skb_queue_walk_safe(queue, skb, tmp) \
  2300. for (skb = (queue)->next, tmp = skb->next; \
  2301. skb != (struct sk_buff *)(queue); \
  2302. skb = tmp, tmp = skb->next)
  2303. #define skb_queue_walk_from(queue, skb) \
  2304. for (; skb != (struct sk_buff *)(queue); \
  2305. skb = skb->next)
  2306. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2307. for (tmp = skb->next; \
  2308. skb != (struct sk_buff *)(queue); \
  2309. skb = tmp, tmp = skb->next)
  2310. #define skb_queue_reverse_walk(queue, skb) \
  2311. for (skb = (queue)->prev; \
  2312. skb != (struct sk_buff *)(queue); \
  2313. skb = skb->prev)
  2314. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2315. for (skb = (queue)->prev, tmp = skb->prev; \
  2316. skb != (struct sk_buff *)(queue); \
  2317. skb = tmp, tmp = skb->prev)
  2318. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2319. for (tmp = skb->prev; \
  2320. skb != (struct sk_buff *)(queue); \
  2321. skb = tmp, tmp = skb->prev)
  2322. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2323. {
  2324. return skb_shinfo(skb)->frag_list != NULL;
  2325. }
  2326. static inline void skb_frag_list_init(struct sk_buff *skb)
  2327. {
  2328. skb_shinfo(skb)->frag_list = NULL;
  2329. }
  2330. static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
  2331. {
  2332. frag->next = skb_shinfo(skb)->frag_list;
  2333. skb_shinfo(skb)->frag_list = frag;
  2334. }
  2335. #define skb_walk_frags(skb, iter) \
  2336. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2337. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2338. int *peeked, int *off, int *err);
  2339. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2340. int *err);
  2341. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2342. struct poll_table_struct *wait);
  2343. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2344. struct iov_iter *to, int size);
  2345. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2346. struct msghdr *msg, int size)
  2347. {
  2348. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2349. }
  2350. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2351. struct msghdr *msg);
  2352. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2353. struct iov_iter *from, int len);
  2354. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2355. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2356. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2357. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2358. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2359. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2360. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2361. int len, __wsum csum);
  2362. int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
  2363. struct pipe_inode_info *pipe, unsigned int len,
  2364. unsigned int flags);
  2365. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2366. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2367. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2368. int len, int hlen);
  2369. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2370. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2371. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2372. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2373. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2374. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2375. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2376. int skb_vlan_pop(struct sk_buff *skb);
  2377. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2378. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2379. {
  2380. return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2381. }
  2382. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2383. {
  2384. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2385. }
  2386. struct skb_checksum_ops {
  2387. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2388. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2389. };
  2390. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2391. __wsum csum, const struct skb_checksum_ops *ops);
  2392. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2393. __wsum csum);
  2394. static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
  2395. int len, void *data, int hlen, void *buffer)
  2396. {
  2397. if (hlen - offset >= len)
  2398. return data + offset;
  2399. if (!skb ||
  2400. skb_copy_bits(skb, offset, buffer, len) < 0)
  2401. return NULL;
  2402. return buffer;
  2403. }
  2404. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  2405. int len, void *buffer)
  2406. {
  2407. return __skb_header_pointer(skb, offset, len, skb->data,
  2408. skb_headlen(skb), buffer);
  2409. }
  2410. /**
  2411. * skb_needs_linearize - check if we need to linearize a given skb
  2412. * depending on the given device features.
  2413. * @skb: socket buffer to check
  2414. * @features: net device features
  2415. *
  2416. * Returns true if either:
  2417. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2418. * 2. skb is fragmented and the device does not support SG.
  2419. */
  2420. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2421. netdev_features_t features)
  2422. {
  2423. return skb_is_nonlinear(skb) &&
  2424. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2425. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2426. }
  2427. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2428. void *to,
  2429. const unsigned int len)
  2430. {
  2431. memcpy(to, skb->data, len);
  2432. }
  2433. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2434. const int offset, void *to,
  2435. const unsigned int len)
  2436. {
  2437. memcpy(to, skb->data + offset, len);
  2438. }
  2439. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2440. const void *from,
  2441. const unsigned int len)
  2442. {
  2443. memcpy(skb->data, from, len);
  2444. }
  2445. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2446. const int offset,
  2447. const void *from,
  2448. const unsigned int len)
  2449. {
  2450. memcpy(skb->data + offset, from, len);
  2451. }
  2452. void skb_init(void);
  2453. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2454. {
  2455. return skb->tstamp;
  2456. }
  2457. /**
  2458. * skb_get_timestamp - get timestamp from a skb
  2459. * @skb: skb to get stamp from
  2460. * @stamp: pointer to struct timeval to store stamp in
  2461. *
  2462. * Timestamps are stored in the skb as offsets to a base timestamp.
  2463. * This function converts the offset back to a struct timeval and stores
  2464. * it in stamp.
  2465. */
  2466. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2467. struct timeval *stamp)
  2468. {
  2469. *stamp = ktime_to_timeval(skb->tstamp);
  2470. }
  2471. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2472. struct timespec *stamp)
  2473. {
  2474. *stamp = ktime_to_timespec(skb->tstamp);
  2475. }
  2476. static inline void __net_timestamp(struct sk_buff *skb)
  2477. {
  2478. skb->tstamp = ktime_get_real();
  2479. }
  2480. static inline ktime_t net_timedelta(ktime_t t)
  2481. {
  2482. return ktime_sub(ktime_get_real(), t);
  2483. }
  2484. static inline ktime_t net_invalid_timestamp(void)
  2485. {
  2486. return ktime_set(0, 0);
  2487. }
  2488. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2489. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2490. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2491. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2492. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2493. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2494. {
  2495. }
  2496. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2497. {
  2498. return false;
  2499. }
  2500. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2501. /**
  2502. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2503. *
  2504. * PHY drivers may accept clones of transmitted packets for
  2505. * timestamping via their phy_driver.txtstamp method. These drivers
  2506. * must call this function to return the skb back to the stack, with
  2507. * or without a timestamp.
  2508. *
  2509. * @skb: clone of the the original outgoing packet
  2510. * @hwtstamps: hardware time stamps, may be NULL if not available
  2511. *
  2512. */
  2513. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2514. struct skb_shared_hwtstamps *hwtstamps);
  2515. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2516. struct skb_shared_hwtstamps *hwtstamps,
  2517. struct sock *sk, int tstype);
  2518. /**
  2519. * skb_tstamp_tx - queue clone of skb with send time stamps
  2520. * @orig_skb: the original outgoing packet
  2521. * @hwtstamps: hardware time stamps, may be NULL if not available
  2522. *
  2523. * If the skb has a socket associated, then this function clones the
  2524. * skb (thus sharing the actual data and optional structures), stores
  2525. * the optional hardware time stamping information (if non NULL) or
  2526. * generates a software time stamp (otherwise), then queues the clone
  2527. * to the error queue of the socket. Errors are silently ignored.
  2528. */
  2529. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2530. struct skb_shared_hwtstamps *hwtstamps);
  2531. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2532. {
  2533. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2534. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2535. skb_tstamp_tx(skb, NULL);
  2536. }
  2537. /**
  2538. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2539. *
  2540. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2541. * function immediately before giving the sk_buff to the MAC hardware.
  2542. *
  2543. * Specifically, one should make absolutely sure that this function is
  2544. * called before TX completion of this packet can trigger. Otherwise
  2545. * the packet could potentially already be freed.
  2546. *
  2547. * @skb: A socket buffer.
  2548. */
  2549. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2550. {
  2551. skb_clone_tx_timestamp(skb);
  2552. sw_tx_timestamp(skb);
  2553. }
  2554. /**
  2555. * skb_complete_wifi_ack - deliver skb with wifi status
  2556. *
  2557. * @skb: the original outgoing packet
  2558. * @acked: ack status
  2559. *
  2560. */
  2561. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2562. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2563. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2564. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2565. {
  2566. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2567. skb->csum_valid ||
  2568. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2569. skb_checksum_start_offset(skb) >= 0));
  2570. }
  2571. /**
  2572. * skb_checksum_complete - Calculate checksum of an entire packet
  2573. * @skb: packet to process
  2574. *
  2575. * This function calculates the checksum over the entire packet plus
  2576. * the value of skb->csum. The latter can be used to supply the
  2577. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2578. * checksum.
  2579. *
  2580. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2581. * this function can be used to verify that checksum on received
  2582. * packets. In that case the function should return zero if the
  2583. * checksum is correct. In particular, this function will return zero
  2584. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2585. * hardware has already verified the correctness of the checksum.
  2586. */
  2587. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2588. {
  2589. return skb_csum_unnecessary(skb) ?
  2590. 0 : __skb_checksum_complete(skb);
  2591. }
  2592. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2593. {
  2594. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2595. if (skb->csum_level == 0)
  2596. skb->ip_summed = CHECKSUM_NONE;
  2597. else
  2598. skb->csum_level--;
  2599. }
  2600. }
  2601. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2602. {
  2603. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2604. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2605. skb->csum_level++;
  2606. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2607. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2608. skb->csum_level = 0;
  2609. }
  2610. }
  2611. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2612. {
  2613. /* Mark current checksum as bad (typically called from GRO
  2614. * path). In the case that ip_summed is CHECKSUM_NONE
  2615. * this must be the first checksum encountered in the packet.
  2616. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2617. * checksum after the last one validated. For UDP, a zero
  2618. * checksum can not be marked as bad.
  2619. */
  2620. if (skb->ip_summed == CHECKSUM_NONE ||
  2621. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2622. skb->csum_bad = 1;
  2623. }
  2624. /* Check if we need to perform checksum complete validation.
  2625. *
  2626. * Returns true if checksum complete is needed, false otherwise
  2627. * (either checksum is unnecessary or zero checksum is allowed).
  2628. */
  2629. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2630. bool zero_okay,
  2631. __sum16 check)
  2632. {
  2633. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2634. skb->csum_valid = 1;
  2635. __skb_decr_checksum_unnecessary(skb);
  2636. return false;
  2637. }
  2638. return true;
  2639. }
  2640. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2641. * in checksum_init.
  2642. */
  2643. #define CHECKSUM_BREAK 76
  2644. /* Validate (init) checksum based on checksum complete.
  2645. *
  2646. * Return values:
  2647. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2648. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2649. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2650. * non-zero: value of invalid checksum
  2651. *
  2652. */
  2653. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2654. bool complete,
  2655. __wsum psum)
  2656. {
  2657. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2658. if (!csum_fold(csum_add(psum, skb->csum))) {
  2659. skb->csum_valid = 1;
  2660. return 0;
  2661. }
  2662. } else if (skb->csum_bad) {
  2663. /* ip_summed == CHECKSUM_NONE in this case */
  2664. return 1;
  2665. }
  2666. skb->csum = psum;
  2667. if (complete || skb->len <= CHECKSUM_BREAK) {
  2668. __sum16 csum;
  2669. csum = __skb_checksum_complete(skb);
  2670. skb->csum_valid = !csum;
  2671. return csum;
  2672. }
  2673. return 0;
  2674. }
  2675. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2676. {
  2677. return 0;
  2678. }
  2679. /* Perform checksum validate (init). Note that this is a macro since we only
  2680. * want to calculate the pseudo header which is an input function if necessary.
  2681. * First we try to validate without any computation (checksum unnecessary) and
  2682. * then calculate based on checksum complete calling the function to compute
  2683. * pseudo header.
  2684. *
  2685. * Return values:
  2686. * 0: checksum is validated or try to in skb_checksum_complete
  2687. * non-zero: value of invalid checksum
  2688. */
  2689. #define __skb_checksum_validate(skb, proto, complete, \
  2690. zero_okay, check, compute_pseudo) \
  2691. ({ \
  2692. __sum16 __ret = 0; \
  2693. skb->csum_valid = 0; \
  2694. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2695. __ret = __skb_checksum_validate_complete(skb, \
  2696. complete, compute_pseudo(skb, proto)); \
  2697. __ret; \
  2698. })
  2699. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2700. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2701. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2702. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2703. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2704. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2705. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2706. compute_pseudo) \
  2707. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  2708. #define skb_checksum_simple_validate(skb) \
  2709. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2710. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2711. {
  2712. return (skb->ip_summed == CHECKSUM_NONE &&
  2713. skb->csum_valid && !skb->csum_bad);
  2714. }
  2715. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2716. __sum16 check, __wsum pseudo)
  2717. {
  2718. skb->csum = ~pseudo;
  2719. skb->ip_summed = CHECKSUM_COMPLETE;
  2720. }
  2721. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  2722. do { \
  2723. if (__skb_checksum_convert_check(skb)) \
  2724. __skb_checksum_convert(skb, check, \
  2725. compute_pseudo(skb, proto)); \
  2726. } while (0)
  2727. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  2728. u16 start, u16 offset)
  2729. {
  2730. skb->ip_summed = CHECKSUM_PARTIAL;
  2731. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  2732. skb->csum_offset = offset - start;
  2733. }
  2734. /* Update skbuf and packet to reflect the remote checksum offload operation.
  2735. * When called, ptr indicates the starting point for skb->csum when
  2736. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  2737. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  2738. */
  2739. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  2740. int start, int offset, bool nopartial)
  2741. {
  2742. __wsum delta;
  2743. if (!nopartial) {
  2744. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  2745. return;
  2746. }
  2747. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  2748. __skb_checksum_complete(skb);
  2749. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  2750. }
  2751. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  2752. /* Adjust skb->csum since we changed the packet */
  2753. skb->csum = csum_add(skb->csum, delta);
  2754. }
  2755. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2756. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2757. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2758. {
  2759. if (nfct && atomic_dec_and_test(&nfct->use))
  2760. nf_conntrack_destroy(nfct);
  2761. }
  2762. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2763. {
  2764. if (nfct)
  2765. atomic_inc(&nfct->use);
  2766. }
  2767. #endif
  2768. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2769. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  2770. {
  2771. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  2772. kfree(nf_bridge);
  2773. }
  2774. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  2775. {
  2776. if (nf_bridge)
  2777. atomic_inc(&nf_bridge->use);
  2778. }
  2779. #endif /* CONFIG_BRIDGE_NETFILTER */
  2780. static inline void nf_reset(struct sk_buff *skb)
  2781. {
  2782. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2783. nf_conntrack_put(skb->nfct);
  2784. skb->nfct = NULL;
  2785. #endif
  2786. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2787. nf_bridge_put(skb->nf_bridge);
  2788. skb->nf_bridge = NULL;
  2789. #endif
  2790. }
  2791. static inline void nf_reset_trace(struct sk_buff *skb)
  2792. {
  2793. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2794. skb->nf_trace = 0;
  2795. #endif
  2796. }
  2797. /* Note: This doesn't put any conntrack and bridge info in dst. */
  2798. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  2799. bool copy)
  2800. {
  2801. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2802. dst->nfct = src->nfct;
  2803. nf_conntrack_get(src->nfct);
  2804. if (copy)
  2805. dst->nfctinfo = src->nfctinfo;
  2806. #endif
  2807. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2808. dst->nf_bridge = src->nf_bridge;
  2809. nf_bridge_get(src->nf_bridge);
  2810. #endif
  2811. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2812. if (copy)
  2813. dst->nf_trace = src->nf_trace;
  2814. #endif
  2815. }
  2816. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2817. {
  2818. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2819. nf_conntrack_put(dst->nfct);
  2820. #endif
  2821. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2822. nf_bridge_put(dst->nf_bridge);
  2823. #endif
  2824. __nf_copy(dst, src, true);
  2825. }
  2826. #ifdef CONFIG_NETWORK_SECMARK
  2827. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2828. {
  2829. to->secmark = from->secmark;
  2830. }
  2831. static inline void skb_init_secmark(struct sk_buff *skb)
  2832. {
  2833. skb->secmark = 0;
  2834. }
  2835. #else
  2836. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2837. { }
  2838. static inline void skb_init_secmark(struct sk_buff *skb)
  2839. { }
  2840. #endif
  2841. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  2842. {
  2843. return !skb->destructor &&
  2844. #if IS_ENABLED(CONFIG_XFRM)
  2845. !skb->sp &&
  2846. #endif
  2847. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  2848. !skb->nfct &&
  2849. #endif
  2850. !skb->_skb_refdst &&
  2851. !skb_has_frag_list(skb);
  2852. }
  2853. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  2854. {
  2855. skb->queue_mapping = queue_mapping;
  2856. }
  2857. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  2858. {
  2859. return skb->queue_mapping;
  2860. }
  2861. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  2862. {
  2863. to->queue_mapping = from->queue_mapping;
  2864. }
  2865. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  2866. {
  2867. skb->queue_mapping = rx_queue + 1;
  2868. }
  2869. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  2870. {
  2871. return skb->queue_mapping - 1;
  2872. }
  2873. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  2874. {
  2875. return skb->queue_mapping != 0;
  2876. }
  2877. u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
  2878. unsigned int num_tx_queues);
  2879. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  2880. {
  2881. #ifdef CONFIG_XFRM
  2882. return skb->sp;
  2883. #else
  2884. return NULL;
  2885. #endif
  2886. }
  2887. /* Keeps track of mac header offset relative to skb->head.
  2888. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  2889. * For non-tunnel skb it points to skb_mac_header() and for
  2890. * tunnel skb it points to outer mac header.
  2891. * Keeps track of level of encapsulation of network headers.
  2892. */
  2893. struct skb_gso_cb {
  2894. int mac_offset;
  2895. int encap_level;
  2896. __u16 csum_start;
  2897. };
  2898. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
  2899. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  2900. {
  2901. return (skb_mac_header(inner_skb) - inner_skb->head) -
  2902. SKB_GSO_CB(inner_skb)->mac_offset;
  2903. }
  2904. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  2905. {
  2906. int new_headroom, headroom;
  2907. int ret;
  2908. headroom = skb_headroom(skb);
  2909. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  2910. if (ret)
  2911. return ret;
  2912. new_headroom = skb_headroom(skb);
  2913. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  2914. return 0;
  2915. }
  2916. /* Compute the checksum for a gso segment. First compute the checksum value
  2917. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  2918. * then add in skb->csum (checksum from csum_start to end of packet).
  2919. * skb->csum and csum_start are then updated to reflect the checksum of the
  2920. * resultant packet starting from the transport header-- the resultant checksum
  2921. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  2922. * header.
  2923. */
  2924. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  2925. {
  2926. int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
  2927. skb_transport_offset(skb);
  2928. __u16 csum;
  2929. csum = csum_fold(csum_partial(skb_transport_header(skb),
  2930. plen, skb->csum));
  2931. skb->csum = res;
  2932. SKB_GSO_CB(skb)->csum_start -= plen;
  2933. return csum;
  2934. }
  2935. static inline bool skb_is_gso(const struct sk_buff *skb)
  2936. {
  2937. return skb_shinfo(skb)->gso_size;
  2938. }
  2939. /* Note: Should be called only if skb_is_gso(skb) is true */
  2940. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  2941. {
  2942. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  2943. }
  2944. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  2945. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  2946. {
  2947. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  2948. * wanted then gso_type will be set. */
  2949. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  2950. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  2951. unlikely(shinfo->gso_type == 0)) {
  2952. __skb_warn_lro_forwarding(skb);
  2953. return true;
  2954. }
  2955. return false;
  2956. }
  2957. static inline void skb_forward_csum(struct sk_buff *skb)
  2958. {
  2959. /* Unfortunately we don't support this one. Any brave souls? */
  2960. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2961. skb->ip_summed = CHECKSUM_NONE;
  2962. }
  2963. /**
  2964. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  2965. * @skb: skb to check
  2966. *
  2967. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  2968. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  2969. * use this helper, to document places where we make this assertion.
  2970. */
  2971. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  2972. {
  2973. #ifdef DEBUG
  2974. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  2975. #endif
  2976. }
  2977. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  2978. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  2979. u32 skb_get_poff(const struct sk_buff *skb);
  2980. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  2981. const struct flow_keys *keys, int hlen);
  2982. /**
  2983. * skb_head_is_locked - Determine if the skb->head is locked down
  2984. * @skb: skb to check
  2985. *
  2986. * The head on skbs build around a head frag can be removed if they are
  2987. * not cloned. This function returns true if the skb head is locked down
  2988. * due to either being allocated via kmalloc, or by being a clone with
  2989. * multiple references to the head.
  2990. */
  2991. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  2992. {
  2993. return !skb->head_frag || skb_cloned(skb);
  2994. }
  2995. /**
  2996. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  2997. *
  2998. * @skb: GSO skb
  2999. *
  3000. * skb_gso_network_seglen is used to determine the real size of the
  3001. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3002. *
  3003. * The MAC/L2 header is not accounted for.
  3004. */
  3005. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3006. {
  3007. unsigned int hdr_len = skb_transport_header(skb) -
  3008. skb_network_header(skb);
  3009. return hdr_len + skb_gso_transport_seglen(skb);
  3010. }
  3011. #endif /* __KERNEL__ */
  3012. #endif /* _LINUX_SKBUFF_H */