skbuff.h 94 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460246124622463246424652466246724682469247024712472247324742475247624772478247924802481248224832484248524862487248824892490249124922493249424952496249724982499250025012502250325042505250625072508250925102511251225132514251525162517251825192520252125222523252425252526252725282529253025312532253325342535253625372538253925402541254225432544254525462547254825492550255125522553255425552556255725582559256025612562256325642565256625672568256925702571257225732574257525762577257825792580258125822583258425852586258725882589259025912592259325942595259625972598259926002601260226032604260526062607260826092610261126122613261426152616261726182619262026212622262326242625262626272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697269826992700270127022703270427052706270727082709271027112712271327142715271627172718271927202721272227232724272527262727272827292730273127322733273427352736273727382739274027412742274327442745274627472748274927502751275227532754275527562757275827592760276127622763276427652766276727682769277027712772277327742775277627772778277927802781278227832784278527862787278827892790279127922793279427952796279727982799280028012802280328042805280628072808280928102811281228132814281528162817281828192820282128222823282428252826282728282829283028312832283328342835283628372838283928402841284228432844284528462847284828492850285128522853285428552856285728582859286028612862286328642865286628672868286928702871287228732874287528762877287828792880288128822883288428852886288728882889289028912892289328942895289628972898289929002901290229032904290529062907290829092910291129122913291429152916291729182919292029212922292329242925292629272928292929302931293229332934293529362937293829392940294129422943294429452946294729482949295029512952295329542955295629572958295929602961296229632964296529662967296829692970297129722973297429752976297729782979298029812982298329842985298629872988298929902991299229932994299529962997299829993000300130023003300430053006300730083009301030113012301330143015301630173018301930203021302230233024302530263027302830293030303130323033303430353036303730383039304030413042304330443045304630473048304930503051305230533054305530563057305830593060306130623063306430653066306730683069307030713072307330743075307630773078307930803081308230833084308530863087308830893090309130923093309430953096309730983099310031013102310331043105310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135313631373138313931403141314231433144314531463147314831493150315131523153315431553156315731583159316031613162316331643165316631673168316931703171317231733174317531763177317831793180318131823183318431853186318731883189319031913192319331943195319631973198319932003201320232033204320532063207320832093210321132123213321432153216321732183219322032213222322332243225322632273228322932303231323232333234323532363237323832393240324132423243324432453246324732483249325032513252325332543255325632573258325932603261326232633264326532663267326832693270327132723273327432753276327732783279328032813282328332843285328632873288328932903291329232933294329532963297329832993300330133023303330433053306330733083309331033113312331333143315331633173318331933203321332233233324332533263327332833293330333133323333333433353336333733383339334033413342334333443345334633473348334933503351335233533354335533563357335833593360336133623363
  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. * This is identical to the case for output below. This may occur on a packet
  83. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  84. * on the same host. The packet can be treated in the same way as
  85. * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
  86. * checksum must be filled in by the OS or the hardware.
  87. *
  88. * B. Checksumming on output.
  89. *
  90. * CHECKSUM_NONE:
  91. *
  92. * The skb was already checksummed by the protocol, or a checksum is not
  93. * required.
  94. *
  95. * CHECKSUM_PARTIAL:
  96. *
  97. * The device is required to checksum the packet as seen by hard_start_xmit()
  98. * from skb->csum_start up to the end, and to record/write the checksum at
  99. * offset skb->csum_start + skb->csum_offset.
  100. *
  101. * The device must show its capabilities in dev->features, set up at device
  102. * setup time, e.g. netdev_features.h:
  103. *
  104. * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
  105. * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
  106. * IPv4. Sigh. Vendors like this way for an unknown reason.
  107. * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
  108. * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
  109. * NETIF_F_... - Well, you get the picture.
  110. *
  111. * CHECKSUM_UNNECESSARY:
  112. *
  113. * Normally, the device will do per protocol specific checksumming. Protocol
  114. * implementations that do not want the NIC to perform the checksum
  115. * calculation should use this flag in their outgoing skbs.
  116. *
  117. * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
  118. * offload. Correspondingly, the FCoE protocol driver
  119. * stack should use CHECKSUM_UNNECESSARY.
  120. *
  121. * Any questions? No questions, good. --ANK
  122. */
  123. /* Don't change this without changing skb_csum_unnecessary! */
  124. #define CHECKSUM_NONE 0
  125. #define CHECKSUM_UNNECESSARY 1
  126. #define CHECKSUM_COMPLETE 2
  127. #define CHECKSUM_PARTIAL 3
  128. /* Maximum value in skb->csum_level */
  129. #define SKB_MAX_CSUM_LEVEL 3
  130. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  131. #define SKB_WITH_OVERHEAD(X) \
  132. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  133. #define SKB_MAX_ORDER(X, ORDER) \
  134. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  135. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  136. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  137. /* return minimum truesize of one skb containing X bytes of data */
  138. #define SKB_TRUESIZE(X) ((X) + \
  139. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  140. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  141. struct net_device;
  142. struct scatterlist;
  143. struct pipe_inode_info;
  144. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  145. struct nf_conntrack {
  146. atomic_t use;
  147. };
  148. #endif
  149. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  150. struct nf_bridge_info {
  151. atomic_t use;
  152. unsigned int mask;
  153. struct net_device *physindev;
  154. struct net_device *physoutdev;
  155. unsigned long data[32 / sizeof(unsigned long)];
  156. };
  157. #endif
  158. struct sk_buff_head {
  159. /* These two members must be first. */
  160. struct sk_buff *next;
  161. struct sk_buff *prev;
  162. __u32 qlen;
  163. spinlock_t lock;
  164. };
  165. struct sk_buff;
  166. /* To allow 64K frame to be packed as single skb without frag_list we
  167. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  168. * buffers which do not start on a page boundary.
  169. *
  170. * Since GRO uses frags we allocate at least 16 regardless of page
  171. * size.
  172. */
  173. #if (65536/PAGE_SIZE + 1) < 16
  174. #define MAX_SKB_FRAGS 16UL
  175. #else
  176. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  177. #endif
  178. typedef struct skb_frag_struct skb_frag_t;
  179. struct skb_frag_struct {
  180. struct {
  181. struct page *p;
  182. } page;
  183. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  184. __u32 page_offset;
  185. __u32 size;
  186. #else
  187. __u16 page_offset;
  188. __u16 size;
  189. #endif
  190. };
  191. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  192. {
  193. return frag->size;
  194. }
  195. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  196. {
  197. frag->size = size;
  198. }
  199. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  200. {
  201. frag->size += delta;
  202. }
  203. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  204. {
  205. frag->size -= delta;
  206. }
  207. #define HAVE_HW_TIME_STAMP
  208. /**
  209. * struct skb_shared_hwtstamps - hardware time stamps
  210. * @hwtstamp: hardware time stamp transformed into duration
  211. * since arbitrary point in time
  212. *
  213. * Software time stamps generated by ktime_get_real() are stored in
  214. * skb->tstamp.
  215. *
  216. * hwtstamps can only be compared against other hwtstamps from
  217. * the same device.
  218. *
  219. * This structure is attached to packets as part of the
  220. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  221. */
  222. struct skb_shared_hwtstamps {
  223. ktime_t hwtstamp;
  224. };
  225. /* Definitions for tx_flags in struct skb_shared_info */
  226. enum {
  227. /* generate hardware time stamp */
  228. SKBTX_HW_TSTAMP = 1 << 0,
  229. /* generate software time stamp when queueing packet to NIC */
  230. SKBTX_SW_TSTAMP = 1 << 1,
  231. /* device driver is going to provide hardware time stamp */
  232. SKBTX_IN_PROGRESS = 1 << 2,
  233. /* device driver supports TX zero-copy buffers */
  234. SKBTX_DEV_ZEROCOPY = 1 << 3,
  235. /* generate wifi status information (where possible) */
  236. SKBTX_WIFI_STATUS = 1 << 4,
  237. /* This indicates at least one fragment might be overwritten
  238. * (as in vmsplice(), sendfile() ...)
  239. * If we need to compute a TX checksum, we'll need to copy
  240. * all frags to avoid possible bad checksum
  241. */
  242. SKBTX_SHARED_FRAG = 1 << 5,
  243. /* generate software time stamp when entering packet scheduling */
  244. SKBTX_SCHED_TSTAMP = 1 << 6,
  245. /* generate software timestamp on peer data acknowledgment */
  246. SKBTX_ACK_TSTAMP = 1 << 7,
  247. };
  248. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  249. SKBTX_SCHED_TSTAMP | \
  250. SKBTX_ACK_TSTAMP)
  251. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  252. /*
  253. * The callback notifies userspace to release buffers when skb DMA is done in
  254. * lower device, the skb last reference should be 0 when calling this.
  255. * The zerocopy_success argument is true if zero copy transmit occurred,
  256. * false on data copy or out of memory error caused by data copy attempt.
  257. * The ctx field is used to track device context.
  258. * The desc field is used to track userspace buffer index.
  259. */
  260. struct ubuf_info {
  261. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  262. void *ctx;
  263. unsigned long desc;
  264. };
  265. /* This data is invariant across clones and lives at
  266. * the end of the header data, ie. at skb->end.
  267. */
  268. struct skb_shared_info {
  269. unsigned char nr_frags;
  270. __u8 tx_flags;
  271. unsigned short gso_size;
  272. /* Warning: this field is not always filled in (UFO)! */
  273. unsigned short gso_segs;
  274. unsigned short gso_type;
  275. struct sk_buff *frag_list;
  276. struct skb_shared_hwtstamps hwtstamps;
  277. u32 tskey;
  278. __be32 ip6_frag_id;
  279. /*
  280. * Warning : all fields before dataref are cleared in __alloc_skb()
  281. */
  282. atomic_t dataref;
  283. /* Intermediate layers must ensure that destructor_arg
  284. * remains valid until skb destructor */
  285. void * destructor_arg;
  286. /* must be last field, see pskb_expand_head() */
  287. skb_frag_t frags[MAX_SKB_FRAGS];
  288. };
  289. /* We divide dataref into two halves. The higher 16 bits hold references
  290. * to the payload part of skb->data. The lower 16 bits hold references to
  291. * the entire skb->data. A clone of a headerless skb holds the length of
  292. * the header in skb->hdr_len.
  293. *
  294. * All users must obey the rule that the skb->data reference count must be
  295. * greater than or equal to the payload reference count.
  296. *
  297. * Holding a reference to the payload part means that the user does not
  298. * care about modifications to the header part of skb->data.
  299. */
  300. #define SKB_DATAREF_SHIFT 16
  301. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  302. enum {
  303. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  304. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  305. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  306. SKB_FCLONE_FREE, /* this companion fclone skb is available */
  307. };
  308. enum {
  309. SKB_GSO_TCPV4 = 1 << 0,
  310. SKB_GSO_UDP = 1 << 1,
  311. /* This indicates the skb is from an untrusted source. */
  312. SKB_GSO_DODGY = 1 << 2,
  313. /* This indicates the tcp segment has CWR set. */
  314. SKB_GSO_TCP_ECN = 1 << 3,
  315. SKB_GSO_TCPV6 = 1 << 4,
  316. SKB_GSO_FCOE = 1 << 5,
  317. SKB_GSO_GRE = 1 << 6,
  318. SKB_GSO_GRE_CSUM = 1 << 7,
  319. SKB_GSO_IPIP = 1 << 8,
  320. SKB_GSO_SIT = 1 << 9,
  321. SKB_GSO_UDP_TUNNEL = 1 << 10,
  322. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  323. SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
  324. };
  325. #if BITS_PER_LONG > 32
  326. #define NET_SKBUFF_DATA_USES_OFFSET 1
  327. #endif
  328. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  329. typedef unsigned int sk_buff_data_t;
  330. #else
  331. typedef unsigned char *sk_buff_data_t;
  332. #endif
  333. /**
  334. * struct skb_mstamp - multi resolution time stamps
  335. * @stamp_us: timestamp in us resolution
  336. * @stamp_jiffies: timestamp in jiffies
  337. */
  338. struct skb_mstamp {
  339. union {
  340. u64 v64;
  341. struct {
  342. u32 stamp_us;
  343. u32 stamp_jiffies;
  344. };
  345. };
  346. };
  347. /**
  348. * skb_mstamp_get - get current timestamp
  349. * @cl: place to store timestamps
  350. */
  351. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  352. {
  353. u64 val = local_clock();
  354. do_div(val, NSEC_PER_USEC);
  355. cl->stamp_us = (u32)val;
  356. cl->stamp_jiffies = (u32)jiffies;
  357. }
  358. /**
  359. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  360. * @t1: pointer to newest sample
  361. * @t0: pointer to oldest sample
  362. */
  363. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  364. const struct skb_mstamp *t0)
  365. {
  366. s32 delta_us = t1->stamp_us - t0->stamp_us;
  367. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  368. /* If delta_us is negative, this might be because interval is too big,
  369. * or local_clock() drift is too big : fallback using jiffies.
  370. */
  371. if (delta_us <= 0 ||
  372. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  373. delta_us = jiffies_to_usecs(delta_jiffies);
  374. return delta_us;
  375. }
  376. /**
  377. * struct sk_buff - socket buffer
  378. * @next: Next buffer in list
  379. * @prev: Previous buffer in list
  380. * @tstamp: Time we arrived/left
  381. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  382. * @sk: Socket we are owned by
  383. * @dev: Device we arrived on/are leaving by
  384. * @cb: Control buffer. Free for use by every layer. Put private vars here
  385. * @_skb_refdst: destination entry (with norefcount bit)
  386. * @sp: the security path, used for xfrm
  387. * @len: Length of actual data
  388. * @data_len: Data length
  389. * @mac_len: Length of link layer header
  390. * @hdr_len: writable header length of cloned skb
  391. * @csum: Checksum (must include start/offset pair)
  392. * @csum_start: Offset from skb->head where checksumming should start
  393. * @csum_offset: Offset from csum_start where checksum should be stored
  394. * @priority: Packet queueing priority
  395. * @ignore_df: allow local fragmentation
  396. * @cloned: Head may be cloned (check refcnt to be sure)
  397. * @ip_summed: Driver fed us an IP checksum
  398. * @nohdr: Payload reference only, must not modify header
  399. * @nfctinfo: Relationship of this skb to the connection
  400. * @pkt_type: Packet class
  401. * @fclone: skbuff clone status
  402. * @ipvs_property: skbuff is owned by ipvs
  403. * @peeked: this packet has been seen already, so stats have been
  404. * done for it, don't do them again
  405. * @nf_trace: netfilter packet trace flag
  406. * @protocol: Packet protocol from driver
  407. * @destructor: Destruct function
  408. * @nfct: Associated connection, if any
  409. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  410. * @skb_iif: ifindex of device we arrived on
  411. * @tc_index: Traffic control index
  412. * @tc_verd: traffic control verdict
  413. * @hash: the packet hash
  414. * @queue_mapping: Queue mapping for multiqueue devices
  415. * @xmit_more: More SKBs are pending for this queue
  416. * @ndisc_nodetype: router type (from link layer)
  417. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  418. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  419. * ports.
  420. * @sw_hash: indicates hash was computed in software stack
  421. * @wifi_acked_valid: wifi_acked was set
  422. * @wifi_acked: whether frame was acked on wifi or not
  423. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  424. * @napi_id: id of the NAPI struct this skb came from
  425. * @secmark: security marking
  426. * @mark: Generic packet mark
  427. * @dropcount: total number of sk_receive_queue overflows
  428. * @vlan_proto: vlan encapsulation protocol
  429. * @vlan_tci: vlan tag control information
  430. * @inner_protocol: Protocol (encapsulation)
  431. * @inner_transport_header: Inner transport layer header (encapsulation)
  432. * @inner_network_header: Network layer header (encapsulation)
  433. * @inner_mac_header: Link layer header (encapsulation)
  434. * @transport_header: Transport layer header
  435. * @network_header: Network layer header
  436. * @mac_header: Link layer header
  437. * @tail: Tail pointer
  438. * @end: End pointer
  439. * @head: Head of buffer
  440. * @data: Data head pointer
  441. * @truesize: Buffer size
  442. * @users: User count - see {datagram,tcp}.c
  443. */
  444. struct sk_buff {
  445. union {
  446. struct {
  447. /* These two members must be first. */
  448. struct sk_buff *next;
  449. struct sk_buff *prev;
  450. union {
  451. ktime_t tstamp;
  452. struct skb_mstamp skb_mstamp;
  453. };
  454. };
  455. struct rb_node rbnode; /* used in netem & tcp stack */
  456. };
  457. struct sock *sk;
  458. struct net_device *dev;
  459. /*
  460. * This is the control buffer. It is free to use for every
  461. * layer. Please put your private variables there. If you
  462. * want to keep them across layers you have to do a skb_clone()
  463. * first. This is owned by whoever has the skb queued ATM.
  464. */
  465. char cb[48] __aligned(8);
  466. unsigned long _skb_refdst;
  467. void (*destructor)(struct sk_buff *skb);
  468. #ifdef CONFIG_XFRM
  469. struct sec_path *sp;
  470. #endif
  471. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  472. struct nf_conntrack *nfct;
  473. #endif
  474. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  475. struct nf_bridge_info *nf_bridge;
  476. #endif
  477. unsigned int len,
  478. data_len;
  479. __u16 mac_len,
  480. hdr_len;
  481. /* Following fields are _not_ copied in __copy_skb_header()
  482. * Note that queue_mapping is here mostly to fill a hole.
  483. */
  484. kmemcheck_bitfield_begin(flags1);
  485. __u16 queue_mapping;
  486. __u8 cloned:1,
  487. nohdr:1,
  488. fclone:2,
  489. peeked:1,
  490. head_frag:1,
  491. xmit_more:1;
  492. /* one bit hole */
  493. kmemcheck_bitfield_end(flags1);
  494. /* fields enclosed in headers_start/headers_end are copied
  495. * using a single memcpy() in __copy_skb_header()
  496. */
  497. /* private: */
  498. __u32 headers_start[0];
  499. /* public: */
  500. /* if you move pkt_type around you also must adapt those constants */
  501. #ifdef __BIG_ENDIAN_BITFIELD
  502. #define PKT_TYPE_MAX (7 << 5)
  503. #else
  504. #define PKT_TYPE_MAX 7
  505. #endif
  506. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  507. __u8 __pkt_type_offset[0];
  508. __u8 pkt_type:3;
  509. __u8 pfmemalloc:1;
  510. __u8 ignore_df:1;
  511. __u8 nfctinfo:3;
  512. __u8 nf_trace:1;
  513. __u8 ip_summed:2;
  514. __u8 ooo_okay:1;
  515. __u8 l4_hash:1;
  516. __u8 sw_hash:1;
  517. __u8 wifi_acked_valid:1;
  518. __u8 wifi_acked:1;
  519. __u8 no_fcs:1;
  520. /* Indicates the inner headers are valid in the skbuff. */
  521. __u8 encapsulation:1;
  522. __u8 encap_hdr_csum:1;
  523. __u8 csum_valid:1;
  524. __u8 csum_complete_sw:1;
  525. __u8 csum_level:2;
  526. __u8 csum_bad:1;
  527. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  528. __u8 ndisc_nodetype:2;
  529. #endif
  530. __u8 ipvs_property:1;
  531. __u8 inner_protocol_type:1;
  532. __u8 remcsum_offload:1;
  533. /* 3 or 5 bit hole */
  534. #ifdef CONFIG_NET_SCHED
  535. __u16 tc_index; /* traffic control index */
  536. #ifdef CONFIG_NET_CLS_ACT
  537. __u16 tc_verd; /* traffic control verdict */
  538. #endif
  539. #endif
  540. union {
  541. __wsum csum;
  542. struct {
  543. __u16 csum_start;
  544. __u16 csum_offset;
  545. };
  546. };
  547. __u32 priority;
  548. int skb_iif;
  549. __u32 hash;
  550. __be16 vlan_proto;
  551. __u16 vlan_tci;
  552. #ifdef CONFIG_NET_RX_BUSY_POLL
  553. unsigned int napi_id;
  554. #endif
  555. #ifdef CONFIG_NETWORK_SECMARK
  556. __u32 secmark;
  557. #endif
  558. union {
  559. __u32 mark;
  560. __u32 dropcount;
  561. __u32 reserved_tailroom;
  562. };
  563. union {
  564. __be16 inner_protocol;
  565. __u8 inner_ipproto;
  566. };
  567. __u16 inner_transport_header;
  568. __u16 inner_network_header;
  569. __u16 inner_mac_header;
  570. __be16 protocol;
  571. __u16 transport_header;
  572. __u16 network_header;
  573. __u16 mac_header;
  574. /* private: */
  575. __u32 headers_end[0];
  576. /* public: */
  577. /* These elements must be at the end, see alloc_skb() for details. */
  578. sk_buff_data_t tail;
  579. sk_buff_data_t end;
  580. unsigned char *head,
  581. *data;
  582. unsigned int truesize;
  583. atomic_t users;
  584. };
  585. #ifdef __KERNEL__
  586. /*
  587. * Handling routines are only of interest to the kernel
  588. */
  589. #include <linux/slab.h>
  590. #define SKB_ALLOC_FCLONE 0x01
  591. #define SKB_ALLOC_RX 0x02
  592. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  593. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  594. {
  595. return unlikely(skb->pfmemalloc);
  596. }
  597. /*
  598. * skb might have a dst pointer attached, refcounted or not.
  599. * _skb_refdst low order bit is set if refcount was _not_ taken
  600. */
  601. #define SKB_DST_NOREF 1UL
  602. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  603. /**
  604. * skb_dst - returns skb dst_entry
  605. * @skb: buffer
  606. *
  607. * Returns skb dst_entry, regardless of reference taken or not.
  608. */
  609. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  610. {
  611. /* If refdst was not refcounted, check we still are in a
  612. * rcu_read_lock section
  613. */
  614. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  615. !rcu_read_lock_held() &&
  616. !rcu_read_lock_bh_held());
  617. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  618. }
  619. /**
  620. * skb_dst_set - sets skb dst
  621. * @skb: buffer
  622. * @dst: dst entry
  623. *
  624. * Sets skb dst, assuming a reference was taken on dst and should
  625. * be released by skb_dst_drop()
  626. */
  627. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  628. {
  629. skb->_skb_refdst = (unsigned long)dst;
  630. }
  631. void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
  632. bool force);
  633. /**
  634. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  635. * @skb: buffer
  636. * @dst: dst entry
  637. *
  638. * Sets skb dst, assuming a reference was not taken on dst.
  639. * If dst entry is cached, we do not take reference and dst_release
  640. * will be avoided by refdst_drop. If dst entry is not cached, we take
  641. * reference, so that last dst_release can destroy the dst immediately.
  642. */
  643. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  644. {
  645. __skb_dst_set_noref(skb, dst, false);
  646. }
  647. /**
  648. * skb_dst_set_noref_force - sets skb dst, without taking reference
  649. * @skb: buffer
  650. * @dst: dst entry
  651. *
  652. * Sets skb dst, assuming a reference was not taken on dst.
  653. * No reference is taken and no dst_release will be called. While for
  654. * cached dsts deferred reclaim is a basic feature, for entries that are
  655. * not cached it is caller's job to guarantee that last dst_release for
  656. * provided dst happens when nobody uses it, eg. after a RCU grace period.
  657. */
  658. static inline void skb_dst_set_noref_force(struct sk_buff *skb,
  659. struct dst_entry *dst)
  660. {
  661. __skb_dst_set_noref(skb, dst, true);
  662. }
  663. /**
  664. * skb_dst_is_noref - Test if skb dst isn't refcounted
  665. * @skb: buffer
  666. */
  667. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  668. {
  669. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  670. }
  671. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  672. {
  673. return (struct rtable *)skb_dst(skb);
  674. }
  675. void kfree_skb(struct sk_buff *skb);
  676. void kfree_skb_list(struct sk_buff *segs);
  677. void skb_tx_error(struct sk_buff *skb);
  678. void consume_skb(struct sk_buff *skb);
  679. void __kfree_skb(struct sk_buff *skb);
  680. extern struct kmem_cache *skbuff_head_cache;
  681. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  682. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  683. bool *fragstolen, int *delta_truesize);
  684. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  685. int node);
  686. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  687. static inline struct sk_buff *alloc_skb(unsigned int size,
  688. gfp_t priority)
  689. {
  690. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  691. }
  692. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  693. unsigned long data_len,
  694. int max_page_order,
  695. int *errcode,
  696. gfp_t gfp_mask);
  697. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  698. struct sk_buff_fclones {
  699. struct sk_buff skb1;
  700. struct sk_buff skb2;
  701. atomic_t fclone_ref;
  702. };
  703. /**
  704. * skb_fclone_busy - check if fclone is busy
  705. * @skb: buffer
  706. *
  707. * Returns true is skb is a fast clone, and its clone is not freed.
  708. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  709. * so we also check that this didnt happen.
  710. */
  711. static inline bool skb_fclone_busy(const struct sock *sk,
  712. const struct sk_buff *skb)
  713. {
  714. const struct sk_buff_fclones *fclones;
  715. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  716. return skb->fclone == SKB_FCLONE_ORIG &&
  717. fclones->skb2.fclone == SKB_FCLONE_CLONE &&
  718. fclones->skb2.sk == sk;
  719. }
  720. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  721. gfp_t priority)
  722. {
  723. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  724. }
  725. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  726. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  727. {
  728. return __alloc_skb_head(priority, -1);
  729. }
  730. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  731. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  732. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  733. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  734. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  735. gfp_t gfp_mask, bool fclone);
  736. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  737. gfp_t gfp_mask)
  738. {
  739. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  740. }
  741. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  742. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  743. unsigned int headroom);
  744. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  745. int newtailroom, gfp_t priority);
  746. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  747. int offset, int len);
  748. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  749. int len);
  750. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  751. int skb_pad(struct sk_buff *skb, int pad);
  752. #define dev_kfree_skb(a) consume_skb(a)
  753. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  754. int getfrag(void *from, char *to, int offset,
  755. int len, int odd, struct sk_buff *skb),
  756. void *from, int length);
  757. struct skb_seq_state {
  758. __u32 lower_offset;
  759. __u32 upper_offset;
  760. __u32 frag_idx;
  761. __u32 stepped_offset;
  762. struct sk_buff *root_skb;
  763. struct sk_buff *cur_skb;
  764. __u8 *frag_data;
  765. };
  766. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  767. unsigned int to, struct skb_seq_state *st);
  768. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  769. struct skb_seq_state *st);
  770. void skb_abort_seq_read(struct skb_seq_state *st);
  771. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  772. unsigned int to, struct ts_config *config,
  773. struct ts_state *state);
  774. /*
  775. * Packet hash types specify the type of hash in skb_set_hash.
  776. *
  777. * Hash types refer to the protocol layer addresses which are used to
  778. * construct a packet's hash. The hashes are used to differentiate or identify
  779. * flows of the protocol layer for the hash type. Hash types are either
  780. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  781. *
  782. * Properties of hashes:
  783. *
  784. * 1) Two packets in different flows have different hash values
  785. * 2) Two packets in the same flow should have the same hash value
  786. *
  787. * A hash at a higher layer is considered to be more specific. A driver should
  788. * set the most specific hash possible.
  789. *
  790. * A driver cannot indicate a more specific hash than the layer at which a hash
  791. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  792. *
  793. * A driver may indicate a hash level which is less specific than the
  794. * actual layer the hash was computed on. For instance, a hash computed
  795. * at L4 may be considered an L3 hash. This should only be done if the
  796. * driver can't unambiguously determine that the HW computed the hash at
  797. * the higher layer. Note that the "should" in the second property above
  798. * permits this.
  799. */
  800. enum pkt_hash_types {
  801. PKT_HASH_TYPE_NONE, /* Undefined type */
  802. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  803. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  804. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  805. };
  806. static inline void
  807. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  808. {
  809. skb->l4_hash = (type == PKT_HASH_TYPE_L4);
  810. skb->sw_hash = 0;
  811. skb->hash = hash;
  812. }
  813. void __skb_get_hash(struct sk_buff *skb);
  814. static inline __u32 skb_get_hash(struct sk_buff *skb)
  815. {
  816. if (!skb->l4_hash && !skb->sw_hash)
  817. __skb_get_hash(skb);
  818. return skb->hash;
  819. }
  820. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  821. {
  822. return skb->hash;
  823. }
  824. static inline void skb_clear_hash(struct sk_buff *skb)
  825. {
  826. skb->hash = 0;
  827. skb->sw_hash = 0;
  828. skb->l4_hash = 0;
  829. }
  830. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  831. {
  832. if (!skb->l4_hash)
  833. skb_clear_hash(skb);
  834. }
  835. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  836. {
  837. to->hash = from->hash;
  838. to->sw_hash = from->sw_hash;
  839. to->l4_hash = from->l4_hash;
  840. };
  841. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  842. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  843. {
  844. return skb->head + skb->end;
  845. }
  846. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  847. {
  848. return skb->end;
  849. }
  850. #else
  851. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  852. {
  853. return skb->end;
  854. }
  855. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  856. {
  857. return skb->end - skb->head;
  858. }
  859. #endif
  860. /* Internal */
  861. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  862. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  863. {
  864. return &skb_shinfo(skb)->hwtstamps;
  865. }
  866. /**
  867. * skb_queue_empty - check if a queue is empty
  868. * @list: queue head
  869. *
  870. * Returns true if the queue is empty, false otherwise.
  871. */
  872. static inline int skb_queue_empty(const struct sk_buff_head *list)
  873. {
  874. return list->next == (const struct sk_buff *) list;
  875. }
  876. /**
  877. * skb_queue_is_last - check if skb is the last entry in the queue
  878. * @list: queue head
  879. * @skb: buffer
  880. *
  881. * Returns true if @skb is the last buffer on the list.
  882. */
  883. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  884. const struct sk_buff *skb)
  885. {
  886. return skb->next == (const struct sk_buff *) list;
  887. }
  888. /**
  889. * skb_queue_is_first - check if skb is the first entry in the queue
  890. * @list: queue head
  891. * @skb: buffer
  892. *
  893. * Returns true if @skb is the first buffer on the list.
  894. */
  895. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  896. const struct sk_buff *skb)
  897. {
  898. return skb->prev == (const struct sk_buff *) list;
  899. }
  900. /**
  901. * skb_queue_next - return the next packet in the queue
  902. * @list: queue head
  903. * @skb: current buffer
  904. *
  905. * Return the next packet in @list after @skb. It is only valid to
  906. * call this if skb_queue_is_last() evaluates to false.
  907. */
  908. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  909. const struct sk_buff *skb)
  910. {
  911. /* This BUG_ON may seem severe, but if we just return then we
  912. * are going to dereference garbage.
  913. */
  914. BUG_ON(skb_queue_is_last(list, skb));
  915. return skb->next;
  916. }
  917. /**
  918. * skb_queue_prev - return the prev packet in the queue
  919. * @list: queue head
  920. * @skb: current buffer
  921. *
  922. * Return the prev packet in @list before @skb. It is only valid to
  923. * call this if skb_queue_is_first() evaluates to false.
  924. */
  925. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  926. const struct sk_buff *skb)
  927. {
  928. /* This BUG_ON may seem severe, but if we just return then we
  929. * are going to dereference garbage.
  930. */
  931. BUG_ON(skb_queue_is_first(list, skb));
  932. return skb->prev;
  933. }
  934. /**
  935. * skb_get - reference buffer
  936. * @skb: buffer to reference
  937. *
  938. * Makes another reference to a socket buffer and returns a pointer
  939. * to the buffer.
  940. */
  941. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  942. {
  943. atomic_inc(&skb->users);
  944. return skb;
  945. }
  946. /*
  947. * If users == 1, we are the only owner and are can avoid redundant
  948. * atomic change.
  949. */
  950. /**
  951. * skb_cloned - is the buffer a clone
  952. * @skb: buffer to check
  953. *
  954. * Returns true if the buffer was generated with skb_clone() and is
  955. * one of multiple shared copies of the buffer. Cloned buffers are
  956. * shared data so must not be written to under normal circumstances.
  957. */
  958. static inline int skb_cloned(const struct sk_buff *skb)
  959. {
  960. return skb->cloned &&
  961. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  962. }
  963. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  964. {
  965. might_sleep_if(pri & __GFP_WAIT);
  966. if (skb_cloned(skb))
  967. return pskb_expand_head(skb, 0, 0, pri);
  968. return 0;
  969. }
  970. /**
  971. * skb_header_cloned - is the header a clone
  972. * @skb: buffer to check
  973. *
  974. * Returns true if modifying the header part of the buffer requires
  975. * the data to be copied.
  976. */
  977. static inline int skb_header_cloned(const struct sk_buff *skb)
  978. {
  979. int dataref;
  980. if (!skb->cloned)
  981. return 0;
  982. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  983. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  984. return dataref != 1;
  985. }
  986. /**
  987. * skb_header_release - release reference to header
  988. * @skb: buffer to operate on
  989. *
  990. * Drop a reference to the header part of the buffer. This is done
  991. * by acquiring a payload reference. You must not read from the header
  992. * part of skb->data after this.
  993. * Note : Check if you can use __skb_header_release() instead.
  994. */
  995. static inline void skb_header_release(struct sk_buff *skb)
  996. {
  997. BUG_ON(skb->nohdr);
  998. skb->nohdr = 1;
  999. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1000. }
  1001. /**
  1002. * __skb_header_release - release reference to header
  1003. * @skb: buffer to operate on
  1004. *
  1005. * Variant of skb_header_release() assuming skb is private to caller.
  1006. * We can avoid one atomic operation.
  1007. */
  1008. static inline void __skb_header_release(struct sk_buff *skb)
  1009. {
  1010. skb->nohdr = 1;
  1011. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1012. }
  1013. /**
  1014. * skb_shared - is the buffer shared
  1015. * @skb: buffer to check
  1016. *
  1017. * Returns true if more than one person has a reference to this
  1018. * buffer.
  1019. */
  1020. static inline int skb_shared(const struct sk_buff *skb)
  1021. {
  1022. return atomic_read(&skb->users) != 1;
  1023. }
  1024. /**
  1025. * skb_share_check - check if buffer is shared and if so clone it
  1026. * @skb: buffer to check
  1027. * @pri: priority for memory allocation
  1028. *
  1029. * If the buffer is shared the buffer is cloned and the old copy
  1030. * drops a reference. A new clone with a single reference is returned.
  1031. * If the buffer is not shared the original buffer is returned. When
  1032. * being called from interrupt status or with spinlocks held pri must
  1033. * be GFP_ATOMIC.
  1034. *
  1035. * NULL is returned on a memory allocation failure.
  1036. */
  1037. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1038. {
  1039. might_sleep_if(pri & __GFP_WAIT);
  1040. if (skb_shared(skb)) {
  1041. struct sk_buff *nskb = skb_clone(skb, pri);
  1042. if (likely(nskb))
  1043. consume_skb(skb);
  1044. else
  1045. kfree_skb(skb);
  1046. skb = nskb;
  1047. }
  1048. return skb;
  1049. }
  1050. /*
  1051. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1052. * packets to handle cases where we have a local reader and forward
  1053. * and a couple of other messy ones. The normal one is tcpdumping
  1054. * a packet thats being forwarded.
  1055. */
  1056. /**
  1057. * skb_unshare - make a copy of a shared buffer
  1058. * @skb: buffer to check
  1059. * @pri: priority for memory allocation
  1060. *
  1061. * If the socket buffer is a clone then this function creates a new
  1062. * copy of the data, drops a reference count on the old copy and returns
  1063. * the new copy with the reference count at 1. If the buffer is not a clone
  1064. * the original buffer is returned. When called with a spinlock held or
  1065. * from interrupt state @pri must be %GFP_ATOMIC
  1066. *
  1067. * %NULL is returned on a memory allocation failure.
  1068. */
  1069. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1070. gfp_t pri)
  1071. {
  1072. might_sleep_if(pri & __GFP_WAIT);
  1073. if (skb_cloned(skb)) {
  1074. struct sk_buff *nskb = skb_copy(skb, pri);
  1075. /* Free our shared copy */
  1076. if (likely(nskb))
  1077. consume_skb(skb);
  1078. else
  1079. kfree_skb(skb);
  1080. skb = nskb;
  1081. }
  1082. return skb;
  1083. }
  1084. /**
  1085. * skb_peek - peek at the head of an &sk_buff_head
  1086. * @list_: list to peek at
  1087. *
  1088. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1089. * be careful with this one. A peek leaves the buffer on the
  1090. * list and someone else may run off with it. You must hold
  1091. * the appropriate locks or have a private queue to do this.
  1092. *
  1093. * Returns %NULL for an empty list or a pointer to the head element.
  1094. * The reference count is not incremented and the reference is therefore
  1095. * volatile. Use with caution.
  1096. */
  1097. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1098. {
  1099. struct sk_buff *skb = list_->next;
  1100. if (skb == (struct sk_buff *)list_)
  1101. skb = NULL;
  1102. return skb;
  1103. }
  1104. /**
  1105. * skb_peek_next - peek skb following the given one from a queue
  1106. * @skb: skb to start from
  1107. * @list_: list to peek at
  1108. *
  1109. * Returns %NULL when the end of the list is met or a pointer to the
  1110. * next element. The reference count is not incremented and the
  1111. * reference is therefore volatile. Use with caution.
  1112. */
  1113. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1114. const struct sk_buff_head *list_)
  1115. {
  1116. struct sk_buff *next = skb->next;
  1117. if (next == (struct sk_buff *)list_)
  1118. next = NULL;
  1119. return next;
  1120. }
  1121. /**
  1122. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1123. * @list_: list to peek at
  1124. *
  1125. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1126. * be careful with this one. A peek leaves the buffer on the
  1127. * list and someone else may run off with it. You must hold
  1128. * the appropriate locks or have a private queue to do this.
  1129. *
  1130. * Returns %NULL for an empty list or a pointer to the tail element.
  1131. * The reference count is not incremented and the reference is therefore
  1132. * volatile. Use with caution.
  1133. */
  1134. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1135. {
  1136. struct sk_buff *skb = list_->prev;
  1137. if (skb == (struct sk_buff *)list_)
  1138. skb = NULL;
  1139. return skb;
  1140. }
  1141. /**
  1142. * skb_queue_len - get queue length
  1143. * @list_: list to measure
  1144. *
  1145. * Return the length of an &sk_buff queue.
  1146. */
  1147. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1148. {
  1149. return list_->qlen;
  1150. }
  1151. /**
  1152. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1153. * @list: queue to initialize
  1154. *
  1155. * This initializes only the list and queue length aspects of
  1156. * an sk_buff_head object. This allows to initialize the list
  1157. * aspects of an sk_buff_head without reinitializing things like
  1158. * the spinlock. It can also be used for on-stack sk_buff_head
  1159. * objects where the spinlock is known to not be used.
  1160. */
  1161. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1162. {
  1163. list->prev = list->next = (struct sk_buff *)list;
  1164. list->qlen = 0;
  1165. }
  1166. /*
  1167. * This function creates a split out lock class for each invocation;
  1168. * this is needed for now since a whole lot of users of the skb-queue
  1169. * infrastructure in drivers have different locking usage (in hardirq)
  1170. * than the networking core (in softirq only). In the long run either the
  1171. * network layer or drivers should need annotation to consolidate the
  1172. * main types of usage into 3 classes.
  1173. */
  1174. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1175. {
  1176. spin_lock_init(&list->lock);
  1177. __skb_queue_head_init(list);
  1178. }
  1179. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1180. struct lock_class_key *class)
  1181. {
  1182. skb_queue_head_init(list);
  1183. lockdep_set_class(&list->lock, class);
  1184. }
  1185. /*
  1186. * Insert an sk_buff on a list.
  1187. *
  1188. * The "__skb_xxxx()" functions are the non-atomic ones that
  1189. * can only be called with interrupts disabled.
  1190. */
  1191. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1192. struct sk_buff_head *list);
  1193. static inline void __skb_insert(struct sk_buff *newsk,
  1194. struct sk_buff *prev, struct sk_buff *next,
  1195. struct sk_buff_head *list)
  1196. {
  1197. newsk->next = next;
  1198. newsk->prev = prev;
  1199. next->prev = prev->next = newsk;
  1200. list->qlen++;
  1201. }
  1202. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1203. struct sk_buff *prev,
  1204. struct sk_buff *next)
  1205. {
  1206. struct sk_buff *first = list->next;
  1207. struct sk_buff *last = list->prev;
  1208. first->prev = prev;
  1209. prev->next = first;
  1210. last->next = next;
  1211. next->prev = last;
  1212. }
  1213. /**
  1214. * skb_queue_splice - join two skb lists, this is designed for stacks
  1215. * @list: the new list to add
  1216. * @head: the place to add it in the first list
  1217. */
  1218. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1219. struct sk_buff_head *head)
  1220. {
  1221. if (!skb_queue_empty(list)) {
  1222. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1223. head->qlen += list->qlen;
  1224. }
  1225. }
  1226. /**
  1227. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1228. * @list: the new list to add
  1229. * @head: the place to add it in the first list
  1230. *
  1231. * The list at @list is reinitialised
  1232. */
  1233. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1234. struct sk_buff_head *head)
  1235. {
  1236. if (!skb_queue_empty(list)) {
  1237. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1238. head->qlen += list->qlen;
  1239. __skb_queue_head_init(list);
  1240. }
  1241. }
  1242. /**
  1243. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1244. * @list: the new list to add
  1245. * @head: the place to add it in the first list
  1246. */
  1247. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1248. struct sk_buff_head *head)
  1249. {
  1250. if (!skb_queue_empty(list)) {
  1251. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1252. head->qlen += list->qlen;
  1253. }
  1254. }
  1255. /**
  1256. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1257. * @list: the new list to add
  1258. * @head: the place to add it in the first list
  1259. *
  1260. * Each of the lists is a queue.
  1261. * The list at @list is reinitialised
  1262. */
  1263. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1264. struct sk_buff_head *head)
  1265. {
  1266. if (!skb_queue_empty(list)) {
  1267. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1268. head->qlen += list->qlen;
  1269. __skb_queue_head_init(list);
  1270. }
  1271. }
  1272. /**
  1273. * __skb_queue_after - queue a buffer at the list head
  1274. * @list: list to use
  1275. * @prev: place after this buffer
  1276. * @newsk: buffer to queue
  1277. *
  1278. * Queue a buffer int the middle of a list. This function takes no locks
  1279. * and you must therefore hold required locks before calling it.
  1280. *
  1281. * A buffer cannot be placed on two lists at the same time.
  1282. */
  1283. static inline void __skb_queue_after(struct sk_buff_head *list,
  1284. struct sk_buff *prev,
  1285. struct sk_buff *newsk)
  1286. {
  1287. __skb_insert(newsk, prev, prev->next, list);
  1288. }
  1289. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1290. struct sk_buff_head *list);
  1291. static inline void __skb_queue_before(struct sk_buff_head *list,
  1292. struct sk_buff *next,
  1293. struct sk_buff *newsk)
  1294. {
  1295. __skb_insert(newsk, next->prev, next, list);
  1296. }
  1297. /**
  1298. * __skb_queue_head - queue a buffer at the list head
  1299. * @list: list to use
  1300. * @newsk: buffer to queue
  1301. *
  1302. * Queue a buffer at the start of a list. This function takes no locks
  1303. * and you must therefore hold required locks before calling it.
  1304. *
  1305. * A buffer cannot be placed on two lists at the same time.
  1306. */
  1307. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1308. static inline void __skb_queue_head(struct sk_buff_head *list,
  1309. struct sk_buff *newsk)
  1310. {
  1311. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1312. }
  1313. /**
  1314. * __skb_queue_tail - queue a buffer at the list tail
  1315. * @list: list to use
  1316. * @newsk: buffer to queue
  1317. *
  1318. * Queue a buffer at the end of a list. This function takes no locks
  1319. * and you must therefore hold required locks before calling it.
  1320. *
  1321. * A buffer cannot be placed on two lists at the same time.
  1322. */
  1323. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1324. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1325. struct sk_buff *newsk)
  1326. {
  1327. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1328. }
  1329. /*
  1330. * remove sk_buff from list. _Must_ be called atomically, and with
  1331. * the list known..
  1332. */
  1333. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1334. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1335. {
  1336. struct sk_buff *next, *prev;
  1337. list->qlen--;
  1338. next = skb->next;
  1339. prev = skb->prev;
  1340. skb->next = skb->prev = NULL;
  1341. next->prev = prev;
  1342. prev->next = next;
  1343. }
  1344. /**
  1345. * __skb_dequeue - remove from the head of the queue
  1346. * @list: list to dequeue from
  1347. *
  1348. * Remove the head of the list. This function does not take any locks
  1349. * so must be used with appropriate locks held only. The head item is
  1350. * returned or %NULL if the list is empty.
  1351. */
  1352. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1353. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1354. {
  1355. struct sk_buff *skb = skb_peek(list);
  1356. if (skb)
  1357. __skb_unlink(skb, list);
  1358. return skb;
  1359. }
  1360. /**
  1361. * __skb_dequeue_tail - remove from the tail of the queue
  1362. * @list: list to dequeue from
  1363. *
  1364. * Remove the tail of the list. This function does not take any locks
  1365. * so must be used with appropriate locks held only. The tail item is
  1366. * returned or %NULL if the list is empty.
  1367. */
  1368. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1369. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1370. {
  1371. struct sk_buff *skb = skb_peek_tail(list);
  1372. if (skb)
  1373. __skb_unlink(skb, list);
  1374. return skb;
  1375. }
  1376. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1377. {
  1378. return skb->data_len;
  1379. }
  1380. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1381. {
  1382. return skb->len - skb->data_len;
  1383. }
  1384. static inline int skb_pagelen(const struct sk_buff *skb)
  1385. {
  1386. int i, len = 0;
  1387. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1388. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1389. return len + skb_headlen(skb);
  1390. }
  1391. /**
  1392. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1393. * @skb: buffer containing fragment to be initialised
  1394. * @i: paged fragment index to initialise
  1395. * @page: the page to use for this fragment
  1396. * @off: the offset to the data with @page
  1397. * @size: the length of the data
  1398. *
  1399. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1400. * offset @off within @page.
  1401. *
  1402. * Does not take any additional reference on the fragment.
  1403. */
  1404. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1405. struct page *page, int off, int size)
  1406. {
  1407. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1408. /*
  1409. * Propagate page->pfmemalloc to the skb if we can. The problem is
  1410. * that not all callers have unique ownership of the page. If
  1411. * pfmemalloc is set, we check the mapping as a mapping implies
  1412. * page->index is set (index and pfmemalloc share space).
  1413. * If it's a valid mapping, we cannot use page->pfmemalloc but we
  1414. * do not lose pfmemalloc information as the pages would not be
  1415. * allocated using __GFP_MEMALLOC.
  1416. */
  1417. frag->page.p = page;
  1418. frag->page_offset = off;
  1419. skb_frag_size_set(frag, size);
  1420. page = compound_head(page);
  1421. if (page->pfmemalloc && !page->mapping)
  1422. skb->pfmemalloc = true;
  1423. }
  1424. /**
  1425. * skb_fill_page_desc - initialise a paged fragment in an skb
  1426. * @skb: buffer containing fragment to be initialised
  1427. * @i: paged fragment index to initialise
  1428. * @page: the page to use for this fragment
  1429. * @off: the offset to the data with @page
  1430. * @size: the length of the data
  1431. *
  1432. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1433. * @skb to point to @size bytes at offset @off within @page. In
  1434. * addition updates @skb such that @i is the last fragment.
  1435. *
  1436. * Does not take any additional reference on the fragment.
  1437. */
  1438. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1439. struct page *page, int off, int size)
  1440. {
  1441. __skb_fill_page_desc(skb, i, page, off, size);
  1442. skb_shinfo(skb)->nr_frags = i + 1;
  1443. }
  1444. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1445. int size, unsigned int truesize);
  1446. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1447. unsigned int truesize);
  1448. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1449. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1450. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1451. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1452. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1453. {
  1454. return skb->head + skb->tail;
  1455. }
  1456. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1457. {
  1458. skb->tail = skb->data - skb->head;
  1459. }
  1460. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1461. {
  1462. skb_reset_tail_pointer(skb);
  1463. skb->tail += offset;
  1464. }
  1465. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1466. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1467. {
  1468. return skb->tail;
  1469. }
  1470. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1471. {
  1472. skb->tail = skb->data;
  1473. }
  1474. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1475. {
  1476. skb->tail = skb->data + offset;
  1477. }
  1478. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1479. /*
  1480. * Add data to an sk_buff
  1481. */
  1482. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1483. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1484. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1485. {
  1486. unsigned char *tmp = skb_tail_pointer(skb);
  1487. SKB_LINEAR_ASSERT(skb);
  1488. skb->tail += len;
  1489. skb->len += len;
  1490. return tmp;
  1491. }
  1492. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1493. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1494. {
  1495. skb->data -= len;
  1496. skb->len += len;
  1497. return skb->data;
  1498. }
  1499. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1500. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1501. {
  1502. skb->len -= len;
  1503. BUG_ON(skb->len < skb->data_len);
  1504. return skb->data += len;
  1505. }
  1506. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1507. {
  1508. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1509. }
  1510. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1511. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1512. {
  1513. if (len > skb_headlen(skb) &&
  1514. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1515. return NULL;
  1516. skb->len -= len;
  1517. return skb->data += len;
  1518. }
  1519. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1520. {
  1521. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1522. }
  1523. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1524. {
  1525. if (likely(len <= skb_headlen(skb)))
  1526. return 1;
  1527. if (unlikely(len > skb->len))
  1528. return 0;
  1529. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1530. }
  1531. /**
  1532. * skb_headroom - bytes at buffer head
  1533. * @skb: buffer to check
  1534. *
  1535. * Return the number of bytes of free space at the head of an &sk_buff.
  1536. */
  1537. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1538. {
  1539. return skb->data - skb->head;
  1540. }
  1541. /**
  1542. * skb_tailroom - bytes at buffer end
  1543. * @skb: buffer to check
  1544. *
  1545. * Return the number of bytes of free space at the tail of an sk_buff
  1546. */
  1547. static inline int skb_tailroom(const struct sk_buff *skb)
  1548. {
  1549. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1550. }
  1551. /**
  1552. * skb_availroom - bytes at buffer end
  1553. * @skb: buffer to check
  1554. *
  1555. * Return the number of bytes of free space at the tail of an sk_buff
  1556. * allocated by sk_stream_alloc()
  1557. */
  1558. static inline int skb_availroom(const struct sk_buff *skb)
  1559. {
  1560. if (skb_is_nonlinear(skb))
  1561. return 0;
  1562. return skb->end - skb->tail - skb->reserved_tailroom;
  1563. }
  1564. /**
  1565. * skb_reserve - adjust headroom
  1566. * @skb: buffer to alter
  1567. * @len: bytes to move
  1568. *
  1569. * Increase the headroom of an empty &sk_buff by reducing the tail
  1570. * room. This is only allowed for an empty buffer.
  1571. */
  1572. static inline void skb_reserve(struct sk_buff *skb, int len)
  1573. {
  1574. skb->data += len;
  1575. skb->tail += len;
  1576. }
  1577. #define ENCAP_TYPE_ETHER 0
  1578. #define ENCAP_TYPE_IPPROTO 1
  1579. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1580. __be16 protocol)
  1581. {
  1582. skb->inner_protocol = protocol;
  1583. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1584. }
  1585. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1586. __u8 ipproto)
  1587. {
  1588. skb->inner_ipproto = ipproto;
  1589. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1590. }
  1591. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1592. {
  1593. skb->inner_mac_header = skb->mac_header;
  1594. skb->inner_network_header = skb->network_header;
  1595. skb->inner_transport_header = skb->transport_header;
  1596. }
  1597. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1598. {
  1599. skb->mac_len = skb->network_header - skb->mac_header;
  1600. }
  1601. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1602. *skb)
  1603. {
  1604. return skb->head + skb->inner_transport_header;
  1605. }
  1606. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1607. {
  1608. skb->inner_transport_header = skb->data - skb->head;
  1609. }
  1610. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1611. const int offset)
  1612. {
  1613. skb_reset_inner_transport_header(skb);
  1614. skb->inner_transport_header += offset;
  1615. }
  1616. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1617. {
  1618. return skb->head + skb->inner_network_header;
  1619. }
  1620. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1621. {
  1622. skb->inner_network_header = skb->data - skb->head;
  1623. }
  1624. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1625. const int offset)
  1626. {
  1627. skb_reset_inner_network_header(skb);
  1628. skb->inner_network_header += offset;
  1629. }
  1630. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1631. {
  1632. return skb->head + skb->inner_mac_header;
  1633. }
  1634. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1635. {
  1636. skb->inner_mac_header = skb->data - skb->head;
  1637. }
  1638. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1639. const int offset)
  1640. {
  1641. skb_reset_inner_mac_header(skb);
  1642. skb->inner_mac_header += offset;
  1643. }
  1644. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1645. {
  1646. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1647. }
  1648. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1649. {
  1650. return skb->head + skb->transport_header;
  1651. }
  1652. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1653. {
  1654. skb->transport_header = skb->data - skb->head;
  1655. }
  1656. static inline void skb_set_transport_header(struct sk_buff *skb,
  1657. const int offset)
  1658. {
  1659. skb_reset_transport_header(skb);
  1660. skb->transport_header += offset;
  1661. }
  1662. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1663. {
  1664. return skb->head + skb->network_header;
  1665. }
  1666. static inline void skb_reset_network_header(struct sk_buff *skb)
  1667. {
  1668. skb->network_header = skb->data - skb->head;
  1669. }
  1670. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1671. {
  1672. skb_reset_network_header(skb);
  1673. skb->network_header += offset;
  1674. }
  1675. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1676. {
  1677. return skb->head + skb->mac_header;
  1678. }
  1679. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1680. {
  1681. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1682. }
  1683. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1684. {
  1685. skb->mac_header = skb->data - skb->head;
  1686. }
  1687. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1688. {
  1689. skb_reset_mac_header(skb);
  1690. skb->mac_header += offset;
  1691. }
  1692. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1693. {
  1694. skb->mac_header = skb->network_header;
  1695. }
  1696. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1697. const int offset_hint)
  1698. {
  1699. struct flow_keys keys;
  1700. if (skb_transport_header_was_set(skb))
  1701. return;
  1702. else if (skb_flow_dissect(skb, &keys))
  1703. skb_set_transport_header(skb, keys.thoff);
  1704. else
  1705. skb_set_transport_header(skb, offset_hint);
  1706. }
  1707. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1708. {
  1709. if (skb_mac_header_was_set(skb)) {
  1710. const unsigned char *old_mac = skb_mac_header(skb);
  1711. skb_set_mac_header(skb, -skb->mac_len);
  1712. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1713. }
  1714. }
  1715. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1716. {
  1717. return skb->csum_start - skb_headroom(skb);
  1718. }
  1719. static inline int skb_transport_offset(const struct sk_buff *skb)
  1720. {
  1721. return skb_transport_header(skb) - skb->data;
  1722. }
  1723. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1724. {
  1725. return skb->transport_header - skb->network_header;
  1726. }
  1727. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1728. {
  1729. return skb->inner_transport_header - skb->inner_network_header;
  1730. }
  1731. static inline int skb_network_offset(const struct sk_buff *skb)
  1732. {
  1733. return skb_network_header(skb) - skb->data;
  1734. }
  1735. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1736. {
  1737. return skb_inner_network_header(skb) - skb->data;
  1738. }
  1739. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1740. {
  1741. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1742. }
  1743. /*
  1744. * CPUs often take a performance hit when accessing unaligned memory
  1745. * locations. The actual performance hit varies, it can be small if the
  1746. * hardware handles it or large if we have to take an exception and fix it
  1747. * in software.
  1748. *
  1749. * Since an ethernet header is 14 bytes network drivers often end up with
  1750. * the IP header at an unaligned offset. The IP header can be aligned by
  1751. * shifting the start of the packet by 2 bytes. Drivers should do this
  1752. * with:
  1753. *
  1754. * skb_reserve(skb, NET_IP_ALIGN);
  1755. *
  1756. * The downside to this alignment of the IP header is that the DMA is now
  1757. * unaligned. On some architectures the cost of an unaligned DMA is high
  1758. * and this cost outweighs the gains made by aligning the IP header.
  1759. *
  1760. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1761. * to be overridden.
  1762. */
  1763. #ifndef NET_IP_ALIGN
  1764. #define NET_IP_ALIGN 2
  1765. #endif
  1766. /*
  1767. * The networking layer reserves some headroom in skb data (via
  1768. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1769. * the header has to grow. In the default case, if the header has to grow
  1770. * 32 bytes or less we avoid the reallocation.
  1771. *
  1772. * Unfortunately this headroom changes the DMA alignment of the resulting
  1773. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1774. * on some architectures. An architecture can override this value,
  1775. * perhaps setting it to a cacheline in size (since that will maintain
  1776. * cacheline alignment of the DMA). It must be a power of 2.
  1777. *
  1778. * Various parts of the networking layer expect at least 32 bytes of
  1779. * headroom, you should not reduce this.
  1780. *
  1781. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1782. * to reduce average number of cache lines per packet.
  1783. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1784. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1785. */
  1786. #ifndef NET_SKB_PAD
  1787. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1788. #endif
  1789. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1790. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1791. {
  1792. if (unlikely(skb_is_nonlinear(skb))) {
  1793. WARN_ON(1);
  1794. return;
  1795. }
  1796. skb->len = len;
  1797. skb_set_tail_pointer(skb, len);
  1798. }
  1799. void skb_trim(struct sk_buff *skb, unsigned int len);
  1800. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1801. {
  1802. if (skb->data_len)
  1803. return ___pskb_trim(skb, len);
  1804. __skb_trim(skb, len);
  1805. return 0;
  1806. }
  1807. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1808. {
  1809. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1810. }
  1811. /**
  1812. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1813. * @skb: buffer to alter
  1814. * @len: new length
  1815. *
  1816. * This is identical to pskb_trim except that the caller knows that
  1817. * the skb is not cloned so we should never get an error due to out-
  1818. * of-memory.
  1819. */
  1820. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  1821. {
  1822. int err = pskb_trim(skb, len);
  1823. BUG_ON(err);
  1824. }
  1825. /**
  1826. * skb_orphan - orphan a buffer
  1827. * @skb: buffer to orphan
  1828. *
  1829. * If a buffer currently has an owner then we call the owner's
  1830. * destructor function and make the @skb unowned. The buffer continues
  1831. * to exist but is no longer charged to its former owner.
  1832. */
  1833. static inline void skb_orphan(struct sk_buff *skb)
  1834. {
  1835. if (skb->destructor) {
  1836. skb->destructor(skb);
  1837. skb->destructor = NULL;
  1838. skb->sk = NULL;
  1839. } else {
  1840. BUG_ON(skb->sk);
  1841. }
  1842. }
  1843. /**
  1844. * skb_orphan_frags - orphan the frags contained in a buffer
  1845. * @skb: buffer to orphan frags from
  1846. * @gfp_mask: allocation mask for replacement pages
  1847. *
  1848. * For each frag in the SKB which needs a destructor (i.e. has an
  1849. * owner) create a copy of that frag and release the original
  1850. * page by calling the destructor.
  1851. */
  1852. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  1853. {
  1854. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  1855. return 0;
  1856. return skb_copy_ubufs(skb, gfp_mask);
  1857. }
  1858. /**
  1859. * __skb_queue_purge - empty a list
  1860. * @list: list to empty
  1861. *
  1862. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  1863. * the list and one reference dropped. This function does not take the
  1864. * list lock and the caller must hold the relevant locks to use it.
  1865. */
  1866. void skb_queue_purge(struct sk_buff_head *list);
  1867. static inline void __skb_queue_purge(struct sk_buff_head *list)
  1868. {
  1869. struct sk_buff *skb;
  1870. while ((skb = __skb_dequeue(list)) != NULL)
  1871. kfree_skb(skb);
  1872. }
  1873. #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
  1874. #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
  1875. #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
  1876. void *netdev_alloc_frag(unsigned int fragsz);
  1877. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  1878. gfp_t gfp_mask);
  1879. /**
  1880. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  1881. * @dev: network device to receive on
  1882. * @length: length to allocate
  1883. *
  1884. * Allocate a new &sk_buff and assign it a usage count of one. The
  1885. * buffer has unspecified headroom built in. Users should allocate
  1886. * the headroom they think they need without accounting for the
  1887. * built in space. The built in space is used for optimisations.
  1888. *
  1889. * %NULL is returned if there is no free memory. Although this function
  1890. * allocates memory it can be called from an interrupt.
  1891. */
  1892. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  1893. unsigned int length)
  1894. {
  1895. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  1896. }
  1897. /* legacy helper around __netdev_alloc_skb() */
  1898. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  1899. gfp_t gfp_mask)
  1900. {
  1901. return __netdev_alloc_skb(NULL, length, gfp_mask);
  1902. }
  1903. /* legacy helper around netdev_alloc_skb() */
  1904. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  1905. {
  1906. return netdev_alloc_skb(NULL, length);
  1907. }
  1908. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  1909. unsigned int length, gfp_t gfp)
  1910. {
  1911. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  1912. if (NET_IP_ALIGN && skb)
  1913. skb_reserve(skb, NET_IP_ALIGN);
  1914. return skb;
  1915. }
  1916. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  1917. unsigned int length)
  1918. {
  1919. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  1920. }
  1921. /**
  1922. * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
  1923. * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
  1924. * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
  1925. * @order: size of the allocation
  1926. *
  1927. * Allocate a new page.
  1928. *
  1929. * %NULL is returned if there is no free memory.
  1930. */
  1931. static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
  1932. struct sk_buff *skb,
  1933. unsigned int order)
  1934. {
  1935. struct page *page;
  1936. gfp_mask |= __GFP_COLD;
  1937. if (!(gfp_mask & __GFP_NOMEMALLOC))
  1938. gfp_mask |= __GFP_MEMALLOC;
  1939. page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  1940. if (skb && page && page->pfmemalloc)
  1941. skb->pfmemalloc = true;
  1942. return page;
  1943. }
  1944. /**
  1945. * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
  1946. * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
  1947. * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
  1948. *
  1949. * Allocate a new page.
  1950. *
  1951. * %NULL is returned if there is no free memory.
  1952. */
  1953. static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
  1954. struct sk_buff *skb)
  1955. {
  1956. return __skb_alloc_pages(gfp_mask, skb, 0);
  1957. }
  1958. /**
  1959. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  1960. * @page: The page that was allocated from skb_alloc_page
  1961. * @skb: The skb that may need pfmemalloc set
  1962. */
  1963. static inline void skb_propagate_pfmemalloc(struct page *page,
  1964. struct sk_buff *skb)
  1965. {
  1966. if (page && page->pfmemalloc)
  1967. skb->pfmemalloc = true;
  1968. }
  1969. /**
  1970. * skb_frag_page - retrieve the page referred to by a paged fragment
  1971. * @frag: the paged fragment
  1972. *
  1973. * Returns the &struct page associated with @frag.
  1974. */
  1975. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  1976. {
  1977. return frag->page.p;
  1978. }
  1979. /**
  1980. * __skb_frag_ref - take an addition reference on a paged fragment.
  1981. * @frag: the paged fragment
  1982. *
  1983. * Takes an additional reference on the paged fragment @frag.
  1984. */
  1985. static inline void __skb_frag_ref(skb_frag_t *frag)
  1986. {
  1987. get_page(skb_frag_page(frag));
  1988. }
  1989. /**
  1990. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  1991. * @skb: the buffer
  1992. * @f: the fragment offset.
  1993. *
  1994. * Takes an additional reference on the @f'th paged fragment of @skb.
  1995. */
  1996. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  1997. {
  1998. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  1999. }
  2000. /**
  2001. * __skb_frag_unref - release a reference on a paged fragment.
  2002. * @frag: the paged fragment
  2003. *
  2004. * Releases a reference on the paged fragment @frag.
  2005. */
  2006. static inline void __skb_frag_unref(skb_frag_t *frag)
  2007. {
  2008. put_page(skb_frag_page(frag));
  2009. }
  2010. /**
  2011. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2012. * @skb: the buffer
  2013. * @f: the fragment offset
  2014. *
  2015. * Releases a reference on the @f'th paged fragment of @skb.
  2016. */
  2017. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2018. {
  2019. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2020. }
  2021. /**
  2022. * skb_frag_address - gets the address of the data contained in a paged fragment
  2023. * @frag: the paged fragment buffer
  2024. *
  2025. * Returns the address of the data within @frag. The page must already
  2026. * be mapped.
  2027. */
  2028. static inline void *skb_frag_address(const skb_frag_t *frag)
  2029. {
  2030. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2031. }
  2032. /**
  2033. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2034. * @frag: the paged fragment buffer
  2035. *
  2036. * Returns the address of the data within @frag. Checks that the page
  2037. * is mapped and returns %NULL otherwise.
  2038. */
  2039. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2040. {
  2041. void *ptr = page_address(skb_frag_page(frag));
  2042. if (unlikely(!ptr))
  2043. return NULL;
  2044. return ptr + frag->page_offset;
  2045. }
  2046. /**
  2047. * __skb_frag_set_page - sets the page contained in a paged fragment
  2048. * @frag: the paged fragment
  2049. * @page: the page to set
  2050. *
  2051. * Sets the fragment @frag to contain @page.
  2052. */
  2053. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2054. {
  2055. frag->page.p = page;
  2056. }
  2057. /**
  2058. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2059. * @skb: the buffer
  2060. * @f: the fragment offset
  2061. * @page: the page to set
  2062. *
  2063. * Sets the @f'th fragment of @skb to contain @page.
  2064. */
  2065. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2066. struct page *page)
  2067. {
  2068. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2069. }
  2070. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2071. /**
  2072. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2073. * @dev: the device to map the fragment to
  2074. * @frag: the paged fragment to map
  2075. * @offset: the offset within the fragment (starting at the
  2076. * fragment's own offset)
  2077. * @size: the number of bytes to map
  2078. * @dir: the direction of the mapping (%PCI_DMA_*)
  2079. *
  2080. * Maps the page associated with @frag to @device.
  2081. */
  2082. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2083. const skb_frag_t *frag,
  2084. size_t offset, size_t size,
  2085. enum dma_data_direction dir)
  2086. {
  2087. return dma_map_page(dev, skb_frag_page(frag),
  2088. frag->page_offset + offset, size, dir);
  2089. }
  2090. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2091. gfp_t gfp_mask)
  2092. {
  2093. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2094. }
  2095. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2096. gfp_t gfp_mask)
  2097. {
  2098. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2099. }
  2100. /**
  2101. * skb_clone_writable - is the header of a clone writable
  2102. * @skb: buffer to check
  2103. * @len: length up to which to write
  2104. *
  2105. * Returns true if modifying the header part of the cloned buffer
  2106. * does not requires the data to be copied.
  2107. */
  2108. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2109. {
  2110. return !skb_header_cloned(skb) &&
  2111. skb_headroom(skb) + len <= skb->hdr_len;
  2112. }
  2113. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2114. int cloned)
  2115. {
  2116. int delta = 0;
  2117. if (headroom > skb_headroom(skb))
  2118. delta = headroom - skb_headroom(skb);
  2119. if (delta || cloned)
  2120. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2121. GFP_ATOMIC);
  2122. return 0;
  2123. }
  2124. /**
  2125. * skb_cow - copy header of skb when it is required
  2126. * @skb: buffer to cow
  2127. * @headroom: needed headroom
  2128. *
  2129. * If the skb passed lacks sufficient headroom or its data part
  2130. * is shared, data is reallocated. If reallocation fails, an error
  2131. * is returned and original skb is not changed.
  2132. *
  2133. * The result is skb with writable area skb->head...skb->tail
  2134. * and at least @headroom of space at head.
  2135. */
  2136. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2137. {
  2138. return __skb_cow(skb, headroom, skb_cloned(skb));
  2139. }
  2140. /**
  2141. * skb_cow_head - skb_cow but only making the head writable
  2142. * @skb: buffer to cow
  2143. * @headroom: needed headroom
  2144. *
  2145. * This function is identical to skb_cow except that we replace the
  2146. * skb_cloned check by skb_header_cloned. It should be used when
  2147. * you only need to push on some header and do not need to modify
  2148. * the data.
  2149. */
  2150. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2151. {
  2152. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2153. }
  2154. /**
  2155. * skb_padto - pad an skbuff up to a minimal size
  2156. * @skb: buffer to pad
  2157. * @len: minimal length
  2158. *
  2159. * Pads up a buffer to ensure the trailing bytes exist and are
  2160. * blanked. If the buffer already contains sufficient data it
  2161. * is untouched. Otherwise it is extended. Returns zero on
  2162. * success. The skb is freed on error.
  2163. */
  2164. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2165. {
  2166. unsigned int size = skb->len;
  2167. if (likely(size >= len))
  2168. return 0;
  2169. return skb_pad(skb, len - size);
  2170. }
  2171. static inline int skb_add_data(struct sk_buff *skb,
  2172. char __user *from, int copy)
  2173. {
  2174. const int off = skb->len;
  2175. if (skb->ip_summed == CHECKSUM_NONE) {
  2176. int err = 0;
  2177. __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
  2178. copy, 0, &err);
  2179. if (!err) {
  2180. skb->csum = csum_block_add(skb->csum, csum, off);
  2181. return 0;
  2182. }
  2183. } else if (!copy_from_user(skb_put(skb, copy), from, copy))
  2184. return 0;
  2185. __skb_trim(skb, off);
  2186. return -EFAULT;
  2187. }
  2188. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2189. const struct page *page, int off)
  2190. {
  2191. if (i) {
  2192. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2193. return page == skb_frag_page(frag) &&
  2194. off == frag->page_offset + skb_frag_size(frag);
  2195. }
  2196. return false;
  2197. }
  2198. static inline int __skb_linearize(struct sk_buff *skb)
  2199. {
  2200. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2201. }
  2202. /**
  2203. * skb_linearize - convert paged skb to linear one
  2204. * @skb: buffer to linarize
  2205. *
  2206. * If there is no free memory -ENOMEM is returned, otherwise zero
  2207. * is returned and the old skb data released.
  2208. */
  2209. static inline int skb_linearize(struct sk_buff *skb)
  2210. {
  2211. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2212. }
  2213. /**
  2214. * skb_has_shared_frag - can any frag be overwritten
  2215. * @skb: buffer to test
  2216. *
  2217. * Return true if the skb has at least one frag that might be modified
  2218. * by an external entity (as in vmsplice()/sendfile())
  2219. */
  2220. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2221. {
  2222. return skb_is_nonlinear(skb) &&
  2223. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2224. }
  2225. /**
  2226. * skb_linearize_cow - make sure skb is linear and writable
  2227. * @skb: buffer to process
  2228. *
  2229. * If there is no free memory -ENOMEM is returned, otherwise zero
  2230. * is returned and the old skb data released.
  2231. */
  2232. static inline int skb_linearize_cow(struct sk_buff *skb)
  2233. {
  2234. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2235. __skb_linearize(skb) : 0;
  2236. }
  2237. /**
  2238. * skb_postpull_rcsum - update checksum for received skb after pull
  2239. * @skb: buffer to update
  2240. * @start: start of data before pull
  2241. * @len: length of data pulled
  2242. *
  2243. * After doing a pull on a received packet, you need to call this to
  2244. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2245. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2246. */
  2247. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2248. const void *start, unsigned int len)
  2249. {
  2250. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2251. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2252. }
  2253. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2254. /**
  2255. * pskb_trim_rcsum - trim received skb and update checksum
  2256. * @skb: buffer to trim
  2257. * @len: new length
  2258. *
  2259. * This is exactly the same as pskb_trim except that it ensures the
  2260. * checksum of received packets are still valid after the operation.
  2261. */
  2262. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2263. {
  2264. if (likely(len >= skb->len))
  2265. return 0;
  2266. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2267. skb->ip_summed = CHECKSUM_NONE;
  2268. return __pskb_trim(skb, len);
  2269. }
  2270. #define skb_queue_walk(queue, skb) \
  2271. for (skb = (queue)->next; \
  2272. skb != (struct sk_buff *)(queue); \
  2273. skb = skb->next)
  2274. #define skb_queue_walk_safe(queue, skb, tmp) \
  2275. for (skb = (queue)->next, tmp = skb->next; \
  2276. skb != (struct sk_buff *)(queue); \
  2277. skb = tmp, tmp = skb->next)
  2278. #define skb_queue_walk_from(queue, skb) \
  2279. for (; skb != (struct sk_buff *)(queue); \
  2280. skb = skb->next)
  2281. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2282. for (tmp = skb->next; \
  2283. skb != (struct sk_buff *)(queue); \
  2284. skb = tmp, tmp = skb->next)
  2285. #define skb_queue_reverse_walk(queue, skb) \
  2286. for (skb = (queue)->prev; \
  2287. skb != (struct sk_buff *)(queue); \
  2288. skb = skb->prev)
  2289. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2290. for (skb = (queue)->prev, tmp = skb->prev; \
  2291. skb != (struct sk_buff *)(queue); \
  2292. skb = tmp, tmp = skb->prev)
  2293. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2294. for (tmp = skb->prev; \
  2295. skb != (struct sk_buff *)(queue); \
  2296. skb = tmp, tmp = skb->prev)
  2297. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2298. {
  2299. return skb_shinfo(skb)->frag_list != NULL;
  2300. }
  2301. static inline void skb_frag_list_init(struct sk_buff *skb)
  2302. {
  2303. skb_shinfo(skb)->frag_list = NULL;
  2304. }
  2305. static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
  2306. {
  2307. frag->next = skb_shinfo(skb)->frag_list;
  2308. skb_shinfo(skb)->frag_list = frag;
  2309. }
  2310. #define skb_walk_frags(skb, iter) \
  2311. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2312. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2313. int *peeked, int *off, int *err);
  2314. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2315. int *err);
  2316. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2317. struct poll_table_struct *wait);
  2318. int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
  2319. struct iovec *to, int size);
  2320. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2321. struct msghdr *msg, int size)
  2322. {
  2323. return skb_copy_datagram_iovec(from, offset, msg->msg_iov, size);
  2324. }
  2325. int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
  2326. struct iovec *iov);
  2327. int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
  2328. const struct iovec *from, int from_offset,
  2329. int len);
  2330. int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
  2331. int offset, size_t count);
  2332. int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
  2333. const struct iovec *to, int to_offset,
  2334. int size);
  2335. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2336. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2337. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2338. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2339. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2340. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2341. int len, __wsum csum);
  2342. int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
  2343. struct pipe_inode_info *pipe, unsigned int len,
  2344. unsigned int flags);
  2345. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2346. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2347. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2348. int len, int hlen);
  2349. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2350. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2351. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2352. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2353. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2354. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2355. struct skb_checksum_ops {
  2356. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2357. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2358. };
  2359. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2360. __wsum csum, const struct skb_checksum_ops *ops);
  2361. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2362. __wsum csum);
  2363. static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
  2364. int len, void *data, int hlen, void *buffer)
  2365. {
  2366. if (hlen - offset >= len)
  2367. return data + offset;
  2368. if (!skb ||
  2369. skb_copy_bits(skb, offset, buffer, len) < 0)
  2370. return NULL;
  2371. return buffer;
  2372. }
  2373. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  2374. int len, void *buffer)
  2375. {
  2376. return __skb_header_pointer(skb, offset, len, skb->data,
  2377. skb_headlen(skb), buffer);
  2378. }
  2379. /**
  2380. * skb_needs_linearize - check if we need to linearize a given skb
  2381. * depending on the given device features.
  2382. * @skb: socket buffer to check
  2383. * @features: net device features
  2384. *
  2385. * Returns true if either:
  2386. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2387. * 2. skb is fragmented and the device does not support SG.
  2388. */
  2389. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2390. netdev_features_t features)
  2391. {
  2392. return skb_is_nonlinear(skb) &&
  2393. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2394. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2395. }
  2396. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2397. void *to,
  2398. const unsigned int len)
  2399. {
  2400. memcpy(to, skb->data, len);
  2401. }
  2402. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2403. const int offset, void *to,
  2404. const unsigned int len)
  2405. {
  2406. memcpy(to, skb->data + offset, len);
  2407. }
  2408. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2409. const void *from,
  2410. const unsigned int len)
  2411. {
  2412. memcpy(skb->data, from, len);
  2413. }
  2414. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2415. const int offset,
  2416. const void *from,
  2417. const unsigned int len)
  2418. {
  2419. memcpy(skb->data + offset, from, len);
  2420. }
  2421. void skb_init(void);
  2422. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2423. {
  2424. return skb->tstamp;
  2425. }
  2426. /**
  2427. * skb_get_timestamp - get timestamp from a skb
  2428. * @skb: skb to get stamp from
  2429. * @stamp: pointer to struct timeval to store stamp in
  2430. *
  2431. * Timestamps are stored in the skb as offsets to a base timestamp.
  2432. * This function converts the offset back to a struct timeval and stores
  2433. * it in stamp.
  2434. */
  2435. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2436. struct timeval *stamp)
  2437. {
  2438. *stamp = ktime_to_timeval(skb->tstamp);
  2439. }
  2440. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2441. struct timespec *stamp)
  2442. {
  2443. *stamp = ktime_to_timespec(skb->tstamp);
  2444. }
  2445. static inline void __net_timestamp(struct sk_buff *skb)
  2446. {
  2447. skb->tstamp = ktime_get_real();
  2448. }
  2449. static inline ktime_t net_timedelta(ktime_t t)
  2450. {
  2451. return ktime_sub(ktime_get_real(), t);
  2452. }
  2453. static inline ktime_t net_invalid_timestamp(void)
  2454. {
  2455. return ktime_set(0, 0);
  2456. }
  2457. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2458. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2459. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2460. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2461. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2462. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2463. {
  2464. }
  2465. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2466. {
  2467. return false;
  2468. }
  2469. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2470. /**
  2471. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2472. *
  2473. * PHY drivers may accept clones of transmitted packets for
  2474. * timestamping via their phy_driver.txtstamp method. These drivers
  2475. * must call this function to return the skb back to the stack, with
  2476. * or without a timestamp.
  2477. *
  2478. * @skb: clone of the the original outgoing packet
  2479. * @hwtstamps: hardware time stamps, may be NULL if not available
  2480. *
  2481. */
  2482. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2483. struct skb_shared_hwtstamps *hwtstamps);
  2484. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2485. struct skb_shared_hwtstamps *hwtstamps,
  2486. struct sock *sk, int tstype);
  2487. /**
  2488. * skb_tstamp_tx - queue clone of skb with send time stamps
  2489. * @orig_skb: the original outgoing packet
  2490. * @hwtstamps: hardware time stamps, may be NULL if not available
  2491. *
  2492. * If the skb has a socket associated, then this function clones the
  2493. * skb (thus sharing the actual data and optional structures), stores
  2494. * the optional hardware time stamping information (if non NULL) or
  2495. * generates a software time stamp (otherwise), then queues the clone
  2496. * to the error queue of the socket. Errors are silently ignored.
  2497. */
  2498. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2499. struct skb_shared_hwtstamps *hwtstamps);
  2500. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2501. {
  2502. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2503. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2504. skb_tstamp_tx(skb, NULL);
  2505. }
  2506. /**
  2507. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2508. *
  2509. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2510. * function immediately before giving the sk_buff to the MAC hardware.
  2511. *
  2512. * Specifically, one should make absolutely sure that this function is
  2513. * called before TX completion of this packet can trigger. Otherwise
  2514. * the packet could potentially already be freed.
  2515. *
  2516. * @skb: A socket buffer.
  2517. */
  2518. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2519. {
  2520. skb_clone_tx_timestamp(skb);
  2521. sw_tx_timestamp(skb);
  2522. }
  2523. /**
  2524. * skb_complete_wifi_ack - deliver skb with wifi status
  2525. *
  2526. * @skb: the original outgoing packet
  2527. * @acked: ack status
  2528. *
  2529. */
  2530. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2531. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2532. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2533. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2534. {
  2535. return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
  2536. }
  2537. /**
  2538. * skb_checksum_complete - Calculate checksum of an entire packet
  2539. * @skb: packet to process
  2540. *
  2541. * This function calculates the checksum over the entire packet plus
  2542. * the value of skb->csum. The latter can be used to supply the
  2543. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2544. * checksum.
  2545. *
  2546. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2547. * this function can be used to verify that checksum on received
  2548. * packets. In that case the function should return zero if the
  2549. * checksum is correct. In particular, this function will return zero
  2550. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2551. * hardware has already verified the correctness of the checksum.
  2552. */
  2553. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2554. {
  2555. return skb_csum_unnecessary(skb) ?
  2556. 0 : __skb_checksum_complete(skb);
  2557. }
  2558. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2559. {
  2560. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2561. if (skb->csum_level == 0)
  2562. skb->ip_summed = CHECKSUM_NONE;
  2563. else
  2564. skb->csum_level--;
  2565. }
  2566. }
  2567. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2568. {
  2569. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2570. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2571. skb->csum_level++;
  2572. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2573. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2574. skb->csum_level = 0;
  2575. }
  2576. }
  2577. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2578. {
  2579. /* Mark current checksum as bad (typically called from GRO
  2580. * path). In the case that ip_summed is CHECKSUM_NONE
  2581. * this must be the first checksum encountered in the packet.
  2582. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2583. * checksum after the last one validated. For UDP, a zero
  2584. * checksum can not be marked as bad.
  2585. */
  2586. if (skb->ip_summed == CHECKSUM_NONE ||
  2587. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2588. skb->csum_bad = 1;
  2589. }
  2590. /* Check if we need to perform checksum complete validation.
  2591. *
  2592. * Returns true if checksum complete is needed, false otherwise
  2593. * (either checksum is unnecessary or zero checksum is allowed).
  2594. */
  2595. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2596. bool zero_okay,
  2597. __sum16 check)
  2598. {
  2599. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2600. skb->csum_valid = 1;
  2601. __skb_decr_checksum_unnecessary(skb);
  2602. return false;
  2603. }
  2604. return true;
  2605. }
  2606. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2607. * in checksum_init.
  2608. */
  2609. #define CHECKSUM_BREAK 76
  2610. /* Validate (init) checksum based on checksum complete.
  2611. *
  2612. * Return values:
  2613. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2614. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2615. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2616. * non-zero: value of invalid checksum
  2617. *
  2618. */
  2619. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2620. bool complete,
  2621. __wsum psum)
  2622. {
  2623. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2624. if (!csum_fold(csum_add(psum, skb->csum))) {
  2625. skb->csum_valid = 1;
  2626. return 0;
  2627. }
  2628. } else if (skb->csum_bad) {
  2629. /* ip_summed == CHECKSUM_NONE in this case */
  2630. return 1;
  2631. }
  2632. skb->csum = psum;
  2633. if (complete || skb->len <= CHECKSUM_BREAK) {
  2634. __sum16 csum;
  2635. csum = __skb_checksum_complete(skb);
  2636. skb->csum_valid = !csum;
  2637. return csum;
  2638. }
  2639. return 0;
  2640. }
  2641. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2642. {
  2643. return 0;
  2644. }
  2645. /* Perform checksum validate (init). Note that this is a macro since we only
  2646. * want to calculate the pseudo header which is an input function if necessary.
  2647. * First we try to validate without any computation (checksum unnecessary) and
  2648. * then calculate based on checksum complete calling the function to compute
  2649. * pseudo header.
  2650. *
  2651. * Return values:
  2652. * 0: checksum is validated or try to in skb_checksum_complete
  2653. * non-zero: value of invalid checksum
  2654. */
  2655. #define __skb_checksum_validate(skb, proto, complete, \
  2656. zero_okay, check, compute_pseudo) \
  2657. ({ \
  2658. __sum16 __ret = 0; \
  2659. skb->csum_valid = 0; \
  2660. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2661. __ret = __skb_checksum_validate_complete(skb, \
  2662. complete, compute_pseudo(skb, proto)); \
  2663. __ret; \
  2664. })
  2665. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2666. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2667. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2668. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2669. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2670. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2671. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2672. compute_pseudo) \
  2673. __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
  2674. #define skb_checksum_simple_validate(skb) \
  2675. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2676. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2677. {
  2678. return (skb->ip_summed == CHECKSUM_NONE &&
  2679. skb->csum_valid && !skb->csum_bad);
  2680. }
  2681. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2682. __sum16 check, __wsum pseudo)
  2683. {
  2684. skb->csum = ~pseudo;
  2685. skb->ip_summed = CHECKSUM_COMPLETE;
  2686. }
  2687. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  2688. do { \
  2689. if (__skb_checksum_convert_check(skb)) \
  2690. __skb_checksum_convert(skb, check, \
  2691. compute_pseudo(skb, proto)); \
  2692. } while (0)
  2693. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2694. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2695. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2696. {
  2697. if (nfct && atomic_dec_and_test(&nfct->use))
  2698. nf_conntrack_destroy(nfct);
  2699. }
  2700. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2701. {
  2702. if (nfct)
  2703. atomic_inc(&nfct->use);
  2704. }
  2705. #endif
  2706. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2707. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  2708. {
  2709. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  2710. kfree(nf_bridge);
  2711. }
  2712. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  2713. {
  2714. if (nf_bridge)
  2715. atomic_inc(&nf_bridge->use);
  2716. }
  2717. #endif /* CONFIG_BRIDGE_NETFILTER */
  2718. static inline void nf_reset(struct sk_buff *skb)
  2719. {
  2720. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2721. nf_conntrack_put(skb->nfct);
  2722. skb->nfct = NULL;
  2723. #endif
  2724. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2725. nf_bridge_put(skb->nf_bridge);
  2726. skb->nf_bridge = NULL;
  2727. #endif
  2728. }
  2729. static inline void nf_reset_trace(struct sk_buff *skb)
  2730. {
  2731. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2732. skb->nf_trace = 0;
  2733. #endif
  2734. }
  2735. /* Note: This doesn't put any conntrack and bridge info in dst. */
  2736. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  2737. bool copy)
  2738. {
  2739. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2740. dst->nfct = src->nfct;
  2741. nf_conntrack_get(src->nfct);
  2742. if (copy)
  2743. dst->nfctinfo = src->nfctinfo;
  2744. #endif
  2745. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2746. dst->nf_bridge = src->nf_bridge;
  2747. nf_bridge_get(src->nf_bridge);
  2748. #endif
  2749. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2750. if (copy)
  2751. dst->nf_trace = src->nf_trace;
  2752. #endif
  2753. }
  2754. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2755. {
  2756. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2757. nf_conntrack_put(dst->nfct);
  2758. #endif
  2759. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2760. nf_bridge_put(dst->nf_bridge);
  2761. #endif
  2762. __nf_copy(dst, src, true);
  2763. }
  2764. #ifdef CONFIG_NETWORK_SECMARK
  2765. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2766. {
  2767. to->secmark = from->secmark;
  2768. }
  2769. static inline void skb_init_secmark(struct sk_buff *skb)
  2770. {
  2771. skb->secmark = 0;
  2772. }
  2773. #else
  2774. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2775. { }
  2776. static inline void skb_init_secmark(struct sk_buff *skb)
  2777. { }
  2778. #endif
  2779. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  2780. {
  2781. return !skb->destructor &&
  2782. #if IS_ENABLED(CONFIG_XFRM)
  2783. !skb->sp &&
  2784. #endif
  2785. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  2786. !skb->nfct &&
  2787. #endif
  2788. !skb->_skb_refdst &&
  2789. !skb_has_frag_list(skb);
  2790. }
  2791. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  2792. {
  2793. skb->queue_mapping = queue_mapping;
  2794. }
  2795. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  2796. {
  2797. return skb->queue_mapping;
  2798. }
  2799. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  2800. {
  2801. to->queue_mapping = from->queue_mapping;
  2802. }
  2803. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  2804. {
  2805. skb->queue_mapping = rx_queue + 1;
  2806. }
  2807. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  2808. {
  2809. return skb->queue_mapping - 1;
  2810. }
  2811. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  2812. {
  2813. return skb->queue_mapping != 0;
  2814. }
  2815. u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
  2816. unsigned int num_tx_queues);
  2817. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  2818. {
  2819. #ifdef CONFIG_XFRM
  2820. return skb->sp;
  2821. #else
  2822. return NULL;
  2823. #endif
  2824. }
  2825. /* Keeps track of mac header offset relative to skb->head.
  2826. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  2827. * For non-tunnel skb it points to skb_mac_header() and for
  2828. * tunnel skb it points to outer mac header.
  2829. * Keeps track of level of encapsulation of network headers.
  2830. */
  2831. struct skb_gso_cb {
  2832. int mac_offset;
  2833. int encap_level;
  2834. __u16 csum_start;
  2835. };
  2836. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
  2837. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  2838. {
  2839. return (skb_mac_header(inner_skb) - inner_skb->head) -
  2840. SKB_GSO_CB(inner_skb)->mac_offset;
  2841. }
  2842. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  2843. {
  2844. int new_headroom, headroom;
  2845. int ret;
  2846. headroom = skb_headroom(skb);
  2847. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  2848. if (ret)
  2849. return ret;
  2850. new_headroom = skb_headroom(skb);
  2851. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  2852. return 0;
  2853. }
  2854. /* Compute the checksum for a gso segment. First compute the checksum value
  2855. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  2856. * then add in skb->csum (checksum from csum_start to end of packet).
  2857. * skb->csum and csum_start are then updated to reflect the checksum of the
  2858. * resultant packet starting from the transport header-- the resultant checksum
  2859. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  2860. * header.
  2861. */
  2862. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  2863. {
  2864. int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
  2865. skb_transport_offset(skb);
  2866. __u16 csum;
  2867. csum = csum_fold(csum_partial(skb_transport_header(skb),
  2868. plen, skb->csum));
  2869. skb->csum = res;
  2870. SKB_GSO_CB(skb)->csum_start -= plen;
  2871. return csum;
  2872. }
  2873. static inline bool skb_is_gso(const struct sk_buff *skb)
  2874. {
  2875. return skb_shinfo(skb)->gso_size;
  2876. }
  2877. /* Note: Should be called only if skb_is_gso(skb) is true */
  2878. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  2879. {
  2880. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  2881. }
  2882. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  2883. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  2884. {
  2885. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  2886. * wanted then gso_type will be set. */
  2887. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  2888. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  2889. unlikely(shinfo->gso_type == 0)) {
  2890. __skb_warn_lro_forwarding(skb);
  2891. return true;
  2892. }
  2893. return false;
  2894. }
  2895. static inline void skb_forward_csum(struct sk_buff *skb)
  2896. {
  2897. /* Unfortunately we don't support this one. Any brave souls? */
  2898. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2899. skb->ip_summed = CHECKSUM_NONE;
  2900. }
  2901. /**
  2902. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  2903. * @skb: skb to check
  2904. *
  2905. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  2906. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  2907. * use this helper, to document places where we make this assertion.
  2908. */
  2909. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  2910. {
  2911. #ifdef DEBUG
  2912. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  2913. #endif
  2914. }
  2915. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  2916. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  2917. u32 skb_get_poff(const struct sk_buff *skb);
  2918. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  2919. const struct flow_keys *keys, int hlen);
  2920. /**
  2921. * skb_head_is_locked - Determine if the skb->head is locked down
  2922. * @skb: skb to check
  2923. *
  2924. * The head on skbs build around a head frag can be removed if they are
  2925. * not cloned. This function returns true if the skb head is locked down
  2926. * due to either being allocated via kmalloc, or by being a clone with
  2927. * multiple references to the head.
  2928. */
  2929. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  2930. {
  2931. return !skb->head_frag || skb_cloned(skb);
  2932. }
  2933. /**
  2934. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  2935. *
  2936. * @skb: GSO skb
  2937. *
  2938. * skb_gso_network_seglen is used to determine the real size of the
  2939. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  2940. *
  2941. * The MAC/L2 header is not accounted for.
  2942. */
  2943. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  2944. {
  2945. unsigned int hdr_len = skb_transport_header(skb) -
  2946. skb_network_header(skb);
  2947. return hdr_len + skb_gso_transport_seglen(skb);
  2948. }
  2949. #endif /* __KERNEL__ */
  2950. #endif /* _LINUX_SKBUFF_H */