skbuff.h 97 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963296429652966296729682969297029712972297329742975297629772978297929802981298229832984298529862987298829892990299129922993299429952996299729982999300030013002300330043005300630073008300930103011301230133014301530163017301830193020302130223023302430253026302730283029303030313032303330343035303630373038303930403041304230433044304530463047304830493050305130523053305430553056305730583059306030613062306330643065306630673068306930703071307230733074307530763077307830793080308130823083308430853086308730883089309030913092309330943095309630973098309931003101310231033104310531063107310831093110311131123113311431153116311731183119312031213122312331243125312631273128312931303131313231333134313531363137313831393140314131423143314431453146314731483149315031513152315331543155315631573158315931603161316231633164316531663167316831693170317131723173317431753176317731783179318031813182318331843185318631873188318931903191319231933194319531963197319831993200320132023203320432053206320732083209321032113212321332143215321632173218321932203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278327932803281328232833284328532863287328832893290329132923293329432953296329732983299330033013302330333043305330633073308330933103311331233133314331533163317331833193320332133223323332433253326332733283329333033313332333333343335333633373338333933403341334233433344334533463347334833493350335133523353335433553356335733583359336033613362336333643365336633673368336933703371337233733374337533763377337833793380338133823383338433853386338733883389339033913392339333943395339633973398339934003401340234033404340534063407340834093410341134123413341434153416341734183419342034213422342334243425342634273428342934303431343234333434343534363437343834393440344134423443344434453446344734483449345034513452345334543455
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/kernel.h>
  16. #include <linux/kmemcheck.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/bug.h>
  20. #include <linux/cache.h>
  21. #include <linux/rbtree.h>
  22. #include <linux/socket.h>
  23. #include <linux/atomic.h>
  24. #include <asm/types.h>
  25. #include <linux/spinlock.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #include <linux/rcupdate.h>
  30. #include <linux/hrtimer.h>
  31. #include <linux/dma-mapping.h>
  32. #include <linux/netdev_features.h>
  33. #include <linux/sched.h>
  34. #include <net/flow_keys.h>
  35. /* A. Checksumming of received packets by device.
  36. *
  37. * CHECKSUM_NONE:
  38. *
  39. * Device failed to checksum this packet e.g. due to lack of capabilities.
  40. * The packet contains full (though not verified) checksum in packet but
  41. * not in skb->csum. Thus, skb->csum is undefined in this case.
  42. *
  43. * CHECKSUM_UNNECESSARY:
  44. *
  45. * The hardware you're dealing with doesn't calculate the full checksum
  46. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  47. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  48. * if their checksums are okay. skb->csum is still undefined in this case
  49. * though. It is a bad option, but, unfortunately, nowadays most vendors do
  50. * this. Apparently with the secret goal to sell you new devices, when you
  51. * will add new protocol to your host, f.e. IPv6 8)
  52. *
  53. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  54. * TCP: IPv6 and IPv4.
  55. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  56. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  57. * may perform further validation in this case.
  58. * GRE: only if the checksum is present in the header.
  59. * SCTP: indicates the CRC in SCTP header has been validated.
  60. *
  61. * skb->csum_level indicates the number of consecutive checksums found in
  62. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  63. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  64. * and a device is able to verify the checksums for UDP (possibly zero),
  65. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  66. * two. If the device were only able to verify the UDP checksum and not
  67. * GRE, either because it doesn't support GRE checksum of because GRE
  68. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  69. * not considered in this case).
  70. *
  71. * CHECKSUM_COMPLETE:
  72. *
  73. * This is the most generic way. The device supplied checksum of the _whole_
  74. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  75. * hardware doesn't need to parse L3/L4 headers to implement this.
  76. *
  77. * Note: Even if device supports only some protocols, but is able to produce
  78. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  79. *
  80. * CHECKSUM_PARTIAL:
  81. *
  82. * A checksum is set up to be offloaded to a device as described in the
  83. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  84. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  85. * on the same host, or it may be set in the input path in GRO or remote
  86. * checksum offload. For the purposes of checksum verification, the checksum
  87. * referred to by skb->csum_start + skb->csum_offset and any preceding
  88. * checksums in the packet are considered verified. Any checksums in the
  89. * packet that are after the checksum being offloaded are not considered to
  90. * be verified.
  91. *
  92. * B. Checksumming on output.
  93. *
  94. * CHECKSUM_NONE:
  95. *
  96. * The skb was already checksummed by the protocol, or a checksum is not
  97. * required.
  98. *
  99. * CHECKSUM_PARTIAL:
  100. *
  101. * The device is required to checksum the packet as seen by hard_start_xmit()
  102. * from skb->csum_start up to the end, and to record/write the checksum at
  103. * offset skb->csum_start + skb->csum_offset.
  104. *
  105. * The device must show its capabilities in dev->features, set up at device
  106. * setup time, e.g. netdev_features.h:
  107. *
  108. * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
  109. * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
  110. * IPv4. Sigh. Vendors like this way for an unknown reason.
  111. * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
  112. * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
  113. * NETIF_F_... - Well, you get the picture.
  114. *
  115. * CHECKSUM_UNNECESSARY:
  116. *
  117. * Normally, the device will do per protocol specific checksumming. Protocol
  118. * implementations that do not want the NIC to perform the checksum
  119. * calculation should use this flag in their outgoing skbs.
  120. *
  121. * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
  122. * offload. Correspondingly, the FCoE protocol driver
  123. * stack should use CHECKSUM_UNNECESSARY.
  124. *
  125. * Any questions? No questions, good. --ANK
  126. */
  127. /* Don't change this without changing skb_csum_unnecessary! */
  128. #define CHECKSUM_NONE 0
  129. #define CHECKSUM_UNNECESSARY 1
  130. #define CHECKSUM_COMPLETE 2
  131. #define CHECKSUM_PARTIAL 3
  132. /* Maximum value in skb->csum_level */
  133. #define SKB_MAX_CSUM_LEVEL 3
  134. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  135. #define SKB_WITH_OVERHEAD(X) \
  136. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  137. #define SKB_MAX_ORDER(X, ORDER) \
  138. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  139. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  140. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  141. /* return minimum truesize of one skb containing X bytes of data */
  142. #define SKB_TRUESIZE(X) ((X) + \
  143. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  144. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  145. struct net_device;
  146. struct scatterlist;
  147. struct pipe_inode_info;
  148. struct iov_iter;
  149. struct napi_struct;
  150. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  151. struct nf_conntrack {
  152. atomic_t use;
  153. };
  154. #endif
  155. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  156. struct nf_bridge_info {
  157. atomic_t use;
  158. enum {
  159. BRNF_PROTO_UNCHANGED,
  160. BRNF_PROTO_8021Q,
  161. BRNF_PROTO_PPPOE
  162. } orig_proto;
  163. bool pkt_otherhost;
  164. unsigned int mask;
  165. struct net_device *physindev;
  166. struct net_device *physoutdev;
  167. char neigh_header[8];
  168. };
  169. #endif
  170. struct sk_buff_head {
  171. /* These two members must be first. */
  172. struct sk_buff *next;
  173. struct sk_buff *prev;
  174. __u32 qlen;
  175. spinlock_t lock;
  176. };
  177. struct sk_buff;
  178. /* To allow 64K frame to be packed as single skb without frag_list we
  179. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  180. * buffers which do not start on a page boundary.
  181. *
  182. * Since GRO uses frags we allocate at least 16 regardless of page
  183. * size.
  184. */
  185. #if (65536/PAGE_SIZE + 1) < 16
  186. #define MAX_SKB_FRAGS 16UL
  187. #else
  188. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  189. #endif
  190. typedef struct skb_frag_struct skb_frag_t;
  191. struct skb_frag_struct {
  192. struct {
  193. struct page *p;
  194. } page;
  195. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  196. __u32 page_offset;
  197. __u32 size;
  198. #else
  199. __u16 page_offset;
  200. __u16 size;
  201. #endif
  202. };
  203. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  204. {
  205. return frag->size;
  206. }
  207. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  208. {
  209. frag->size = size;
  210. }
  211. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  212. {
  213. frag->size += delta;
  214. }
  215. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  216. {
  217. frag->size -= delta;
  218. }
  219. #define HAVE_HW_TIME_STAMP
  220. /**
  221. * struct skb_shared_hwtstamps - hardware time stamps
  222. * @hwtstamp: hardware time stamp transformed into duration
  223. * since arbitrary point in time
  224. *
  225. * Software time stamps generated by ktime_get_real() are stored in
  226. * skb->tstamp.
  227. *
  228. * hwtstamps can only be compared against other hwtstamps from
  229. * the same device.
  230. *
  231. * This structure is attached to packets as part of the
  232. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  233. */
  234. struct skb_shared_hwtstamps {
  235. ktime_t hwtstamp;
  236. };
  237. /* Definitions for tx_flags in struct skb_shared_info */
  238. enum {
  239. /* generate hardware time stamp */
  240. SKBTX_HW_TSTAMP = 1 << 0,
  241. /* generate software time stamp when queueing packet to NIC */
  242. SKBTX_SW_TSTAMP = 1 << 1,
  243. /* device driver is going to provide hardware time stamp */
  244. SKBTX_IN_PROGRESS = 1 << 2,
  245. /* device driver supports TX zero-copy buffers */
  246. SKBTX_DEV_ZEROCOPY = 1 << 3,
  247. /* generate wifi status information (where possible) */
  248. SKBTX_WIFI_STATUS = 1 << 4,
  249. /* This indicates at least one fragment might be overwritten
  250. * (as in vmsplice(), sendfile() ...)
  251. * If we need to compute a TX checksum, we'll need to copy
  252. * all frags to avoid possible bad checksum
  253. */
  254. SKBTX_SHARED_FRAG = 1 << 5,
  255. /* generate software time stamp when entering packet scheduling */
  256. SKBTX_SCHED_TSTAMP = 1 << 6,
  257. /* generate software timestamp on peer data acknowledgment */
  258. SKBTX_ACK_TSTAMP = 1 << 7,
  259. };
  260. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  261. SKBTX_SCHED_TSTAMP | \
  262. SKBTX_ACK_TSTAMP)
  263. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  264. /*
  265. * The callback notifies userspace to release buffers when skb DMA is done in
  266. * lower device, the skb last reference should be 0 when calling this.
  267. * The zerocopy_success argument is true if zero copy transmit occurred,
  268. * false on data copy or out of memory error caused by data copy attempt.
  269. * The ctx field is used to track device context.
  270. * The desc field is used to track userspace buffer index.
  271. */
  272. struct ubuf_info {
  273. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  274. void *ctx;
  275. unsigned long desc;
  276. };
  277. /* This data is invariant across clones and lives at
  278. * the end of the header data, ie. at skb->end.
  279. */
  280. struct skb_shared_info {
  281. unsigned char nr_frags;
  282. __u8 tx_flags;
  283. unsigned short gso_size;
  284. /* Warning: this field is not always filled in (UFO)! */
  285. unsigned short gso_segs;
  286. unsigned short gso_type;
  287. struct sk_buff *frag_list;
  288. struct skb_shared_hwtstamps hwtstamps;
  289. u32 tskey;
  290. __be32 ip6_frag_id;
  291. /*
  292. * Warning : all fields before dataref are cleared in __alloc_skb()
  293. */
  294. atomic_t dataref;
  295. /* Intermediate layers must ensure that destructor_arg
  296. * remains valid until skb destructor */
  297. void * destructor_arg;
  298. /* must be last field, see pskb_expand_head() */
  299. skb_frag_t frags[MAX_SKB_FRAGS];
  300. };
  301. /* We divide dataref into two halves. The higher 16 bits hold references
  302. * to the payload part of skb->data. The lower 16 bits hold references to
  303. * the entire skb->data. A clone of a headerless skb holds the length of
  304. * the header in skb->hdr_len.
  305. *
  306. * All users must obey the rule that the skb->data reference count must be
  307. * greater than or equal to the payload reference count.
  308. *
  309. * Holding a reference to the payload part means that the user does not
  310. * care about modifications to the header part of skb->data.
  311. */
  312. #define SKB_DATAREF_SHIFT 16
  313. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  314. enum {
  315. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  316. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  317. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  318. };
  319. enum {
  320. SKB_GSO_TCPV4 = 1 << 0,
  321. SKB_GSO_UDP = 1 << 1,
  322. /* This indicates the skb is from an untrusted source. */
  323. SKB_GSO_DODGY = 1 << 2,
  324. /* This indicates the tcp segment has CWR set. */
  325. SKB_GSO_TCP_ECN = 1 << 3,
  326. SKB_GSO_TCPV6 = 1 << 4,
  327. SKB_GSO_FCOE = 1 << 5,
  328. SKB_GSO_GRE = 1 << 6,
  329. SKB_GSO_GRE_CSUM = 1 << 7,
  330. SKB_GSO_IPIP = 1 << 8,
  331. SKB_GSO_SIT = 1 << 9,
  332. SKB_GSO_UDP_TUNNEL = 1 << 10,
  333. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  334. SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
  335. };
  336. #if BITS_PER_LONG > 32
  337. #define NET_SKBUFF_DATA_USES_OFFSET 1
  338. #endif
  339. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  340. typedef unsigned int sk_buff_data_t;
  341. #else
  342. typedef unsigned char *sk_buff_data_t;
  343. #endif
  344. /**
  345. * struct skb_mstamp - multi resolution time stamps
  346. * @stamp_us: timestamp in us resolution
  347. * @stamp_jiffies: timestamp in jiffies
  348. */
  349. struct skb_mstamp {
  350. union {
  351. u64 v64;
  352. struct {
  353. u32 stamp_us;
  354. u32 stamp_jiffies;
  355. };
  356. };
  357. };
  358. /**
  359. * skb_mstamp_get - get current timestamp
  360. * @cl: place to store timestamps
  361. */
  362. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  363. {
  364. u64 val = local_clock();
  365. do_div(val, NSEC_PER_USEC);
  366. cl->stamp_us = (u32)val;
  367. cl->stamp_jiffies = (u32)jiffies;
  368. }
  369. /**
  370. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  371. * @t1: pointer to newest sample
  372. * @t0: pointer to oldest sample
  373. */
  374. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  375. const struct skb_mstamp *t0)
  376. {
  377. s32 delta_us = t1->stamp_us - t0->stamp_us;
  378. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  379. /* If delta_us is negative, this might be because interval is too big,
  380. * or local_clock() drift is too big : fallback using jiffies.
  381. */
  382. if (delta_us <= 0 ||
  383. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  384. delta_us = jiffies_to_usecs(delta_jiffies);
  385. return delta_us;
  386. }
  387. /**
  388. * struct sk_buff - socket buffer
  389. * @next: Next buffer in list
  390. * @prev: Previous buffer in list
  391. * @tstamp: Time we arrived/left
  392. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  393. * @sk: Socket we are owned by
  394. * @dev: Device we arrived on/are leaving by
  395. * @cb: Control buffer. Free for use by every layer. Put private vars here
  396. * @_skb_refdst: destination entry (with norefcount bit)
  397. * @sp: the security path, used for xfrm
  398. * @len: Length of actual data
  399. * @data_len: Data length
  400. * @mac_len: Length of link layer header
  401. * @hdr_len: writable header length of cloned skb
  402. * @csum: Checksum (must include start/offset pair)
  403. * @csum_start: Offset from skb->head where checksumming should start
  404. * @csum_offset: Offset from csum_start where checksum should be stored
  405. * @priority: Packet queueing priority
  406. * @ignore_df: allow local fragmentation
  407. * @cloned: Head may be cloned (check refcnt to be sure)
  408. * @ip_summed: Driver fed us an IP checksum
  409. * @nohdr: Payload reference only, must not modify header
  410. * @nfctinfo: Relationship of this skb to the connection
  411. * @pkt_type: Packet class
  412. * @fclone: skbuff clone status
  413. * @ipvs_property: skbuff is owned by ipvs
  414. * @peeked: this packet has been seen already, so stats have been
  415. * done for it, don't do them again
  416. * @nf_trace: netfilter packet trace flag
  417. * @protocol: Packet protocol from driver
  418. * @destructor: Destruct function
  419. * @nfct: Associated connection, if any
  420. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  421. * @skb_iif: ifindex of device we arrived on
  422. * @tc_index: Traffic control index
  423. * @tc_verd: traffic control verdict
  424. * @hash: the packet hash
  425. * @queue_mapping: Queue mapping for multiqueue devices
  426. * @xmit_more: More SKBs are pending for this queue
  427. * @ndisc_nodetype: router type (from link layer)
  428. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  429. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  430. * ports.
  431. * @sw_hash: indicates hash was computed in software stack
  432. * @wifi_acked_valid: wifi_acked was set
  433. * @wifi_acked: whether frame was acked on wifi or not
  434. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  435. * @napi_id: id of the NAPI struct this skb came from
  436. * @secmark: security marking
  437. * @mark: Generic packet mark
  438. * @vlan_proto: vlan encapsulation protocol
  439. * @vlan_tci: vlan tag control information
  440. * @inner_protocol: Protocol (encapsulation)
  441. * @inner_transport_header: Inner transport layer header (encapsulation)
  442. * @inner_network_header: Network layer header (encapsulation)
  443. * @inner_mac_header: Link layer header (encapsulation)
  444. * @transport_header: Transport layer header
  445. * @network_header: Network layer header
  446. * @mac_header: Link layer header
  447. * @tail: Tail pointer
  448. * @end: End pointer
  449. * @head: Head of buffer
  450. * @data: Data head pointer
  451. * @truesize: Buffer size
  452. * @users: User count - see {datagram,tcp}.c
  453. */
  454. struct sk_buff {
  455. union {
  456. struct {
  457. /* These two members must be first. */
  458. struct sk_buff *next;
  459. struct sk_buff *prev;
  460. union {
  461. ktime_t tstamp;
  462. struct skb_mstamp skb_mstamp;
  463. };
  464. };
  465. struct rb_node rbnode; /* used in netem & tcp stack */
  466. };
  467. struct sock *sk;
  468. struct net_device *dev;
  469. /*
  470. * This is the control buffer. It is free to use for every
  471. * layer. Please put your private variables there. If you
  472. * want to keep them across layers you have to do a skb_clone()
  473. * first. This is owned by whoever has the skb queued ATM.
  474. */
  475. char cb[48] __aligned(8);
  476. unsigned long _skb_refdst;
  477. void (*destructor)(struct sk_buff *skb);
  478. #ifdef CONFIG_XFRM
  479. struct sec_path *sp;
  480. #endif
  481. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  482. struct nf_conntrack *nfct;
  483. #endif
  484. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  485. struct nf_bridge_info *nf_bridge;
  486. #endif
  487. unsigned int len,
  488. data_len;
  489. __u16 mac_len,
  490. hdr_len;
  491. /* Following fields are _not_ copied in __copy_skb_header()
  492. * Note that queue_mapping is here mostly to fill a hole.
  493. */
  494. kmemcheck_bitfield_begin(flags1);
  495. __u16 queue_mapping;
  496. __u8 cloned:1,
  497. nohdr:1,
  498. fclone:2,
  499. peeked:1,
  500. head_frag:1,
  501. xmit_more:1;
  502. /* one bit hole */
  503. kmemcheck_bitfield_end(flags1);
  504. /* fields enclosed in headers_start/headers_end are copied
  505. * using a single memcpy() in __copy_skb_header()
  506. */
  507. /* private: */
  508. __u32 headers_start[0];
  509. /* public: */
  510. /* if you move pkt_type around you also must adapt those constants */
  511. #ifdef __BIG_ENDIAN_BITFIELD
  512. #define PKT_TYPE_MAX (7 << 5)
  513. #else
  514. #define PKT_TYPE_MAX 7
  515. #endif
  516. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  517. __u8 __pkt_type_offset[0];
  518. __u8 pkt_type:3;
  519. __u8 pfmemalloc:1;
  520. __u8 ignore_df:1;
  521. __u8 nfctinfo:3;
  522. __u8 nf_trace:1;
  523. __u8 ip_summed:2;
  524. __u8 ooo_okay:1;
  525. __u8 l4_hash:1;
  526. __u8 sw_hash:1;
  527. __u8 wifi_acked_valid:1;
  528. __u8 wifi_acked:1;
  529. __u8 no_fcs:1;
  530. /* Indicates the inner headers are valid in the skbuff. */
  531. __u8 encapsulation:1;
  532. __u8 encap_hdr_csum:1;
  533. __u8 csum_valid:1;
  534. __u8 csum_complete_sw:1;
  535. __u8 csum_level:2;
  536. __u8 csum_bad:1;
  537. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  538. __u8 ndisc_nodetype:2;
  539. #endif
  540. __u8 ipvs_property:1;
  541. __u8 inner_protocol_type:1;
  542. __u8 remcsum_offload:1;
  543. /* 3 or 5 bit hole */
  544. #ifdef CONFIG_NET_SCHED
  545. __u16 tc_index; /* traffic control index */
  546. #ifdef CONFIG_NET_CLS_ACT
  547. __u16 tc_verd; /* traffic control verdict */
  548. #endif
  549. #endif
  550. union {
  551. __wsum csum;
  552. struct {
  553. __u16 csum_start;
  554. __u16 csum_offset;
  555. };
  556. };
  557. __u32 priority;
  558. int skb_iif;
  559. __u32 hash;
  560. __be16 vlan_proto;
  561. __u16 vlan_tci;
  562. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  563. union {
  564. unsigned int napi_id;
  565. unsigned int sender_cpu;
  566. };
  567. #endif
  568. #ifdef CONFIG_NETWORK_SECMARK
  569. __u32 secmark;
  570. #endif
  571. union {
  572. __u32 mark;
  573. __u32 reserved_tailroom;
  574. };
  575. union {
  576. __be16 inner_protocol;
  577. __u8 inner_ipproto;
  578. };
  579. __u16 inner_transport_header;
  580. __u16 inner_network_header;
  581. __u16 inner_mac_header;
  582. __be16 protocol;
  583. __u16 transport_header;
  584. __u16 network_header;
  585. __u16 mac_header;
  586. /* private: */
  587. __u32 headers_end[0];
  588. /* public: */
  589. /* These elements must be at the end, see alloc_skb() for details. */
  590. sk_buff_data_t tail;
  591. sk_buff_data_t end;
  592. unsigned char *head,
  593. *data;
  594. unsigned int truesize;
  595. atomic_t users;
  596. };
  597. #ifdef __KERNEL__
  598. /*
  599. * Handling routines are only of interest to the kernel
  600. */
  601. #include <linux/slab.h>
  602. #define SKB_ALLOC_FCLONE 0x01
  603. #define SKB_ALLOC_RX 0x02
  604. #define SKB_ALLOC_NAPI 0x04
  605. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  606. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  607. {
  608. return unlikely(skb->pfmemalloc);
  609. }
  610. /*
  611. * skb might have a dst pointer attached, refcounted or not.
  612. * _skb_refdst low order bit is set if refcount was _not_ taken
  613. */
  614. #define SKB_DST_NOREF 1UL
  615. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  616. /**
  617. * skb_dst - returns skb dst_entry
  618. * @skb: buffer
  619. *
  620. * Returns skb dst_entry, regardless of reference taken or not.
  621. */
  622. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  623. {
  624. /* If refdst was not refcounted, check we still are in a
  625. * rcu_read_lock section
  626. */
  627. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  628. !rcu_read_lock_held() &&
  629. !rcu_read_lock_bh_held());
  630. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  631. }
  632. /**
  633. * skb_dst_set - sets skb dst
  634. * @skb: buffer
  635. * @dst: dst entry
  636. *
  637. * Sets skb dst, assuming a reference was taken on dst and should
  638. * be released by skb_dst_drop()
  639. */
  640. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  641. {
  642. skb->_skb_refdst = (unsigned long)dst;
  643. }
  644. /**
  645. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  646. * @skb: buffer
  647. * @dst: dst entry
  648. *
  649. * Sets skb dst, assuming a reference was not taken on dst.
  650. * If dst entry is cached, we do not take reference and dst_release
  651. * will be avoided by refdst_drop. If dst entry is not cached, we take
  652. * reference, so that last dst_release can destroy the dst immediately.
  653. */
  654. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  655. {
  656. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  657. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  658. }
  659. /**
  660. * skb_dst_is_noref - Test if skb dst isn't refcounted
  661. * @skb: buffer
  662. */
  663. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  664. {
  665. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  666. }
  667. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  668. {
  669. return (struct rtable *)skb_dst(skb);
  670. }
  671. void kfree_skb(struct sk_buff *skb);
  672. void kfree_skb_list(struct sk_buff *segs);
  673. void skb_tx_error(struct sk_buff *skb);
  674. void consume_skb(struct sk_buff *skb);
  675. void __kfree_skb(struct sk_buff *skb);
  676. extern struct kmem_cache *skbuff_head_cache;
  677. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  678. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  679. bool *fragstolen, int *delta_truesize);
  680. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  681. int node);
  682. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  683. static inline struct sk_buff *alloc_skb(unsigned int size,
  684. gfp_t priority)
  685. {
  686. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  687. }
  688. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  689. unsigned long data_len,
  690. int max_page_order,
  691. int *errcode,
  692. gfp_t gfp_mask);
  693. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  694. struct sk_buff_fclones {
  695. struct sk_buff skb1;
  696. struct sk_buff skb2;
  697. atomic_t fclone_ref;
  698. };
  699. /**
  700. * skb_fclone_busy - check if fclone is busy
  701. * @skb: buffer
  702. *
  703. * Returns true is skb is a fast clone, and its clone is not freed.
  704. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  705. * so we also check that this didnt happen.
  706. */
  707. static inline bool skb_fclone_busy(const struct sock *sk,
  708. const struct sk_buff *skb)
  709. {
  710. const struct sk_buff_fclones *fclones;
  711. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  712. return skb->fclone == SKB_FCLONE_ORIG &&
  713. atomic_read(&fclones->fclone_ref) > 1 &&
  714. fclones->skb2.sk == sk;
  715. }
  716. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  717. gfp_t priority)
  718. {
  719. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  720. }
  721. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  722. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  723. {
  724. return __alloc_skb_head(priority, -1);
  725. }
  726. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  727. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  728. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  729. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  730. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  731. gfp_t gfp_mask, bool fclone);
  732. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  733. gfp_t gfp_mask)
  734. {
  735. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  736. }
  737. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  738. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  739. unsigned int headroom);
  740. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  741. int newtailroom, gfp_t priority);
  742. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  743. int offset, int len);
  744. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  745. int len);
  746. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  747. int skb_pad(struct sk_buff *skb, int pad);
  748. #define dev_kfree_skb(a) consume_skb(a)
  749. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  750. int getfrag(void *from, char *to, int offset,
  751. int len, int odd, struct sk_buff *skb),
  752. void *from, int length);
  753. struct skb_seq_state {
  754. __u32 lower_offset;
  755. __u32 upper_offset;
  756. __u32 frag_idx;
  757. __u32 stepped_offset;
  758. struct sk_buff *root_skb;
  759. struct sk_buff *cur_skb;
  760. __u8 *frag_data;
  761. };
  762. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  763. unsigned int to, struct skb_seq_state *st);
  764. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  765. struct skb_seq_state *st);
  766. void skb_abort_seq_read(struct skb_seq_state *st);
  767. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  768. unsigned int to, struct ts_config *config);
  769. /*
  770. * Packet hash types specify the type of hash in skb_set_hash.
  771. *
  772. * Hash types refer to the protocol layer addresses which are used to
  773. * construct a packet's hash. The hashes are used to differentiate or identify
  774. * flows of the protocol layer for the hash type. Hash types are either
  775. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  776. *
  777. * Properties of hashes:
  778. *
  779. * 1) Two packets in different flows have different hash values
  780. * 2) Two packets in the same flow should have the same hash value
  781. *
  782. * A hash at a higher layer is considered to be more specific. A driver should
  783. * set the most specific hash possible.
  784. *
  785. * A driver cannot indicate a more specific hash than the layer at which a hash
  786. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  787. *
  788. * A driver may indicate a hash level which is less specific than the
  789. * actual layer the hash was computed on. For instance, a hash computed
  790. * at L4 may be considered an L3 hash. This should only be done if the
  791. * driver can't unambiguously determine that the HW computed the hash at
  792. * the higher layer. Note that the "should" in the second property above
  793. * permits this.
  794. */
  795. enum pkt_hash_types {
  796. PKT_HASH_TYPE_NONE, /* Undefined type */
  797. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  798. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  799. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  800. };
  801. static inline void
  802. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  803. {
  804. skb->l4_hash = (type == PKT_HASH_TYPE_L4);
  805. skb->sw_hash = 0;
  806. skb->hash = hash;
  807. }
  808. void __skb_get_hash(struct sk_buff *skb);
  809. static inline __u32 skb_get_hash(struct sk_buff *skb)
  810. {
  811. if (!skb->l4_hash && !skb->sw_hash)
  812. __skb_get_hash(skb);
  813. return skb->hash;
  814. }
  815. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  816. {
  817. return skb->hash;
  818. }
  819. static inline void skb_clear_hash(struct sk_buff *skb)
  820. {
  821. skb->hash = 0;
  822. skb->sw_hash = 0;
  823. skb->l4_hash = 0;
  824. }
  825. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  826. {
  827. if (!skb->l4_hash)
  828. skb_clear_hash(skb);
  829. }
  830. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  831. {
  832. to->hash = from->hash;
  833. to->sw_hash = from->sw_hash;
  834. to->l4_hash = from->l4_hash;
  835. };
  836. static inline void skb_sender_cpu_clear(struct sk_buff *skb)
  837. {
  838. #ifdef CONFIG_XPS
  839. skb->sender_cpu = 0;
  840. #endif
  841. }
  842. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  843. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  844. {
  845. return skb->head + skb->end;
  846. }
  847. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  848. {
  849. return skb->end;
  850. }
  851. #else
  852. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  853. {
  854. return skb->end;
  855. }
  856. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  857. {
  858. return skb->end - skb->head;
  859. }
  860. #endif
  861. /* Internal */
  862. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  863. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  864. {
  865. return &skb_shinfo(skb)->hwtstamps;
  866. }
  867. /**
  868. * skb_queue_empty - check if a queue is empty
  869. * @list: queue head
  870. *
  871. * Returns true if the queue is empty, false otherwise.
  872. */
  873. static inline int skb_queue_empty(const struct sk_buff_head *list)
  874. {
  875. return list->next == (const struct sk_buff *) list;
  876. }
  877. /**
  878. * skb_queue_is_last - check if skb is the last entry in the queue
  879. * @list: queue head
  880. * @skb: buffer
  881. *
  882. * Returns true if @skb is the last buffer on the list.
  883. */
  884. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  885. const struct sk_buff *skb)
  886. {
  887. return skb->next == (const struct sk_buff *) list;
  888. }
  889. /**
  890. * skb_queue_is_first - check if skb is the first entry in the queue
  891. * @list: queue head
  892. * @skb: buffer
  893. *
  894. * Returns true if @skb is the first buffer on the list.
  895. */
  896. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  897. const struct sk_buff *skb)
  898. {
  899. return skb->prev == (const struct sk_buff *) list;
  900. }
  901. /**
  902. * skb_queue_next - return the next packet in the queue
  903. * @list: queue head
  904. * @skb: current buffer
  905. *
  906. * Return the next packet in @list after @skb. It is only valid to
  907. * call this if skb_queue_is_last() evaluates to false.
  908. */
  909. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  910. const struct sk_buff *skb)
  911. {
  912. /* This BUG_ON may seem severe, but if we just return then we
  913. * are going to dereference garbage.
  914. */
  915. BUG_ON(skb_queue_is_last(list, skb));
  916. return skb->next;
  917. }
  918. /**
  919. * skb_queue_prev - return the prev packet in the queue
  920. * @list: queue head
  921. * @skb: current buffer
  922. *
  923. * Return the prev packet in @list before @skb. It is only valid to
  924. * call this if skb_queue_is_first() evaluates to false.
  925. */
  926. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  927. const struct sk_buff *skb)
  928. {
  929. /* This BUG_ON may seem severe, but if we just return then we
  930. * are going to dereference garbage.
  931. */
  932. BUG_ON(skb_queue_is_first(list, skb));
  933. return skb->prev;
  934. }
  935. /**
  936. * skb_get - reference buffer
  937. * @skb: buffer to reference
  938. *
  939. * Makes another reference to a socket buffer and returns a pointer
  940. * to the buffer.
  941. */
  942. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  943. {
  944. atomic_inc(&skb->users);
  945. return skb;
  946. }
  947. /*
  948. * If users == 1, we are the only owner and are can avoid redundant
  949. * atomic change.
  950. */
  951. /**
  952. * skb_cloned - is the buffer a clone
  953. * @skb: buffer to check
  954. *
  955. * Returns true if the buffer was generated with skb_clone() and is
  956. * one of multiple shared copies of the buffer. Cloned buffers are
  957. * shared data so must not be written to under normal circumstances.
  958. */
  959. static inline int skb_cloned(const struct sk_buff *skb)
  960. {
  961. return skb->cloned &&
  962. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  963. }
  964. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  965. {
  966. might_sleep_if(pri & __GFP_WAIT);
  967. if (skb_cloned(skb))
  968. return pskb_expand_head(skb, 0, 0, pri);
  969. return 0;
  970. }
  971. /**
  972. * skb_header_cloned - is the header a clone
  973. * @skb: buffer to check
  974. *
  975. * Returns true if modifying the header part of the buffer requires
  976. * the data to be copied.
  977. */
  978. static inline int skb_header_cloned(const struct sk_buff *skb)
  979. {
  980. int dataref;
  981. if (!skb->cloned)
  982. return 0;
  983. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  984. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  985. return dataref != 1;
  986. }
  987. /**
  988. * skb_header_release - release reference to header
  989. * @skb: buffer to operate on
  990. *
  991. * Drop a reference to the header part of the buffer. This is done
  992. * by acquiring a payload reference. You must not read from the header
  993. * part of skb->data after this.
  994. * Note : Check if you can use __skb_header_release() instead.
  995. */
  996. static inline void skb_header_release(struct sk_buff *skb)
  997. {
  998. BUG_ON(skb->nohdr);
  999. skb->nohdr = 1;
  1000. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1001. }
  1002. /**
  1003. * __skb_header_release - release reference to header
  1004. * @skb: buffer to operate on
  1005. *
  1006. * Variant of skb_header_release() assuming skb is private to caller.
  1007. * We can avoid one atomic operation.
  1008. */
  1009. static inline void __skb_header_release(struct sk_buff *skb)
  1010. {
  1011. skb->nohdr = 1;
  1012. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1013. }
  1014. /**
  1015. * skb_shared - is the buffer shared
  1016. * @skb: buffer to check
  1017. *
  1018. * Returns true if more than one person has a reference to this
  1019. * buffer.
  1020. */
  1021. static inline int skb_shared(const struct sk_buff *skb)
  1022. {
  1023. return atomic_read(&skb->users) != 1;
  1024. }
  1025. /**
  1026. * skb_share_check - check if buffer is shared and if so clone it
  1027. * @skb: buffer to check
  1028. * @pri: priority for memory allocation
  1029. *
  1030. * If the buffer is shared the buffer is cloned and the old copy
  1031. * drops a reference. A new clone with a single reference is returned.
  1032. * If the buffer is not shared the original buffer is returned. When
  1033. * being called from interrupt status or with spinlocks held pri must
  1034. * be GFP_ATOMIC.
  1035. *
  1036. * NULL is returned on a memory allocation failure.
  1037. */
  1038. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1039. {
  1040. might_sleep_if(pri & __GFP_WAIT);
  1041. if (skb_shared(skb)) {
  1042. struct sk_buff *nskb = skb_clone(skb, pri);
  1043. if (likely(nskb))
  1044. consume_skb(skb);
  1045. else
  1046. kfree_skb(skb);
  1047. skb = nskb;
  1048. }
  1049. return skb;
  1050. }
  1051. /*
  1052. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1053. * packets to handle cases where we have a local reader and forward
  1054. * and a couple of other messy ones. The normal one is tcpdumping
  1055. * a packet thats being forwarded.
  1056. */
  1057. /**
  1058. * skb_unshare - make a copy of a shared buffer
  1059. * @skb: buffer to check
  1060. * @pri: priority for memory allocation
  1061. *
  1062. * If the socket buffer is a clone then this function creates a new
  1063. * copy of the data, drops a reference count on the old copy and returns
  1064. * the new copy with the reference count at 1. If the buffer is not a clone
  1065. * the original buffer is returned. When called with a spinlock held or
  1066. * from interrupt state @pri must be %GFP_ATOMIC
  1067. *
  1068. * %NULL is returned on a memory allocation failure.
  1069. */
  1070. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1071. gfp_t pri)
  1072. {
  1073. might_sleep_if(pri & __GFP_WAIT);
  1074. if (skb_cloned(skb)) {
  1075. struct sk_buff *nskb = skb_copy(skb, pri);
  1076. /* Free our shared copy */
  1077. if (likely(nskb))
  1078. consume_skb(skb);
  1079. else
  1080. kfree_skb(skb);
  1081. skb = nskb;
  1082. }
  1083. return skb;
  1084. }
  1085. /**
  1086. * skb_peek - peek at the head of an &sk_buff_head
  1087. * @list_: list to peek at
  1088. *
  1089. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1090. * be careful with this one. A peek leaves the buffer on the
  1091. * list and someone else may run off with it. You must hold
  1092. * the appropriate locks or have a private queue to do this.
  1093. *
  1094. * Returns %NULL for an empty list or a pointer to the head element.
  1095. * The reference count is not incremented and the reference is therefore
  1096. * volatile. Use with caution.
  1097. */
  1098. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1099. {
  1100. struct sk_buff *skb = list_->next;
  1101. if (skb == (struct sk_buff *)list_)
  1102. skb = NULL;
  1103. return skb;
  1104. }
  1105. /**
  1106. * skb_peek_next - peek skb following the given one from a queue
  1107. * @skb: skb to start from
  1108. * @list_: list to peek at
  1109. *
  1110. * Returns %NULL when the end of the list is met or a pointer to the
  1111. * next element. The reference count is not incremented and the
  1112. * reference is therefore volatile. Use with caution.
  1113. */
  1114. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1115. const struct sk_buff_head *list_)
  1116. {
  1117. struct sk_buff *next = skb->next;
  1118. if (next == (struct sk_buff *)list_)
  1119. next = NULL;
  1120. return next;
  1121. }
  1122. /**
  1123. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1124. * @list_: list to peek at
  1125. *
  1126. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1127. * be careful with this one. A peek leaves the buffer on the
  1128. * list and someone else may run off with it. You must hold
  1129. * the appropriate locks or have a private queue to do this.
  1130. *
  1131. * Returns %NULL for an empty list or a pointer to the tail element.
  1132. * The reference count is not incremented and the reference is therefore
  1133. * volatile. Use with caution.
  1134. */
  1135. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1136. {
  1137. struct sk_buff *skb = list_->prev;
  1138. if (skb == (struct sk_buff *)list_)
  1139. skb = NULL;
  1140. return skb;
  1141. }
  1142. /**
  1143. * skb_queue_len - get queue length
  1144. * @list_: list to measure
  1145. *
  1146. * Return the length of an &sk_buff queue.
  1147. */
  1148. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1149. {
  1150. return list_->qlen;
  1151. }
  1152. /**
  1153. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1154. * @list: queue to initialize
  1155. *
  1156. * This initializes only the list and queue length aspects of
  1157. * an sk_buff_head object. This allows to initialize the list
  1158. * aspects of an sk_buff_head without reinitializing things like
  1159. * the spinlock. It can also be used for on-stack sk_buff_head
  1160. * objects where the spinlock is known to not be used.
  1161. */
  1162. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1163. {
  1164. list->prev = list->next = (struct sk_buff *)list;
  1165. list->qlen = 0;
  1166. }
  1167. /*
  1168. * This function creates a split out lock class for each invocation;
  1169. * this is needed for now since a whole lot of users of the skb-queue
  1170. * infrastructure in drivers have different locking usage (in hardirq)
  1171. * than the networking core (in softirq only). In the long run either the
  1172. * network layer or drivers should need annotation to consolidate the
  1173. * main types of usage into 3 classes.
  1174. */
  1175. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1176. {
  1177. spin_lock_init(&list->lock);
  1178. __skb_queue_head_init(list);
  1179. }
  1180. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1181. struct lock_class_key *class)
  1182. {
  1183. skb_queue_head_init(list);
  1184. lockdep_set_class(&list->lock, class);
  1185. }
  1186. /*
  1187. * Insert an sk_buff on a list.
  1188. *
  1189. * The "__skb_xxxx()" functions are the non-atomic ones that
  1190. * can only be called with interrupts disabled.
  1191. */
  1192. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1193. struct sk_buff_head *list);
  1194. static inline void __skb_insert(struct sk_buff *newsk,
  1195. struct sk_buff *prev, struct sk_buff *next,
  1196. struct sk_buff_head *list)
  1197. {
  1198. newsk->next = next;
  1199. newsk->prev = prev;
  1200. next->prev = prev->next = newsk;
  1201. list->qlen++;
  1202. }
  1203. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1204. struct sk_buff *prev,
  1205. struct sk_buff *next)
  1206. {
  1207. struct sk_buff *first = list->next;
  1208. struct sk_buff *last = list->prev;
  1209. first->prev = prev;
  1210. prev->next = first;
  1211. last->next = next;
  1212. next->prev = last;
  1213. }
  1214. /**
  1215. * skb_queue_splice - join two skb lists, this is designed for stacks
  1216. * @list: the new list to add
  1217. * @head: the place to add it in the first list
  1218. */
  1219. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1220. struct sk_buff_head *head)
  1221. {
  1222. if (!skb_queue_empty(list)) {
  1223. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1224. head->qlen += list->qlen;
  1225. }
  1226. }
  1227. /**
  1228. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1229. * @list: the new list to add
  1230. * @head: the place to add it in the first list
  1231. *
  1232. * The list at @list is reinitialised
  1233. */
  1234. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1235. struct sk_buff_head *head)
  1236. {
  1237. if (!skb_queue_empty(list)) {
  1238. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1239. head->qlen += list->qlen;
  1240. __skb_queue_head_init(list);
  1241. }
  1242. }
  1243. /**
  1244. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1245. * @list: the new list to add
  1246. * @head: the place to add it in the first list
  1247. */
  1248. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1249. struct sk_buff_head *head)
  1250. {
  1251. if (!skb_queue_empty(list)) {
  1252. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1253. head->qlen += list->qlen;
  1254. }
  1255. }
  1256. /**
  1257. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1258. * @list: the new list to add
  1259. * @head: the place to add it in the first list
  1260. *
  1261. * Each of the lists is a queue.
  1262. * The list at @list is reinitialised
  1263. */
  1264. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1265. struct sk_buff_head *head)
  1266. {
  1267. if (!skb_queue_empty(list)) {
  1268. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1269. head->qlen += list->qlen;
  1270. __skb_queue_head_init(list);
  1271. }
  1272. }
  1273. /**
  1274. * __skb_queue_after - queue a buffer at the list head
  1275. * @list: list to use
  1276. * @prev: place after this buffer
  1277. * @newsk: buffer to queue
  1278. *
  1279. * Queue a buffer int the middle of a list. This function takes no locks
  1280. * and you must therefore hold required locks before calling it.
  1281. *
  1282. * A buffer cannot be placed on two lists at the same time.
  1283. */
  1284. static inline void __skb_queue_after(struct sk_buff_head *list,
  1285. struct sk_buff *prev,
  1286. struct sk_buff *newsk)
  1287. {
  1288. __skb_insert(newsk, prev, prev->next, list);
  1289. }
  1290. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1291. struct sk_buff_head *list);
  1292. static inline void __skb_queue_before(struct sk_buff_head *list,
  1293. struct sk_buff *next,
  1294. struct sk_buff *newsk)
  1295. {
  1296. __skb_insert(newsk, next->prev, next, list);
  1297. }
  1298. /**
  1299. * __skb_queue_head - queue a buffer at the list head
  1300. * @list: list to use
  1301. * @newsk: buffer to queue
  1302. *
  1303. * Queue a buffer at the start of a list. This function takes no locks
  1304. * and you must therefore hold required locks before calling it.
  1305. *
  1306. * A buffer cannot be placed on two lists at the same time.
  1307. */
  1308. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1309. static inline void __skb_queue_head(struct sk_buff_head *list,
  1310. struct sk_buff *newsk)
  1311. {
  1312. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1313. }
  1314. /**
  1315. * __skb_queue_tail - queue a buffer at the list tail
  1316. * @list: list to use
  1317. * @newsk: buffer to queue
  1318. *
  1319. * Queue a buffer at the end of a list. This function takes no locks
  1320. * and you must therefore hold required locks before calling it.
  1321. *
  1322. * A buffer cannot be placed on two lists at the same time.
  1323. */
  1324. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1325. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1326. struct sk_buff *newsk)
  1327. {
  1328. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1329. }
  1330. /*
  1331. * remove sk_buff from list. _Must_ be called atomically, and with
  1332. * the list known..
  1333. */
  1334. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1335. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1336. {
  1337. struct sk_buff *next, *prev;
  1338. list->qlen--;
  1339. next = skb->next;
  1340. prev = skb->prev;
  1341. skb->next = skb->prev = NULL;
  1342. next->prev = prev;
  1343. prev->next = next;
  1344. }
  1345. /**
  1346. * __skb_dequeue - remove from the head of the queue
  1347. * @list: list to dequeue from
  1348. *
  1349. * Remove the head of the list. This function does not take any locks
  1350. * so must be used with appropriate locks held only. The head item is
  1351. * returned or %NULL if the list is empty.
  1352. */
  1353. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1354. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1355. {
  1356. struct sk_buff *skb = skb_peek(list);
  1357. if (skb)
  1358. __skb_unlink(skb, list);
  1359. return skb;
  1360. }
  1361. /**
  1362. * __skb_dequeue_tail - remove from the tail of the queue
  1363. * @list: list to dequeue from
  1364. *
  1365. * Remove the tail of the list. This function does not take any locks
  1366. * so must be used with appropriate locks held only. The tail item is
  1367. * returned or %NULL if the list is empty.
  1368. */
  1369. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1370. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1371. {
  1372. struct sk_buff *skb = skb_peek_tail(list);
  1373. if (skb)
  1374. __skb_unlink(skb, list);
  1375. return skb;
  1376. }
  1377. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1378. {
  1379. return skb->data_len;
  1380. }
  1381. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1382. {
  1383. return skb->len - skb->data_len;
  1384. }
  1385. static inline int skb_pagelen(const struct sk_buff *skb)
  1386. {
  1387. int i, len = 0;
  1388. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1389. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1390. return len + skb_headlen(skb);
  1391. }
  1392. /**
  1393. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1394. * @skb: buffer containing fragment to be initialised
  1395. * @i: paged fragment index to initialise
  1396. * @page: the page to use for this fragment
  1397. * @off: the offset to the data with @page
  1398. * @size: the length of the data
  1399. *
  1400. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1401. * offset @off within @page.
  1402. *
  1403. * Does not take any additional reference on the fragment.
  1404. */
  1405. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1406. struct page *page, int off, int size)
  1407. {
  1408. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1409. /*
  1410. * Propagate page->pfmemalloc to the skb if we can. The problem is
  1411. * that not all callers have unique ownership of the page. If
  1412. * pfmemalloc is set, we check the mapping as a mapping implies
  1413. * page->index is set (index and pfmemalloc share space).
  1414. * If it's a valid mapping, we cannot use page->pfmemalloc but we
  1415. * do not lose pfmemalloc information as the pages would not be
  1416. * allocated using __GFP_MEMALLOC.
  1417. */
  1418. frag->page.p = page;
  1419. frag->page_offset = off;
  1420. skb_frag_size_set(frag, size);
  1421. page = compound_head(page);
  1422. if (page->pfmemalloc && !page->mapping)
  1423. skb->pfmemalloc = true;
  1424. }
  1425. /**
  1426. * skb_fill_page_desc - initialise a paged fragment in an skb
  1427. * @skb: buffer containing fragment to be initialised
  1428. * @i: paged fragment index to initialise
  1429. * @page: the page to use for this fragment
  1430. * @off: the offset to the data with @page
  1431. * @size: the length of the data
  1432. *
  1433. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1434. * @skb to point to @size bytes at offset @off within @page. In
  1435. * addition updates @skb such that @i is the last fragment.
  1436. *
  1437. * Does not take any additional reference on the fragment.
  1438. */
  1439. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1440. struct page *page, int off, int size)
  1441. {
  1442. __skb_fill_page_desc(skb, i, page, off, size);
  1443. skb_shinfo(skb)->nr_frags = i + 1;
  1444. }
  1445. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1446. int size, unsigned int truesize);
  1447. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1448. unsigned int truesize);
  1449. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1450. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1451. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1452. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1453. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1454. {
  1455. return skb->head + skb->tail;
  1456. }
  1457. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1458. {
  1459. skb->tail = skb->data - skb->head;
  1460. }
  1461. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1462. {
  1463. skb_reset_tail_pointer(skb);
  1464. skb->tail += offset;
  1465. }
  1466. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1467. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1468. {
  1469. return skb->tail;
  1470. }
  1471. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1472. {
  1473. skb->tail = skb->data;
  1474. }
  1475. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1476. {
  1477. skb->tail = skb->data + offset;
  1478. }
  1479. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1480. /*
  1481. * Add data to an sk_buff
  1482. */
  1483. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1484. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1485. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1486. {
  1487. unsigned char *tmp = skb_tail_pointer(skb);
  1488. SKB_LINEAR_ASSERT(skb);
  1489. skb->tail += len;
  1490. skb->len += len;
  1491. return tmp;
  1492. }
  1493. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1494. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1495. {
  1496. skb->data -= len;
  1497. skb->len += len;
  1498. return skb->data;
  1499. }
  1500. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1501. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1502. {
  1503. skb->len -= len;
  1504. BUG_ON(skb->len < skb->data_len);
  1505. return skb->data += len;
  1506. }
  1507. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1508. {
  1509. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1510. }
  1511. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1512. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1513. {
  1514. if (len > skb_headlen(skb) &&
  1515. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1516. return NULL;
  1517. skb->len -= len;
  1518. return skb->data += len;
  1519. }
  1520. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1521. {
  1522. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1523. }
  1524. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1525. {
  1526. if (likely(len <= skb_headlen(skb)))
  1527. return 1;
  1528. if (unlikely(len > skb->len))
  1529. return 0;
  1530. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1531. }
  1532. /**
  1533. * skb_headroom - bytes at buffer head
  1534. * @skb: buffer to check
  1535. *
  1536. * Return the number of bytes of free space at the head of an &sk_buff.
  1537. */
  1538. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1539. {
  1540. return skb->data - skb->head;
  1541. }
  1542. /**
  1543. * skb_tailroom - bytes at buffer end
  1544. * @skb: buffer to check
  1545. *
  1546. * Return the number of bytes of free space at the tail of an sk_buff
  1547. */
  1548. static inline int skb_tailroom(const struct sk_buff *skb)
  1549. {
  1550. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1551. }
  1552. /**
  1553. * skb_availroom - bytes at buffer end
  1554. * @skb: buffer to check
  1555. *
  1556. * Return the number of bytes of free space at the tail of an sk_buff
  1557. * allocated by sk_stream_alloc()
  1558. */
  1559. static inline int skb_availroom(const struct sk_buff *skb)
  1560. {
  1561. if (skb_is_nonlinear(skb))
  1562. return 0;
  1563. return skb->end - skb->tail - skb->reserved_tailroom;
  1564. }
  1565. /**
  1566. * skb_reserve - adjust headroom
  1567. * @skb: buffer to alter
  1568. * @len: bytes to move
  1569. *
  1570. * Increase the headroom of an empty &sk_buff by reducing the tail
  1571. * room. This is only allowed for an empty buffer.
  1572. */
  1573. static inline void skb_reserve(struct sk_buff *skb, int len)
  1574. {
  1575. skb->data += len;
  1576. skb->tail += len;
  1577. }
  1578. #define ENCAP_TYPE_ETHER 0
  1579. #define ENCAP_TYPE_IPPROTO 1
  1580. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1581. __be16 protocol)
  1582. {
  1583. skb->inner_protocol = protocol;
  1584. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1585. }
  1586. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1587. __u8 ipproto)
  1588. {
  1589. skb->inner_ipproto = ipproto;
  1590. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1591. }
  1592. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1593. {
  1594. skb->inner_mac_header = skb->mac_header;
  1595. skb->inner_network_header = skb->network_header;
  1596. skb->inner_transport_header = skb->transport_header;
  1597. }
  1598. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1599. {
  1600. skb->mac_len = skb->network_header - skb->mac_header;
  1601. }
  1602. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1603. *skb)
  1604. {
  1605. return skb->head + skb->inner_transport_header;
  1606. }
  1607. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1608. {
  1609. skb->inner_transport_header = skb->data - skb->head;
  1610. }
  1611. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1612. const int offset)
  1613. {
  1614. skb_reset_inner_transport_header(skb);
  1615. skb->inner_transport_header += offset;
  1616. }
  1617. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1618. {
  1619. return skb->head + skb->inner_network_header;
  1620. }
  1621. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1622. {
  1623. skb->inner_network_header = skb->data - skb->head;
  1624. }
  1625. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1626. const int offset)
  1627. {
  1628. skb_reset_inner_network_header(skb);
  1629. skb->inner_network_header += offset;
  1630. }
  1631. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1632. {
  1633. return skb->head + skb->inner_mac_header;
  1634. }
  1635. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1636. {
  1637. skb->inner_mac_header = skb->data - skb->head;
  1638. }
  1639. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1640. const int offset)
  1641. {
  1642. skb_reset_inner_mac_header(skb);
  1643. skb->inner_mac_header += offset;
  1644. }
  1645. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1646. {
  1647. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1648. }
  1649. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1650. {
  1651. return skb->head + skb->transport_header;
  1652. }
  1653. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1654. {
  1655. skb->transport_header = skb->data - skb->head;
  1656. }
  1657. static inline void skb_set_transport_header(struct sk_buff *skb,
  1658. const int offset)
  1659. {
  1660. skb_reset_transport_header(skb);
  1661. skb->transport_header += offset;
  1662. }
  1663. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1664. {
  1665. return skb->head + skb->network_header;
  1666. }
  1667. static inline void skb_reset_network_header(struct sk_buff *skb)
  1668. {
  1669. skb->network_header = skb->data - skb->head;
  1670. }
  1671. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1672. {
  1673. skb_reset_network_header(skb);
  1674. skb->network_header += offset;
  1675. }
  1676. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1677. {
  1678. return skb->head + skb->mac_header;
  1679. }
  1680. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1681. {
  1682. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1683. }
  1684. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1685. {
  1686. skb->mac_header = skb->data - skb->head;
  1687. }
  1688. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1689. {
  1690. skb_reset_mac_header(skb);
  1691. skb->mac_header += offset;
  1692. }
  1693. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1694. {
  1695. skb->mac_header = skb->network_header;
  1696. }
  1697. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1698. const int offset_hint)
  1699. {
  1700. struct flow_keys keys;
  1701. if (skb_transport_header_was_set(skb))
  1702. return;
  1703. else if (skb_flow_dissect(skb, &keys))
  1704. skb_set_transport_header(skb, keys.thoff);
  1705. else
  1706. skb_set_transport_header(skb, offset_hint);
  1707. }
  1708. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1709. {
  1710. if (skb_mac_header_was_set(skb)) {
  1711. const unsigned char *old_mac = skb_mac_header(skb);
  1712. skb_set_mac_header(skb, -skb->mac_len);
  1713. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1714. }
  1715. }
  1716. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1717. {
  1718. return skb->csum_start - skb_headroom(skb);
  1719. }
  1720. static inline int skb_transport_offset(const struct sk_buff *skb)
  1721. {
  1722. return skb_transport_header(skb) - skb->data;
  1723. }
  1724. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1725. {
  1726. return skb->transport_header - skb->network_header;
  1727. }
  1728. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1729. {
  1730. return skb->inner_transport_header - skb->inner_network_header;
  1731. }
  1732. static inline int skb_network_offset(const struct sk_buff *skb)
  1733. {
  1734. return skb_network_header(skb) - skb->data;
  1735. }
  1736. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1737. {
  1738. return skb_inner_network_header(skb) - skb->data;
  1739. }
  1740. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1741. {
  1742. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1743. }
  1744. /*
  1745. * CPUs often take a performance hit when accessing unaligned memory
  1746. * locations. The actual performance hit varies, it can be small if the
  1747. * hardware handles it or large if we have to take an exception and fix it
  1748. * in software.
  1749. *
  1750. * Since an ethernet header is 14 bytes network drivers often end up with
  1751. * the IP header at an unaligned offset. The IP header can be aligned by
  1752. * shifting the start of the packet by 2 bytes. Drivers should do this
  1753. * with:
  1754. *
  1755. * skb_reserve(skb, NET_IP_ALIGN);
  1756. *
  1757. * The downside to this alignment of the IP header is that the DMA is now
  1758. * unaligned. On some architectures the cost of an unaligned DMA is high
  1759. * and this cost outweighs the gains made by aligning the IP header.
  1760. *
  1761. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1762. * to be overridden.
  1763. */
  1764. #ifndef NET_IP_ALIGN
  1765. #define NET_IP_ALIGN 2
  1766. #endif
  1767. /*
  1768. * The networking layer reserves some headroom in skb data (via
  1769. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1770. * the header has to grow. In the default case, if the header has to grow
  1771. * 32 bytes or less we avoid the reallocation.
  1772. *
  1773. * Unfortunately this headroom changes the DMA alignment of the resulting
  1774. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1775. * on some architectures. An architecture can override this value,
  1776. * perhaps setting it to a cacheline in size (since that will maintain
  1777. * cacheline alignment of the DMA). It must be a power of 2.
  1778. *
  1779. * Various parts of the networking layer expect at least 32 bytes of
  1780. * headroom, you should not reduce this.
  1781. *
  1782. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1783. * to reduce average number of cache lines per packet.
  1784. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1785. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1786. */
  1787. #ifndef NET_SKB_PAD
  1788. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1789. #endif
  1790. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1791. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1792. {
  1793. if (unlikely(skb_is_nonlinear(skb))) {
  1794. WARN_ON(1);
  1795. return;
  1796. }
  1797. skb->len = len;
  1798. skb_set_tail_pointer(skb, len);
  1799. }
  1800. void skb_trim(struct sk_buff *skb, unsigned int len);
  1801. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1802. {
  1803. if (skb->data_len)
  1804. return ___pskb_trim(skb, len);
  1805. __skb_trim(skb, len);
  1806. return 0;
  1807. }
  1808. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1809. {
  1810. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1811. }
  1812. /**
  1813. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1814. * @skb: buffer to alter
  1815. * @len: new length
  1816. *
  1817. * This is identical to pskb_trim except that the caller knows that
  1818. * the skb is not cloned so we should never get an error due to out-
  1819. * of-memory.
  1820. */
  1821. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  1822. {
  1823. int err = pskb_trim(skb, len);
  1824. BUG_ON(err);
  1825. }
  1826. /**
  1827. * skb_orphan - orphan a buffer
  1828. * @skb: buffer to orphan
  1829. *
  1830. * If a buffer currently has an owner then we call the owner's
  1831. * destructor function and make the @skb unowned. The buffer continues
  1832. * to exist but is no longer charged to its former owner.
  1833. */
  1834. static inline void skb_orphan(struct sk_buff *skb)
  1835. {
  1836. if (skb->destructor) {
  1837. skb->destructor(skb);
  1838. skb->destructor = NULL;
  1839. skb->sk = NULL;
  1840. } else {
  1841. BUG_ON(skb->sk);
  1842. }
  1843. }
  1844. /**
  1845. * skb_orphan_frags - orphan the frags contained in a buffer
  1846. * @skb: buffer to orphan frags from
  1847. * @gfp_mask: allocation mask for replacement pages
  1848. *
  1849. * For each frag in the SKB which needs a destructor (i.e. has an
  1850. * owner) create a copy of that frag and release the original
  1851. * page by calling the destructor.
  1852. */
  1853. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  1854. {
  1855. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  1856. return 0;
  1857. return skb_copy_ubufs(skb, gfp_mask);
  1858. }
  1859. /**
  1860. * __skb_queue_purge - empty a list
  1861. * @list: list to empty
  1862. *
  1863. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  1864. * the list and one reference dropped. This function does not take the
  1865. * list lock and the caller must hold the relevant locks to use it.
  1866. */
  1867. void skb_queue_purge(struct sk_buff_head *list);
  1868. static inline void __skb_queue_purge(struct sk_buff_head *list)
  1869. {
  1870. struct sk_buff *skb;
  1871. while ((skb = __skb_dequeue(list)) != NULL)
  1872. kfree_skb(skb);
  1873. }
  1874. #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
  1875. #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
  1876. #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
  1877. void *netdev_alloc_frag(unsigned int fragsz);
  1878. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  1879. gfp_t gfp_mask);
  1880. /**
  1881. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  1882. * @dev: network device to receive on
  1883. * @length: length to allocate
  1884. *
  1885. * Allocate a new &sk_buff and assign it a usage count of one. The
  1886. * buffer has unspecified headroom built in. Users should allocate
  1887. * the headroom they think they need without accounting for the
  1888. * built in space. The built in space is used for optimisations.
  1889. *
  1890. * %NULL is returned if there is no free memory. Although this function
  1891. * allocates memory it can be called from an interrupt.
  1892. */
  1893. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  1894. unsigned int length)
  1895. {
  1896. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  1897. }
  1898. /* legacy helper around __netdev_alloc_skb() */
  1899. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  1900. gfp_t gfp_mask)
  1901. {
  1902. return __netdev_alloc_skb(NULL, length, gfp_mask);
  1903. }
  1904. /* legacy helper around netdev_alloc_skb() */
  1905. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  1906. {
  1907. return netdev_alloc_skb(NULL, length);
  1908. }
  1909. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  1910. unsigned int length, gfp_t gfp)
  1911. {
  1912. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  1913. if (NET_IP_ALIGN && skb)
  1914. skb_reserve(skb, NET_IP_ALIGN);
  1915. return skb;
  1916. }
  1917. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  1918. unsigned int length)
  1919. {
  1920. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  1921. }
  1922. void *napi_alloc_frag(unsigned int fragsz);
  1923. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  1924. unsigned int length, gfp_t gfp_mask);
  1925. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  1926. unsigned int length)
  1927. {
  1928. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  1929. }
  1930. /**
  1931. * __dev_alloc_pages - allocate page for network Rx
  1932. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1933. * @order: size of the allocation
  1934. *
  1935. * Allocate a new page.
  1936. *
  1937. * %NULL is returned if there is no free memory.
  1938. */
  1939. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  1940. unsigned int order)
  1941. {
  1942. /* This piece of code contains several assumptions.
  1943. * 1. This is for device Rx, therefor a cold page is preferred.
  1944. * 2. The expectation is the user wants a compound page.
  1945. * 3. If requesting a order 0 page it will not be compound
  1946. * due to the check to see if order has a value in prep_new_page
  1947. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  1948. * code in gfp_to_alloc_flags that should be enforcing this.
  1949. */
  1950. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  1951. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  1952. }
  1953. static inline struct page *dev_alloc_pages(unsigned int order)
  1954. {
  1955. return __dev_alloc_pages(GFP_ATOMIC, order);
  1956. }
  1957. /**
  1958. * __dev_alloc_page - allocate a page for network Rx
  1959. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1960. *
  1961. * Allocate a new page.
  1962. *
  1963. * %NULL is returned if there is no free memory.
  1964. */
  1965. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  1966. {
  1967. return __dev_alloc_pages(gfp_mask, 0);
  1968. }
  1969. static inline struct page *dev_alloc_page(void)
  1970. {
  1971. return __dev_alloc_page(GFP_ATOMIC);
  1972. }
  1973. /**
  1974. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  1975. * @page: The page that was allocated from skb_alloc_page
  1976. * @skb: The skb that may need pfmemalloc set
  1977. */
  1978. static inline void skb_propagate_pfmemalloc(struct page *page,
  1979. struct sk_buff *skb)
  1980. {
  1981. if (page && page->pfmemalloc)
  1982. skb->pfmemalloc = true;
  1983. }
  1984. /**
  1985. * skb_frag_page - retrieve the page referred to by a paged fragment
  1986. * @frag: the paged fragment
  1987. *
  1988. * Returns the &struct page associated with @frag.
  1989. */
  1990. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  1991. {
  1992. return frag->page.p;
  1993. }
  1994. /**
  1995. * __skb_frag_ref - take an addition reference on a paged fragment.
  1996. * @frag: the paged fragment
  1997. *
  1998. * Takes an additional reference on the paged fragment @frag.
  1999. */
  2000. static inline void __skb_frag_ref(skb_frag_t *frag)
  2001. {
  2002. get_page(skb_frag_page(frag));
  2003. }
  2004. /**
  2005. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2006. * @skb: the buffer
  2007. * @f: the fragment offset.
  2008. *
  2009. * Takes an additional reference on the @f'th paged fragment of @skb.
  2010. */
  2011. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2012. {
  2013. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2014. }
  2015. /**
  2016. * __skb_frag_unref - release a reference on a paged fragment.
  2017. * @frag: the paged fragment
  2018. *
  2019. * Releases a reference on the paged fragment @frag.
  2020. */
  2021. static inline void __skb_frag_unref(skb_frag_t *frag)
  2022. {
  2023. put_page(skb_frag_page(frag));
  2024. }
  2025. /**
  2026. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2027. * @skb: the buffer
  2028. * @f: the fragment offset
  2029. *
  2030. * Releases a reference on the @f'th paged fragment of @skb.
  2031. */
  2032. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2033. {
  2034. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2035. }
  2036. /**
  2037. * skb_frag_address - gets the address of the data contained in a paged fragment
  2038. * @frag: the paged fragment buffer
  2039. *
  2040. * Returns the address of the data within @frag. The page must already
  2041. * be mapped.
  2042. */
  2043. static inline void *skb_frag_address(const skb_frag_t *frag)
  2044. {
  2045. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2046. }
  2047. /**
  2048. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2049. * @frag: the paged fragment buffer
  2050. *
  2051. * Returns the address of the data within @frag. Checks that the page
  2052. * is mapped and returns %NULL otherwise.
  2053. */
  2054. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2055. {
  2056. void *ptr = page_address(skb_frag_page(frag));
  2057. if (unlikely(!ptr))
  2058. return NULL;
  2059. return ptr + frag->page_offset;
  2060. }
  2061. /**
  2062. * __skb_frag_set_page - sets the page contained in a paged fragment
  2063. * @frag: the paged fragment
  2064. * @page: the page to set
  2065. *
  2066. * Sets the fragment @frag to contain @page.
  2067. */
  2068. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2069. {
  2070. frag->page.p = page;
  2071. }
  2072. /**
  2073. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2074. * @skb: the buffer
  2075. * @f: the fragment offset
  2076. * @page: the page to set
  2077. *
  2078. * Sets the @f'th fragment of @skb to contain @page.
  2079. */
  2080. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2081. struct page *page)
  2082. {
  2083. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2084. }
  2085. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2086. /**
  2087. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2088. * @dev: the device to map the fragment to
  2089. * @frag: the paged fragment to map
  2090. * @offset: the offset within the fragment (starting at the
  2091. * fragment's own offset)
  2092. * @size: the number of bytes to map
  2093. * @dir: the direction of the mapping (%PCI_DMA_*)
  2094. *
  2095. * Maps the page associated with @frag to @device.
  2096. */
  2097. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2098. const skb_frag_t *frag,
  2099. size_t offset, size_t size,
  2100. enum dma_data_direction dir)
  2101. {
  2102. return dma_map_page(dev, skb_frag_page(frag),
  2103. frag->page_offset + offset, size, dir);
  2104. }
  2105. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2106. gfp_t gfp_mask)
  2107. {
  2108. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2109. }
  2110. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2111. gfp_t gfp_mask)
  2112. {
  2113. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2114. }
  2115. /**
  2116. * skb_clone_writable - is the header of a clone writable
  2117. * @skb: buffer to check
  2118. * @len: length up to which to write
  2119. *
  2120. * Returns true if modifying the header part of the cloned buffer
  2121. * does not requires the data to be copied.
  2122. */
  2123. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2124. {
  2125. return !skb_header_cloned(skb) &&
  2126. skb_headroom(skb) + len <= skb->hdr_len;
  2127. }
  2128. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2129. int cloned)
  2130. {
  2131. int delta = 0;
  2132. if (headroom > skb_headroom(skb))
  2133. delta = headroom - skb_headroom(skb);
  2134. if (delta || cloned)
  2135. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2136. GFP_ATOMIC);
  2137. return 0;
  2138. }
  2139. /**
  2140. * skb_cow - copy header of skb when it is required
  2141. * @skb: buffer to cow
  2142. * @headroom: needed headroom
  2143. *
  2144. * If the skb passed lacks sufficient headroom or its data part
  2145. * is shared, data is reallocated. If reallocation fails, an error
  2146. * is returned and original skb is not changed.
  2147. *
  2148. * The result is skb with writable area skb->head...skb->tail
  2149. * and at least @headroom of space at head.
  2150. */
  2151. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2152. {
  2153. return __skb_cow(skb, headroom, skb_cloned(skb));
  2154. }
  2155. /**
  2156. * skb_cow_head - skb_cow but only making the head writable
  2157. * @skb: buffer to cow
  2158. * @headroom: needed headroom
  2159. *
  2160. * This function is identical to skb_cow except that we replace the
  2161. * skb_cloned check by skb_header_cloned. It should be used when
  2162. * you only need to push on some header and do not need to modify
  2163. * the data.
  2164. */
  2165. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2166. {
  2167. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2168. }
  2169. /**
  2170. * skb_padto - pad an skbuff up to a minimal size
  2171. * @skb: buffer to pad
  2172. * @len: minimal length
  2173. *
  2174. * Pads up a buffer to ensure the trailing bytes exist and are
  2175. * blanked. If the buffer already contains sufficient data it
  2176. * is untouched. Otherwise it is extended. Returns zero on
  2177. * success. The skb is freed on error.
  2178. */
  2179. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2180. {
  2181. unsigned int size = skb->len;
  2182. if (likely(size >= len))
  2183. return 0;
  2184. return skb_pad(skb, len - size);
  2185. }
  2186. /**
  2187. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2188. * @skb: buffer to pad
  2189. * @len: minimal length
  2190. *
  2191. * Pads up a buffer to ensure the trailing bytes exist and are
  2192. * blanked. If the buffer already contains sufficient data it
  2193. * is untouched. Otherwise it is extended. Returns zero on
  2194. * success. The skb is freed on error.
  2195. */
  2196. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2197. {
  2198. unsigned int size = skb->len;
  2199. if (unlikely(size < len)) {
  2200. len -= size;
  2201. if (skb_pad(skb, len))
  2202. return -ENOMEM;
  2203. __skb_put(skb, len);
  2204. }
  2205. return 0;
  2206. }
  2207. static inline int skb_add_data(struct sk_buff *skb,
  2208. struct iov_iter *from, int copy)
  2209. {
  2210. const int off = skb->len;
  2211. if (skb->ip_summed == CHECKSUM_NONE) {
  2212. __wsum csum = 0;
  2213. if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
  2214. &csum, from) == copy) {
  2215. skb->csum = csum_block_add(skb->csum, csum, off);
  2216. return 0;
  2217. }
  2218. } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
  2219. return 0;
  2220. __skb_trim(skb, off);
  2221. return -EFAULT;
  2222. }
  2223. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2224. const struct page *page, int off)
  2225. {
  2226. if (i) {
  2227. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2228. return page == skb_frag_page(frag) &&
  2229. off == frag->page_offset + skb_frag_size(frag);
  2230. }
  2231. return false;
  2232. }
  2233. static inline int __skb_linearize(struct sk_buff *skb)
  2234. {
  2235. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2236. }
  2237. /**
  2238. * skb_linearize - convert paged skb to linear one
  2239. * @skb: buffer to linarize
  2240. *
  2241. * If there is no free memory -ENOMEM is returned, otherwise zero
  2242. * is returned and the old skb data released.
  2243. */
  2244. static inline int skb_linearize(struct sk_buff *skb)
  2245. {
  2246. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2247. }
  2248. /**
  2249. * skb_has_shared_frag - can any frag be overwritten
  2250. * @skb: buffer to test
  2251. *
  2252. * Return true if the skb has at least one frag that might be modified
  2253. * by an external entity (as in vmsplice()/sendfile())
  2254. */
  2255. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2256. {
  2257. return skb_is_nonlinear(skb) &&
  2258. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2259. }
  2260. /**
  2261. * skb_linearize_cow - make sure skb is linear and writable
  2262. * @skb: buffer to process
  2263. *
  2264. * If there is no free memory -ENOMEM is returned, otherwise zero
  2265. * is returned and the old skb data released.
  2266. */
  2267. static inline int skb_linearize_cow(struct sk_buff *skb)
  2268. {
  2269. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2270. __skb_linearize(skb) : 0;
  2271. }
  2272. /**
  2273. * skb_postpull_rcsum - update checksum for received skb after pull
  2274. * @skb: buffer to update
  2275. * @start: start of data before pull
  2276. * @len: length of data pulled
  2277. *
  2278. * After doing a pull on a received packet, you need to call this to
  2279. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2280. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2281. */
  2282. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2283. const void *start, unsigned int len)
  2284. {
  2285. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2286. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2287. }
  2288. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2289. /**
  2290. * pskb_trim_rcsum - trim received skb and update checksum
  2291. * @skb: buffer to trim
  2292. * @len: new length
  2293. *
  2294. * This is exactly the same as pskb_trim except that it ensures the
  2295. * checksum of received packets are still valid after the operation.
  2296. */
  2297. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2298. {
  2299. if (likely(len >= skb->len))
  2300. return 0;
  2301. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2302. skb->ip_summed = CHECKSUM_NONE;
  2303. return __pskb_trim(skb, len);
  2304. }
  2305. #define skb_queue_walk(queue, skb) \
  2306. for (skb = (queue)->next; \
  2307. skb != (struct sk_buff *)(queue); \
  2308. skb = skb->next)
  2309. #define skb_queue_walk_safe(queue, skb, tmp) \
  2310. for (skb = (queue)->next, tmp = skb->next; \
  2311. skb != (struct sk_buff *)(queue); \
  2312. skb = tmp, tmp = skb->next)
  2313. #define skb_queue_walk_from(queue, skb) \
  2314. for (; skb != (struct sk_buff *)(queue); \
  2315. skb = skb->next)
  2316. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2317. for (tmp = skb->next; \
  2318. skb != (struct sk_buff *)(queue); \
  2319. skb = tmp, tmp = skb->next)
  2320. #define skb_queue_reverse_walk(queue, skb) \
  2321. for (skb = (queue)->prev; \
  2322. skb != (struct sk_buff *)(queue); \
  2323. skb = skb->prev)
  2324. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2325. for (skb = (queue)->prev, tmp = skb->prev; \
  2326. skb != (struct sk_buff *)(queue); \
  2327. skb = tmp, tmp = skb->prev)
  2328. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2329. for (tmp = skb->prev; \
  2330. skb != (struct sk_buff *)(queue); \
  2331. skb = tmp, tmp = skb->prev)
  2332. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2333. {
  2334. return skb_shinfo(skb)->frag_list != NULL;
  2335. }
  2336. static inline void skb_frag_list_init(struct sk_buff *skb)
  2337. {
  2338. skb_shinfo(skb)->frag_list = NULL;
  2339. }
  2340. static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
  2341. {
  2342. frag->next = skb_shinfo(skb)->frag_list;
  2343. skb_shinfo(skb)->frag_list = frag;
  2344. }
  2345. #define skb_walk_frags(skb, iter) \
  2346. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2347. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2348. int *peeked, int *off, int *err);
  2349. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2350. int *err);
  2351. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2352. struct poll_table_struct *wait);
  2353. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2354. struct iov_iter *to, int size);
  2355. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2356. struct msghdr *msg, int size)
  2357. {
  2358. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2359. }
  2360. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2361. struct msghdr *msg);
  2362. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2363. struct iov_iter *from, int len);
  2364. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2365. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2366. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2367. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2368. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2369. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2370. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2371. int len, __wsum csum);
  2372. int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
  2373. struct pipe_inode_info *pipe, unsigned int len,
  2374. unsigned int flags);
  2375. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2376. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2377. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2378. int len, int hlen);
  2379. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2380. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2381. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2382. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2383. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2384. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2385. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2386. int skb_vlan_pop(struct sk_buff *skb);
  2387. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2388. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2389. {
  2390. return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2391. }
  2392. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2393. {
  2394. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2395. }
  2396. struct skb_checksum_ops {
  2397. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2398. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2399. };
  2400. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2401. __wsum csum, const struct skb_checksum_ops *ops);
  2402. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2403. __wsum csum);
  2404. static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
  2405. int len, void *data, int hlen, void *buffer)
  2406. {
  2407. if (hlen - offset >= len)
  2408. return data + offset;
  2409. if (!skb ||
  2410. skb_copy_bits(skb, offset, buffer, len) < 0)
  2411. return NULL;
  2412. return buffer;
  2413. }
  2414. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  2415. int len, void *buffer)
  2416. {
  2417. return __skb_header_pointer(skb, offset, len, skb->data,
  2418. skb_headlen(skb), buffer);
  2419. }
  2420. /**
  2421. * skb_needs_linearize - check if we need to linearize a given skb
  2422. * depending on the given device features.
  2423. * @skb: socket buffer to check
  2424. * @features: net device features
  2425. *
  2426. * Returns true if either:
  2427. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2428. * 2. skb is fragmented and the device does not support SG.
  2429. */
  2430. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2431. netdev_features_t features)
  2432. {
  2433. return skb_is_nonlinear(skb) &&
  2434. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2435. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2436. }
  2437. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2438. void *to,
  2439. const unsigned int len)
  2440. {
  2441. memcpy(to, skb->data, len);
  2442. }
  2443. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2444. const int offset, void *to,
  2445. const unsigned int len)
  2446. {
  2447. memcpy(to, skb->data + offset, len);
  2448. }
  2449. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2450. const void *from,
  2451. const unsigned int len)
  2452. {
  2453. memcpy(skb->data, from, len);
  2454. }
  2455. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2456. const int offset,
  2457. const void *from,
  2458. const unsigned int len)
  2459. {
  2460. memcpy(skb->data + offset, from, len);
  2461. }
  2462. void skb_init(void);
  2463. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2464. {
  2465. return skb->tstamp;
  2466. }
  2467. /**
  2468. * skb_get_timestamp - get timestamp from a skb
  2469. * @skb: skb to get stamp from
  2470. * @stamp: pointer to struct timeval to store stamp in
  2471. *
  2472. * Timestamps are stored in the skb as offsets to a base timestamp.
  2473. * This function converts the offset back to a struct timeval and stores
  2474. * it in stamp.
  2475. */
  2476. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2477. struct timeval *stamp)
  2478. {
  2479. *stamp = ktime_to_timeval(skb->tstamp);
  2480. }
  2481. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2482. struct timespec *stamp)
  2483. {
  2484. *stamp = ktime_to_timespec(skb->tstamp);
  2485. }
  2486. static inline void __net_timestamp(struct sk_buff *skb)
  2487. {
  2488. skb->tstamp = ktime_get_real();
  2489. }
  2490. static inline ktime_t net_timedelta(ktime_t t)
  2491. {
  2492. return ktime_sub(ktime_get_real(), t);
  2493. }
  2494. static inline ktime_t net_invalid_timestamp(void)
  2495. {
  2496. return ktime_set(0, 0);
  2497. }
  2498. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2499. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2500. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2501. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2502. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2503. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2504. {
  2505. }
  2506. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2507. {
  2508. return false;
  2509. }
  2510. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2511. /**
  2512. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2513. *
  2514. * PHY drivers may accept clones of transmitted packets for
  2515. * timestamping via their phy_driver.txtstamp method. These drivers
  2516. * must call this function to return the skb back to the stack, with
  2517. * or without a timestamp.
  2518. *
  2519. * @skb: clone of the the original outgoing packet
  2520. * @hwtstamps: hardware time stamps, may be NULL if not available
  2521. *
  2522. */
  2523. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2524. struct skb_shared_hwtstamps *hwtstamps);
  2525. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2526. struct skb_shared_hwtstamps *hwtstamps,
  2527. struct sock *sk, int tstype);
  2528. /**
  2529. * skb_tstamp_tx - queue clone of skb with send time stamps
  2530. * @orig_skb: the original outgoing packet
  2531. * @hwtstamps: hardware time stamps, may be NULL if not available
  2532. *
  2533. * If the skb has a socket associated, then this function clones the
  2534. * skb (thus sharing the actual data and optional structures), stores
  2535. * the optional hardware time stamping information (if non NULL) or
  2536. * generates a software time stamp (otherwise), then queues the clone
  2537. * to the error queue of the socket. Errors are silently ignored.
  2538. */
  2539. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2540. struct skb_shared_hwtstamps *hwtstamps);
  2541. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2542. {
  2543. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2544. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2545. skb_tstamp_tx(skb, NULL);
  2546. }
  2547. /**
  2548. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2549. *
  2550. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2551. * function immediately before giving the sk_buff to the MAC hardware.
  2552. *
  2553. * Specifically, one should make absolutely sure that this function is
  2554. * called before TX completion of this packet can trigger. Otherwise
  2555. * the packet could potentially already be freed.
  2556. *
  2557. * @skb: A socket buffer.
  2558. */
  2559. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2560. {
  2561. skb_clone_tx_timestamp(skb);
  2562. sw_tx_timestamp(skb);
  2563. }
  2564. /**
  2565. * skb_complete_wifi_ack - deliver skb with wifi status
  2566. *
  2567. * @skb: the original outgoing packet
  2568. * @acked: ack status
  2569. *
  2570. */
  2571. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2572. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2573. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2574. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2575. {
  2576. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2577. skb->csum_valid ||
  2578. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2579. skb_checksum_start_offset(skb) >= 0));
  2580. }
  2581. /**
  2582. * skb_checksum_complete - Calculate checksum of an entire packet
  2583. * @skb: packet to process
  2584. *
  2585. * This function calculates the checksum over the entire packet plus
  2586. * the value of skb->csum. The latter can be used to supply the
  2587. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2588. * checksum.
  2589. *
  2590. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2591. * this function can be used to verify that checksum on received
  2592. * packets. In that case the function should return zero if the
  2593. * checksum is correct. In particular, this function will return zero
  2594. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2595. * hardware has already verified the correctness of the checksum.
  2596. */
  2597. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2598. {
  2599. return skb_csum_unnecessary(skb) ?
  2600. 0 : __skb_checksum_complete(skb);
  2601. }
  2602. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2603. {
  2604. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2605. if (skb->csum_level == 0)
  2606. skb->ip_summed = CHECKSUM_NONE;
  2607. else
  2608. skb->csum_level--;
  2609. }
  2610. }
  2611. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2612. {
  2613. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2614. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2615. skb->csum_level++;
  2616. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2617. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2618. skb->csum_level = 0;
  2619. }
  2620. }
  2621. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2622. {
  2623. /* Mark current checksum as bad (typically called from GRO
  2624. * path). In the case that ip_summed is CHECKSUM_NONE
  2625. * this must be the first checksum encountered in the packet.
  2626. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2627. * checksum after the last one validated. For UDP, a zero
  2628. * checksum can not be marked as bad.
  2629. */
  2630. if (skb->ip_summed == CHECKSUM_NONE ||
  2631. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2632. skb->csum_bad = 1;
  2633. }
  2634. /* Check if we need to perform checksum complete validation.
  2635. *
  2636. * Returns true if checksum complete is needed, false otherwise
  2637. * (either checksum is unnecessary or zero checksum is allowed).
  2638. */
  2639. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2640. bool zero_okay,
  2641. __sum16 check)
  2642. {
  2643. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2644. skb->csum_valid = 1;
  2645. __skb_decr_checksum_unnecessary(skb);
  2646. return false;
  2647. }
  2648. return true;
  2649. }
  2650. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2651. * in checksum_init.
  2652. */
  2653. #define CHECKSUM_BREAK 76
  2654. /* Unset checksum-complete
  2655. *
  2656. * Unset checksum complete can be done when packet is being modified
  2657. * (uncompressed for instance) and checksum-complete value is
  2658. * invalidated.
  2659. */
  2660. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2661. {
  2662. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2663. skb->ip_summed = CHECKSUM_NONE;
  2664. }
  2665. /* Validate (init) checksum based on checksum complete.
  2666. *
  2667. * Return values:
  2668. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2669. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2670. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2671. * non-zero: value of invalid checksum
  2672. *
  2673. */
  2674. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2675. bool complete,
  2676. __wsum psum)
  2677. {
  2678. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2679. if (!csum_fold(csum_add(psum, skb->csum))) {
  2680. skb->csum_valid = 1;
  2681. return 0;
  2682. }
  2683. } else if (skb->csum_bad) {
  2684. /* ip_summed == CHECKSUM_NONE in this case */
  2685. return 1;
  2686. }
  2687. skb->csum = psum;
  2688. if (complete || skb->len <= CHECKSUM_BREAK) {
  2689. __sum16 csum;
  2690. csum = __skb_checksum_complete(skb);
  2691. skb->csum_valid = !csum;
  2692. return csum;
  2693. }
  2694. return 0;
  2695. }
  2696. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2697. {
  2698. return 0;
  2699. }
  2700. /* Perform checksum validate (init). Note that this is a macro since we only
  2701. * want to calculate the pseudo header which is an input function if necessary.
  2702. * First we try to validate without any computation (checksum unnecessary) and
  2703. * then calculate based on checksum complete calling the function to compute
  2704. * pseudo header.
  2705. *
  2706. * Return values:
  2707. * 0: checksum is validated or try to in skb_checksum_complete
  2708. * non-zero: value of invalid checksum
  2709. */
  2710. #define __skb_checksum_validate(skb, proto, complete, \
  2711. zero_okay, check, compute_pseudo) \
  2712. ({ \
  2713. __sum16 __ret = 0; \
  2714. skb->csum_valid = 0; \
  2715. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2716. __ret = __skb_checksum_validate_complete(skb, \
  2717. complete, compute_pseudo(skb, proto)); \
  2718. __ret; \
  2719. })
  2720. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2721. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2722. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2723. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2724. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2725. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2726. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2727. compute_pseudo) \
  2728. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  2729. #define skb_checksum_simple_validate(skb) \
  2730. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2731. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2732. {
  2733. return (skb->ip_summed == CHECKSUM_NONE &&
  2734. skb->csum_valid && !skb->csum_bad);
  2735. }
  2736. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2737. __sum16 check, __wsum pseudo)
  2738. {
  2739. skb->csum = ~pseudo;
  2740. skb->ip_summed = CHECKSUM_COMPLETE;
  2741. }
  2742. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  2743. do { \
  2744. if (__skb_checksum_convert_check(skb)) \
  2745. __skb_checksum_convert(skb, check, \
  2746. compute_pseudo(skb, proto)); \
  2747. } while (0)
  2748. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  2749. u16 start, u16 offset)
  2750. {
  2751. skb->ip_summed = CHECKSUM_PARTIAL;
  2752. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  2753. skb->csum_offset = offset - start;
  2754. }
  2755. /* Update skbuf and packet to reflect the remote checksum offload operation.
  2756. * When called, ptr indicates the starting point for skb->csum when
  2757. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  2758. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  2759. */
  2760. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  2761. int start, int offset, bool nopartial)
  2762. {
  2763. __wsum delta;
  2764. if (!nopartial) {
  2765. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  2766. return;
  2767. }
  2768. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  2769. __skb_checksum_complete(skb);
  2770. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  2771. }
  2772. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  2773. /* Adjust skb->csum since we changed the packet */
  2774. skb->csum = csum_add(skb->csum, delta);
  2775. }
  2776. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2777. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2778. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2779. {
  2780. if (nfct && atomic_dec_and_test(&nfct->use))
  2781. nf_conntrack_destroy(nfct);
  2782. }
  2783. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2784. {
  2785. if (nfct)
  2786. atomic_inc(&nfct->use);
  2787. }
  2788. #endif
  2789. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2790. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  2791. {
  2792. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  2793. kfree(nf_bridge);
  2794. }
  2795. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  2796. {
  2797. if (nf_bridge)
  2798. atomic_inc(&nf_bridge->use);
  2799. }
  2800. #endif /* CONFIG_BRIDGE_NETFILTER */
  2801. static inline void nf_reset(struct sk_buff *skb)
  2802. {
  2803. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2804. nf_conntrack_put(skb->nfct);
  2805. skb->nfct = NULL;
  2806. #endif
  2807. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2808. nf_bridge_put(skb->nf_bridge);
  2809. skb->nf_bridge = NULL;
  2810. #endif
  2811. }
  2812. static inline void nf_reset_trace(struct sk_buff *skb)
  2813. {
  2814. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2815. skb->nf_trace = 0;
  2816. #endif
  2817. }
  2818. /* Note: This doesn't put any conntrack and bridge info in dst. */
  2819. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  2820. bool copy)
  2821. {
  2822. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2823. dst->nfct = src->nfct;
  2824. nf_conntrack_get(src->nfct);
  2825. if (copy)
  2826. dst->nfctinfo = src->nfctinfo;
  2827. #endif
  2828. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2829. dst->nf_bridge = src->nf_bridge;
  2830. nf_bridge_get(src->nf_bridge);
  2831. #endif
  2832. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2833. if (copy)
  2834. dst->nf_trace = src->nf_trace;
  2835. #endif
  2836. }
  2837. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2838. {
  2839. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2840. nf_conntrack_put(dst->nfct);
  2841. #endif
  2842. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2843. nf_bridge_put(dst->nf_bridge);
  2844. #endif
  2845. __nf_copy(dst, src, true);
  2846. }
  2847. #ifdef CONFIG_NETWORK_SECMARK
  2848. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2849. {
  2850. to->secmark = from->secmark;
  2851. }
  2852. static inline void skb_init_secmark(struct sk_buff *skb)
  2853. {
  2854. skb->secmark = 0;
  2855. }
  2856. #else
  2857. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2858. { }
  2859. static inline void skb_init_secmark(struct sk_buff *skb)
  2860. { }
  2861. #endif
  2862. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  2863. {
  2864. return !skb->destructor &&
  2865. #if IS_ENABLED(CONFIG_XFRM)
  2866. !skb->sp &&
  2867. #endif
  2868. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  2869. !skb->nfct &&
  2870. #endif
  2871. !skb->_skb_refdst &&
  2872. !skb_has_frag_list(skb);
  2873. }
  2874. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  2875. {
  2876. skb->queue_mapping = queue_mapping;
  2877. }
  2878. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  2879. {
  2880. return skb->queue_mapping;
  2881. }
  2882. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  2883. {
  2884. to->queue_mapping = from->queue_mapping;
  2885. }
  2886. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  2887. {
  2888. skb->queue_mapping = rx_queue + 1;
  2889. }
  2890. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  2891. {
  2892. return skb->queue_mapping - 1;
  2893. }
  2894. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  2895. {
  2896. return skb->queue_mapping != 0;
  2897. }
  2898. u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
  2899. unsigned int num_tx_queues);
  2900. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  2901. {
  2902. #ifdef CONFIG_XFRM
  2903. return skb->sp;
  2904. #else
  2905. return NULL;
  2906. #endif
  2907. }
  2908. /* Keeps track of mac header offset relative to skb->head.
  2909. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  2910. * For non-tunnel skb it points to skb_mac_header() and for
  2911. * tunnel skb it points to outer mac header.
  2912. * Keeps track of level of encapsulation of network headers.
  2913. */
  2914. struct skb_gso_cb {
  2915. int mac_offset;
  2916. int encap_level;
  2917. __u16 csum_start;
  2918. };
  2919. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
  2920. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  2921. {
  2922. return (skb_mac_header(inner_skb) - inner_skb->head) -
  2923. SKB_GSO_CB(inner_skb)->mac_offset;
  2924. }
  2925. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  2926. {
  2927. int new_headroom, headroom;
  2928. int ret;
  2929. headroom = skb_headroom(skb);
  2930. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  2931. if (ret)
  2932. return ret;
  2933. new_headroom = skb_headroom(skb);
  2934. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  2935. return 0;
  2936. }
  2937. /* Compute the checksum for a gso segment. First compute the checksum value
  2938. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  2939. * then add in skb->csum (checksum from csum_start to end of packet).
  2940. * skb->csum and csum_start are then updated to reflect the checksum of the
  2941. * resultant packet starting from the transport header-- the resultant checksum
  2942. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  2943. * header.
  2944. */
  2945. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  2946. {
  2947. int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
  2948. skb_transport_offset(skb);
  2949. __u16 csum;
  2950. csum = csum_fold(csum_partial(skb_transport_header(skb),
  2951. plen, skb->csum));
  2952. skb->csum = res;
  2953. SKB_GSO_CB(skb)->csum_start -= plen;
  2954. return csum;
  2955. }
  2956. static inline bool skb_is_gso(const struct sk_buff *skb)
  2957. {
  2958. return skb_shinfo(skb)->gso_size;
  2959. }
  2960. /* Note: Should be called only if skb_is_gso(skb) is true */
  2961. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  2962. {
  2963. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  2964. }
  2965. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  2966. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  2967. {
  2968. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  2969. * wanted then gso_type will be set. */
  2970. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  2971. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  2972. unlikely(shinfo->gso_type == 0)) {
  2973. __skb_warn_lro_forwarding(skb);
  2974. return true;
  2975. }
  2976. return false;
  2977. }
  2978. static inline void skb_forward_csum(struct sk_buff *skb)
  2979. {
  2980. /* Unfortunately we don't support this one. Any brave souls? */
  2981. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2982. skb->ip_summed = CHECKSUM_NONE;
  2983. }
  2984. /**
  2985. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  2986. * @skb: skb to check
  2987. *
  2988. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  2989. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  2990. * use this helper, to document places where we make this assertion.
  2991. */
  2992. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  2993. {
  2994. #ifdef DEBUG
  2995. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  2996. #endif
  2997. }
  2998. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  2999. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3000. u32 skb_get_poff(const struct sk_buff *skb);
  3001. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  3002. const struct flow_keys *keys, int hlen);
  3003. /**
  3004. * skb_head_is_locked - Determine if the skb->head is locked down
  3005. * @skb: skb to check
  3006. *
  3007. * The head on skbs build around a head frag can be removed if they are
  3008. * not cloned. This function returns true if the skb head is locked down
  3009. * due to either being allocated via kmalloc, or by being a clone with
  3010. * multiple references to the head.
  3011. */
  3012. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3013. {
  3014. return !skb->head_frag || skb_cloned(skb);
  3015. }
  3016. /**
  3017. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3018. *
  3019. * @skb: GSO skb
  3020. *
  3021. * skb_gso_network_seglen is used to determine the real size of the
  3022. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3023. *
  3024. * The MAC/L2 header is not accounted for.
  3025. */
  3026. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3027. {
  3028. unsigned int hdr_len = skb_transport_header(skb) -
  3029. skb_network_header(skb);
  3030. return hdr_len + skb_gso_transport_seglen(skb);
  3031. }
  3032. #endif /* __KERNEL__ */
  3033. #endif /* _LINUX_SKBUFF_H */