skbuff.h 95 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963296429652966296729682969297029712972297329742975297629772978297929802981298229832984298529862987298829892990299129922993299429952996299729982999300030013002300330043005300630073008300930103011301230133014301530163017301830193020302130223023302430253026302730283029303030313032303330343035303630373038303930403041304230433044304530463047304830493050305130523053305430553056305730583059306030613062306330643065306630673068306930703071307230733074307530763077307830793080308130823083308430853086308730883089309030913092309330943095309630973098309931003101310231033104310531063107310831093110311131123113311431153116311731183119312031213122312331243125312631273128312931303131313231333134313531363137313831393140314131423143314431453146314731483149315031513152315331543155315631573158315931603161316231633164316531663167316831693170317131723173317431753176317731783179318031813182318331843185318631873188318931903191319231933194319531963197319831993200320132023203320432053206320732083209321032113212321332143215321632173218321932203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278327932803281328232833284328532863287328832893290329132923293329432953296329732983299330033013302330333043305330633073308330933103311331233133314331533163317331833193320332133223323332433253326332733283329333033313332333333343335333633373338333933403341334233433344334533463347334833493350335133523353335433553356335733583359336033613362336333643365336633673368336933703371337233733374337533763377337833793380338133823383338433853386338733883389
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/kernel.h>
  16. #include <linux/kmemcheck.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/bug.h>
  20. #include <linux/cache.h>
  21. #include <linux/rbtree.h>
  22. #include <linux/socket.h>
  23. #include <linux/atomic.h>
  24. #include <asm/types.h>
  25. #include <linux/spinlock.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #include <linux/rcupdate.h>
  30. #include <linux/hrtimer.h>
  31. #include <linux/dma-mapping.h>
  32. #include <linux/netdev_features.h>
  33. #include <linux/sched.h>
  34. #include <net/flow_keys.h>
  35. /* A. Checksumming of received packets by device.
  36. *
  37. * CHECKSUM_NONE:
  38. *
  39. * Device failed to checksum this packet e.g. due to lack of capabilities.
  40. * The packet contains full (though not verified) checksum in packet but
  41. * not in skb->csum. Thus, skb->csum is undefined in this case.
  42. *
  43. * CHECKSUM_UNNECESSARY:
  44. *
  45. * The hardware you're dealing with doesn't calculate the full checksum
  46. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  47. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  48. * if their checksums are okay. skb->csum is still undefined in this case
  49. * though. It is a bad option, but, unfortunately, nowadays most vendors do
  50. * this. Apparently with the secret goal to sell you new devices, when you
  51. * will add new protocol to your host, f.e. IPv6 8)
  52. *
  53. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  54. * TCP: IPv6 and IPv4.
  55. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  56. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  57. * may perform further validation in this case.
  58. * GRE: only if the checksum is present in the header.
  59. * SCTP: indicates the CRC in SCTP header has been validated.
  60. *
  61. * skb->csum_level indicates the number of consecutive checksums found in
  62. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  63. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  64. * and a device is able to verify the checksums for UDP (possibly zero),
  65. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  66. * two. If the device were only able to verify the UDP checksum and not
  67. * GRE, either because it doesn't support GRE checksum of because GRE
  68. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  69. * not considered in this case).
  70. *
  71. * CHECKSUM_COMPLETE:
  72. *
  73. * This is the most generic way. The device supplied checksum of the _whole_
  74. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  75. * hardware doesn't need to parse L3/L4 headers to implement this.
  76. *
  77. * Note: Even if device supports only some protocols, but is able to produce
  78. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  79. *
  80. * CHECKSUM_PARTIAL:
  81. *
  82. * This is identical to the case for output below. This may occur on a packet
  83. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  84. * on the same host. The packet can be treated in the same way as
  85. * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
  86. * checksum must be filled in by the OS or the hardware.
  87. *
  88. * B. Checksumming on output.
  89. *
  90. * CHECKSUM_NONE:
  91. *
  92. * The skb was already checksummed by the protocol, or a checksum is not
  93. * required.
  94. *
  95. * CHECKSUM_PARTIAL:
  96. *
  97. * The device is required to checksum the packet as seen by hard_start_xmit()
  98. * from skb->csum_start up to the end, and to record/write the checksum at
  99. * offset skb->csum_start + skb->csum_offset.
  100. *
  101. * The device must show its capabilities in dev->features, set up at device
  102. * setup time, e.g. netdev_features.h:
  103. *
  104. * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
  105. * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
  106. * IPv4. Sigh. Vendors like this way for an unknown reason.
  107. * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
  108. * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
  109. * NETIF_F_... - Well, you get the picture.
  110. *
  111. * CHECKSUM_UNNECESSARY:
  112. *
  113. * Normally, the device will do per protocol specific checksumming. Protocol
  114. * implementations that do not want the NIC to perform the checksum
  115. * calculation should use this flag in their outgoing skbs.
  116. *
  117. * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
  118. * offload. Correspondingly, the FCoE protocol driver
  119. * stack should use CHECKSUM_UNNECESSARY.
  120. *
  121. * Any questions? No questions, good. --ANK
  122. */
  123. /* Don't change this without changing skb_csum_unnecessary! */
  124. #define CHECKSUM_NONE 0
  125. #define CHECKSUM_UNNECESSARY 1
  126. #define CHECKSUM_COMPLETE 2
  127. #define CHECKSUM_PARTIAL 3
  128. /* Maximum value in skb->csum_level */
  129. #define SKB_MAX_CSUM_LEVEL 3
  130. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  131. #define SKB_WITH_OVERHEAD(X) \
  132. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  133. #define SKB_MAX_ORDER(X, ORDER) \
  134. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  135. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  136. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  137. /* return minimum truesize of one skb containing X bytes of data */
  138. #define SKB_TRUESIZE(X) ((X) + \
  139. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  140. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  141. struct net_device;
  142. struct scatterlist;
  143. struct pipe_inode_info;
  144. struct iov_iter;
  145. struct napi_struct;
  146. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  147. struct nf_conntrack {
  148. atomic_t use;
  149. };
  150. #endif
  151. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  152. struct nf_bridge_info {
  153. atomic_t use;
  154. unsigned int mask;
  155. struct net_device *physindev;
  156. struct net_device *physoutdev;
  157. unsigned long data[32 / sizeof(unsigned long)];
  158. };
  159. #endif
  160. struct sk_buff_head {
  161. /* These two members must be first. */
  162. struct sk_buff *next;
  163. struct sk_buff *prev;
  164. __u32 qlen;
  165. spinlock_t lock;
  166. };
  167. struct sk_buff;
  168. /* To allow 64K frame to be packed as single skb without frag_list we
  169. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  170. * buffers which do not start on a page boundary.
  171. *
  172. * Since GRO uses frags we allocate at least 16 regardless of page
  173. * size.
  174. */
  175. #if (65536/PAGE_SIZE + 1) < 16
  176. #define MAX_SKB_FRAGS 16UL
  177. #else
  178. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  179. #endif
  180. typedef struct skb_frag_struct skb_frag_t;
  181. struct skb_frag_struct {
  182. struct {
  183. struct page *p;
  184. } page;
  185. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  186. __u32 page_offset;
  187. __u32 size;
  188. #else
  189. __u16 page_offset;
  190. __u16 size;
  191. #endif
  192. };
  193. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  194. {
  195. return frag->size;
  196. }
  197. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  198. {
  199. frag->size = size;
  200. }
  201. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  202. {
  203. frag->size += delta;
  204. }
  205. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  206. {
  207. frag->size -= delta;
  208. }
  209. #define HAVE_HW_TIME_STAMP
  210. /**
  211. * struct skb_shared_hwtstamps - hardware time stamps
  212. * @hwtstamp: hardware time stamp transformed into duration
  213. * since arbitrary point in time
  214. *
  215. * Software time stamps generated by ktime_get_real() are stored in
  216. * skb->tstamp.
  217. *
  218. * hwtstamps can only be compared against other hwtstamps from
  219. * the same device.
  220. *
  221. * This structure is attached to packets as part of the
  222. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  223. */
  224. struct skb_shared_hwtstamps {
  225. ktime_t hwtstamp;
  226. };
  227. /* Definitions for tx_flags in struct skb_shared_info */
  228. enum {
  229. /* generate hardware time stamp */
  230. SKBTX_HW_TSTAMP = 1 << 0,
  231. /* generate software time stamp when queueing packet to NIC */
  232. SKBTX_SW_TSTAMP = 1 << 1,
  233. /* device driver is going to provide hardware time stamp */
  234. SKBTX_IN_PROGRESS = 1 << 2,
  235. /* device driver supports TX zero-copy buffers */
  236. SKBTX_DEV_ZEROCOPY = 1 << 3,
  237. /* generate wifi status information (where possible) */
  238. SKBTX_WIFI_STATUS = 1 << 4,
  239. /* This indicates at least one fragment might be overwritten
  240. * (as in vmsplice(), sendfile() ...)
  241. * If we need to compute a TX checksum, we'll need to copy
  242. * all frags to avoid possible bad checksum
  243. */
  244. SKBTX_SHARED_FRAG = 1 << 5,
  245. /* generate software time stamp when entering packet scheduling */
  246. SKBTX_SCHED_TSTAMP = 1 << 6,
  247. /* generate software timestamp on peer data acknowledgment */
  248. SKBTX_ACK_TSTAMP = 1 << 7,
  249. };
  250. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  251. SKBTX_SCHED_TSTAMP | \
  252. SKBTX_ACK_TSTAMP)
  253. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  254. /*
  255. * The callback notifies userspace to release buffers when skb DMA is done in
  256. * lower device, the skb last reference should be 0 when calling this.
  257. * The zerocopy_success argument is true if zero copy transmit occurred,
  258. * false on data copy or out of memory error caused by data copy attempt.
  259. * The ctx field is used to track device context.
  260. * The desc field is used to track userspace buffer index.
  261. */
  262. struct ubuf_info {
  263. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  264. void *ctx;
  265. unsigned long desc;
  266. };
  267. /* This data is invariant across clones and lives at
  268. * the end of the header data, ie. at skb->end.
  269. */
  270. struct skb_shared_info {
  271. unsigned char nr_frags;
  272. __u8 tx_flags;
  273. unsigned short gso_size;
  274. /* Warning: this field is not always filled in (UFO)! */
  275. unsigned short gso_segs;
  276. unsigned short gso_type;
  277. struct sk_buff *frag_list;
  278. struct skb_shared_hwtstamps hwtstamps;
  279. u32 tskey;
  280. __be32 ip6_frag_id;
  281. /*
  282. * Warning : all fields before dataref are cleared in __alloc_skb()
  283. */
  284. atomic_t dataref;
  285. /* Intermediate layers must ensure that destructor_arg
  286. * remains valid until skb destructor */
  287. void * destructor_arg;
  288. /* must be last field, see pskb_expand_head() */
  289. skb_frag_t frags[MAX_SKB_FRAGS];
  290. };
  291. /* We divide dataref into two halves. The higher 16 bits hold references
  292. * to the payload part of skb->data. The lower 16 bits hold references to
  293. * the entire skb->data. A clone of a headerless skb holds the length of
  294. * the header in skb->hdr_len.
  295. *
  296. * All users must obey the rule that the skb->data reference count must be
  297. * greater than or equal to the payload reference count.
  298. *
  299. * Holding a reference to the payload part means that the user does not
  300. * care about modifications to the header part of skb->data.
  301. */
  302. #define SKB_DATAREF_SHIFT 16
  303. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  304. enum {
  305. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  306. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  307. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  308. };
  309. enum {
  310. SKB_GSO_TCPV4 = 1 << 0,
  311. SKB_GSO_UDP = 1 << 1,
  312. /* This indicates the skb is from an untrusted source. */
  313. SKB_GSO_DODGY = 1 << 2,
  314. /* This indicates the tcp segment has CWR set. */
  315. SKB_GSO_TCP_ECN = 1 << 3,
  316. SKB_GSO_TCPV6 = 1 << 4,
  317. SKB_GSO_FCOE = 1 << 5,
  318. SKB_GSO_GRE = 1 << 6,
  319. SKB_GSO_GRE_CSUM = 1 << 7,
  320. SKB_GSO_IPIP = 1 << 8,
  321. SKB_GSO_SIT = 1 << 9,
  322. SKB_GSO_UDP_TUNNEL = 1 << 10,
  323. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  324. SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
  325. };
  326. #if BITS_PER_LONG > 32
  327. #define NET_SKBUFF_DATA_USES_OFFSET 1
  328. #endif
  329. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  330. typedef unsigned int sk_buff_data_t;
  331. #else
  332. typedef unsigned char *sk_buff_data_t;
  333. #endif
  334. /**
  335. * struct skb_mstamp - multi resolution time stamps
  336. * @stamp_us: timestamp in us resolution
  337. * @stamp_jiffies: timestamp in jiffies
  338. */
  339. struct skb_mstamp {
  340. union {
  341. u64 v64;
  342. struct {
  343. u32 stamp_us;
  344. u32 stamp_jiffies;
  345. };
  346. };
  347. };
  348. /**
  349. * skb_mstamp_get - get current timestamp
  350. * @cl: place to store timestamps
  351. */
  352. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  353. {
  354. u64 val = local_clock();
  355. do_div(val, NSEC_PER_USEC);
  356. cl->stamp_us = (u32)val;
  357. cl->stamp_jiffies = (u32)jiffies;
  358. }
  359. /**
  360. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  361. * @t1: pointer to newest sample
  362. * @t0: pointer to oldest sample
  363. */
  364. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  365. const struct skb_mstamp *t0)
  366. {
  367. s32 delta_us = t1->stamp_us - t0->stamp_us;
  368. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  369. /* If delta_us is negative, this might be because interval is too big,
  370. * or local_clock() drift is too big : fallback using jiffies.
  371. */
  372. if (delta_us <= 0 ||
  373. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  374. delta_us = jiffies_to_usecs(delta_jiffies);
  375. return delta_us;
  376. }
  377. /**
  378. * struct sk_buff - socket buffer
  379. * @next: Next buffer in list
  380. * @prev: Previous buffer in list
  381. * @tstamp: Time we arrived/left
  382. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  383. * @sk: Socket we are owned by
  384. * @dev: Device we arrived on/are leaving by
  385. * @cb: Control buffer. Free for use by every layer. Put private vars here
  386. * @_skb_refdst: destination entry (with norefcount bit)
  387. * @sp: the security path, used for xfrm
  388. * @len: Length of actual data
  389. * @data_len: Data length
  390. * @mac_len: Length of link layer header
  391. * @hdr_len: writable header length of cloned skb
  392. * @csum: Checksum (must include start/offset pair)
  393. * @csum_start: Offset from skb->head where checksumming should start
  394. * @csum_offset: Offset from csum_start where checksum should be stored
  395. * @priority: Packet queueing priority
  396. * @ignore_df: allow local fragmentation
  397. * @cloned: Head may be cloned (check refcnt to be sure)
  398. * @ip_summed: Driver fed us an IP checksum
  399. * @nohdr: Payload reference only, must not modify header
  400. * @nfctinfo: Relationship of this skb to the connection
  401. * @pkt_type: Packet class
  402. * @fclone: skbuff clone status
  403. * @ipvs_property: skbuff is owned by ipvs
  404. * @peeked: this packet has been seen already, so stats have been
  405. * done for it, don't do them again
  406. * @nf_trace: netfilter packet trace flag
  407. * @protocol: Packet protocol from driver
  408. * @destructor: Destruct function
  409. * @nfct: Associated connection, if any
  410. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  411. * @skb_iif: ifindex of device we arrived on
  412. * @tc_index: Traffic control index
  413. * @tc_verd: traffic control verdict
  414. * @hash: the packet hash
  415. * @queue_mapping: Queue mapping for multiqueue devices
  416. * @xmit_more: More SKBs are pending for this queue
  417. * @ndisc_nodetype: router type (from link layer)
  418. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  419. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  420. * ports.
  421. * @sw_hash: indicates hash was computed in software stack
  422. * @wifi_acked_valid: wifi_acked was set
  423. * @wifi_acked: whether frame was acked on wifi or not
  424. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  425. * @napi_id: id of the NAPI struct this skb came from
  426. * @secmark: security marking
  427. * @mark: Generic packet mark
  428. * @dropcount: total number of sk_receive_queue overflows
  429. * @vlan_proto: vlan encapsulation protocol
  430. * @vlan_tci: vlan tag control information
  431. * @inner_protocol: Protocol (encapsulation)
  432. * @inner_transport_header: Inner transport layer header (encapsulation)
  433. * @inner_network_header: Network layer header (encapsulation)
  434. * @inner_mac_header: Link layer header (encapsulation)
  435. * @transport_header: Transport layer header
  436. * @network_header: Network layer header
  437. * @mac_header: Link layer header
  438. * @tail: Tail pointer
  439. * @end: End pointer
  440. * @head: Head of buffer
  441. * @data: Data head pointer
  442. * @truesize: Buffer size
  443. * @users: User count - see {datagram,tcp}.c
  444. */
  445. struct sk_buff {
  446. union {
  447. struct {
  448. /* These two members must be first. */
  449. struct sk_buff *next;
  450. struct sk_buff *prev;
  451. union {
  452. ktime_t tstamp;
  453. struct skb_mstamp skb_mstamp;
  454. };
  455. };
  456. struct rb_node rbnode; /* used in netem & tcp stack */
  457. };
  458. struct sock *sk;
  459. struct net_device *dev;
  460. /*
  461. * This is the control buffer. It is free to use for every
  462. * layer. Please put your private variables there. If you
  463. * want to keep them across layers you have to do a skb_clone()
  464. * first. This is owned by whoever has the skb queued ATM.
  465. */
  466. char cb[48] __aligned(8);
  467. unsigned long _skb_refdst;
  468. void (*destructor)(struct sk_buff *skb);
  469. #ifdef CONFIG_XFRM
  470. struct sec_path *sp;
  471. #endif
  472. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  473. struct nf_conntrack *nfct;
  474. #endif
  475. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  476. struct nf_bridge_info *nf_bridge;
  477. #endif
  478. unsigned int len,
  479. data_len;
  480. __u16 mac_len,
  481. hdr_len;
  482. /* Following fields are _not_ copied in __copy_skb_header()
  483. * Note that queue_mapping is here mostly to fill a hole.
  484. */
  485. kmemcheck_bitfield_begin(flags1);
  486. __u16 queue_mapping;
  487. __u8 cloned:1,
  488. nohdr:1,
  489. fclone:2,
  490. peeked:1,
  491. head_frag:1,
  492. xmit_more:1;
  493. /* one bit hole */
  494. kmemcheck_bitfield_end(flags1);
  495. /* fields enclosed in headers_start/headers_end are copied
  496. * using a single memcpy() in __copy_skb_header()
  497. */
  498. /* private: */
  499. __u32 headers_start[0];
  500. /* public: */
  501. /* if you move pkt_type around you also must adapt those constants */
  502. #ifdef __BIG_ENDIAN_BITFIELD
  503. #define PKT_TYPE_MAX (7 << 5)
  504. #else
  505. #define PKT_TYPE_MAX 7
  506. #endif
  507. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  508. __u8 __pkt_type_offset[0];
  509. __u8 pkt_type:3;
  510. __u8 pfmemalloc:1;
  511. __u8 ignore_df:1;
  512. __u8 nfctinfo:3;
  513. __u8 nf_trace:1;
  514. __u8 ip_summed:2;
  515. __u8 ooo_okay:1;
  516. __u8 l4_hash:1;
  517. __u8 sw_hash:1;
  518. __u8 wifi_acked_valid:1;
  519. __u8 wifi_acked:1;
  520. __u8 no_fcs:1;
  521. /* Indicates the inner headers are valid in the skbuff. */
  522. __u8 encapsulation:1;
  523. __u8 encap_hdr_csum:1;
  524. __u8 csum_valid:1;
  525. __u8 csum_complete_sw:1;
  526. __u8 csum_level:2;
  527. __u8 csum_bad:1;
  528. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  529. __u8 ndisc_nodetype:2;
  530. #endif
  531. __u8 ipvs_property:1;
  532. __u8 inner_protocol_type:1;
  533. __u8 remcsum_offload:1;
  534. /* 3 or 5 bit hole */
  535. #ifdef CONFIG_NET_SCHED
  536. __u16 tc_index; /* traffic control index */
  537. #ifdef CONFIG_NET_CLS_ACT
  538. __u16 tc_verd; /* traffic control verdict */
  539. #endif
  540. #endif
  541. union {
  542. __wsum csum;
  543. struct {
  544. __u16 csum_start;
  545. __u16 csum_offset;
  546. };
  547. };
  548. __u32 priority;
  549. int skb_iif;
  550. __u32 hash;
  551. __be16 vlan_proto;
  552. __u16 vlan_tci;
  553. #ifdef CONFIG_NET_RX_BUSY_POLL
  554. unsigned int napi_id;
  555. #endif
  556. #ifdef CONFIG_NETWORK_SECMARK
  557. __u32 secmark;
  558. #endif
  559. union {
  560. __u32 mark;
  561. __u32 dropcount;
  562. __u32 reserved_tailroom;
  563. };
  564. union {
  565. __be16 inner_protocol;
  566. __u8 inner_ipproto;
  567. };
  568. __u16 inner_transport_header;
  569. __u16 inner_network_header;
  570. __u16 inner_mac_header;
  571. __be16 protocol;
  572. __u16 transport_header;
  573. __u16 network_header;
  574. __u16 mac_header;
  575. /* private: */
  576. __u32 headers_end[0];
  577. /* public: */
  578. /* These elements must be at the end, see alloc_skb() for details. */
  579. sk_buff_data_t tail;
  580. sk_buff_data_t end;
  581. unsigned char *head,
  582. *data;
  583. unsigned int truesize;
  584. atomic_t users;
  585. };
  586. #ifdef __KERNEL__
  587. /*
  588. * Handling routines are only of interest to the kernel
  589. */
  590. #include <linux/slab.h>
  591. #define SKB_ALLOC_FCLONE 0x01
  592. #define SKB_ALLOC_RX 0x02
  593. #define SKB_ALLOC_NAPI 0x04
  594. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  595. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  596. {
  597. return unlikely(skb->pfmemalloc);
  598. }
  599. /*
  600. * skb might have a dst pointer attached, refcounted or not.
  601. * _skb_refdst low order bit is set if refcount was _not_ taken
  602. */
  603. #define SKB_DST_NOREF 1UL
  604. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  605. /**
  606. * skb_dst - returns skb dst_entry
  607. * @skb: buffer
  608. *
  609. * Returns skb dst_entry, regardless of reference taken or not.
  610. */
  611. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  612. {
  613. /* If refdst was not refcounted, check we still are in a
  614. * rcu_read_lock section
  615. */
  616. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  617. !rcu_read_lock_held() &&
  618. !rcu_read_lock_bh_held());
  619. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  620. }
  621. /**
  622. * skb_dst_set - sets skb dst
  623. * @skb: buffer
  624. * @dst: dst entry
  625. *
  626. * Sets skb dst, assuming a reference was taken on dst and should
  627. * be released by skb_dst_drop()
  628. */
  629. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  630. {
  631. skb->_skb_refdst = (unsigned long)dst;
  632. }
  633. /**
  634. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  635. * @skb: buffer
  636. * @dst: dst entry
  637. *
  638. * Sets skb dst, assuming a reference was not taken on dst.
  639. * If dst entry is cached, we do not take reference and dst_release
  640. * will be avoided by refdst_drop. If dst entry is not cached, we take
  641. * reference, so that last dst_release can destroy the dst immediately.
  642. */
  643. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  644. {
  645. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  646. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  647. }
  648. /**
  649. * skb_dst_is_noref - Test if skb dst isn't refcounted
  650. * @skb: buffer
  651. */
  652. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  653. {
  654. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  655. }
  656. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  657. {
  658. return (struct rtable *)skb_dst(skb);
  659. }
  660. void kfree_skb(struct sk_buff *skb);
  661. void kfree_skb_list(struct sk_buff *segs);
  662. void skb_tx_error(struct sk_buff *skb);
  663. void consume_skb(struct sk_buff *skb);
  664. void __kfree_skb(struct sk_buff *skb);
  665. extern struct kmem_cache *skbuff_head_cache;
  666. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  667. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  668. bool *fragstolen, int *delta_truesize);
  669. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  670. int node);
  671. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  672. static inline struct sk_buff *alloc_skb(unsigned int size,
  673. gfp_t priority)
  674. {
  675. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  676. }
  677. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  678. unsigned long data_len,
  679. int max_page_order,
  680. int *errcode,
  681. gfp_t gfp_mask);
  682. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  683. struct sk_buff_fclones {
  684. struct sk_buff skb1;
  685. struct sk_buff skb2;
  686. atomic_t fclone_ref;
  687. };
  688. /**
  689. * skb_fclone_busy - check if fclone is busy
  690. * @skb: buffer
  691. *
  692. * Returns true is skb is a fast clone, and its clone is not freed.
  693. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  694. * so we also check that this didnt happen.
  695. */
  696. static inline bool skb_fclone_busy(const struct sock *sk,
  697. const struct sk_buff *skb)
  698. {
  699. const struct sk_buff_fclones *fclones;
  700. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  701. return skb->fclone == SKB_FCLONE_ORIG &&
  702. atomic_read(&fclones->fclone_ref) > 1 &&
  703. fclones->skb2.sk == sk;
  704. }
  705. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  706. gfp_t priority)
  707. {
  708. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  709. }
  710. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  711. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  712. {
  713. return __alloc_skb_head(priority, -1);
  714. }
  715. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  716. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  717. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  718. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  719. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  720. gfp_t gfp_mask, bool fclone);
  721. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  722. gfp_t gfp_mask)
  723. {
  724. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  725. }
  726. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  727. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  728. unsigned int headroom);
  729. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  730. int newtailroom, gfp_t priority);
  731. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  732. int offset, int len);
  733. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  734. int len);
  735. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  736. int skb_pad(struct sk_buff *skb, int pad);
  737. #define dev_kfree_skb(a) consume_skb(a)
  738. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  739. int getfrag(void *from, char *to, int offset,
  740. int len, int odd, struct sk_buff *skb),
  741. void *from, int length);
  742. struct skb_seq_state {
  743. __u32 lower_offset;
  744. __u32 upper_offset;
  745. __u32 frag_idx;
  746. __u32 stepped_offset;
  747. struct sk_buff *root_skb;
  748. struct sk_buff *cur_skb;
  749. __u8 *frag_data;
  750. };
  751. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  752. unsigned int to, struct skb_seq_state *st);
  753. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  754. struct skb_seq_state *st);
  755. void skb_abort_seq_read(struct skb_seq_state *st);
  756. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  757. unsigned int to, struct ts_config *config,
  758. struct ts_state *state);
  759. /*
  760. * Packet hash types specify the type of hash in skb_set_hash.
  761. *
  762. * Hash types refer to the protocol layer addresses which are used to
  763. * construct a packet's hash. The hashes are used to differentiate or identify
  764. * flows of the protocol layer for the hash type. Hash types are either
  765. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  766. *
  767. * Properties of hashes:
  768. *
  769. * 1) Two packets in different flows have different hash values
  770. * 2) Two packets in the same flow should have the same hash value
  771. *
  772. * A hash at a higher layer is considered to be more specific. A driver should
  773. * set the most specific hash possible.
  774. *
  775. * A driver cannot indicate a more specific hash than the layer at which a hash
  776. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  777. *
  778. * A driver may indicate a hash level which is less specific than the
  779. * actual layer the hash was computed on. For instance, a hash computed
  780. * at L4 may be considered an L3 hash. This should only be done if the
  781. * driver can't unambiguously determine that the HW computed the hash at
  782. * the higher layer. Note that the "should" in the second property above
  783. * permits this.
  784. */
  785. enum pkt_hash_types {
  786. PKT_HASH_TYPE_NONE, /* Undefined type */
  787. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  788. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  789. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  790. };
  791. static inline void
  792. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  793. {
  794. skb->l4_hash = (type == PKT_HASH_TYPE_L4);
  795. skb->sw_hash = 0;
  796. skb->hash = hash;
  797. }
  798. void __skb_get_hash(struct sk_buff *skb);
  799. static inline __u32 skb_get_hash(struct sk_buff *skb)
  800. {
  801. if (!skb->l4_hash && !skb->sw_hash)
  802. __skb_get_hash(skb);
  803. return skb->hash;
  804. }
  805. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  806. {
  807. return skb->hash;
  808. }
  809. static inline void skb_clear_hash(struct sk_buff *skb)
  810. {
  811. skb->hash = 0;
  812. skb->sw_hash = 0;
  813. skb->l4_hash = 0;
  814. }
  815. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  816. {
  817. if (!skb->l4_hash)
  818. skb_clear_hash(skb);
  819. }
  820. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  821. {
  822. to->hash = from->hash;
  823. to->sw_hash = from->sw_hash;
  824. to->l4_hash = from->l4_hash;
  825. };
  826. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  827. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  828. {
  829. return skb->head + skb->end;
  830. }
  831. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  832. {
  833. return skb->end;
  834. }
  835. #else
  836. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  837. {
  838. return skb->end;
  839. }
  840. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  841. {
  842. return skb->end - skb->head;
  843. }
  844. #endif
  845. /* Internal */
  846. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  847. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  848. {
  849. return &skb_shinfo(skb)->hwtstamps;
  850. }
  851. /**
  852. * skb_queue_empty - check if a queue is empty
  853. * @list: queue head
  854. *
  855. * Returns true if the queue is empty, false otherwise.
  856. */
  857. static inline int skb_queue_empty(const struct sk_buff_head *list)
  858. {
  859. return list->next == (const struct sk_buff *) list;
  860. }
  861. /**
  862. * skb_queue_is_last - check if skb is the last entry in the queue
  863. * @list: queue head
  864. * @skb: buffer
  865. *
  866. * Returns true if @skb is the last buffer on the list.
  867. */
  868. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  869. const struct sk_buff *skb)
  870. {
  871. return skb->next == (const struct sk_buff *) list;
  872. }
  873. /**
  874. * skb_queue_is_first - check if skb is the first entry in the queue
  875. * @list: queue head
  876. * @skb: buffer
  877. *
  878. * Returns true if @skb is the first buffer on the list.
  879. */
  880. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  881. const struct sk_buff *skb)
  882. {
  883. return skb->prev == (const struct sk_buff *) list;
  884. }
  885. /**
  886. * skb_queue_next - return the next packet in the queue
  887. * @list: queue head
  888. * @skb: current buffer
  889. *
  890. * Return the next packet in @list after @skb. It is only valid to
  891. * call this if skb_queue_is_last() evaluates to false.
  892. */
  893. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  894. const struct sk_buff *skb)
  895. {
  896. /* This BUG_ON may seem severe, but if we just return then we
  897. * are going to dereference garbage.
  898. */
  899. BUG_ON(skb_queue_is_last(list, skb));
  900. return skb->next;
  901. }
  902. /**
  903. * skb_queue_prev - return the prev packet in the queue
  904. * @list: queue head
  905. * @skb: current buffer
  906. *
  907. * Return the prev packet in @list before @skb. It is only valid to
  908. * call this if skb_queue_is_first() evaluates to false.
  909. */
  910. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  911. const struct sk_buff *skb)
  912. {
  913. /* This BUG_ON may seem severe, but if we just return then we
  914. * are going to dereference garbage.
  915. */
  916. BUG_ON(skb_queue_is_first(list, skb));
  917. return skb->prev;
  918. }
  919. /**
  920. * skb_get - reference buffer
  921. * @skb: buffer to reference
  922. *
  923. * Makes another reference to a socket buffer and returns a pointer
  924. * to the buffer.
  925. */
  926. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  927. {
  928. atomic_inc(&skb->users);
  929. return skb;
  930. }
  931. /*
  932. * If users == 1, we are the only owner and are can avoid redundant
  933. * atomic change.
  934. */
  935. /**
  936. * skb_cloned - is the buffer a clone
  937. * @skb: buffer to check
  938. *
  939. * Returns true if the buffer was generated with skb_clone() and is
  940. * one of multiple shared copies of the buffer. Cloned buffers are
  941. * shared data so must not be written to under normal circumstances.
  942. */
  943. static inline int skb_cloned(const struct sk_buff *skb)
  944. {
  945. return skb->cloned &&
  946. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  947. }
  948. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  949. {
  950. might_sleep_if(pri & __GFP_WAIT);
  951. if (skb_cloned(skb))
  952. return pskb_expand_head(skb, 0, 0, pri);
  953. return 0;
  954. }
  955. /**
  956. * skb_header_cloned - is the header a clone
  957. * @skb: buffer to check
  958. *
  959. * Returns true if modifying the header part of the buffer requires
  960. * the data to be copied.
  961. */
  962. static inline int skb_header_cloned(const struct sk_buff *skb)
  963. {
  964. int dataref;
  965. if (!skb->cloned)
  966. return 0;
  967. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  968. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  969. return dataref != 1;
  970. }
  971. /**
  972. * skb_header_release - release reference to header
  973. * @skb: buffer to operate on
  974. *
  975. * Drop a reference to the header part of the buffer. This is done
  976. * by acquiring a payload reference. You must not read from the header
  977. * part of skb->data after this.
  978. * Note : Check if you can use __skb_header_release() instead.
  979. */
  980. static inline void skb_header_release(struct sk_buff *skb)
  981. {
  982. BUG_ON(skb->nohdr);
  983. skb->nohdr = 1;
  984. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  985. }
  986. /**
  987. * __skb_header_release - release reference to header
  988. * @skb: buffer to operate on
  989. *
  990. * Variant of skb_header_release() assuming skb is private to caller.
  991. * We can avoid one atomic operation.
  992. */
  993. static inline void __skb_header_release(struct sk_buff *skb)
  994. {
  995. skb->nohdr = 1;
  996. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  997. }
  998. /**
  999. * skb_shared - is the buffer shared
  1000. * @skb: buffer to check
  1001. *
  1002. * Returns true if more than one person has a reference to this
  1003. * buffer.
  1004. */
  1005. static inline int skb_shared(const struct sk_buff *skb)
  1006. {
  1007. return atomic_read(&skb->users) != 1;
  1008. }
  1009. /**
  1010. * skb_share_check - check if buffer is shared and if so clone it
  1011. * @skb: buffer to check
  1012. * @pri: priority for memory allocation
  1013. *
  1014. * If the buffer is shared the buffer is cloned and the old copy
  1015. * drops a reference. A new clone with a single reference is returned.
  1016. * If the buffer is not shared the original buffer is returned. When
  1017. * being called from interrupt status or with spinlocks held pri must
  1018. * be GFP_ATOMIC.
  1019. *
  1020. * NULL is returned on a memory allocation failure.
  1021. */
  1022. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1023. {
  1024. might_sleep_if(pri & __GFP_WAIT);
  1025. if (skb_shared(skb)) {
  1026. struct sk_buff *nskb = skb_clone(skb, pri);
  1027. if (likely(nskb))
  1028. consume_skb(skb);
  1029. else
  1030. kfree_skb(skb);
  1031. skb = nskb;
  1032. }
  1033. return skb;
  1034. }
  1035. /*
  1036. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1037. * packets to handle cases where we have a local reader and forward
  1038. * and a couple of other messy ones. The normal one is tcpdumping
  1039. * a packet thats being forwarded.
  1040. */
  1041. /**
  1042. * skb_unshare - make a copy of a shared buffer
  1043. * @skb: buffer to check
  1044. * @pri: priority for memory allocation
  1045. *
  1046. * If the socket buffer is a clone then this function creates a new
  1047. * copy of the data, drops a reference count on the old copy and returns
  1048. * the new copy with the reference count at 1. If the buffer is not a clone
  1049. * the original buffer is returned. When called with a spinlock held or
  1050. * from interrupt state @pri must be %GFP_ATOMIC
  1051. *
  1052. * %NULL is returned on a memory allocation failure.
  1053. */
  1054. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1055. gfp_t pri)
  1056. {
  1057. might_sleep_if(pri & __GFP_WAIT);
  1058. if (skb_cloned(skb)) {
  1059. struct sk_buff *nskb = skb_copy(skb, pri);
  1060. /* Free our shared copy */
  1061. if (likely(nskb))
  1062. consume_skb(skb);
  1063. else
  1064. kfree_skb(skb);
  1065. skb = nskb;
  1066. }
  1067. return skb;
  1068. }
  1069. /**
  1070. * skb_peek - peek at the head of an &sk_buff_head
  1071. * @list_: list to peek at
  1072. *
  1073. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1074. * be careful with this one. A peek leaves the buffer on the
  1075. * list and someone else may run off with it. You must hold
  1076. * the appropriate locks or have a private queue to do this.
  1077. *
  1078. * Returns %NULL for an empty list or a pointer to the head element.
  1079. * The reference count is not incremented and the reference is therefore
  1080. * volatile. Use with caution.
  1081. */
  1082. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1083. {
  1084. struct sk_buff *skb = list_->next;
  1085. if (skb == (struct sk_buff *)list_)
  1086. skb = NULL;
  1087. return skb;
  1088. }
  1089. /**
  1090. * skb_peek_next - peek skb following the given one from a queue
  1091. * @skb: skb to start from
  1092. * @list_: list to peek at
  1093. *
  1094. * Returns %NULL when the end of the list is met or a pointer to the
  1095. * next element. The reference count is not incremented and the
  1096. * reference is therefore volatile. Use with caution.
  1097. */
  1098. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1099. const struct sk_buff_head *list_)
  1100. {
  1101. struct sk_buff *next = skb->next;
  1102. if (next == (struct sk_buff *)list_)
  1103. next = NULL;
  1104. return next;
  1105. }
  1106. /**
  1107. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1108. * @list_: list to peek at
  1109. *
  1110. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1111. * be careful with this one. A peek leaves the buffer on the
  1112. * list and someone else may run off with it. You must hold
  1113. * the appropriate locks or have a private queue to do this.
  1114. *
  1115. * Returns %NULL for an empty list or a pointer to the tail element.
  1116. * The reference count is not incremented and the reference is therefore
  1117. * volatile. Use with caution.
  1118. */
  1119. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1120. {
  1121. struct sk_buff *skb = list_->prev;
  1122. if (skb == (struct sk_buff *)list_)
  1123. skb = NULL;
  1124. return skb;
  1125. }
  1126. /**
  1127. * skb_queue_len - get queue length
  1128. * @list_: list to measure
  1129. *
  1130. * Return the length of an &sk_buff queue.
  1131. */
  1132. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1133. {
  1134. return list_->qlen;
  1135. }
  1136. /**
  1137. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1138. * @list: queue to initialize
  1139. *
  1140. * This initializes only the list and queue length aspects of
  1141. * an sk_buff_head object. This allows to initialize the list
  1142. * aspects of an sk_buff_head without reinitializing things like
  1143. * the spinlock. It can also be used for on-stack sk_buff_head
  1144. * objects where the spinlock is known to not be used.
  1145. */
  1146. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1147. {
  1148. list->prev = list->next = (struct sk_buff *)list;
  1149. list->qlen = 0;
  1150. }
  1151. /*
  1152. * This function creates a split out lock class for each invocation;
  1153. * this is needed for now since a whole lot of users of the skb-queue
  1154. * infrastructure in drivers have different locking usage (in hardirq)
  1155. * than the networking core (in softirq only). In the long run either the
  1156. * network layer or drivers should need annotation to consolidate the
  1157. * main types of usage into 3 classes.
  1158. */
  1159. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1160. {
  1161. spin_lock_init(&list->lock);
  1162. __skb_queue_head_init(list);
  1163. }
  1164. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1165. struct lock_class_key *class)
  1166. {
  1167. skb_queue_head_init(list);
  1168. lockdep_set_class(&list->lock, class);
  1169. }
  1170. /*
  1171. * Insert an sk_buff on a list.
  1172. *
  1173. * The "__skb_xxxx()" functions are the non-atomic ones that
  1174. * can only be called with interrupts disabled.
  1175. */
  1176. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1177. struct sk_buff_head *list);
  1178. static inline void __skb_insert(struct sk_buff *newsk,
  1179. struct sk_buff *prev, struct sk_buff *next,
  1180. struct sk_buff_head *list)
  1181. {
  1182. newsk->next = next;
  1183. newsk->prev = prev;
  1184. next->prev = prev->next = newsk;
  1185. list->qlen++;
  1186. }
  1187. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1188. struct sk_buff *prev,
  1189. struct sk_buff *next)
  1190. {
  1191. struct sk_buff *first = list->next;
  1192. struct sk_buff *last = list->prev;
  1193. first->prev = prev;
  1194. prev->next = first;
  1195. last->next = next;
  1196. next->prev = last;
  1197. }
  1198. /**
  1199. * skb_queue_splice - join two skb lists, this is designed for stacks
  1200. * @list: the new list to add
  1201. * @head: the place to add it in the first list
  1202. */
  1203. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1204. struct sk_buff_head *head)
  1205. {
  1206. if (!skb_queue_empty(list)) {
  1207. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1208. head->qlen += list->qlen;
  1209. }
  1210. }
  1211. /**
  1212. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1213. * @list: the new list to add
  1214. * @head: the place to add it in the first list
  1215. *
  1216. * The list at @list is reinitialised
  1217. */
  1218. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1219. struct sk_buff_head *head)
  1220. {
  1221. if (!skb_queue_empty(list)) {
  1222. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1223. head->qlen += list->qlen;
  1224. __skb_queue_head_init(list);
  1225. }
  1226. }
  1227. /**
  1228. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1229. * @list: the new list to add
  1230. * @head: the place to add it in the first list
  1231. */
  1232. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1233. struct sk_buff_head *head)
  1234. {
  1235. if (!skb_queue_empty(list)) {
  1236. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1237. head->qlen += list->qlen;
  1238. }
  1239. }
  1240. /**
  1241. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1242. * @list: the new list to add
  1243. * @head: the place to add it in the first list
  1244. *
  1245. * Each of the lists is a queue.
  1246. * The list at @list is reinitialised
  1247. */
  1248. static inline void skb_queue_splice_tail_init(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. __skb_queue_head_init(list);
  1255. }
  1256. }
  1257. /**
  1258. * __skb_queue_after - queue a buffer at the list head
  1259. * @list: list to use
  1260. * @prev: place after this buffer
  1261. * @newsk: buffer to queue
  1262. *
  1263. * Queue a buffer int the middle of a list. This function takes no locks
  1264. * and you must therefore hold required locks before calling it.
  1265. *
  1266. * A buffer cannot be placed on two lists at the same time.
  1267. */
  1268. static inline void __skb_queue_after(struct sk_buff_head *list,
  1269. struct sk_buff *prev,
  1270. struct sk_buff *newsk)
  1271. {
  1272. __skb_insert(newsk, prev, prev->next, list);
  1273. }
  1274. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1275. struct sk_buff_head *list);
  1276. static inline void __skb_queue_before(struct sk_buff_head *list,
  1277. struct sk_buff *next,
  1278. struct sk_buff *newsk)
  1279. {
  1280. __skb_insert(newsk, next->prev, next, list);
  1281. }
  1282. /**
  1283. * __skb_queue_head - queue a buffer at the list head
  1284. * @list: list to use
  1285. * @newsk: buffer to queue
  1286. *
  1287. * Queue a buffer at the start of a list. This function takes no locks
  1288. * and you must therefore hold required locks before calling it.
  1289. *
  1290. * A buffer cannot be placed on two lists at the same time.
  1291. */
  1292. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1293. static inline void __skb_queue_head(struct sk_buff_head *list,
  1294. struct sk_buff *newsk)
  1295. {
  1296. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1297. }
  1298. /**
  1299. * __skb_queue_tail - queue a buffer at the list tail
  1300. * @list: list to use
  1301. * @newsk: buffer to queue
  1302. *
  1303. * Queue a buffer at the end 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_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1309. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1310. struct sk_buff *newsk)
  1311. {
  1312. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1313. }
  1314. /*
  1315. * remove sk_buff from list. _Must_ be called atomically, and with
  1316. * the list known..
  1317. */
  1318. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1319. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1320. {
  1321. struct sk_buff *next, *prev;
  1322. list->qlen--;
  1323. next = skb->next;
  1324. prev = skb->prev;
  1325. skb->next = skb->prev = NULL;
  1326. next->prev = prev;
  1327. prev->next = next;
  1328. }
  1329. /**
  1330. * __skb_dequeue - remove from the head of the queue
  1331. * @list: list to dequeue from
  1332. *
  1333. * Remove the head of the list. This function does not take any locks
  1334. * so must be used with appropriate locks held only. The head item is
  1335. * returned or %NULL if the list is empty.
  1336. */
  1337. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1338. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1339. {
  1340. struct sk_buff *skb = skb_peek(list);
  1341. if (skb)
  1342. __skb_unlink(skb, list);
  1343. return skb;
  1344. }
  1345. /**
  1346. * __skb_dequeue_tail - remove from the tail of the queue
  1347. * @list: list to dequeue from
  1348. *
  1349. * Remove the tail of the list. This function does not take any locks
  1350. * so must be used with appropriate locks held only. The tail item is
  1351. * returned or %NULL if the list is empty.
  1352. */
  1353. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1354. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1355. {
  1356. struct sk_buff *skb = skb_peek_tail(list);
  1357. if (skb)
  1358. __skb_unlink(skb, list);
  1359. return skb;
  1360. }
  1361. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1362. {
  1363. return skb->data_len;
  1364. }
  1365. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1366. {
  1367. return skb->len - skb->data_len;
  1368. }
  1369. static inline int skb_pagelen(const struct sk_buff *skb)
  1370. {
  1371. int i, len = 0;
  1372. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1373. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1374. return len + skb_headlen(skb);
  1375. }
  1376. /**
  1377. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1378. * @skb: buffer containing fragment to be initialised
  1379. * @i: paged fragment index to initialise
  1380. * @page: the page to use for this fragment
  1381. * @off: the offset to the data with @page
  1382. * @size: the length of the data
  1383. *
  1384. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1385. * offset @off within @page.
  1386. *
  1387. * Does not take any additional reference on the fragment.
  1388. */
  1389. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1390. struct page *page, int off, int size)
  1391. {
  1392. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1393. /*
  1394. * Propagate page->pfmemalloc to the skb if we can. The problem is
  1395. * that not all callers have unique ownership of the page. If
  1396. * pfmemalloc is set, we check the mapping as a mapping implies
  1397. * page->index is set (index and pfmemalloc share space).
  1398. * If it's a valid mapping, we cannot use page->pfmemalloc but we
  1399. * do not lose pfmemalloc information as the pages would not be
  1400. * allocated using __GFP_MEMALLOC.
  1401. */
  1402. frag->page.p = page;
  1403. frag->page_offset = off;
  1404. skb_frag_size_set(frag, size);
  1405. page = compound_head(page);
  1406. if (page->pfmemalloc && !page->mapping)
  1407. skb->pfmemalloc = true;
  1408. }
  1409. /**
  1410. * skb_fill_page_desc - initialise a paged fragment in an skb
  1411. * @skb: buffer containing fragment to be initialised
  1412. * @i: paged fragment index to initialise
  1413. * @page: the page to use for this fragment
  1414. * @off: the offset to the data with @page
  1415. * @size: the length of the data
  1416. *
  1417. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1418. * @skb to point to @size bytes at offset @off within @page. In
  1419. * addition updates @skb such that @i is the last fragment.
  1420. *
  1421. * Does not take any additional reference on the fragment.
  1422. */
  1423. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1424. struct page *page, int off, int size)
  1425. {
  1426. __skb_fill_page_desc(skb, i, page, off, size);
  1427. skb_shinfo(skb)->nr_frags = i + 1;
  1428. }
  1429. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1430. int size, unsigned int truesize);
  1431. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1432. unsigned int truesize);
  1433. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1434. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1435. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1436. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1437. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1438. {
  1439. return skb->head + skb->tail;
  1440. }
  1441. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1442. {
  1443. skb->tail = skb->data - skb->head;
  1444. }
  1445. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1446. {
  1447. skb_reset_tail_pointer(skb);
  1448. skb->tail += offset;
  1449. }
  1450. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1451. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1452. {
  1453. return skb->tail;
  1454. }
  1455. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1456. {
  1457. skb->tail = skb->data;
  1458. }
  1459. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1460. {
  1461. skb->tail = skb->data + offset;
  1462. }
  1463. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1464. /*
  1465. * Add data to an sk_buff
  1466. */
  1467. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1468. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1469. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1470. {
  1471. unsigned char *tmp = skb_tail_pointer(skb);
  1472. SKB_LINEAR_ASSERT(skb);
  1473. skb->tail += len;
  1474. skb->len += len;
  1475. return tmp;
  1476. }
  1477. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1478. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1479. {
  1480. skb->data -= len;
  1481. skb->len += len;
  1482. return skb->data;
  1483. }
  1484. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1485. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1486. {
  1487. skb->len -= len;
  1488. BUG_ON(skb->len < skb->data_len);
  1489. return skb->data += len;
  1490. }
  1491. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1492. {
  1493. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1494. }
  1495. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1496. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1497. {
  1498. if (len > skb_headlen(skb) &&
  1499. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1500. return NULL;
  1501. skb->len -= len;
  1502. return skb->data += len;
  1503. }
  1504. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1505. {
  1506. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1507. }
  1508. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1509. {
  1510. if (likely(len <= skb_headlen(skb)))
  1511. return 1;
  1512. if (unlikely(len > skb->len))
  1513. return 0;
  1514. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1515. }
  1516. /**
  1517. * skb_headroom - bytes at buffer head
  1518. * @skb: buffer to check
  1519. *
  1520. * Return the number of bytes of free space at the head of an &sk_buff.
  1521. */
  1522. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1523. {
  1524. return skb->data - skb->head;
  1525. }
  1526. /**
  1527. * skb_tailroom - bytes at buffer end
  1528. * @skb: buffer to check
  1529. *
  1530. * Return the number of bytes of free space at the tail of an sk_buff
  1531. */
  1532. static inline int skb_tailroom(const struct sk_buff *skb)
  1533. {
  1534. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1535. }
  1536. /**
  1537. * skb_availroom - bytes at buffer end
  1538. * @skb: buffer to check
  1539. *
  1540. * Return the number of bytes of free space at the tail of an sk_buff
  1541. * allocated by sk_stream_alloc()
  1542. */
  1543. static inline int skb_availroom(const struct sk_buff *skb)
  1544. {
  1545. if (skb_is_nonlinear(skb))
  1546. return 0;
  1547. return skb->end - skb->tail - skb->reserved_tailroom;
  1548. }
  1549. /**
  1550. * skb_reserve - adjust headroom
  1551. * @skb: buffer to alter
  1552. * @len: bytes to move
  1553. *
  1554. * Increase the headroom of an empty &sk_buff by reducing the tail
  1555. * room. This is only allowed for an empty buffer.
  1556. */
  1557. static inline void skb_reserve(struct sk_buff *skb, int len)
  1558. {
  1559. skb->data += len;
  1560. skb->tail += len;
  1561. }
  1562. #define ENCAP_TYPE_ETHER 0
  1563. #define ENCAP_TYPE_IPPROTO 1
  1564. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1565. __be16 protocol)
  1566. {
  1567. skb->inner_protocol = protocol;
  1568. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1569. }
  1570. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1571. __u8 ipproto)
  1572. {
  1573. skb->inner_ipproto = ipproto;
  1574. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1575. }
  1576. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1577. {
  1578. skb->inner_mac_header = skb->mac_header;
  1579. skb->inner_network_header = skb->network_header;
  1580. skb->inner_transport_header = skb->transport_header;
  1581. }
  1582. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1583. {
  1584. skb->mac_len = skb->network_header - skb->mac_header;
  1585. }
  1586. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1587. *skb)
  1588. {
  1589. return skb->head + skb->inner_transport_header;
  1590. }
  1591. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1592. {
  1593. skb->inner_transport_header = skb->data - skb->head;
  1594. }
  1595. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1596. const int offset)
  1597. {
  1598. skb_reset_inner_transport_header(skb);
  1599. skb->inner_transport_header += offset;
  1600. }
  1601. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1602. {
  1603. return skb->head + skb->inner_network_header;
  1604. }
  1605. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1606. {
  1607. skb->inner_network_header = skb->data - skb->head;
  1608. }
  1609. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1610. const int offset)
  1611. {
  1612. skb_reset_inner_network_header(skb);
  1613. skb->inner_network_header += offset;
  1614. }
  1615. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1616. {
  1617. return skb->head + skb->inner_mac_header;
  1618. }
  1619. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1620. {
  1621. skb->inner_mac_header = skb->data - skb->head;
  1622. }
  1623. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1624. const int offset)
  1625. {
  1626. skb_reset_inner_mac_header(skb);
  1627. skb->inner_mac_header += offset;
  1628. }
  1629. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1630. {
  1631. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1632. }
  1633. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1634. {
  1635. return skb->head + skb->transport_header;
  1636. }
  1637. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1638. {
  1639. skb->transport_header = skb->data - skb->head;
  1640. }
  1641. static inline void skb_set_transport_header(struct sk_buff *skb,
  1642. const int offset)
  1643. {
  1644. skb_reset_transport_header(skb);
  1645. skb->transport_header += offset;
  1646. }
  1647. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1648. {
  1649. return skb->head + skb->network_header;
  1650. }
  1651. static inline void skb_reset_network_header(struct sk_buff *skb)
  1652. {
  1653. skb->network_header = skb->data - skb->head;
  1654. }
  1655. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1656. {
  1657. skb_reset_network_header(skb);
  1658. skb->network_header += offset;
  1659. }
  1660. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1661. {
  1662. return skb->head + skb->mac_header;
  1663. }
  1664. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1665. {
  1666. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1667. }
  1668. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1669. {
  1670. skb->mac_header = skb->data - skb->head;
  1671. }
  1672. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1673. {
  1674. skb_reset_mac_header(skb);
  1675. skb->mac_header += offset;
  1676. }
  1677. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1678. {
  1679. skb->mac_header = skb->network_header;
  1680. }
  1681. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1682. const int offset_hint)
  1683. {
  1684. struct flow_keys keys;
  1685. if (skb_transport_header_was_set(skb))
  1686. return;
  1687. else if (skb_flow_dissect(skb, &keys))
  1688. skb_set_transport_header(skb, keys.thoff);
  1689. else
  1690. skb_set_transport_header(skb, offset_hint);
  1691. }
  1692. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1693. {
  1694. if (skb_mac_header_was_set(skb)) {
  1695. const unsigned char *old_mac = skb_mac_header(skb);
  1696. skb_set_mac_header(skb, -skb->mac_len);
  1697. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1698. }
  1699. }
  1700. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1701. {
  1702. return skb->csum_start - skb_headroom(skb);
  1703. }
  1704. static inline int skb_transport_offset(const struct sk_buff *skb)
  1705. {
  1706. return skb_transport_header(skb) - skb->data;
  1707. }
  1708. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1709. {
  1710. return skb->transport_header - skb->network_header;
  1711. }
  1712. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1713. {
  1714. return skb->inner_transport_header - skb->inner_network_header;
  1715. }
  1716. static inline int skb_network_offset(const struct sk_buff *skb)
  1717. {
  1718. return skb_network_header(skb) - skb->data;
  1719. }
  1720. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1721. {
  1722. return skb_inner_network_header(skb) - skb->data;
  1723. }
  1724. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1725. {
  1726. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1727. }
  1728. /*
  1729. * CPUs often take a performance hit when accessing unaligned memory
  1730. * locations. The actual performance hit varies, it can be small if the
  1731. * hardware handles it or large if we have to take an exception and fix it
  1732. * in software.
  1733. *
  1734. * Since an ethernet header is 14 bytes network drivers often end up with
  1735. * the IP header at an unaligned offset. The IP header can be aligned by
  1736. * shifting the start of the packet by 2 bytes. Drivers should do this
  1737. * with:
  1738. *
  1739. * skb_reserve(skb, NET_IP_ALIGN);
  1740. *
  1741. * The downside to this alignment of the IP header is that the DMA is now
  1742. * unaligned. On some architectures the cost of an unaligned DMA is high
  1743. * and this cost outweighs the gains made by aligning the IP header.
  1744. *
  1745. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1746. * to be overridden.
  1747. */
  1748. #ifndef NET_IP_ALIGN
  1749. #define NET_IP_ALIGN 2
  1750. #endif
  1751. /*
  1752. * The networking layer reserves some headroom in skb data (via
  1753. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1754. * the header has to grow. In the default case, if the header has to grow
  1755. * 32 bytes or less we avoid the reallocation.
  1756. *
  1757. * Unfortunately this headroom changes the DMA alignment of the resulting
  1758. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1759. * on some architectures. An architecture can override this value,
  1760. * perhaps setting it to a cacheline in size (since that will maintain
  1761. * cacheline alignment of the DMA). It must be a power of 2.
  1762. *
  1763. * Various parts of the networking layer expect at least 32 bytes of
  1764. * headroom, you should not reduce this.
  1765. *
  1766. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1767. * to reduce average number of cache lines per packet.
  1768. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1769. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1770. */
  1771. #ifndef NET_SKB_PAD
  1772. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1773. #endif
  1774. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1775. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1776. {
  1777. if (unlikely(skb_is_nonlinear(skb))) {
  1778. WARN_ON(1);
  1779. return;
  1780. }
  1781. skb->len = len;
  1782. skb_set_tail_pointer(skb, len);
  1783. }
  1784. void skb_trim(struct sk_buff *skb, unsigned int len);
  1785. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1786. {
  1787. if (skb->data_len)
  1788. return ___pskb_trim(skb, len);
  1789. __skb_trim(skb, len);
  1790. return 0;
  1791. }
  1792. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1793. {
  1794. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1795. }
  1796. /**
  1797. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1798. * @skb: buffer to alter
  1799. * @len: new length
  1800. *
  1801. * This is identical to pskb_trim except that the caller knows that
  1802. * the skb is not cloned so we should never get an error due to out-
  1803. * of-memory.
  1804. */
  1805. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  1806. {
  1807. int err = pskb_trim(skb, len);
  1808. BUG_ON(err);
  1809. }
  1810. /**
  1811. * skb_orphan - orphan a buffer
  1812. * @skb: buffer to orphan
  1813. *
  1814. * If a buffer currently has an owner then we call the owner's
  1815. * destructor function and make the @skb unowned. The buffer continues
  1816. * to exist but is no longer charged to its former owner.
  1817. */
  1818. static inline void skb_orphan(struct sk_buff *skb)
  1819. {
  1820. if (skb->destructor) {
  1821. skb->destructor(skb);
  1822. skb->destructor = NULL;
  1823. skb->sk = NULL;
  1824. } else {
  1825. BUG_ON(skb->sk);
  1826. }
  1827. }
  1828. /**
  1829. * skb_orphan_frags - orphan the frags contained in a buffer
  1830. * @skb: buffer to orphan frags from
  1831. * @gfp_mask: allocation mask for replacement pages
  1832. *
  1833. * For each frag in the SKB which needs a destructor (i.e. has an
  1834. * owner) create a copy of that frag and release the original
  1835. * page by calling the destructor.
  1836. */
  1837. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  1838. {
  1839. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  1840. return 0;
  1841. return skb_copy_ubufs(skb, gfp_mask);
  1842. }
  1843. /**
  1844. * __skb_queue_purge - empty a list
  1845. * @list: list to empty
  1846. *
  1847. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  1848. * the list and one reference dropped. This function does not take the
  1849. * list lock and the caller must hold the relevant locks to use it.
  1850. */
  1851. void skb_queue_purge(struct sk_buff_head *list);
  1852. static inline void __skb_queue_purge(struct sk_buff_head *list)
  1853. {
  1854. struct sk_buff *skb;
  1855. while ((skb = __skb_dequeue(list)) != NULL)
  1856. kfree_skb(skb);
  1857. }
  1858. #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
  1859. #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
  1860. #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
  1861. void *netdev_alloc_frag(unsigned int fragsz);
  1862. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  1863. gfp_t gfp_mask);
  1864. /**
  1865. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  1866. * @dev: network device to receive on
  1867. * @length: length to allocate
  1868. *
  1869. * Allocate a new &sk_buff and assign it a usage count of one. The
  1870. * buffer has unspecified headroom built in. Users should allocate
  1871. * the headroom they think they need without accounting for the
  1872. * built in space. The built in space is used for optimisations.
  1873. *
  1874. * %NULL is returned if there is no free memory. Although this function
  1875. * allocates memory it can be called from an interrupt.
  1876. */
  1877. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  1878. unsigned int length)
  1879. {
  1880. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  1881. }
  1882. /* legacy helper around __netdev_alloc_skb() */
  1883. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  1884. gfp_t gfp_mask)
  1885. {
  1886. return __netdev_alloc_skb(NULL, length, gfp_mask);
  1887. }
  1888. /* legacy helper around netdev_alloc_skb() */
  1889. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  1890. {
  1891. return netdev_alloc_skb(NULL, length);
  1892. }
  1893. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  1894. unsigned int length, gfp_t gfp)
  1895. {
  1896. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  1897. if (NET_IP_ALIGN && skb)
  1898. skb_reserve(skb, NET_IP_ALIGN);
  1899. return skb;
  1900. }
  1901. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  1902. unsigned int length)
  1903. {
  1904. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  1905. }
  1906. void *napi_alloc_frag(unsigned int fragsz);
  1907. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  1908. unsigned int length, gfp_t gfp_mask);
  1909. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  1910. unsigned int length)
  1911. {
  1912. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  1913. }
  1914. /**
  1915. * __dev_alloc_pages - allocate page for network Rx
  1916. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1917. * @order: size of the allocation
  1918. *
  1919. * Allocate a new page.
  1920. *
  1921. * %NULL is returned if there is no free memory.
  1922. */
  1923. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  1924. unsigned int order)
  1925. {
  1926. /* This piece of code contains several assumptions.
  1927. * 1. This is for device Rx, therefor a cold page is preferred.
  1928. * 2. The expectation is the user wants a compound page.
  1929. * 3. If requesting a order 0 page it will not be compound
  1930. * due to the check to see if order has a value in prep_new_page
  1931. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  1932. * code in gfp_to_alloc_flags that should be enforcing this.
  1933. */
  1934. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  1935. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  1936. }
  1937. static inline struct page *dev_alloc_pages(unsigned int order)
  1938. {
  1939. return __dev_alloc_pages(GFP_ATOMIC, order);
  1940. }
  1941. /**
  1942. * __dev_alloc_page - allocate a page for network Rx
  1943. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1944. *
  1945. * Allocate a new page.
  1946. *
  1947. * %NULL is returned if there is no free memory.
  1948. */
  1949. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  1950. {
  1951. return __dev_alloc_pages(gfp_mask, 0);
  1952. }
  1953. static inline struct page *dev_alloc_page(void)
  1954. {
  1955. return __dev_alloc_page(GFP_ATOMIC);
  1956. }
  1957. /**
  1958. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  1959. * @page: The page that was allocated from skb_alloc_page
  1960. * @skb: The skb that may need pfmemalloc set
  1961. */
  1962. static inline void skb_propagate_pfmemalloc(struct page *page,
  1963. struct sk_buff *skb)
  1964. {
  1965. if (page && page->pfmemalloc)
  1966. skb->pfmemalloc = true;
  1967. }
  1968. /**
  1969. * skb_frag_page - retrieve the page referred to by a paged fragment
  1970. * @frag: the paged fragment
  1971. *
  1972. * Returns the &struct page associated with @frag.
  1973. */
  1974. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  1975. {
  1976. return frag->page.p;
  1977. }
  1978. /**
  1979. * __skb_frag_ref - take an addition reference on a paged fragment.
  1980. * @frag: the paged fragment
  1981. *
  1982. * Takes an additional reference on the paged fragment @frag.
  1983. */
  1984. static inline void __skb_frag_ref(skb_frag_t *frag)
  1985. {
  1986. get_page(skb_frag_page(frag));
  1987. }
  1988. /**
  1989. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  1990. * @skb: the buffer
  1991. * @f: the fragment offset.
  1992. *
  1993. * Takes an additional reference on the @f'th paged fragment of @skb.
  1994. */
  1995. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  1996. {
  1997. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  1998. }
  1999. /**
  2000. * __skb_frag_unref - release a reference on a paged fragment.
  2001. * @frag: the paged fragment
  2002. *
  2003. * Releases a reference on the paged fragment @frag.
  2004. */
  2005. static inline void __skb_frag_unref(skb_frag_t *frag)
  2006. {
  2007. put_page(skb_frag_page(frag));
  2008. }
  2009. /**
  2010. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2011. * @skb: the buffer
  2012. * @f: the fragment offset
  2013. *
  2014. * Releases a reference on the @f'th paged fragment of @skb.
  2015. */
  2016. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2017. {
  2018. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2019. }
  2020. /**
  2021. * skb_frag_address - gets the address of the data contained in a paged fragment
  2022. * @frag: the paged fragment buffer
  2023. *
  2024. * Returns the address of the data within @frag. The page must already
  2025. * be mapped.
  2026. */
  2027. static inline void *skb_frag_address(const skb_frag_t *frag)
  2028. {
  2029. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2030. }
  2031. /**
  2032. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2033. * @frag: the paged fragment buffer
  2034. *
  2035. * Returns the address of the data within @frag. Checks that the page
  2036. * is mapped and returns %NULL otherwise.
  2037. */
  2038. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2039. {
  2040. void *ptr = page_address(skb_frag_page(frag));
  2041. if (unlikely(!ptr))
  2042. return NULL;
  2043. return ptr + frag->page_offset;
  2044. }
  2045. /**
  2046. * __skb_frag_set_page - sets the page contained in a paged fragment
  2047. * @frag: the paged fragment
  2048. * @page: the page to set
  2049. *
  2050. * Sets the fragment @frag to contain @page.
  2051. */
  2052. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2053. {
  2054. frag->page.p = page;
  2055. }
  2056. /**
  2057. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2058. * @skb: the buffer
  2059. * @f: the fragment offset
  2060. * @page: the page to set
  2061. *
  2062. * Sets the @f'th fragment of @skb to contain @page.
  2063. */
  2064. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2065. struct page *page)
  2066. {
  2067. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2068. }
  2069. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2070. /**
  2071. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2072. * @dev: the device to map the fragment to
  2073. * @frag: the paged fragment to map
  2074. * @offset: the offset within the fragment (starting at the
  2075. * fragment's own offset)
  2076. * @size: the number of bytes to map
  2077. * @dir: the direction of the mapping (%PCI_DMA_*)
  2078. *
  2079. * Maps the page associated with @frag to @device.
  2080. */
  2081. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2082. const skb_frag_t *frag,
  2083. size_t offset, size_t size,
  2084. enum dma_data_direction dir)
  2085. {
  2086. return dma_map_page(dev, skb_frag_page(frag),
  2087. frag->page_offset + offset, size, dir);
  2088. }
  2089. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2090. gfp_t gfp_mask)
  2091. {
  2092. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2093. }
  2094. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2095. gfp_t gfp_mask)
  2096. {
  2097. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2098. }
  2099. /**
  2100. * skb_clone_writable - is the header of a clone writable
  2101. * @skb: buffer to check
  2102. * @len: length up to which to write
  2103. *
  2104. * Returns true if modifying the header part of the cloned buffer
  2105. * does not requires the data to be copied.
  2106. */
  2107. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2108. {
  2109. return !skb_header_cloned(skb) &&
  2110. skb_headroom(skb) + len <= skb->hdr_len;
  2111. }
  2112. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2113. int cloned)
  2114. {
  2115. int delta = 0;
  2116. if (headroom > skb_headroom(skb))
  2117. delta = headroom - skb_headroom(skb);
  2118. if (delta || cloned)
  2119. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2120. GFP_ATOMIC);
  2121. return 0;
  2122. }
  2123. /**
  2124. * skb_cow - copy header of skb when it is required
  2125. * @skb: buffer to cow
  2126. * @headroom: needed headroom
  2127. *
  2128. * If the skb passed lacks sufficient headroom or its data part
  2129. * is shared, data is reallocated. If reallocation fails, an error
  2130. * is returned and original skb is not changed.
  2131. *
  2132. * The result is skb with writable area skb->head...skb->tail
  2133. * and at least @headroom of space at head.
  2134. */
  2135. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2136. {
  2137. return __skb_cow(skb, headroom, skb_cloned(skb));
  2138. }
  2139. /**
  2140. * skb_cow_head - skb_cow but only making the head writable
  2141. * @skb: buffer to cow
  2142. * @headroom: needed headroom
  2143. *
  2144. * This function is identical to skb_cow except that we replace the
  2145. * skb_cloned check by skb_header_cloned. It should be used when
  2146. * you only need to push on some header and do not need to modify
  2147. * the data.
  2148. */
  2149. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2150. {
  2151. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2152. }
  2153. /**
  2154. * skb_padto - pad an skbuff up to a minimal size
  2155. * @skb: buffer to pad
  2156. * @len: minimal length
  2157. *
  2158. * Pads up a buffer to ensure the trailing bytes exist and are
  2159. * blanked. If the buffer already contains sufficient data it
  2160. * is untouched. Otherwise it is extended. Returns zero on
  2161. * success. The skb is freed on error.
  2162. */
  2163. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2164. {
  2165. unsigned int size = skb->len;
  2166. if (likely(size >= len))
  2167. return 0;
  2168. return skb_pad(skb, len - size);
  2169. }
  2170. /**
  2171. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2172. * @skb: buffer to pad
  2173. * @len: minimal length
  2174. *
  2175. * Pads up a buffer to ensure the trailing bytes exist and are
  2176. * blanked. If the buffer already contains sufficient data it
  2177. * is untouched. Otherwise it is extended. Returns zero on
  2178. * success. The skb is freed on error.
  2179. */
  2180. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2181. {
  2182. unsigned int size = skb->len;
  2183. if (unlikely(size < len)) {
  2184. len -= size;
  2185. if (skb_pad(skb, len))
  2186. return -ENOMEM;
  2187. __skb_put(skb, len);
  2188. }
  2189. return 0;
  2190. }
  2191. static inline int skb_add_data(struct sk_buff *skb,
  2192. struct iov_iter *from, int copy)
  2193. {
  2194. const int off = skb->len;
  2195. if (skb->ip_summed == CHECKSUM_NONE) {
  2196. __wsum csum = 0;
  2197. if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
  2198. &csum, from) == copy) {
  2199. skb->csum = csum_block_add(skb->csum, csum, off);
  2200. return 0;
  2201. }
  2202. } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
  2203. return 0;
  2204. __skb_trim(skb, off);
  2205. return -EFAULT;
  2206. }
  2207. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2208. const struct page *page, int off)
  2209. {
  2210. if (i) {
  2211. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2212. return page == skb_frag_page(frag) &&
  2213. off == frag->page_offset + skb_frag_size(frag);
  2214. }
  2215. return false;
  2216. }
  2217. static inline int __skb_linearize(struct sk_buff *skb)
  2218. {
  2219. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2220. }
  2221. /**
  2222. * skb_linearize - convert paged skb to linear one
  2223. * @skb: buffer to linarize
  2224. *
  2225. * If there is no free memory -ENOMEM is returned, otherwise zero
  2226. * is returned and the old skb data released.
  2227. */
  2228. static inline int skb_linearize(struct sk_buff *skb)
  2229. {
  2230. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2231. }
  2232. /**
  2233. * skb_has_shared_frag - can any frag be overwritten
  2234. * @skb: buffer to test
  2235. *
  2236. * Return true if the skb has at least one frag that might be modified
  2237. * by an external entity (as in vmsplice()/sendfile())
  2238. */
  2239. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2240. {
  2241. return skb_is_nonlinear(skb) &&
  2242. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2243. }
  2244. /**
  2245. * skb_linearize_cow - make sure skb is linear and writable
  2246. * @skb: buffer to process
  2247. *
  2248. * If there is no free memory -ENOMEM is returned, otherwise zero
  2249. * is returned and the old skb data released.
  2250. */
  2251. static inline int skb_linearize_cow(struct sk_buff *skb)
  2252. {
  2253. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2254. __skb_linearize(skb) : 0;
  2255. }
  2256. /**
  2257. * skb_postpull_rcsum - update checksum for received skb after pull
  2258. * @skb: buffer to update
  2259. * @start: start of data before pull
  2260. * @len: length of data pulled
  2261. *
  2262. * After doing a pull on a received packet, you need to call this to
  2263. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2264. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2265. */
  2266. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2267. const void *start, unsigned int len)
  2268. {
  2269. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2270. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2271. }
  2272. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2273. /**
  2274. * pskb_trim_rcsum - trim received skb and update checksum
  2275. * @skb: buffer to trim
  2276. * @len: new length
  2277. *
  2278. * This is exactly the same as pskb_trim except that it ensures the
  2279. * checksum of received packets are still valid after the operation.
  2280. */
  2281. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2282. {
  2283. if (likely(len >= skb->len))
  2284. return 0;
  2285. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2286. skb->ip_summed = CHECKSUM_NONE;
  2287. return __pskb_trim(skb, len);
  2288. }
  2289. #define skb_queue_walk(queue, skb) \
  2290. for (skb = (queue)->next; \
  2291. skb != (struct sk_buff *)(queue); \
  2292. skb = skb->next)
  2293. #define skb_queue_walk_safe(queue, skb, tmp) \
  2294. for (skb = (queue)->next, tmp = skb->next; \
  2295. skb != (struct sk_buff *)(queue); \
  2296. skb = tmp, tmp = skb->next)
  2297. #define skb_queue_walk_from(queue, skb) \
  2298. for (; skb != (struct sk_buff *)(queue); \
  2299. skb = skb->next)
  2300. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2301. for (tmp = skb->next; \
  2302. skb != (struct sk_buff *)(queue); \
  2303. skb = tmp, tmp = skb->next)
  2304. #define skb_queue_reverse_walk(queue, skb) \
  2305. for (skb = (queue)->prev; \
  2306. skb != (struct sk_buff *)(queue); \
  2307. skb = skb->prev)
  2308. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2309. for (skb = (queue)->prev, tmp = skb->prev; \
  2310. skb != (struct sk_buff *)(queue); \
  2311. skb = tmp, tmp = skb->prev)
  2312. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2313. for (tmp = skb->prev; \
  2314. skb != (struct sk_buff *)(queue); \
  2315. skb = tmp, tmp = skb->prev)
  2316. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2317. {
  2318. return skb_shinfo(skb)->frag_list != NULL;
  2319. }
  2320. static inline void skb_frag_list_init(struct sk_buff *skb)
  2321. {
  2322. skb_shinfo(skb)->frag_list = NULL;
  2323. }
  2324. static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
  2325. {
  2326. frag->next = skb_shinfo(skb)->frag_list;
  2327. skb_shinfo(skb)->frag_list = frag;
  2328. }
  2329. #define skb_walk_frags(skb, iter) \
  2330. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2331. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2332. int *peeked, int *off, int *err);
  2333. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2334. int *err);
  2335. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2336. struct poll_table_struct *wait);
  2337. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2338. struct iov_iter *to, int size);
  2339. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2340. struct msghdr *msg, int size)
  2341. {
  2342. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2343. }
  2344. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2345. struct msghdr *msg);
  2346. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2347. struct iov_iter *from, int len);
  2348. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2349. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2350. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2351. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2352. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2353. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2354. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2355. int len, __wsum csum);
  2356. int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
  2357. struct pipe_inode_info *pipe, unsigned int len,
  2358. unsigned int flags);
  2359. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2360. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2361. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2362. int len, int hlen);
  2363. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2364. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2365. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2366. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2367. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2368. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2369. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2370. int skb_vlan_pop(struct sk_buff *skb);
  2371. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2372. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2373. {
  2374. return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2375. }
  2376. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2377. {
  2378. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2379. }
  2380. struct skb_checksum_ops {
  2381. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2382. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2383. };
  2384. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2385. __wsum csum, const struct skb_checksum_ops *ops);
  2386. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2387. __wsum csum);
  2388. static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
  2389. int len, void *data, int hlen, void *buffer)
  2390. {
  2391. if (hlen - offset >= len)
  2392. return data + offset;
  2393. if (!skb ||
  2394. skb_copy_bits(skb, offset, buffer, len) < 0)
  2395. return NULL;
  2396. return buffer;
  2397. }
  2398. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  2399. int len, void *buffer)
  2400. {
  2401. return __skb_header_pointer(skb, offset, len, skb->data,
  2402. skb_headlen(skb), buffer);
  2403. }
  2404. /**
  2405. * skb_needs_linearize - check if we need to linearize a given skb
  2406. * depending on the given device features.
  2407. * @skb: socket buffer to check
  2408. * @features: net device features
  2409. *
  2410. * Returns true if either:
  2411. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2412. * 2. skb is fragmented and the device does not support SG.
  2413. */
  2414. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2415. netdev_features_t features)
  2416. {
  2417. return skb_is_nonlinear(skb) &&
  2418. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2419. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2420. }
  2421. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2422. void *to,
  2423. const unsigned int len)
  2424. {
  2425. memcpy(to, skb->data, len);
  2426. }
  2427. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2428. const int offset, void *to,
  2429. const unsigned int len)
  2430. {
  2431. memcpy(to, skb->data + offset, len);
  2432. }
  2433. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2434. const void *from,
  2435. const unsigned int len)
  2436. {
  2437. memcpy(skb->data, from, len);
  2438. }
  2439. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2440. const int offset,
  2441. const void *from,
  2442. const unsigned int len)
  2443. {
  2444. memcpy(skb->data + offset, from, len);
  2445. }
  2446. void skb_init(void);
  2447. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2448. {
  2449. return skb->tstamp;
  2450. }
  2451. /**
  2452. * skb_get_timestamp - get timestamp from a skb
  2453. * @skb: skb to get stamp from
  2454. * @stamp: pointer to struct timeval to store stamp in
  2455. *
  2456. * Timestamps are stored in the skb as offsets to a base timestamp.
  2457. * This function converts the offset back to a struct timeval and stores
  2458. * it in stamp.
  2459. */
  2460. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2461. struct timeval *stamp)
  2462. {
  2463. *stamp = ktime_to_timeval(skb->tstamp);
  2464. }
  2465. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2466. struct timespec *stamp)
  2467. {
  2468. *stamp = ktime_to_timespec(skb->tstamp);
  2469. }
  2470. static inline void __net_timestamp(struct sk_buff *skb)
  2471. {
  2472. skb->tstamp = ktime_get_real();
  2473. }
  2474. static inline ktime_t net_timedelta(ktime_t t)
  2475. {
  2476. return ktime_sub(ktime_get_real(), t);
  2477. }
  2478. static inline ktime_t net_invalid_timestamp(void)
  2479. {
  2480. return ktime_set(0, 0);
  2481. }
  2482. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2483. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2484. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2485. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2486. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2487. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2488. {
  2489. }
  2490. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2491. {
  2492. return false;
  2493. }
  2494. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2495. /**
  2496. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2497. *
  2498. * PHY drivers may accept clones of transmitted packets for
  2499. * timestamping via their phy_driver.txtstamp method. These drivers
  2500. * must call this function to return the skb back to the stack, with
  2501. * or without a timestamp.
  2502. *
  2503. * @skb: clone of the the original outgoing packet
  2504. * @hwtstamps: hardware time stamps, may be NULL if not available
  2505. *
  2506. */
  2507. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2508. struct skb_shared_hwtstamps *hwtstamps);
  2509. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2510. struct skb_shared_hwtstamps *hwtstamps,
  2511. struct sock *sk, int tstype);
  2512. /**
  2513. * skb_tstamp_tx - queue clone of skb with send time stamps
  2514. * @orig_skb: the original outgoing packet
  2515. * @hwtstamps: hardware time stamps, may be NULL if not available
  2516. *
  2517. * If the skb has a socket associated, then this function clones the
  2518. * skb (thus sharing the actual data and optional structures), stores
  2519. * the optional hardware time stamping information (if non NULL) or
  2520. * generates a software time stamp (otherwise), then queues the clone
  2521. * to the error queue of the socket. Errors are silently ignored.
  2522. */
  2523. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2524. struct skb_shared_hwtstamps *hwtstamps);
  2525. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2526. {
  2527. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2528. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2529. skb_tstamp_tx(skb, NULL);
  2530. }
  2531. /**
  2532. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2533. *
  2534. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2535. * function immediately before giving the sk_buff to the MAC hardware.
  2536. *
  2537. * Specifically, one should make absolutely sure that this function is
  2538. * called before TX completion of this packet can trigger. Otherwise
  2539. * the packet could potentially already be freed.
  2540. *
  2541. * @skb: A socket buffer.
  2542. */
  2543. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2544. {
  2545. skb_clone_tx_timestamp(skb);
  2546. sw_tx_timestamp(skb);
  2547. }
  2548. /**
  2549. * skb_complete_wifi_ack - deliver skb with wifi status
  2550. *
  2551. * @skb: the original outgoing packet
  2552. * @acked: ack status
  2553. *
  2554. */
  2555. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2556. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2557. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2558. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2559. {
  2560. return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
  2561. }
  2562. /**
  2563. * skb_checksum_complete - Calculate checksum of an entire packet
  2564. * @skb: packet to process
  2565. *
  2566. * This function calculates the checksum over the entire packet plus
  2567. * the value of skb->csum. The latter can be used to supply the
  2568. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2569. * checksum.
  2570. *
  2571. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2572. * this function can be used to verify that checksum on received
  2573. * packets. In that case the function should return zero if the
  2574. * checksum is correct. In particular, this function will return zero
  2575. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2576. * hardware has already verified the correctness of the checksum.
  2577. */
  2578. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2579. {
  2580. return skb_csum_unnecessary(skb) ?
  2581. 0 : __skb_checksum_complete(skb);
  2582. }
  2583. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2584. {
  2585. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2586. if (skb->csum_level == 0)
  2587. skb->ip_summed = CHECKSUM_NONE;
  2588. else
  2589. skb->csum_level--;
  2590. }
  2591. }
  2592. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2593. {
  2594. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2595. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2596. skb->csum_level++;
  2597. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2598. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2599. skb->csum_level = 0;
  2600. }
  2601. }
  2602. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2603. {
  2604. /* Mark current checksum as bad (typically called from GRO
  2605. * path). In the case that ip_summed is CHECKSUM_NONE
  2606. * this must be the first checksum encountered in the packet.
  2607. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2608. * checksum after the last one validated. For UDP, a zero
  2609. * checksum can not be marked as bad.
  2610. */
  2611. if (skb->ip_summed == CHECKSUM_NONE ||
  2612. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2613. skb->csum_bad = 1;
  2614. }
  2615. /* Check if we need to perform checksum complete validation.
  2616. *
  2617. * Returns true if checksum complete is needed, false otherwise
  2618. * (either checksum is unnecessary or zero checksum is allowed).
  2619. */
  2620. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2621. bool zero_okay,
  2622. __sum16 check)
  2623. {
  2624. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2625. skb->csum_valid = 1;
  2626. __skb_decr_checksum_unnecessary(skb);
  2627. return false;
  2628. }
  2629. return true;
  2630. }
  2631. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2632. * in checksum_init.
  2633. */
  2634. #define CHECKSUM_BREAK 76
  2635. /* Validate (init) checksum based on checksum complete.
  2636. *
  2637. * Return values:
  2638. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2639. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2640. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2641. * non-zero: value of invalid checksum
  2642. *
  2643. */
  2644. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2645. bool complete,
  2646. __wsum psum)
  2647. {
  2648. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2649. if (!csum_fold(csum_add(psum, skb->csum))) {
  2650. skb->csum_valid = 1;
  2651. return 0;
  2652. }
  2653. } else if (skb->csum_bad) {
  2654. /* ip_summed == CHECKSUM_NONE in this case */
  2655. return 1;
  2656. }
  2657. skb->csum = psum;
  2658. if (complete || skb->len <= CHECKSUM_BREAK) {
  2659. __sum16 csum;
  2660. csum = __skb_checksum_complete(skb);
  2661. skb->csum_valid = !csum;
  2662. return csum;
  2663. }
  2664. return 0;
  2665. }
  2666. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2667. {
  2668. return 0;
  2669. }
  2670. /* Perform checksum validate (init). Note that this is a macro since we only
  2671. * want to calculate the pseudo header which is an input function if necessary.
  2672. * First we try to validate without any computation (checksum unnecessary) and
  2673. * then calculate based on checksum complete calling the function to compute
  2674. * pseudo header.
  2675. *
  2676. * Return values:
  2677. * 0: checksum is validated or try to in skb_checksum_complete
  2678. * non-zero: value of invalid checksum
  2679. */
  2680. #define __skb_checksum_validate(skb, proto, complete, \
  2681. zero_okay, check, compute_pseudo) \
  2682. ({ \
  2683. __sum16 __ret = 0; \
  2684. skb->csum_valid = 0; \
  2685. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2686. __ret = __skb_checksum_validate_complete(skb, \
  2687. complete, compute_pseudo(skb, proto)); \
  2688. __ret; \
  2689. })
  2690. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2691. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2692. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2693. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2694. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2695. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2696. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2697. compute_pseudo) \
  2698. __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
  2699. #define skb_checksum_simple_validate(skb) \
  2700. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2701. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2702. {
  2703. return (skb->ip_summed == CHECKSUM_NONE &&
  2704. skb->csum_valid && !skb->csum_bad);
  2705. }
  2706. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2707. __sum16 check, __wsum pseudo)
  2708. {
  2709. skb->csum = ~pseudo;
  2710. skb->ip_summed = CHECKSUM_COMPLETE;
  2711. }
  2712. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  2713. do { \
  2714. if (__skb_checksum_convert_check(skb)) \
  2715. __skb_checksum_convert(skb, check, \
  2716. compute_pseudo(skb, proto)); \
  2717. } while (0)
  2718. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2719. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2720. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2721. {
  2722. if (nfct && atomic_dec_and_test(&nfct->use))
  2723. nf_conntrack_destroy(nfct);
  2724. }
  2725. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2726. {
  2727. if (nfct)
  2728. atomic_inc(&nfct->use);
  2729. }
  2730. #endif
  2731. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2732. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  2733. {
  2734. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  2735. kfree(nf_bridge);
  2736. }
  2737. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  2738. {
  2739. if (nf_bridge)
  2740. atomic_inc(&nf_bridge->use);
  2741. }
  2742. #endif /* CONFIG_BRIDGE_NETFILTER */
  2743. static inline void nf_reset(struct sk_buff *skb)
  2744. {
  2745. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2746. nf_conntrack_put(skb->nfct);
  2747. skb->nfct = NULL;
  2748. #endif
  2749. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2750. nf_bridge_put(skb->nf_bridge);
  2751. skb->nf_bridge = NULL;
  2752. #endif
  2753. }
  2754. static inline void nf_reset_trace(struct sk_buff *skb)
  2755. {
  2756. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2757. skb->nf_trace = 0;
  2758. #endif
  2759. }
  2760. /* Note: This doesn't put any conntrack and bridge info in dst. */
  2761. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  2762. bool copy)
  2763. {
  2764. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2765. dst->nfct = src->nfct;
  2766. nf_conntrack_get(src->nfct);
  2767. if (copy)
  2768. dst->nfctinfo = src->nfctinfo;
  2769. #endif
  2770. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2771. dst->nf_bridge = src->nf_bridge;
  2772. nf_bridge_get(src->nf_bridge);
  2773. #endif
  2774. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2775. if (copy)
  2776. dst->nf_trace = src->nf_trace;
  2777. #endif
  2778. }
  2779. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2780. {
  2781. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2782. nf_conntrack_put(dst->nfct);
  2783. #endif
  2784. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2785. nf_bridge_put(dst->nf_bridge);
  2786. #endif
  2787. __nf_copy(dst, src, true);
  2788. }
  2789. #ifdef CONFIG_NETWORK_SECMARK
  2790. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2791. {
  2792. to->secmark = from->secmark;
  2793. }
  2794. static inline void skb_init_secmark(struct sk_buff *skb)
  2795. {
  2796. skb->secmark = 0;
  2797. }
  2798. #else
  2799. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2800. { }
  2801. static inline void skb_init_secmark(struct sk_buff *skb)
  2802. { }
  2803. #endif
  2804. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  2805. {
  2806. return !skb->destructor &&
  2807. #if IS_ENABLED(CONFIG_XFRM)
  2808. !skb->sp &&
  2809. #endif
  2810. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  2811. !skb->nfct &&
  2812. #endif
  2813. !skb->_skb_refdst &&
  2814. !skb_has_frag_list(skb);
  2815. }
  2816. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  2817. {
  2818. skb->queue_mapping = queue_mapping;
  2819. }
  2820. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  2821. {
  2822. return skb->queue_mapping;
  2823. }
  2824. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  2825. {
  2826. to->queue_mapping = from->queue_mapping;
  2827. }
  2828. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  2829. {
  2830. skb->queue_mapping = rx_queue + 1;
  2831. }
  2832. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  2833. {
  2834. return skb->queue_mapping - 1;
  2835. }
  2836. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  2837. {
  2838. return skb->queue_mapping != 0;
  2839. }
  2840. u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
  2841. unsigned int num_tx_queues);
  2842. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  2843. {
  2844. #ifdef CONFIG_XFRM
  2845. return skb->sp;
  2846. #else
  2847. return NULL;
  2848. #endif
  2849. }
  2850. /* Keeps track of mac header offset relative to skb->head.
  2851. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  2852. * For non-tunnel skb it points to skb_mac_header() and for
  2853. * tunnel skb it points to outer mac header.
  2854. * Keeps track of level of encapsulation of network headers.
  2855. */
  2856. struct skb_gso_cb {
  2857. int mac_offset;
  2858. int encap_level;
  2859. __u16 csum_start;
  2860. };
  2861. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
  2862. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  2863. {
  2864. return (skb_mac_header(inner_skb) - inner_skb->head) -
  2865. SKB_GSO_CB(inner_skb)->mac_offset;
  2866. }
  2867. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  2868. {
  2869. int new_headroom, headroom;
  2870. int ret;
  2871. headroom = skb_headroom(skb);
  2872. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  2873. if (ret)
  2874. return ret;
  2875. new_headroom = skb_headroom(skb);
  2876. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  2877. return 0;
  2878. }
  2879. /* Compute the checksum for a gso segment. First compute the checksum value
  2880. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  2881. * then add in skb->csum (checksum from csum_start to end of packet).
  2882. * skb->csum and csum_start are then updated to reflect the checksum of the
  2883. * resultant packet starting from the transport header-- the resultant checksum
  2884. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  2885. * header.
  2886. */
  2887. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  2888. {
  2889. int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
  2890. skb_transport_offset(skb);
  2891. __u16 csum;
  2892. csum = csum_fold(csum_partial(skb_transport_header(skb),
  2893. plen, skb->csum));
  2894. skb->csum = res;
  2895. SKB_GSO_CB(skb)->csum_start -= plen;
  2896. return csum;
  2897. }
  2898. static inline bool skb_is_gso(const struct sk_buff *skb)
  2899. {
  2900. return skb_shinfo(skb)->gso_size;
  2901. }
  2902. /* Note: Should be called only if skb_is_gso(skb) is true */
  2903. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  2904. {
  2905. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  2906. }
  2907. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  2908. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  2909. {
  2910. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  2911. * wanted then gso_type will be set. */
  2912. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  2913. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  2914. unlikely(shinfo->gso_type == 0)) {
  2915. __skb_warn_lro_forwarding(skb);
  2916. return true;
  2917. }
  2918. return false;
  2919. }
  2920. static inline void skb_forward_csum(struct sk_buff *skb)
  2921. {
  2922. /* Unfortunately we don't support this one. Any brave souls? */
  2923. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2924. skb->ip_summed = CHECKSUM_NONE;
  2925. }
  2926. /**
  2927. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  2928. * @skb: skb to check
  2929. *
  2930. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  2931. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  2932. * use this helper, to document places where we make this assertion.
  2933. */
  2934. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  2935. {
  2936. #ifdef DEBUG
  2937. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  2938. #endif
  2939. }
  2940. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  2941. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  2942. u32 skb_get_poff(const struct sk_buff *skb);
  2943. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  2944. const struct flow_keys *keys, int hlen);
  2945. /**
  2946. * skb_head_is_locked - Determine if the skb->head is locked down
  2947. * @skb: skb to check
  2948. *
  2949. * The head on skbs build around a head frag can be removed if they are
  2950. * not cloned. This function returns true if the skb head is locked down
  2951. * due to either being allocated via kmalloc, or by being a clone with
  2952. * multiple references to the head.
  2953. */
  2954. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  2955. {
  2956. return !skb->head_frag || skb_cloned(skb);
  2957. }
  2958. /**
  2959. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  2960. *
  2961. * @skb: GSO skb
  2962. *
  2963. * skb_gso_network_seglen is used to determine the real size of the
  2964. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  2965. *
  2966. * The MAC/L2 header is not accounted for.
  2967. */
  2968. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  2969. {
  2970. unsigned int hdr_len = skb_transport_header(skb) -
  2971. skb_network_header(skb);
  2972. return hdr_len + skb_gso_transport_seglen(skb);
  2973. }
  2974. #endif /* __KERNEL__ */
  2975. #endif /* _LINUX_SKBUFF_H */