skbuff.h 92 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963296429652966296729682969297029712972297329742975297629772978297929802981298229832984298529862987298829892990299129922993299429952996299729982999300030013002300330043005300630073008300930103011301230133014301530163017301830193020302130223023302430253026302730283029303030313032303330343035303630373038303930403041304230433044304530463047304830493050305130523053305430553056305730583059306030613062306330643065306630673068306930703071307230733074307530763077307830793080308130823083308430853086308730883089309030913092309330943095309630973098309931003101310231033104310531063107310831093110311131123113311431153116311731183119312031213122312331243125312631273128312931303131313231333134313531363137313831393140314131423143314431453146314731483149315031513152315331543155315631573158315931603161316231633164316531663167316831693170317131723173317431753176317731783179318031813182318331843185318631873188318931903191319231933194319531963197319831993200320132023203320432053206320732083209321032113212321332143215321632173218321932203221322232233224322532263227322832293230323132323233323432353236323732383239324032413242324332443245324632473248324932503251325232533254325532563257325832593260326132623263326432653266326732683269327032713272327332743275327632773278327932803281328232833284328532863287
  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/atomic.h>
  22. #include <asm/types.h>
  23. #include <linux/spinlock.h>
  24. #include <linux/net.h>
  25. #include <linux/textsearch.h>
  26. #include <net/checksum.h>
  27. #include <linux/rcupdate.h>
  28. #include <linux/dmaengine.h>
  29. #include <linux/hrtimer.h>
  30. #include <linux/dma-mapping.h>
  31. #include <linux/netdev_features.h>
  32. #include <linux/sched.h>
  33. #include <net/flow_keys.h>
  34. /* A. Checksumming of received packets by device.
  35. *
  36. * CHECKSUM_NONE:
  37. *
  38. * Device failed to checksum this packet e.g. due to lack of capabilities.
  39. * The packet contains full (though not verified) checksum in packet but
  40. * not in skb->csum. Thus, skb->csum is undefined in this case.
  41. *
  42. * CHECKSUM_UNNECESSARY:
  43. *
  44. * The hardware you're dealing with doesn't calculate the full checksum
  45. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  46. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  47. * if their checksums are okay. skb->csum is still undefined in this case
  48. * though. It is a bad option, but, unfortunately, nowadays most vendors do
  49. * this. Apparently with the secret goal to sell you new devices, when you
  50. * will add new protocol to your host, f.e. IPv6 8)
  51. *
  52. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  53. * TCP: IPv6 and IPv4.
  54. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  55. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  56. * may perform further validation in this case.
  57. * GRE: only if the checksum is present in the header.
  58. * SCTP: indicates the CRC in SCTP header has been validated.
  59. *
  60. * skb->csum_level indicates the number of consecutive checksums found in
  61. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  62. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  63. * and a device is able to verify the checksums for UDP (possibly zero),
  64. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  65. * two. If the device were only able to verify the UDP checksum and not
  66. * GRE, either because it doesn't support GRE checksum of because GRE
  67. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  68. * not considered in this case).
  69. *
  70. * CHECKSUM_COMPLETE:
  71. *
  72. * This is the most generic way. The device supplied checksum of the _whole_
  73. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  74. * hardware doesn't need to parse L3/L4 headers to implement this.
  75. *
  76. * Note: Even if device supports only some protocols, but is able to produce
  77. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  78. *
  79. * CHECKSUM_PARTIAL:
  80. *
  81. * This is identical to the case for output below. This may occur on a packet
  82. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  83. * on the same host. The packet can be treated in the same way as
  84. * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
  85. * checksum must be filled in by the OS or the hardware.
  86. *
  87. * B. Checksumming on output.
  88. *
  89. * CHECKSUM_NONE:
  90. *
  91. * The skb was already checksummed by the protocol, or a checksum is not
  92. * required.
  93. *
  94. * CHECKSUM_PARTIAL:
  95. *
  96. * The device is required to checksum the packet as seen by hard_start_xmit()
  97. * from skb->csum_start up to the end, and to record/write the checksum at
  98. * offset skb->csum_start + skb->csum_offset.
  99. *
  100. * The device must show its capabilities in dev->features, set up at device
  101. * setup time, e.g. netdev_features.h:
  102. *
  103. * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
  104. * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
  105. * IPv4. Sigh. Vendors like this way for an unknown reason.
  106. * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
  107. * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
  108. * NETIF_F_... - Well, you get the picture.
  109. *
  110. * CHECKSUM_UNNECESSARY:
  111. *
  112. * Normally, the device will do per protocol specific checksumming. Protocol
  113. * implementations that do not want the NIC to perform the checksum
  114. * calculation should use this flag in their outgoing skbs.
  115. *
  116. * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
  117. * offload. Correspondingly, the FCoE protocol driver
  118. * stack should use CHECKSUM_UNNECESSARY.
  119. *
  120. * Any questions? No questions, good. --ANK
  121. */
  122. /* Don't change this without changing skb_csum_unnecessary! */
  123. #define CHECKSUM_NONE 0
  124. #define CHECKSUM_UNNECESSARY 1
  125. #define CHECKSUM_COMPLETE 2
  126. #define CHECKSUM_PARTIAL 3
  127. /* Maximum value in skb->csum_level */
  128. #define SKB_MAX_CSUM_LEVEL 3
  129. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  130. #define SKB_WITH_OVERHEAD(X) \
  131. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  132. #define SKB_MAX_ORDER(X, ORDER) \
  133. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  134. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  135. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  136. /* return minimum truesize of one skb containing X bytes of data */
  137. #define SKB_TRUESIZE(X) ((X) + \
  138. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  139. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  140. struct net_device;
  141. struct scatterlist;
  142. struct pipe_inode_info;
  143. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  144. struct nf_conntrack {
  145. atomic_t use;
  146. };
  147. #endif
  148. #ifdef CONFIG_BRIDGE_NETFILTER
  149. struct nf_bridge_info {
  150. atomic_t use;
  151. unsigned int mask;
  152. struct net_device *physindev;
  153. struct net_device *physoutdev;
  154. unsigned long data[32 / sizeof(unsigned long)];
  155. };
  156. #endif
  157. struct sk_buff_head {
  158. /* These two members must be first. */
  159. struct sk_buff *next;
  160. struct sk_buff *prev;
  161. __u32 qlen;
  162. spinlock_t lock;
  163. };
  164. struct sk_buff;
  165. /* To allow 64K frame to be packed as single skb without frag_list we
  166. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  167. * buffers which do not start on a page boundary.
  168. *
  169. * Since GRO uses frags we allocate at least 16 regardless of page
  170. * size.
  171. */
  172. #if (65536/PAGE_SIZE + 1) < 16
  173. #define MAX_SKB_FRAGS 16UL
  174. #else
  175. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  176. #endif
  177. typedef struct skb_frag_struct skb_frag_t;
  178. struct skb_frag_struct {
  179. struct {
  180. struct page *p;
  181. } page;
  182. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  183. __u32 page_offset;
  184. __u32 size;
  185. #else
  186. __u16 page_offset;
  187. __u16 size;
  188. #endif
  189. };
  190. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  191. {
  192. return frag->size;
  193. }
  194. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  195. {
  196. frag->size = size;
  197. }
  198. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  199. {
  200. frag->size += delta;
  201. }
  202. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  203. {
  204. frag->size -= delta;
  205. }
  206. #define HAVE_HW_TIME_STAMP
  207. /**
  208. * struct skb_shared_hwtstamps - hardware time stamps
  209. * @hwtstamp: hardware time stamp transformed into duration
  210. * since arbitrary point in time
  211. *
  212. * Software time stamps generated by ktime_get_real() are stored in
  213. * skb->tstamp.
  214. *
  215. * hwtstamps can only be compared against other hwtstamps from
  216. * the same device.
  217. *
  218. * This structure is attached to packets as part of the
  219. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  220. */
  221. struct skb_shared_hwtstamps {
  222. ktime_t hwtstamp;
  223. };
  224. /* Definitions for tx_flags in struct skb_shared_info */
  225. enum {
  226. /* generate hardware time stamp */
  227. SKBTX_HW_TSTAMP = 1 << 0,
  228. /* generate software time stamp when queueing packet to NIC */
  229. SKBTX_SW_TSTAMP = 1 << 1,
  230. /* device driver is going to provide hardware time stamp */
  231. SKBTX_IN_PROGRESS = 1 << 2,
  232. /* device driver supports TX zero-copy buffers */
  233. SKBTX_DEV_ZEROCOPY = 1 << 3,
  234. /* generate wifi status information (where possible) */
  235. SKBTX_WIFI_STATUS = 1 << 4,
  236. /* This indicates at least one fragment might be overwritten
  237. * (as in vmsplice(), sendfile() ...)
  238. * If we need to compute a TX checksum, we'll need to copy
  239. * all frags to avoid possible bad checksum
  240. */
  241. SKBTX_SHARED_FRAG = 1 << 5,
  242. /* generate software time stamp when entering packet scheduling */
  243. SKBTX_SCHED_TSTAMP = 1 << 6,
  244. /* generate software timestamp on peer data acknowledgment */
  245. SKBTX_ACK_TSTAMP = 1 << 7,
  246. };
  247. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  248. SKBTX_SCHED_TSTAMP | \
  249. SKBTX_ACK_TSTAMP)
  250. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  251. /*
  252. * The callback notifies userspace to release buffers when skb DMA is done in
  253. * lower device, the skb last reference should be 0 when calling this.
  254. * The zerocopy_success argument is true if zero copy transmit occurred,
  255. * false on data copy or out of memory error caused by data copy attempt.
  256. * The ctx field is used to track device context.
  257. * The desc field is used to track userspace buffer index.
  258. */
  259. struct ubuf_info {
  260. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  261. void *ctx;
  262. unsigned long desc;
  263. };
  264. /* This data is invariant across clones and lives at
  265. * the end of the header data, ie. at skb->end.
  266. */
  267. struct skb_shared_info {
  268. unsigned char nr_frags;
  269. __u8 tx_flags;
  270. unsigned short gso_size;
  271. /* Warning: this field is not always filled in (UFO)! */
  272. unsigned short gso_segs;
  273. unsigned short gso_type;
  274. struct sk_buff *frag_list;
  275. struct skb_shared_hwtstamps hwtstamps;
  276. u32 tskey;
  277. __be32 ip6_frag_id;
  278. /*
  279. * Warning : all fields before dataref are cleared in __alloc_skb()
  280. */
  281. atomic_t dataref;
  282. /* Intermediate layers must ensure that destructor_arg
  283. * remains valid until skb destructor */
  284. void * destructor_arg;
  285. /* must be last field, see pskb_expand_head() */
  286. skb_frag_t frags[MAX_SKB_FRAGS];
  287. };
  288. /* We divide dataref into two halves. The higher 16 bits hold references
  289. * to the payload part of skb->data. The lower 16 bits hold references to
  290. * the entire skb->data. A clone of a headerless skb holds the length of
  291. * the header in skb->hdr_len.
  292. *
  293. * All users must obey the rule that the skb->data reference count must be
  294. * greater than or equal to the payload reference count.
  295. *
  296. * Holding a reference to the payload part means that the user does not
  297. * care about modifications to the header part of skb->data.
  298. */
  299. #define SKB_DATAREF_SHIFT 16
  300. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  301. enum {
  302. SKB_FCLONE_UNAVAILABLE,
  303. SKB_FCLONE_ORIG,
  304. SKB_FCLONE_CLONE,
  305. };
  306. enum {
  307. SKB_GSO_TCPV4 = 1 << 0,
  308. SKB_GSO_UDP = 1 << 1,
  309. /* This indicates the skb is from an untrusted source. */
  310. SKB_GSO_DODGY = 1 << 2,
  311. /* This indicates the tcp segment has CWR set. */
  312. SKB_GSO_TCP_ECN = 1 << 3,
  313. SKB_GSO_TCPV6 = 1 << 4,
  314. SKB_GSO_FCOE = 1 << 5,
  315. SKB_GSO_GRE = 1 << 6,
  316. SKB_GSO_GRE_CSUM = 1 << 7,
  317. SKB_GSO_IPIP = 1 << 8,
  318. SKB_GSO_SIT = 1 << 9,
  319. SKB_GSO_UDP_TUNNEL = 1 << 10,
  320. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  321. SKB_GSO_MPLS = 1 << 12,
  322. };
  323. #if BITS_PER_LONG > 32
  324. #define NET_SKBUFF_DATA_USES_OFFSET 1
  325. #endif
  326. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  327. typedef unsigned int sk_buff_data_t;
  328. #else
  329. typedef unsigned char *sk_buff_data_t;
  330. #endif
  331. /**
  332. * struct skb_mstamp - multi resolution time stamps
  333. * @stamp_us: timestamp in us resolution
  334. * @stamp_jiffies: timestamp in jiffies
  335. */
  336. struct skb_mstamp {
  337. union {
  338. u64 v64;
  339. struct {
  340. u32 stamp_us;
  341. u32 stamp_jiffies;
  342. };
  343. };
  344. };
  345. /**
  346. * skb_mstamp_get - get current timestamp
  347. * @cl: place to store timestamps
  348. */
  349. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  350. {
  351. u64 val = local_clock();
  352. do_div(val, NSEC_PER_USEC);
  353. cl->stamp_us = (u32)val;
  354. cl->stamp_jiffies = (u32)jiffies;
  355. }
  356. /**
  357. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  358. * @t1: pointer to newest sample
  359. * @t0: pointer to oldest sample
  360. */
  361. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  362. const struct skb_mstamp *t0)
  363. {
  364. s32 delta_us = t1->stamp_us - t0->stamp_us;
  365. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  366. /* If delta_us is negative, this might be because interval is too big,
  367. * or local_clock() drift is too big : fallback using jiffies.
  368. */
  369. if (delta_us <= 0 ||
  370. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  371. delta_us = jiffies_to_usecs(delta_jiffies);
  372. return delta_us;
  373. }
  374. /**
  375. * struct sk_buff - socket buffer
  376. * @next: Next buffer in list
  377. * @prev: Previous buffer in list
  378. * @tstamp: Time we arrived/left
  379. * @sk: Socket we are owned by
  380. * @dev: Device we arrived on/are leaving by
  381. * @cb: Control buffer. Free for use by every layer. Put private vars here
  382. * @_skb_refdst: destination entry (with norefcount bit)
  383. * @sp: the security path, used for xfrm
  384. * @len: Length of actual data
  385. * @data_len: Data length
  386. * @mac_len: Length of link layer header
  387. * @hdr_len: writable header length of cloned skb
  388. * @csum: Checksum (must include start/offset pair)
  389. * @csum_start: Offset from skb->head where checksumming should start
  390. * @csum_offset: Offset from csum_start where checksum should be stored
  391. * @priority: Packet queueing priority
  392. * @ignore_df: allow local fragmentation
  393. * @cloned: Head may be cloned (check refcnt to be sure)
  394. * @ip_summed: Driver fed us an IP checksum
  395. * @nohdr: Payload reference only, must not modify header
  396. * @nfctinfo: Relationship of this skb to the connection
  397. * @pkt_type: Packet class
  398. * @fclone: skbuff clone status
  399. * @ipvs_property: skbuff is owned by ipvs
  400. * @peeked: this packet has been seen already, so stats have been
  401. * done for it, don't do them again
  402. * @nf_trace: netfilter packet trace flag
  403. * @protocol: Packet protocol from driver
  404. * @destructor: Destruct function
  405. * @nfct: Associated connection, if any
  406. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  407. * @skb_iif: ifindex of device we arrived on
  408. * @tc_index: Traffic control index
  409. * @tc_verd: traffic control verdict
  410. * @hash: the packet hash
  411. * @queue_mapping: Queue mapping for multiqueue devices
  412. * @xmit_more: More SKBs are pending for this queue
  413. * @ndisc_nodetype: router type (from link layer)
  414. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  415. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  416. * ports.
  417. * @sw_hash: indicates hash was computed in software stack
  418. * @wifi_acked_valid: wifi_acked was set
  419. * @wifi_acked: whether frame was acked on wifi or not
  420. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  421. * @dma_cookie: a cookie to one of several possible DMA operations
  422. * done by skb DMA functions
  423. * @napi_id: id of the NAPI struct this skb came from
  424. * @secmark: security marking
  425. * @mark: Generic packet mark
  426. * @dropcount: total number of sk_receive_queue overflows
  427. * @vlan_proto: vlan encapsulation protocol
  428. * @vlan_tci: vlan tag control information
  429. * @inner_protocol: Protocol (encapsulation)
  430. * @inner_transport_header: Inner transport layer header (encapsulation)
  431. * @inner_network_header: Network layer header (encapsulation)
  432. * @inner_mac_header: Link layer header (encapsulation)
  433. * @transport_header: Transport layer header
  434. * @network_header: Network layer header
  435. * @mac_header: Link layer header
  436. * @tail: Tail pointer
  437. * @end: End pointer
  438. * @head: Head of buffer
  439. * @data: Data head pointer
  440. * @truesize: Buffer size
  441. * @users: User count - see {datagram,tcp}.c
  442. */
  443. struct sk_buff {
  444. /* These two members must be first. */
  445. struct sk_buff *next;
  446. struct sk_buff *prev;
  447. union {
  448. ktime_t tstamp;
  449. struct skb_mstamp skb_mstamp;
  450. };
  451. struct sock *sk;
  452. struct net_device *dev;
  453. /*
  454. * This is the control buffer. It is free to use for every
  455. * layer. Please put your private variables there. If you
  456. * want to keep them across layers you have to do a skb_clone()
  457. * first. This is owned by whoever has the skb queued ATM.
  458. */
  459. char cb[48] __aligned(8);
  460. unsigned long _skb_refdst;
  461. #ifdef CONFIG_XFRM
  462. struct sec_path *sp;
  463. #endif
  464. unsigned int len,
  465. data_len;
  466. __u16 mac_len,
  467. hdr_len;
  468. union {
  469. __wsum csum;
  470. struct {
  471. __u16 csum_start;
  472. __u16 csum_offset;
  473. };
  474. };
  475. __u32 priority;
  476. kmemcheck_bitfield_begin(flags1);
  477. __u8 ignore_df:1,
  478. cloned:1,
  479. ip_summed:2,
  480. nohdr:1,
  481. nfctinfo:3;
  482. /* if you move pkt_type around you also must adapt those constants */
  483. #ifdef __BIG_ENDIAN_BITFIELD
  484. #define PKT_TYPE_MAX (7 << 5)
  485. #else
  486. #define PKT_TYPE_MAX 7
  487. #endif
  488. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  489. __u8 __pkt_type_offset[0];
  490. __u8 pkt_type:3,
  491. fclone:2,
  492. ipvs_property:1,
  493. peeked:1,
  494. nf_trace:1;
  495. kmemcheck_bitfield_end(flags1);
  496. __be16 protocol;
  497. void (*destructor)(struct sk_buff *skb);
  498. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  499. struct nf_conntrack *nfct;
  500. #endif
  501. #ifdef CONFIG_BRIDGE_NETFILTER
  502. struct nf_bridge_info *nf_bridge;
  503. #endif
  504. int skb_iif;
  505. __u32 hash;
  506. __be16 vlan_proto;
  507. __u16 vlan_tci;
  508. #ifdef CONFIG_NET_SCHED
  509. __u16 tc_index; /* traffic control index */
  510. #ifdef CONFIG_NET_CLS_ACT
  511. __u16 tc_verd; /* traffic control verdict */
  512. #endif
  513. #endif
  514. __u16 queue_mapping;
  515. kmemcheck_bitfield_begin(flags2);
  516. __u8 xmit_more:1;
  517. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  518. __u8 ndisc_nodetype:2;
  519. #endif
  520. __u8 pfmemalloc:1;
  521. __u8 ooo_okay:1;
  522. __u8 l4_hash:1;
  523. __u8 sw_hash:1;
  524. __u8 wifi_acked_valid:1;
  525. __u8 wifi_acked:1;
  526. __u8 no_fcs:1;
  527. __u8 head_frag:1;
  528. /* Indicates the inner headers are valid in the skbuff. */
  529. __u8 encapsulation:1;
  530. __u8 encap_hdr_csum:1;
  531. __u8 csum_valid:1;
  532. __u8 csum_complete_sw:1;
  533. /* 1/3 bit hole (depending on ndisc_nodetype presence) */
  534. kmemcheck_bitfield_end(flags2);
  535. #if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
  536. union {
  537. unsigned int napi_id;
  538. dma_cookie_t dma_cookie;
  539. };
  540. #endif
  541. #ifdef CONFIG_NETWORK_SECMARK
  542. __u32 secmark;
  543. #endif
  544. union {
  545. __u32 mark;
  546. __u32 dropcount;
  547. __u32 reserved_tailroom;
  548. };
  549. kmemcheck_bitfield_begin(flags3);
  550. __u8 csum_level:2;
  551. __u8 csum_bad:1;
  552. /* 13 bit hole */
  553. kmemcheck_bitfield_end(flags3);
  554. __be16 inner_protocol;
  555. __u16 inner_transport_header;
  556. __u16 inner_network_header;
  557. __u16 inner_mac_header;
  558. __u16 transport_header;
  559. __u16 network_header;
  560. __u16 mac_header;
  561. /* These elements must be at the end, see alloc_skb() for details. */
  562. sk_buff_data_t tail;
  563. sk_buff_data_t end;
  564. unsigned char *head,
  565. *data;
  566. unsigned int truesize;
  567. atomic_t users;
  568. };
  569. #ifdef __KERNEL__
  570. /*
  571. * Handling routines are only of interest to the kernel
  572. */
  573. #include <linux/slab.h>
  574. #define SKB_ALLOC_FCLONE 0x01
  575. #define SKB_ALLOC_RX 0x02
  576. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  577. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  578. {
  579. return unlikely(skb->pfmemalloc);
  580. }
  581. /*
  582. * skb might have a dst pointer attached, refcounted or not.
  583. * _skb_refdst low order bit is set if refcount was _not_ taken
  584. */
  585. #define SKB_DST_NOREF 1UL
  586. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  587. /**
  588. * skb_dst - returns skb dst_entry
  589. * @skb: buffer
  590. *
  591. * Returns skb dst_entry, regardless of reference taken or not.
  592. */
  593. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  594. {
  595. /* If refdst was not refcounted, check we still are in a
  596. * rcu_read_lock section
  597. */
  598. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  599. !rcu_read_lock_held() &&
  600. !rcu_read_lock_bh_held());
  601. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  602. }
  603. /**
  604. * skb_dst_set - sets skb dst
  605. * @skb: buffer
  606. * @dst: dst entry
  607. *
  608. * Sets skb dst, assuming a reference was taken on dst and should
  609. * be released by skb_dst_drop()
  610. */
  611. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  612. {
  613. skb->_skb_refdst = (unsigned long)dst;
  614. }
  615. void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
  616. bool force);
  617. /**
  618. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  619. * @skb: buffer
  620. * @dst: dst entry
  621. *
  622. * Sets skb dst, assuming a reference was not taken on dst.
  623. * If dst entry is cached, we do not take reference and dst_release
  624. * will be avoided by refdst_drop. If dst entry is not cached, we take
  625. * reference, so that last dst_release can destroy the dst immediately.
  626. */
  627. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  628. {
  629. __skb_dst_set_noref(skb, dst, false);
  630. }
  631. /**
  632. * skb_dst_set_noref_force - sets skb dst, without taking reference
  633. * @skb: buffer
  634. * @dst: dst entry
  635. *
  636. * Sets skb dst, assuming a reference was not taken on dst.
  637. * No reference is taken and no dst_release will be called. While for
  638. * cached dsts deferred reclaim is a basic feature, for entries that are
  639. * not cached it is caller's job to guarantee that last dst_release for
  640. * provided dst happens when nobody uses it, eg. after a RCU grace period.
  641. */
  642. static inline void skb_dst_set_noref_force(struct sk_buff *skb,
  643. struct dst_entry *dst)
  644. {
  645. __skb_dst_set_noref(skb, dst, true);
  646. }
  647. /**
  648. * skb_dst_is_noref - Test if skb dst isn't refcounted
  649. * @skb: buffer
  650. */
  651. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  652. {
  653. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  654. }
  655. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  656. {
  657. return (struct rtable *)skb_dst(skb);
  658. }
  659. void kfree_skb(struct sk_buff *skb);
  660. void kfree_skb_list(struct sk_buff *segs);
  661. void skb_tx_error(struct sk_buff *skb);
  662. void consume_skb(struct sk_buff *skb);
  663. void __kfree_skb(struct sk_buff *skb);
  664. extern struct kmem_cache *skbuff_head_cache;
  665. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  666. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  667. bool *fragstolen, int *delta_truesize);
  668. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  669. int node);
  670. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  671. static inline struct sk_buff *alloc_skb(unsigned int size,
  672. gfp_t priority)
  673. {
  674. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  675. }
  676. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  677. unsigned long data_len,
  678. int max_page_order,
  679. int *errcode,
  680. gfp_t gfp_mask);
  681. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  682. gfp_t priority)
  683. {
  684. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  685. }
  686. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  687. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  688. {
  689. return __alloc_skb_head(priority, -1);
  690. }
  691. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  692. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  693. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  694. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  695. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  696. gfp_t gfp_mask, bool fclone);
  697. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  698. gfp_t gfp_mask)
  699. {
  700. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  701. }
  702. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  703. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  704. unsigned int headroom);
  705. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  706. int newtailroom, gfp_t priority);
  707. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  708. int offset, int len);
  709. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  710. int len);
  711. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  712. int skb_pad(struct sk_buff *skb, int pad);
  713. #define dev_kfree_skb(a) consume_skb(a)
  714. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  715. int getfrag(void *from, char *to, int offset,
  716. int len, int odd, struct sk_buff *skb),
  717. void *from, int length);
  718. struct skb_seq_state {
  719. __u32 lower_offset;
  720. __u32 upper_offset;
  721. __u32 frag_idx;
  722. __u32 stepped_offset;
  723. struct sk_buff *root_skb;
  724. struct sk_buff *cur_skb;
  725. __u8 *frag_data;
  726. };
  727. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  728. unsigned int to, struct skb_seq_state *st);
  729. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  730. struct skb_seq_state *st);
  731. void skb_abort_seq_read(struct skb_seq_state *st);
  732. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  733. unsigned int to, struct ts_config *config,
  734. struct ts_state *state);
  735. /*
  736. * Packet hash types specify the type of hash in skb_set_hash.
  737. *
  738. * Hash types refer to the protocol layer addresses which are used to
  739. * construct a packet's hash. The hashes are used to differentiate or identify
  740. * flows of the protocol layer for the hash type. Hash types are either
  741. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  742. *
  743. * Properties of hashes:
  744. *
  745. * 1) Two packets in different flows have different hash values
  746. * 2) Two packets in the same flow should have the same hash value
  747. *
  748. * A hash at a higher layer is considered to be more specific. A driver should
  749. * set the most specific hash possible.
  750. *
  751. * A driver cannot indicate a more specific hash than the layer at which a hash
  752. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  753. *
  754. * A driver may indicate a hash level which is less specific than the
  755. * actual layer the hash was computed on. For instance, a hash computed
  756. * at L4 may be considered an L3 hash. This should only be done if the
  757. * driver can't unambiguously determine that the HW computed the hash at
  758. * the higher layer. Note that the "should" in the second property above
  759. * permits this.
  760. */
  761. enum pkt_hash_types {
  762. PKT_HASH_TYPE_NONE, /* Undefined type */
  763. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  764. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  765. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  766. };
  767. static inline void
  768. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  769. {
  770. skb->l4_hash = (type == PKT_HASH_TYPE_L4);
  771. skb->sw_hash = 0;
  772. skb->hash = hash;
  773. }
  774. void __skb_get_hash(struct sk_buff *skb);
  775. static inline __u32 skb_get_hash(struct sk_buff *skb)
  776. {
  777. if (!skb->l4_hash && !skb->sw_hash)
  778. __skb_get_hash(skb);
  779. return skb->hash;
  780. }
  781. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  782. {
  783. return skb->hash;
  784. }
  785. static inline void skb_clear_hash(struct sk_buff *skb)
  786. {
  787. skb->hash = 0;
  788. skb->sw_hash = 0;
  789. skb->l4_hash = 0;
  790. }
  791. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  792. {
  793. if (!skb->l4_hash)
  794. skb_clear_hash(skb);
  795. }
  796. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  797. {
  798. to->hash = from->hash;
  799. to->sw_hash = from->sw_hash;
  800. to->l4_hash = from->l4_hash;
  801. };
  802. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  803. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  804. {
  805. return skb->head + skb->end;
  806. }
  807. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  808. {
  809. return skb->end;
  810. }
  811. #else
  812. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  813. {
  814. return skb->end;
  815. }
  816. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  817. {
  818. return skb->end - skb->head;
  819. }
  820. #endif
  821. /* Internal */
  822. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  823. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  824. {
  825. return &skb_shinfo(skb)->hwtstamps;
  826. }
  827. /**
  828. * skb_queue_empty - check if a queue is empty
  829. * @list: queue head
  830. *
  831. * Returns true if the queue is empty, false otherwise.
  832. */
  833. static inline int skb_queue_empty(const struct sk_buff_head *list)
  834. {
  835. return list->next == (const struct sk_buff *) list;
  836. }
  837. /**
  838. * skb_queue_is_last - check if skb is the last entry in the queue
  839. * @list: queue head
  840. * @skb: buffer
  841. *
  842. * Returns true if @skb is the last buffer on the list.
  843. */
  844. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  845. const struct sk_buff *skb)
  846. {
  847. return skb->next == (const struct sk_buff *) list;
  848. }
  849. /**
  850. * skb_queue_is_first - check if skb is the first entry in the queue
  851. * @list: queue head
  852. * @skb: buffer
  853. *
  854. * Returns true if @skb is the first buffer on the list.
  855. */
  856. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  857. const struct sk_buff *skb)
  858. {
  859. return skb->prev == (const struct sk_buff *) list;
  860. }
  861. /**
  862. * skb_queue_next - return the next packet in the queue
  863. * @list: queue head
  864. * @skb: current buffer
  865. *
  866. * Return the next packet in @list after @skb. It is only valid to
  867. * call this if skb_queue_is_last() evaluates to false.
  868. */
  869. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  870. const struct sk_buff *skb)
  871. {
  872. /* This BUG_ON may seem severe, but if we just return then we
  873. * are going to dereference garbage.
  874. */
  875. BUG_ON(skb_queue_is_last(list, skb));
  876. return skb->next;
  877. }
  878. /**
  879. * skb_queue_prev - return the prev packet in the queue
  880. * @list: queue head
  881. * @skb: current buffer
  882. *
  883. * Return the prev packet in @list before @skb. It is only valid to
  884. * call this if skb_queue_is_first() evaluates to false.
  885. */
  886. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  887. const struct sk_buff *skb)
  888. {
  889. /* This BUG_ON may seem severe, but if we just return then we
  890. * are going to dereference garbage.
  891. */
  892. BUG_ON(skb_queue_is_first(list, skb));
  893. return skb->prev;
  894. }
  895. /**
  896. * skb_get - reference buffer
  897. * @skb: buffer to reference
  898. *
  899. * Makes another reference to a socket buffer and returns a pointer
  900. * to the buffer.
  901. */
  902. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  903. {
  904. atomic_inc(&skb->users);
  905. return skb;
  906. }
  907. /*
  908. * If users == 1, we are the only owner and are can avoid redundant
  909. * atomic change.
  910. */
  911. /**
  912. * skb_cloned - is the buffer a clone
  913. * @skb: buffer to check
  914. *
  915. * Returns true if the buffer was generated with skb_clone() and is
  916. * one of multiple shared copies of the buffer. Cloned buffers are
  917. * shared data so must not be written to under normal circumstances.
  918. */
  919. static inline int skb_cloned(const struct sk_buff *skb)
  920. {
  921. return skb->cloned &&
  922. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  923. }
  924. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  925. {
  926. might_sleep_if(pri & __GFP_WAIT);
  927. if (skb_cloned(skb))
  928. return pskb_expand_head(skb, 0, 0, pri);
  929. return 0;
  930. }
  931. /**
  932. * skb_header_cloned - is the header a clone
  933. * @skb: buffer to check
  934. *
  935. * Returns true if modifying the header part of the buffer requires
  936. * the data to be copied.
  937. */
  938. static inline int skb_header_cloned(const struct sk_buff *skb)
  939. {
  940. int dataref;
  941. if (!skb->cloned)
  942. return 0;
  943. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  944. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  945. return dataref != 1;
  946. }
  947. /**
  948. * skb_header_release - release reference to header
  949. * @skb: buffer to operate on
  950. *
  951. * Drop a reference to the header part of the buffer. This is done
  952. * by acquiring a payload reference. You must not read from the header
  953. * part of skb->data after this.
  954. * Note : Check if you can use __skb_header_release() instead.
  955. */
  956. static inline void skb_header_release(struct sk_buff *skb)
  957. {
  958. BUG_ON(skb->nohdr);
  959. skb->nohdr = 1;
  960. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  961. }
  962. /**
  963. * __skb_header_release - release reference to header
  964. * @skb: buffer to operate on
  965. *
  966. * Variant of skb_header_release() assuming skb is private to caller.
  967. * We can avoid one atomic operation.
  968. */
  969. static inline void __skb_header_release(struct sk_buff *skb)
  970. {
  971. skb->nohdr = 1;
  972. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  973. }
  974. /**
  975. * skb_shared - is the buffer shared
  976. * @skb: buffer to check
  977. *
  978. * Returns true if more than one person has a reference to this
  979. * buffer.
  980. */
  981. static inline int skb_shared(const struct sk_buff *skb)
  982. {
  983. return atomic_read(&skb->users) != 1;
  984. }
  985. /**
  986. * skb_share_check - check if buffer is shared and if so clone it
  987. * @skb: buffer to check
  988. * @pri: priority for memory allocation
  989. *
  990. * If the buffer is shared the buffer is cloned and the old copy
  991. * drops a reference. A new clone with a single reference is returned.
  992. * If the buffer is not shared the original buffer is returned. When
  993. * being called from interrupt status or with spinlocks held pri must
  994. * be GFP_ATOMIC.
  995. *
  996. * NULL is returned on a memory allocation failure.
  997. */
  998. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  999. {
  1000. might_sleep_if(pri & __GFP_WAIT);
  1001. if (skb_shared(skb)) {
  1002. struct sk_buff *nskb = skb_clone(skb, pri);
  1003. if (likely(nskb))
  1004. consume_skb(skb);
  1005. else
  1006. kfree_skb(skb);
  1007. skb = nskb;
  1008. }
  1009. return skb;
  1010. }
  1011. /*
  1012. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1013. * packets to handle cases where we have a local reader and forward
  1014. * and a couple of other messy ones. The normal one is tcpdumping
  1015. * a packet thats being forwarded.
  1016. */
  1017. /**
  1018. * skb_unshare - make a copy of a shared buffer
  1019. * @skb: buffer to check
  1020. * @pri: priority for memory allocation
  1021. *
  1022. * If the socket buffer is a clone then this function creates a new
  1023. * copy of the data, drops a reference count on the old copy and returns
  1024. * the new copy with the reference count at 1. If the buffer is not a clone
  1025. * the original buffer is returned. When called with a spinlock held or
  1026. * from interrupt state @pri must be %GFP_ATOMIC
  1027. *
  1028. * %NULL is returned on a memory allocation failure.
  1029. */
  1030. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1031. gfp_t pri)
  1032. {
  1033. might_sleep_if(pri & __GFP_WAIT);
  1034. if (skb_cloned(skb)) {
  1035. struct sk_buff *nskb = skb_copy(skb, pri);
  1036. kfree_skb(skb); /* Free our shared copy */
  1037. skb = nskb;
  1038. }
  1039. return skb;
  1040. }
  1041. /**
  1042. * skb_peek - peek at the head of an &sk_buff_head
  1043. * @list_: list to peek at
  1044. *
  1045. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1046. * be careful with this one. A peek leaves the buffer on the
  1047. * list and someone else may run off with it. You must hold
  1048. * the appropriate locks or have a private queue to do this.
  1049. *
  1050. * Returns %NULL for an empty list or a pointer to the head element.
  1051. * The reference count is not incremented and the reference is therefore
  1052. * volatile. Use with caution.
  1053. */
  1054. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1055. {
  1056. struct sk_buff *skb = list_->next;
  1057. if (skb == (struct sk_buff *)list_)
  1058. skb = NULL;
  1059. return skb;
  1060. }
  1061. /**
  1062. * skb_peek_next - peek skb following the given one from a queue
  1063. * @skb: skb to start from
  1064. * @list_: list to peek at
  1065. *
  1066. * Returns %NULL when the end of the list is met or a pointer to the
  1067. * next element. The reference count is not incremented and the
  1068. * reference is therefore volatile. Use with caution.
  1069. */
  1070. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1071. const struct sk_buff_head *list_)
  1072. {
  1073. struct sk_buff *next = skb->next;
  1074. if (next == (struct sk_buff *)list_)
  1075. next = NULL;
  1076. return next;
  1077. }
  1078. /**
  1079. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1080. * @list_: list to peek at
  1081. *
  1082. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1083. * be careful with this one. A peek leaves the buffer on the
  1084. * list and someone else may run off with it. You must hold
  1085. * the appropriate locks or have a private queue to do this.
  1086. *
  1087. * Returns %NULL for an empty list or a pointer to the tail element.
  1088. * The reference count is not incremented and the reference is therefore
  1089. * volatile. Use with caution.
  1090. */
  1091. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1092. {
  1093. struct sk_buff *skb = list_->prev;
  1094. if (skb == (struct sk_buff *)list_)
  1095. skb = NULL;
  1096. return skb;
  1097. }
  1098. /**
  1099. * skb_queue_len - get queue length
  1100. * @list_: list to measure
  1101. *
  1102. * Return the length of an &sk_buff queue.
  1103. */
  1104. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1105. {
  1106. return list_->qlen;
  1107. }
  1108. /**
  1109. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1110. * @list: queue to initialize
  1111. *
  1112. * This initializes only the list and queue length aspects of
  1113. * an sk_buff_head object. This allows to initialize the list
  1114. * aspects of an sk_buff_head without reinitializing things like
  1115. * the spinlock. It can also be used for on-stack sk_buff_head
  1116. * objects where the spinlock is known to not be used.
  1117. */
  1118. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1119. {
  1120. list->prev = list->next = (struct sk_buff *)list;
  1121. list->qlen = 0;
  1122. }
  1123. /*
  1124. * This function creates a split out lock class for each invocation;
  1125. * this is needed for now since a whole lot of users of the skb-queue
  1126. * infrastructure in drivers have different locking usage (in hardirq)
  1127. * than the networking core (in softirq only). In the long run either the
  1128. * network layer or drivers should need annotation to consolidate the
  1129. * main types of usage into 3 classes.
  1130. */
  1131. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1132. {
  1133. spin_lock_init(&list->lock);
  1134. __skb_queue_head_init(list);
  1135. }
  1136. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1137. struct lock_class_key *class)
  1138. {
  1139. skb_queue_head_init(list);
  1140. lockdep_set_class(&list->lock, class);
  1141. }
  1142. /*
  1143. * Insert an sk_buff on a list.
  1144. *
  1145. * The "__skb_xxxx()" functions are the non-atomic ones that
  1146. * can only be called with interrupts disabled.
  1147. */
  1148. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1149. struct sk_buff_head *list);
  1150. static inline void __skb_insert(struct sk_buff *newsk,
  1151. struct sk_buff *prev, struct sk_buff *next,
  1152. struct sk_buff_head *list)
  1153. {
  1154. newsk->next = next;
  1155. newsk->prev = prev;
  1156. next->prev = prev->next = newsk;
  1157. list->qlen++;
  1158. }
  1159. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1160. struct sk_buff *prev,
  1161. struct sk_buff *next)
  1162. {
  1163. struct sk_buff *first = list->next;
  1164. struct sk_buff *last = list->prev;
  1165. first->prev = prev;
  1166. prev->next = first;
  1167. last->next = next;
  1168. next->prev = last;
  1169. }
  1170. /**
  1171. * skb_queue_splice - join two skb lists, this is designed for stacks
  1172. * @list: the new list to add
  1173. * @head: the place to add it in the first list
  1174. */
  1175. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1176. struct sk_buff_head *head)
  1177. {
  1178. if (!skb_queue_empty(list)) {
  1179. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1180. head->qlen += list->qlen;
  1181. }
  1182. }
  1183. /**
  1184. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1185. * @list: the new list to add
  1186. * @head: the place to add it in the first list
  1187. *
  1188. * The list at @list is reinitialised
  1189. */
  1190. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1191. struct sk_buff_head *head)
  1192. {
  1193. if (!skb_queue_empty(list)) {
  1194. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1195. head->qlen += list->qlen;
  1196. __skb_queue_head_init(list);
  1197. }
  1198. }
  1199. /**
  1200. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1201. * @list: the new list to add
  1202. * @head: the place to add it in the first list
  1203. */
  1204. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1205. struct sk_buff_head *head)
  1206. {
  1207. if (!skb_queue_empty(list)) {
  1208. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1209. head->qlen += list->qlen;
  1210. }
  1211. }
  1212. /**
  1213. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1214. * @list: the new list to add
  1215. * @head: the place to add it in the first list
  1216. *
  1217. * Each of the lists is a queue.
  1218. * The list at @list is reinitialised
  1219. */
  1220. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1221. struct sk_buff_head *head)
  1222. {
  1223. if (!skb_queue_empty(list)) {
  1224. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1225. head->qlen += list->qlen;
  1226. __skb_queue_head_init(list);
  1227. }
  1228. }
  1229. /**
  1230. * __skb_queue_after - queue a buffer at the list head
  1231. * @list: list to use
  1232. * @prev: place after this buffer
  1233. * @newsk: buffer to queue
  1234. *
  1235. * Queue a buffer int the middle of a list. This function takes no locks
  1236. * and you must therefore hold required locks before calling it.
  1237. *
  1238. * A buffer cannot be placed on two lists at the same time.
  1239. */
  1240. static inline void __skb_queue_after(struct sk_buff_head *list,
  1241. struct sk_buff *prev,
  1242. struct sk_buff *newsk)
  1243. {
  1244. __skb_insert(newsk, prev, prev->next, list);
  1245. }
  1246. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1247. struct sk_buff_head *list);
  1248. static inline void __skb_queue_before(struct sk_buff_head *list,
  1249. struct sk_buff *next,
  1250. struct sk_buff *newsk)
  1251. {
  1252. __skb_insert(newsk, next->prev, next, list);
  1253. }
  1254. /**
  1255. * __skb_queue_head - queue a buffer at the list head
  1256. * @list: list to use
  1257. * @newsk: buffer to queue
  1258. *
  1259. * Queue a buffer at the start of a list. This function takes no locks
  1260. * and you must therefore hold required locks before calling it.
  1261. *
  1262. * A buffer cannot be placed on two lists at the same time.
  1263. */
  1264. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1265. static inline void __skb_queue_head(struct sk_buff_head *list,
  1266. struct sk_buff *newsk)
  1267. {
  1268. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1269. }
  1270. /**
  1271. * __skb_queue_tail - queue a buffer at the list tail
  1272. * @list: list to use
  1273. * @newsk: buffer to queue
  1274. *
  1275. * Queue a buffer at the end of a list. This function takes no locks
  1276. * and you must therefore hold required locks before calling it.
  1277. *
  1278. * A buffer cannot be placed on two lists at the same time.
  1279. */
  1280. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1281. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1282. struct sk_buff *newsk)
  1283. {
  1284. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1285. }
  1286. /*
  1287. * remove sk_buff from list. _Must_ be called atomically, and with
  1288. * the list known..
  1289. */
  1290. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1291. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1292. {
  1293. struct sk_buff *next, *prev;
  1294. list->qlen--;
  1295. next = skb->next;
  1296. prev = skb->prev;
  1297. skb->next = skb->prev = NULL;
  1298. next->prev = prev;
  1299. prev->next = next;
  1300. }
  1301. /**
  1302. * __skb_dequeue - remove from the head of the queue
  1303. * @list: list to dequeue from
  1304. *
  1305. * Remove the head of the list. This function does not take any locks
  1306. * so must be used with appropriate locks held only. The head item is
  1307. * returned or %NULL if the list is empty.
  1308. */
  1309. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1310. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1311. {
  1312. struct sk_buff *skb = skb_peek(list);
  1313. if (skb)
  1314. __skb_unlink(skb, list);
  1315. return skb;
  1316. }
  1317. /**
  1318. * __skb_dequeue_tail - remove from the tail of the queue
  1319. * @list: list to dequeue from
  1320. *
  1321. * Remove the tail of the list. This function does not take any locks
  1322. * so must be used with appropriate locks held only. The tail item is
  1323. * returned or %NULL if the list is empty.
  1324. */
  1325. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1326. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1327. {
  1328. struct sk_buff *skb = skb_peek_tail(list);
  1329. if (skb)
  1330. __skb_unlink(skb, list);
  1331. return skb;
  1332. }
  1333. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1334. {
  1335. return skb->data_len;
  1336. }
  1337. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1338. {
  1339. return skb->len - skb->data_len;
  1340. }
  1341. static inline int skb_pagelen(const struct sk_buff *skb)
  1342. {
  1343. int i, len = 0;
  1344. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1345. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1346. return len + skb_headlen(skb);
  1347. }
  1348. /**
  1349. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1350. * @skb: buffer containing fragment to be initialised
  1351. * @i: paged fragment index to initialise
  1352. * @page: the page to use for this fragment
  1353. * @off: the offset to the data with @page
  1354. * @size: the length of the data
  1355. *
  1356. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1357. * offset @off within @page.
  1358. *
  1359. * Does not take any additional reference on the fragment.
  1360. */
  1361. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1362. struct page *page, int off, int size)
  1363. {
  1364. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1365. /*
  1366. * Propagate page->pfmemalloc to the skb if we can. The problem is
  1367. * that not all callers have unique ownership of the page. If
  1368. * pfmemalloc is set, we check the mapping as a mapping implies
  1369. * page->index is set (index and pfmemalloc share space).
  1370. * If it's a valid mapping, we cannot use page->pfmemalloc but we
  1371. * do not lose pfmemalloc information as the pages would not be
  1372. * allocated using __GFP_MEMALLOC.
  1373. */
  1374. frag->page.p = page;
  1375. frag->page_offset = off;
  1376. skb_frag_size_set(frag, size);
  1377. page = compound_head(page);
  1378. if (page->pfmemalloc && !page->mapping)
  1379. skb->pfmemalloc = true;
  1380. }
  1381. /**
  1382. * skb_fill_page_desc - initialise a paged fragment in an skb
  1383. * @skb: buffer containing fragment to be initialised
  1384. * @i: paged fragment index to initialise
  1385. * @page: the page to use for this fragment
  1386. * @off: the offset to the data with @page
  1387. * @size: the length of the data
  1388. *
  1389. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1390. * @skb to point to @size bytes at offset @off within @page. In
  1391. * addition updates @skb such that @i is the last fragment.
  1392. *
  1393. * Does not take any additional reference on the fragment.
  1394. */
  1395. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1396. struct page *page, int off, int size)
  1397. {
  1398. __skb_fill_page_desc(skb, i, page, off, size);
  1399. skb_shinfo(skb)->nr_frags = i + 1;
  1400. }
  1401. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1402. int size, unsigned int truesize);
  1403. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1404. unsigned int truesize);
  1405. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1406. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1407. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1408. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1409. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1410. {
  1411. return skb->head + skb->tail;
  1412. }
  1413. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1414. {
  1415. skb->tail = skb->data - skb->head;
  1416. }
  1417. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1418. {
  1419. skb_reset_tail_pointer(skb);
  1420. skb->tail += offset;
  1421. }
  1422. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1423. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1424. {
  1425. return skb->tail;
  1426. }
  1427. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1428. {
  1429. skb->tail = skb->data;
  1430. }
  1431. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1432. {
  1433. skb->tail = skb->data + offset;
  1434. }
  1435. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1436. /*
  1437. * Add data to an sk_buff
  1438. */
  1439. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1440. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1441. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1442. {
  1443. unsigned char *tmp = skb_tail_pointer(skb);
  1444. SKB_LINEAR_ASSERT(skb);
  1445. skb->tail += len;
  1446. skb->len += len;
  1447. return tmp;
  1448. }
  1449. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1450. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1451. {
  1452. skb->data -= len;
  1453. skb->len += len;
  1454. return skb->data;
  1455. }
  1456. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1457. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1458. {
  1459. skb->len -= len;
  1460. BUG_ON(skb->len < skb->data_len);
  1461. return skb->data += len;
  1462. }
  1463. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1464. {
  1465. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1466. }
  1467. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1468. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1469. {
  1470. if (len > skb_headlen(skb) &&
  1471. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1472. return NULL;
  1473. skb->len -= len;
  1474. return skb->data += len;
  1475. }
  1476. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1477. {
  1478. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1479. }
  1480. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1481. {
  1482. if (likely(len <= skb_headlen(skb)))
  1483. return 1;
  1484. if (unlikely(len > skb->len))
  1485. return 0;
  1486. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1487. }
  1488. /**
  1489. * skb_headroom - bytes at buffer head
  1490. * @skb: buffer to check
  1491. *
  1492. * Return the number of bytes of free space at the head of an &sk_buff.
  1493. */
  1494. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1495. {
  1496. return skb->data - skb->head;
  1497. }
  1498. /**
  1499. * skb_tailroom - bytes at buffer end
  1500. * @skb: buffer to check
  1501. *
  1502. * Return the number of bytes of free space at the tail of an sk_buff
  1503. */
  1504. static inline int skb_tailroom(const struct sk_buff *skb)
  1505. {
  1506. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1507. }
  1508. /**
  1509. * skb_availroom - bytes at buffer end
  1510. * @skb: buffer to check
  1511. *
  1512. * Return the number of bytes of free space at the tail of an sk_buff
  1513. * allocated by sk_stream_alloc()
  1514. */
  1515. static inline int skb_availroom(const struct sk_buff *skb)
  1516. {
  1517. if (skb_is_nonlinear(skb))
  1518. return 0;
  1519. return skb->end - skb->tail - skb->reserved_tailroom;
  1520. }
  1521. /**
  1522. * skb_reserve - adjust headroom
  1523. * @skb: buffer to alter
  1524. * @len: bytes to move
  1525. *
  1526. * Increase the headroom of an empty &sk_buff by reducing the tail
  1527. * room. This is only allowed for an empty buffer.
  1528. */
  1529. static inline void skb_reserve(struct sk_buff *skb, int len)
  1530. {
  1531. skb->data += len;
  1532. skb->tail += len;
  1533. }
  1534. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1535. {
  1536. skb->inner_mac_header = skb->mac_header;
  1537. skb->inner_network_header = skb->network_header;
  1538. skb->inner_transport_header = skb->transport_header;
  1539. }
  1540. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1541. {
  1542. skb->mac_len = skb->network_header - skb->mac_header;
  1543. }
  1544. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1545. *skb)
  1546. {
  1547. return skb->head + skb->inner_transport_header;
  1548. }
  1549. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1550. {
  1551. skb->inner_transport_header = skb->data - skb->head;
  1552. }
  1553. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1554. const int offset)
  1555. {
  1556. skb_reset_inner_transport_header(skb);
  1557. skb->inner_transport_header += offset;
  1558. }
  1559. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1560. {
  1561. return skb->head + skb->inner_network_header;
  1562. }
  1563. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1564. {
  1565. skb->inner_network_header = skb->data - skb->head;
  1566. }
  1567. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1568. const int offset)
  1569. {
  1570. skb_reset_inner_network_header(skb);
  1571. skb->inner_network_header += offset;
  1572. }
  1573. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1574. {
  1575. return skb->head + skb->inner_mac_header;
  1576. }
  1577. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1578. {
  1579. skb->inner_mac_header = skb->data - skb->head;
  1580. }
  1581. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1582. const int offset)
  1583. {
  1584. skb_reset_inner_mac_header(skb);
  1585. skb->inner_mac_header += offset;
  1586. }
  1587. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1588. {
  1589. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1590. }
  1591. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1592. {
  1593. return skb->head + skb->transport_header;
  1594. }
  1595. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1596. {
  1597. skb->transport_header = skb->data - skb->head;
  1598. }
  1599. static inline void skb_set_transport_header(struct sk_buff *skb,
  1600. const int offset)
  1601. {
  1602. skb_reset_transport_header(skb);
  1603. skb->transport_header += offset;
  1604. }
  1605. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1606. {
  1607. return skb->head + skb->network_header;
  1608. }
  1609. static inline void skb_reset_network_header(struct sk_buff *skb)
  1610. {
  1611. skb->network_header = skb->data - skb->head;
  1612. }
  1613. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1614. {
  1615. skb_reset_network_header(skb);
  1616. skb->network_header += offset;
  1617. }
  1618. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1619. {
  1620. return skb->head + skb->mac_header;
  1621. }
  1622. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1623. {
  1624. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1625. }
  1626. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1627. {
  1628. skb->mac_header = skb->data - skb->head;
  1629. }
  1630. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1631. {
  1632. skb_reset_mac_header(skb);
  1633. skb->mac_header += offset;
  1634. }
  1635. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1636. {
  1637. skb->mac_header = skb->network_header;
  1638. }
  1639. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1640. const int offset_hint)
  1641. {
  1642. struct flow_keys keys;
  1643. if (skb_transport_header_was_set(skb))
  1644. return;
  1645. else if (skb_flow_dissect(skb, &keys))
  1646. skb_set_transport_header(skb, keys.thoff);
  1647. else
  1648. skb_set_transport_header(skb, offset_hint);
  1649. }
  1650. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1651. {
  1652. if (skb_mac_header_was_set(skb)) {
  1653. const unsigned char *old_mac = skb_mac_header(skb);
  1654. skb_set_mac_header(skb, -skb->mac_len);
  1655. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1656. }
  1657. }
  1658. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1659. {
  1660. return skb->csum_start - skb_headroom(skb);
  1661. }
  1662. static inline int skb_transport_offset(const struct sk_buff *skb)
  1663. {
  1664. return skb_transport_header(skb) - skb->data;
  1665. }
  1666. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1667. {
  1668. return skb->transport_header - skb->network_header;
  1669. }
  1670. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1671. {
  1672. return skb->inner_transport_header - skb->inner_network_header;
  1673. }
  1674. static inline int skb_network_offset(const struct sk_buff *skb)
  1675. {
  1676. return skb_network_header(skb) - skb->data;
  1677. }
  1678. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1679. {
  1680. return skb_inner_network_header(skb) - skb->data;
  1681. }
  1682. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1683. {
  1684. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1685. }
  1686. /*
  1687. * CPUs often take a performance hit when accessing unaligned memory
  1688. * locations. The actual performance hit varies, it can be small if the
  1689. * hardware handles it or large if we have to take an exception and fix it
  1690. * in software.
  1691. *
  1692. * Since an ethernet header is 14 bytes network drivers often end up with
  1693. * the IP header at an unaligned offset. The IP header can be aligned by
  1694. * shifting the start of the packet by 2 bytes. Drivers should do this
  1695. * with:
  1696. *
  1697. * skb_reserve(skb, NET_IP_ALIGN);
  1698. *
  1699. * The downside to this alignment of the IP header is that the DMA is now
  1700. * unaligned. On some architectures the cost of an unaligned DMA is high
  1701. * and this cost outweighs the gains made by aligning the IP header.
  1702. *
  1703. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1704. * to be overridden.
  1705. */
  1706. #ifndef NET_IP_ALIGN
  1707. #define NET_IP_ALIGN 2
  1708. #endif
  1709. /*
  1710. * The networking layer reserves some headroom in skb data (via
  1711. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1712. * the header has to grow. In the default case, if the header has to grow
  1713. * 32 bytes or less we avoid the reallocation.
  1714. *
  1715. * Unfortunately this headroom changes the DMA alignment of the resulting
  1716. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1717. * on some architectures. An architecture can override this value,
  1718. * perhaps setting it to a cacheline in size (since that will maintain
  1719. * cacheline alignment of the DMA). It must be a power of 2.
  1720. *
  1721. * Various parts of the networking layer expect at least 32 bytes of
  1722. * headroom, you should not reduce this.
  1723. *
  1724. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1725. * to reduce average number of cache lines per packet.
  1726. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1727. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1728. */
  1729. #ifndef NET_SKB_PAD
  1730. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1731. #endif
  1732. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1733. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1734. {
  1735. if (unlikely(skb_is_nonlinear(skb))) {
  1736. WARN_ON(1);
  1737. return;
  1738. }
  1739. skb->len = len;
  1740. skb_set_tail_pointer(skb, len);
  1741. }
  1742. void skb_trim(struct sk_buff *skb, unsigned int len);
  1743. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1744. {
  1745. if (skb->data_len)
  1746. return ___pskb_trim(skb, len);
  1747. __skb_trim(skb, len);
  1748. return 0;
  1749. }
  1750. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1751. {
  1752. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1753. }
  1754. /**
  1755. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1756. * @skb: buffer to alter
  1757. * @len: new length
  1758. *
  1759. * This is identical to pskb_trim except that the caller knows that
  1760. * the skb is not cloned so we should never get an error due to out-
  1761. * of-memory.
  1762. */
  1763. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  1764. {
  1765. int err = pskb_trim(skb, len);
  1766. BUG_ON(err);
  1767. }
  1768. /**
  1769. * skb_orphan - orphan a buffer
  1770. * @skb: buffer to orphan
  1771. *
  1772. * If a buffer currently has an owner then we call the owner's
  1773. * destructor function and make the @skb unowned. The buffer continues
  1774. * to exist but is no longer charged to its former owner.
  1775. */
  1776. static inline void skb_orphan(struct sk_buff *skb)
  1777. {
  1778. if (skb->destructor) {
  1779. skb->destructor(skb);
  1780. skb->destructor = NULL;
  1781. skb->sk = NULL;
  1782. } else {
  1783. BUG_ON(skb->sk);
  1784. }
  1785. }
  1786. /**
  1787. * skb_orphan_frags - orphan the frags contained in a buffer
  1788. * @skb: buffer to orphan frags from
  1789. * @gfp_mask: allocation mask for replacement pages
  1790. *
  1791. * For each frag in the SKB which needs a destructor (i.e. has an
  1792. * owner) create a copy of that frag and release the original
  1793. * page by calling the destructor.
  1794. */
  1795. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  1796. {
  1797. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  1798. return 0;
  1799. return skb_copy_ubufs(skb, gfp_mask);
  1800. }
  1801. /**
  1802. * __skb_queue_purge - empty a list
  1803. * @list: list to empty
  1804. *
  1805. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  1806. * the list and one reference dropped. This function does not take the
  1807. * list lock and the caller must hold the relevant locks to use it.
  1808. */
  1809. void skb_queue_purge(struct sk_buff_head *list);
  1810. static inline void __skb_queue_purge(struct sk_buff_head *list)
  1811. {
  1812. struct sk_buff *skb;
  1813. while ((skb = __skb_dequeue(list)) != NULL)
  1814. kfree_skb(skb);
  1815. }
  1816. #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
  1817. #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
  1818. #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
  1819. void *netdev_alloc_frag(unsigned int fragsz);
  1820. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  1821. gfp_t gfp_mask);
  1822. /**
  1823. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  1824. * @dev: network device to receive on
  1825. * @length: length to allocate
  1826. *
  1827. * Allocate a new &sk_buff and assign it a usage count of one. The
  1828. * buffer has unspecified headroom built in. Users should allocate
  1829. * the headroom they think they need without accounting for the
  1830. * built in space. The built in space is used for optimisations.
  1831. *
  1832. * %NULL is returned if there is no free memory. Although this function
  1833. * allocates memory it can be called from an interrupt.
  1834. */
  1835. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  1836. unsigned int length)
  1837. {
  1838. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  1839. }
  1840. /* legacy helper around __netdev_alloc_skb() */
  1841. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  1842. gfp_t gfp_mask)
  1843. {
  1844. return __netdev_alloc_skb(NULL, length, gfp_mask);
  1845. }
  1846. /* legacy helper around netdev_alloc_skb() */
  1847. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  1848. {
  1849. return netdev_alloc_skb(NULL, length);
  1850. }
  1851. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  1852. unsigned int length, gfp_t gfp)
  1853. {
  1854. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  1855. if (NET_IP_ALIGN && skb)
  1856. skb_reserve(skb, NET_IP_ALIGN);
  1857. return skb;
  1858. }
  1859. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  1860. unsigned int length)
  1861. {
  1862. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  1863. }
  1864. /**
  1865. * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
  1866. * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
  1867. * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
  1868. * @order: size of the allocation
  1869. *
  1870. * Allocate a new page.
  1871. *
  1872. * %NULL is returned if there is no free memory.
  1873. */
  1874. static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
  1875. struct sk_buff *skb,
  1876. unsigned int order)
  1877. {
  1878. struct page *page;
  1879. gfp_mask |= __GFP_COLD;
  1880. if (!(gfp_mask & __GFP_NOMEMALLOC))
  1881. gfp_mask |= __GFP_MEMALLOC;
  1882. page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  1883. if (skb && page && page->pfmemalloc)
  1884. skb->pfmemalloc = true;
  1885. return page;
  1886. }
  1887. /**
  1888. * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
  1889. * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
  1890. * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
  1891. *
  1892. * Allocate a new page.
  1893. *
  1894. * %NULL is returned if there is no free memory.
  1895. */
  1896. static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
  1897. struct sk_buff *skb)
  1898. {
  1899. return __skb_alloc_pages(gfp_mask, skb, 0);
  1900. }
  1901. /**
  1902. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  1903. * @page: The page that was allocated from skb_alloc_page
  1904. * @skb: The skb that may need pfmemalloc set
  1905. */
  1906. static inline void skb_propagate_pfmemalloc(struct page *page,
  1907. struct sk_buff *skb)
  1908. {
  1909. if (page && page->pfmemalloc)
  1910. skb->pfmemalloc = true;
  1911. }
  1912. /**
  1913. * skb_frag_page - retrieve the page referred to by a paged fragment
  1914. * @frag: the paged fragment
  1915. *
  1916. * Returns the &struct page associated with @frag.
  1917. */
  1918. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  1919. {
  1920. return frag->page.p;
  1921. }
  1922. /**
  1923. * __skb_frag_ref - take an addition reference on a paged fragment.
  1924. * @frag: the paged fragment
  1925. *
  1926. * Takes an additional reference on the paged fragment @frag.
  1927. */
  1928. static inline void __skb_frag_ref(skb_frag_t *frag)
  1929. {
  1930. get_page(skb_frag_page(frag));
  1931. }
  1932. /**
  1933. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  1934. * @skb: the buffer
  1935. * @f: the fragment offset.
  1936. *
  1937. * Takes an additional reference on the @f'th paged fragment of @skb.
  1938. */
  1939. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  1940. {
  1941. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  1942. }
  1943. /**
  1944. * __skb_frag_unref - release a reference on a paged fragment.
  1945. * @frag: the paged fragment
  1946. *
  1947. * Releases a reference on the paged fragment @frag.
  1948. */
  1949. static inline void __skb_frag_unref(skb_frag_t *frag)
  1950. {
  1951. put_page(skb_frag_page(frag));
  1952. }
  1953. /**
  1954. * skb_frag_unref - release a reference on a paged fragment of an skb.
  1955. * @skb: the buffer
  1956. * @f: the fragment offset
  1957. *
  1958. * Releases a reference on the @f'th paged fragment of @skb.
  1959. */
  1960. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  1961. {
  1962. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  1963. }
  1964. /**
  1965. * skb_frag_address - gets the address of the data contained in a paged fragment
  1966. * @frag: the paged fragment buffer
  1967. *
  1968. * Returns the address of the data within @frag. The page must already
  1969. * be mapped.
  1970. */
  1971. static inline void *skb_frag_address(const skb_frag_t *frag)
  1972. {
  1973. return page_address(skb_frag_page(frag)) + frag->page_offset;
  1974. }
  1975. /**
  1976. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  1977. * @frag: the paged fragment buffer
  1978. *
  1979. * Returns the address of the data within @frag. Checks that the page
  1980. * is mapped and returns %NULL otherwise.
  1981. */
  1982. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  1983. {
  1984. void *ptr = page_address(skb_frag_page(frag));
  1985. if (unlikely(!ptr))
  1986. return NULL;
  1987. return ptr + frag->page_offset;
  1988. }
  1989. /**
  1990. * __skb_frag_set_page - sets the page contained in a paged fragment
  1991. * @frag: the paged fragment
  1992. * @page: the page to set
  1993. *
  1994. * Sets the fragment @frag to contain @page.
  1995. */
  1996. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  1997. {
  1998. frag->page.p = page;
  1999. }
  2000. /**
  2001. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2002. * @skb: the buffer
  2003. * @f: the fragment offset
  2004. * @page: the page to set
  2005. *
  2006. * Sets the @f'th fragment of @skb to contain @page.
  2007. */
  2008. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2009. struct page *page)
  2010. {
  2011. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2012. }
  2013. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2014. /**
  2015. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2016. * @dev: the device to map the fragment to
  2017. * @frag: the paged fragment to map
  2018. * @offset: the offset within the fragment (starting at the
  2019. * fragment's own offset)
  2020. * @size: the number of bytes to map
  2021. * @dir: the direction of the mapping (%PCI_DMA_*)
  2022. *
  2023. * Maps the page associated with @frag to @device.
  2024. */
  2025. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2026. const skb_frag_t *frag,
  2027. size_t offset, size_t size,
  2028. enum dma_data_direction dir)
  2029. {
  2030. return dma_map_page(dev, skb_frag_page(frag),
  2031. frag->page_offset + offset, size, dir);
  2032. }
  2033. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2034. gfp_t gfp_mask)
  2035. {
  2036. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2037. }
  2038. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2039. gfp_t gfp_mask)
  2040. {
  2041. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2042. }
  2043. /**
  2044. * skb_clone_writable - is the header of a clone writable
  2045. * @skb: buffer to check
  2046. * @len: length up to which to write
  2047. *
  2048. * Returns true if modifying the header part of the cloned buffer
  2049. * does not requires the data to be copied.
  2050. */
  2051. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2052. {
  2053. return !skb_header_cloned(skb) &&
  2054. skb_headroom(skb) + len <= skb->hdr_len;
  2055. }
  2056. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2057. int cloned)
  2058. {
  2059. int delta = 0;
  2060. if (headroom > skb_headroom(skb))
  2061. delta = headroom - skb_headroom(skb);
  2062. if (delta || cloned)
  2063. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2064. GFP_ATOMIC);
  2065. return 0;
  2066. }
  2067. /**
  2068. * skb_cow - copy header of skb when it is required
  2069. * @skb: buffer to cow
  2070. * @headroom: needed headroom
  2071. *
  2072. * If the skb passed lacks sufficient headroom or its data part
  2073. * is shared, data is reallocated. If reallocation fails, an error
  2074. * is returned and original skb is not changed.
  2075. *
  2076. * The result is skb with writable area skb->head...skb->tail
  2077. * and at least @headroom of space at head.
  2078. */
  2079. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2080. {
  2081. return __skb_cow(skb, headroom, skb_cloned(skb));
  2082. }
  2083. /**
  2084. * skb_cow_head - skb_cow but only making the head writable
  2085. * @skb: buffer to cow
  2086. * @headroom: needed headroom
  2087. *
  2088. * This function is identical to skb_cow except that we replace the
  2089. * skb_cloned check by skb_header_cloned. It should be used when
  2090. * you only need to push on some header and do not need to modify
  2091. * the data.
  2092. */
  2093. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2094. {
  2095. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2096. }
  2097. /**
  2098. * skb_padto - pad an skbuff up to a minimal size
  2099. * @skb: buffer to pad
  2100. * @len: minimal length
  2101. *
  2102. * Pads up a buffer to ensure the trailing bytes exist and are
  2103. * blanked. If the buffer already contains sufficient data it
  2104. * is untouched. Otherwise it is extended. Returns zero on
  2105. * success. The skb is freed on error.
  2106. */
  2107. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2108. {
  2109. unsigned int size = skb->len;
  2110. if (likely(size >= len))
  2111. return 0;
  2112. return skb_pad(skb, len - size);
  2113. }
  2114. static inline int skb_add_data(struct sk_buff *skb,
  2115. char __user *from, int copy)
  2116. {
  2117. const int off = skb->len;
  2118. if (skb->ip_summed == CHECKSUM_NONE) {
  2119. int err = 0;
  2120. __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
  2121. copy, 0, &err);
  2122. if (!err) {
  2123. skb->csum = csum_block_add(skb->csum, csum, off);
  2124. return 0;
  2125. }
  2126. } else if (!copy_from_user(skb_put(skb, copy), from, copy))
  2127. return 0;
  2128. __skb_trim(skb, off);
  2129. return -EFAULT;
  2130. }
  2131. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2132. const struct page *page, int off)
  2133. {
  2134. if (i) {
  2135. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2136. return page == skb_frag_page(frag) &&
  2137. off == frag->page_offset + skb_frag_size(frag);
  2138. }
  2139. return false;
  2140. }
  2141. static inline int __skb_linearize(struct sk_buff *skb)
  2142. {
  2143. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2144. }
  2145. /**
  2146. * skb_linearize - convert paged skb to linear one
  2147. * @skb: buffer to linarize
  2148. *
  2149. * If there is no free memory -ENOMEM is returned, otherwise zero
  2150. * is returned and the old skb data released.
  2151. */
  2152. static inline int skb_linearize(struct sk_buff *skb)
  2153. {
  2154. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2155. }
  2156. /**
  2157. * skb_has_shared_frag - can any frag be overwritten
  2158. * @skb: buffer to test
  2159. *
  2160. * Return true if the skb has at least one frag that might be modified
  2161. * by an external entity (as in vmsplice()/sendfile())
  2162. */
  2163. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2164. {
  2165. return skb_is_nonlinear(skb) &&
  2166. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2167. }
  2168. /**
  2169. * skb_linearize_cow - make sure skb is linear and writable
  2170. * @skb: buffer to process
  2171. *
  2172. * If there is no free memory -ENOMEM is returned, otherwise zero
  2173. * is returned and the old skb data released.
  2174. */
  2175. static inline int skb_linearize_cow(struct sk_buff *skb)
  2176. {
  2177. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2178. __skb_linearize(skb) : 0;
  2179. }
  2180. /**
  2181. * skb_postpull_rcsum - update checksum for received skb after pull
  2182. * @skb: buffer to update
  2183. * @start: start of data before pull
  2184. * @len: length of data pulled
  2185. *
  2186. * After doing a pull on a received packet, you need to call this to
  2187. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2188. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2189. */
  2190. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2191. const void *start, unsigned int len)
  2192. {
  2193. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2194. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2195. }
  2196. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2197. /**
  2198. * pskb_trim_rcsum - trim received skb and update checksum
  2199. * @skb: buffer to trim
  2200. * @len: new length
  2201. *
  2202. * This is exactly the same as pskb_trim except that it ensures the
  2203. * checksum of received packets are still valid after the operation.
  2204. */
  2205. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2206. {
  2207. if (likely(len >= skb->len))
  2208. return 0;
  2209. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2210. skb->ip_summed = CHECKSUM_NONE;
  2211. return __pskb_trim(skb, len);
  2212. }
  2213. #define skb_queue_walk(queue, skb) \
  2214. for (skb = (queue)->next; \
  2215. skb != (struct sk_buff *)(queue); \
  2216. skb = skb->next)
  2217. #define skb_queue_walk_safe(queue, skb, tmp) \
  2218. for (skb = (queue)->next, tmp = skb->next; \
  2219. skb != (struct sk_buff *)(queue); \
  2220. skb = tmp, tmp = skb->next)
  2221. #define skb_queue_walk_from(queue, skb) \
  2222. for (; skb != (struct sk_buff *)(queue); \
  2223. skb = skb->next)
  2224. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2225. for (tmp = skb->next; \
  2226. skb != (struct sk_buff *)(queue); \
  2227. skb = tmp, tmp = skb->next)
  2228. #define skb_queue_reverse_walk(queue, skb) \
  2229. for (skb = (queue)->prev; \
  2230. skb != (struct sk_buff *)(queue); \
  2231. skb = skb->prev)
  2232. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2233. for (skb = (queue)->prev, tmp = skb->prev; \
  2234. skb != (struct sk_buff *)(queue); \
  2235. skb = tmp, tmp = skb->prev)
  2236. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2237. for (tmp = skb->prev; \
  2238. skb != (struct sk_buff *)(queue); \
  2239. skb = tmp, tmp = skb->prev)
  2240. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2241. {
  2242. return skb_shinfo(skb)->frag_list != NULL;
  2243. }
  2244. static inline void skb_frag_list_init(struct sk_buff *skb)
  2245. {
  2246. skb_shinfo(skb)->frag_list = NULL;
  2247. }
  2248. static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
  2249. {
  2250. frag->next = skb_shinfo(skb)->frag_list;
  2251. skb_shinfo(skb)->frag_list = frag;
  2252. }
  2253. #define skb_walk_frags(skb, iter) \
  2254. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2255. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2256. int *peeked, int *off, int *err);
  2257. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2258. int *err);
  2259. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2260. struct poll_table_struct *wait);
  2261. int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
  2262. struct iovec *to, int size);
  2263. int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
  2264. struct iovec *iov);
  2265. int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
  2266. const struct iovec *from, int from_offset,
  2267. int len);
  2268. int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
  2269. int offset, size_t count);
  2270. int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
  2271. const struct iovec *to, int to_offset,
  2272. int size);
  2273. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2274. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2275. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2276. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2277. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2278. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2279. int len, __wsum csum);
  2280. int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
  2281. struct pipe_inode_info *pipe, unsigned int len,
  2282. unsigned int flags);
  2283. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2284. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2285. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2286. int len, int hlen);
  2287. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2288. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2289. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2290. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2291. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2292. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2293. struct skb_checksum_ops {
  2294. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2295. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2296. };
  2297. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2298. __wsum csum, const struct skb_checksum_ops *ops);
  2299. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2300. __wsum csum);
  2301. static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
  2302. int len, void *data, int hlen, void *buffer)
  2303. {
  2304. if (hlen - offset >= len)
  2305. return data + offset;
  2306. if (!skb ||
  2307. skb_copy_bits(skb, offset, buffer, len) < 0)
  2308. return NULL;
  2309. return buffer;
  2310. }
  2311. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  2312. int len, void *buffer)
  2313. {
  2314. return __skb_header_pointer(skb, offset, len, skb->data,
  2315. skb_headlen(skb), buffer);
  2316. }
  2317. /**
  2318. * skb_needs_linearize - check if we need to linearize a given skb
  2319. * depending on the given device features.
  2320. * @skb: socket buffer to check
  2321. * @features: net device features
  2322. *
  2323. * Returns true if either:
  2324. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2325. * 2. skb is fragmented and the device does not support SG.
  2326. */
  2327. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2328. netdev_features_t features)
  2329. {
  2330. return skb_is_nonlinear(skb) &&
  2331. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2332. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2333. }
  2334. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2335. void *to,
  2336. const unsigned int len)
  2337. {
  2338. memcpy(to, skb->data, len);
  2339. }
  2340. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2341. const int offset, void *to,
  2342. const unsigned int len)
  2343. {
  2344. memcpy(to, skb->data + offset, len);
  2345. }
  2346. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2347. const void *from,
  2348. const unsigned int len)
  2349. {
  2350. memcpy(skb->data, from, len);
  2351. }
  2352. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2353. const int offset,
  2354. const void *from,
  2355. const unsigned int len)
  2356. {
  2357. memcpy(skb->data + offset, from, len);
  2358. }
  2359. void skb_init(void);
  2360. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2361. {
  2362. return skb->tstamp;
  2363. }
  2364. /**
  2365. * skb_get_timestamp - get timestamp from a skb
  2366. * @skb: skb to get stamp from
  2367. * @stamp: pointer to struct timeval to store stamp in
  2368. *
  2369. * Timestamps are stored in the skb as offsets to a base timestamp.
  2370. * This function converts the offset back to a struct timeval and stores
  2371. * it in stamp.
  2372. */
  2373. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2374. struct timeval *stamp)
  2375. {
  2376. *stamp = ktime_to_timeval(skb->tstamp);
  2377. }
  2378. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2379. struct timespec *stamp)
  2380. {
  2381. *stamp = ktime_to_timespec(skb->tstamp);
  2382. }
  2383. static inline void __net_timestamp(struct sk_buff *skb)
  2384. {
  2385. skb->tstamp = ktime_get_real();
  2386. }
  2387. static inline ktime_t net_timedelta(ktime_t t)
  2388. {
  2389. return ktime_sub(ktime_get_real(), t);
  2390. }
  2391. static inline ktime_t net_invalid_timestamp(void)
  2392. {
  2393. return ktime_set(0, 0);
  2394. }
  2395. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2396. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2397. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2398. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2399. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2400. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2401. {
  2402. }
  2403. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2404. {
  2405. return false;
  2406. }
  2407. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2408. /**
  2409. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2410. *
  2411. * PHY drivers may accept clones of transmitted packets for
  2412. * timestamping via their phy_driver.txtstamp method. These drivers
  2413. * must call this function to return the skb back to the stack, with
  2414. * or without a timestamp.
  2415. *
  2416. * @skb: clone of the the original outgoing packet
  2417. * @hwtstamps: hardware time stamps, may be NULL if not available
  2418. *
  2419. */
  2420. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2421. struct skb_shared_hwtstamps *hwtstamps);
  2422. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2423. struct skb_shared_hwtstamps *hwtstamps,
  2424. struct sock *sk, int tstype);
  2425. /**
  2426. * skb_tstamp_tx - queue clone of skb with send time stamps
  2427. * @orig_skb: the original outgoing packet
  2428. * @hwtstamps: hardware time stamps, may be NULL if not available
  2429. *
  2430. * If the skb has a socket associated, then this function clones the
  2431. * skb (thus sharing the actual data and optional structures), stores
  2432. * the optional hardware time stamping information (if non NULL) or
  2433. * generates a software time stamp (otherwise), then queues the clone
  2434. * to the error queue of the socket. Errors are silently ignored.
  2435. */
  2436. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2437. struct skb_shared_hwtstamps *hwtstamps);
  2438. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2439. {
  2440. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2441. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2442. skb_tstamp_tx(skb, NULL);
  2443. }
  2444. /**
  2445. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2446. *
  2447. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2448. * function immediately before giving the sk_buff to the MAC hardware.
  2449. *
  2450. * Specifically, one should make absolutely sure that this function is
  2451. * called before TX completion of this packet can trigger. Otherwise
  2452. * the packet could potentially already be freed.
  2453. *
  2454. * @skb: A socket buffer.
  2455. */
  2456. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2457. {
  2458. skb_clone_tx_timestamp(skb);
  2459. sw_tx_timestamp(skb);
  2460. }
  2461. /**
  2462. * skb_complete_wifi_ack - deliver skb with wifi status
  2463. *
  2464. * @skb: the original outgoing packet
  2465. * @acked: ack status
  2466. *
  2467. */
  2468. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2469. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2470. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2471. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2472. {
  2473. return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
  2474. }
  2475. /**
  2476. * skb_checksum_complete - Calculate checksum of an entire packet
  2477. * @skb: packet to process
  2478. *
  2479. * This function calculates the checksum over the entire packet plus
  2480. * the value of skb->csum. The latter can be used to supply the
  2481. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2482. * checksum.
  2483. *
  2484. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2485. * this function can be used to verify that checksum on received
  2486. * packets. In that case the function should return zero if the
  2487. * checksum is correct. In particular, this function will return zero
  2488. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2489. * hardware has already verified the correctness of the checksum.
  2490. */
  2491. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2492. {
  2493. return skb_csum_unnecessary(skb) ?
  2494. 0 : __skb_checksum_complete(skb);
  2495. }
  2496. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2497. {
  2498. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2499. if (skb->csum_level == 0)
  2500. skb->ip_summed = CHECKSUM_NONE;
  2501. else
  2502. skb->csum_level--;
  2503. }
  2504. }
  2505. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2506. {
  2507. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2508. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2509. skb->csum_level++;
  2510. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2511. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2512. skb->csum_level = 0;
  2513. }
  2514. }
  2515. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2516. {
  2517. /* Mark current checksum as bad (typically called from GRO
  2518. * path). In the case that ip_summed is CHECKSUM_NONE
  2519. * this must be the first checksum encountered in the packet.
  2520. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2521. * checksum after the last one validated. For UDP, a zero
  2522. * checksum can not be marked as bad.
  2523. */
  2524. if (skb->ip_summed == CHECKSUM_NONE ||
  2525. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2526. skb->csum_bad = 1;
  2527. }
  2528. /* Check if we need to perform checksum complete validation.
  2529. *
  2530. * Returns true if checksum complete is needed, false otherwise
  2531. * (either checksum is unnecessary or zero checksum is allowed).
  2532. */
  2533. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2534. bool zero_okay,
  2535. __sum16 check)
  2536. {
  2537. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2538. skb->csum_valid = 1;
  2539. __skb_decr_checksum_unnecessary(skb);
  2540. return false;
  2541. }
  2542. return true;
  2543. }
  2544. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2545. * in checksum_init.
  2546. */
  2547. #define CHECKSUM_BREAK 76
  2548. /* Validate (init) checksum based on checksum complete.
  2549. *
  2550. * Return values:
  2551. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2552. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2553. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2554. * non-zero: value of invalid checksum
  2555. *
  2556. */
  2557. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2558. bool complete,
  2559. __wsum psum)
  2560. {
  2561. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2562. if (!csum_fold(csum_add(psum, skb->csum))) {
  2563. skb->csum_valid = 1;
  2564. return 0;
  2565. }
  2566. } else if (skb->csum_bad) {
  2567. /* ip_summed == CHECKSUM_NONE in this case */
  2568. return 1;
  2569. }
  2570. skb->csum = psum;
  2571. if (complete || skb->len <= CHECKSUM_BREAK) {
  2572. __sum16 csum;
  2573. csum = __skb_checksum_complete(skb);
  2574. skb->csum_valid = !csum;
  2575. return csum;
  2576. }
  2577. return 0;
  2578. }
  2579. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2580. {
  2581. return 0;
  2582. }
  2583. /* Perform checksum validate (init). Note that this is a macro since we only
  2584. * want to calculate the pseudo header which is an input function if necessary.
  2585. * First we try to validate without any computation (checksum unnecessary) and
  2586. * then calculate based on checksum complete calling the function to compute
  2587. * pseudo header.
  2588. *
  2589. * Return values:
  2590. * 0: checksum is validated or try to in skb_checksum_complete
  2591. * non-zero: value of invalid checksum
  2592. */
  2593. #define __skb_checksum_validate(skb, proto, complete, \
  2594. zero_okay, check, compute_pseudo) \
  2595. ({ \
  2596. __sum16 __ret = 0; \
  2597. skb->csum_valid = 0; \
  2598. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2599. __ret = __skb_checksum_validate_complete(skb, \
  2600. complete, compute_pseudo(skb, proto)); \
  2601. __ret; \
  2602. })
  2603. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2604. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2605. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2606. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2607. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2608. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2609. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2610. compute_pseudo) \
  2611. __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
  2612. #define skb_checksum_simple_validate(skb) \
  2613. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2614. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2615. {
  2616. return (skb->ip_summed == CHECKSUM_NONE &&
  2617. skb->csum_valid && !skb->csum_bad);
  2618. }
  2619. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2620. __sum16 check, __wsum pseudo)
  2621. {
  2622. skb->csum = ~pseudo;
  2623. skb->ip_summed = CHECKSUM_COMPLETE;
  2624. }
  2625. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  2626. do { \
  2627. if (__skb_checksum_convert_check(skb)) \
  2628. __skb_checksum_convert(skb, check, \
  2629. compute_pseudo(skb, proto)); \
  2630. } while (0)
  2631. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2632. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2633. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2634. {
  2635. if (nfct && atomic_dec_and_test(&nfct->use))
  2636. nf_conntrack_destroy(nfct);
  2637. }
  2638. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2639. {
  2640. if (nfct)
  2641. atomic_inc(&nfct->use);
  2642. }
  2643. #endif
  2644. #ifdef CONFIG_BRIDGE_NETFILTER
  2645. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  2646. {
  2647. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  2648. kfree(nf_bridge);
  2649. }
  2650. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  2651. {
  2652. if (nf_bridge)
  2653. atomic_inc(&nf_bridge->use);
  2654. }
  2655. #endif /* CONFIG_BRIDGE_NETFILTER */
  2656. static inline void nf_reset(struct sk_buff *skb)
  2657. {
  2658. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2659. nf_conntrack_put(skb->nfct);
  2660. skb->nfct = NULL;
  2661. #endif
  2662. #ifdef CONFIG_BRIDGE_NETFILTER
  2663. nf_bridge_put(skb->nf_bridge);
  2664. skb->nf_bridge = NULL;
  2665. #endif
  2666. }
  2667. static inline void nf_reset_trace(struct sk_buff *skb)
  2668. {
  2669. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2670. skb->nf_trace = 0;
  2671. #endif
  2672. }
  2673. /* Note: This doesn't put any conntrack and bridge info in dst. */
  2674. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2675. {
  2676. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2677. dst->nfct = src->nfct;
  2678. nf_conntrack_get(src->nfct);
  2679. dst->nfctinfo = src->nfctinfo;
  2680. #endif
  2681. #ifdef CONFIG_BRIDGE_NETFILTER
  2682. dst->nf_bridge = src->nf_bridge;
  2683. nf_bridge_get(src->nf_bridge);
  2684. #endif
  2685. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2686. dst->nf_trace = src->nf_trace;
  2687. #endif
  2688. }
  2689. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2690. {
  2691. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2692. nf_conntrack_put(dst->nfct);
  2693. #endif
  2694. #ifdef CONFIG_BRIDGE_NETFILTER
  2695. nf_bridge_put(dst->nf_bridge);
  2696. #endif
  2697. __nf_copy(dst, src);
  2698. }
  2699. #ifdef CONFIG_NETWORK_SECMARK
  2700. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2701. {
  2702. to->secmark = from->secmark;
  2703. }
  2704. static inline void skb_init_secmark(struct sk_buff *skb)
  2705. {
  2706. skb->secmark = 0;
  2707. }
  2708. #else
  2709. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2710. { }
  2711. static inline void skb_init_secmark(struct sk_buff *skb)
  2712. { }
  2713. #endif
  2714. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  2715. {
  2716. return !skb->destructor &&
  2717. #if IS_ENABLED(CONFIG_XFRM)
  2718. !skb->sp &&
  2719. #endif
  2720. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  2721. !skb->nfct &&
  2722. #endif
  2723. !skb->_skb_refdst &&
  2724. !skb_has_frag_list(skb);
  2725. }
  2726. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  2727. {
  2728. skb->queue_mapping = queue_mapping;
  2729. }
  2730. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  2731. {
  2732. return skb->queue_mapping;
  2733. }
  2734. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  2735. {
  2736. to->queue_mapping = from->queue_mapping;
  2737. }
  2738. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  2739. {
  2740. skb->queue_mapping = rx_queue + 1;
  2741. }
  2742. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  2743. {
  2744. return skb->queue_mapping - 1;
  2745. }
  2746. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  2747. {
  2748. return skb->queue_mapping != 0;
  2749. }
  2750. u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
  2751. unsigned int num_tx_queues);
  2752. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  2753. {
  2754. #ifdef CONFIG_XFRM
  2755. return skb->sp;
  2756. #else
  2757. return NULL;
  2758. #endif
  2759. }
  2760. /* Keeps track of mac header offset relative to skb->head.
  2761. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  2762. * For non-tunnel skb it points to skb_mac_header() and for
  2763. * tunnel skb it points to outer mac header.
  2764. * Keeps track of level of encapsulation of network headers.
  2765. */
  2766. struct skb_gso_cb {
  2767. int mac_offset;
  2768. int encap_level;
  2769. __u16 csum_start;
  2770. };
  2771. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
  2772. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  2773. {
  2774. return (skb_mac_header(inner_skb) - inner_skb->head) -
  2775. SKB_GSO_CB(inner_skb)->mac_offset;
  2776. }
  2777. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  2778. {
  2779. int new_headroom, headroom;
  2780. int ret;
  2781. headroom = skb_headroom(skb);
  2782. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  2783. if (ret)
  2784. return ret;
  2785. new_headroom = skb_headroom(skb);
  2786. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  2787. return 0;
  2788. }
  2789. /* Compute the checksum for a gso segment. First compute the checksum value
  2790. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  2791. * then add in skb->csum (checksum from csum_start to end of packet).
  2792. * skb->csum and csum_start are then updated to reflect the checksum of the
  2793. * resultant packet starting from the transport header-- the resultant checksum
  2794. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  2795. * header.
  2796. */
  2797. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  2798. {
  2799. int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
  2800. skb_transport_offset(skb);
  2801. __u16 csum;
  2802. csum = csum_fold(csum_partial(skb_transport_header(skb),
  2803. plen, skb->csum));
  2804. skb->csum = res;
  2805. SKB_GSO_CB(skb)->csum_start -= plen;
  2806. return csum;
  2807. }
  2808. static inline bool skb_is_gso(const struct sk_buff *skb)
  2809. {
  2810. return skb_shinfo(skb)->gso_size;
  2811. }
  2812. /* Note: Should be called only if skb_is_gso(skb) is true */
  2813. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  2814. {
  2815. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  2816. }
  2817. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  2818. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  2819. {
  2820. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  2821. * wanted then gso_type will be set. */
  2822. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  2823. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  2824. unlikely(shinfo->gso_type == 0)) {
  2825. __skb_warn_lro_forwarding(skb);
  2826. return true;
  2827. }
  2828. return false;
  2829. }
  2830. static inline void skb_forward_csum(struct sk_buff *skb)
  2831. {
  2832. /* Unfortunately we don't support this one. Any brave souls? */
  2833. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2834. skb->ip_summed = CHECKSUM_NONE;
  2835. }
  2836. /**
  2837. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  2838. * @skb: skb to check
  2839. *
  2840. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  2841. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  2842. * use this helper, to document places where we make this assertion.
  2843. */
  2844. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  2845. {
  2846. #ifdef DEBUG
  2847. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  2848. #endif
  2849. }
  2850. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  2851. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  2852. u32 skb_get_poff(const struct sk_buff *skb);
  2853. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  2854. const struct flow_keys *keys, int hlen);
  2855. /**
  2856. * skb_head_is_locked - Determine if the skb->head is locked down
  2857. * @skb: skb to check
  2858. *
  2859. * The head on skbs build around a head frag can be removed if they are
  2860. * not cloned. This function returns true if the skb head is locked down
  2861. * due to either being allocated via kmalloc, or by being a clone with
  2862. * multiple references to the head.
  2863. */
  2864. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  2865. {
  2866. return !skb->head_frag || skb_cloned(skb);
  2867. }
  2868. /**
  2869. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  2870. *
  2871. * @skb: GSO skb
  2872. *
  2873. * skb_gso_network_seglen is used to determine the real size of the
  2874. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  2875. *
  2876. * The MAC/L2 header is not accounted for.
  2877. */
  2878. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  2879. {
  2880. unsigned int hdr_len = skb_transport_header(skb) -
  2881. skb_network_header(skb);
  2882. return hdr_len + skb_gso_transport_seglen(skb);
  2883. }
  2884. #endif /* __KERNEL__ */
  2885. #endif /* _LINUX_SKBUFF_H */