1.\" Copyright (c) 2000 FreeBSD Inc. 2.\" All rights reserved. 3.\" 4.\" Redistribution and use in source and binary forms, with or without 5.\" modification, are permitted provided that the following conditions 6.\" are met: 7.\" 1. Redistributions of source code must retain the above copyright 8.\" notice, this list of conditions and the following disclaimer. 9.\" 2. Redistributions in binary form must reproduce the above copyright 10.\" notice, this list of conditions and the following disclaimer in the 11.\" documentation and/or other materials provided with the distribution. 12.\" 13.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16.\" ARE DISCLAIMED. IN NO EVENT SHALL [your name] OR CONTRIBUTORS BE LIABLE 17.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23.\" SUCH DAMAGE. 24.\" 25.\" $FreeBSD$ 26.\" 27.Dd October 21, 2014 28.Dt MBUF 9 29.Os 30.\" 31.Sh NAME 32.Nm mbuf 33.Nd "memory management in the kernel IPC subsystem" 34.\" 35.Sh SYNOPSIS 36.In sys/param.h 37.In sys/systm.h 38.In sys/mbuf.h 39.\" 40.Ss Mbuf allocation macros 41.Fn MGET "struct mbuf *mbuf" "int how" "short type" 42.Fn MGETHDR "struct mbuf *mbuf" "int how" "short type" 43.Fn MCLGET "struct mbuf *mbuf" "int how" 44.Fo MEXTADD 45.Fa "struct mbuf *mbuf" 46.Fa "caddr_t buf" 47.Fa "u_int size" 48.Fa "void (*free)(void *opt_arg1, void *opt_arg2)" 49.Fa "void *opt_arg1" 50.Fa "void *opt_arg2" 51.Fa "short flags" 52.Fa "int type" 53.Fc 54.Fn MEXTFREE "struct mbuf *mbuf" 55.Fn MFREE "struct mbuf *mbuf" "struct mbuf *successor" 56.\" 57.Ss Mbuf utility macros 58.Fn mtod "struct mbuf *mbuf" "type" 59.Fn M_ALIGN "struct mbuf *mbuf" "u_int len" 60.Fn MH_ALIGN "struct mbuf *mbuf" "u_int len" 61.Ft int 62.Fn M_LEADINGSPACE "struct mbuf *mbuf" 63.Ft int 64.Fn M_TRAILINGSPACE "struct mbuf *mbuf" 65.Fn M_MOVE_PKTHDR "struct mbuf *to" "struct mbuf *from" 66.Fn M_PREPEND "struct mbuf *mbuf" "int len" "int how" 67.Fn MCHTYPE "struct mbuf *mbuf" "short type" 68.Ft int 69.Fn M_WRITABLE "struct mbuf *mbuf" 70.\" 71.Ss Mbuf allocation functions 72.Ft struct mbuf * 73.Fn m_get "int how" "short type" 74.Ft struct mbuf * 75.Fn m_get2 "int size" "int how" "short type" "int flags" 76.Ft struct mbuf * 77.Fn m_getm "struct mbuf *orig" "int len" "int how" "short type" 78.Ft struct mbuf * 79.Fn m_getjcl "int how" "short type" "int flags" "int size" 80.Ft struct mbuf * 81.Fn m_getcl "int how" "short type" "int flags" 82.Ft struct mbuf * 83.Fn m_getclr "int how" "short type" 84.Ft struct mbuf * 85.Fn m_gethdr "int how" "short type" 86.Ft struct mbuf * 87.Fn m_free "struct mbuf *mbuf" 88.Ft void 89.Fn m_freem "struct mbuf *mbuf" 90.\" 91.Ss Mbuf utility functions 92.Ft void 93.Fn m_adj "struct mbuf *mbuf" "int len" 94.Ft void 95.Fn m_align "struct mbuf *mbuf" "int len" 96.Ft int 97.Fn m_append "struct mbuf *mbuf" "int len" "c_caddr_t cp" 98.Ft struct mbuf * 99.Fn m_prepend "struct mbuf *mbuf" "int len" "int how" 100.Ft struct mbuf * 101.Fn m_copyup "struct mbuf *mbuf" "int len" "int dstoff" 102.Ft struct mbuf * 103.Fn m_pullup "struct mbuf *mbuf" "int len" 104.Ft struct mbuf * 105.Fn m_pulldown "struct mbuf *mbuf" "int offset" "int len" "int *offsetp" 106.Ft struct mbuf * 107.Fn m_copym "struct mbuf *mbuf" "int offset" "int len" "int how" 108.Ft struct mbuf * 109.Fn m_copypacket "struct mbuf *mbuf" "int how" 110.Ft struct mbuf * 111.Fn m_dup "struct mbuf *mbuf" "int how" 112.Ft void 113.Fn m_copydata "const struct mbuf *mbuf" "int offset" "int len" "caddr_t buf" 114.Ft void 115.Fn m_copyback "struct mbuf *mbuf" "int offset" "int len" "caddr_t buf" 116.Ft struct mbuf * 117.Fo m_devget 118.Fa "char *buf" 119.Fa "int len" 120.Fa "int offset" 121.Fa "struct ifnet *ifp" 122.Fa "void (*copy)(char *from, caddr_t to, u_int len)" 123.Fc 124.Ft void 125.Fn m_cat "struct mbuf *m" "struct mbuf *n" 126.Ft u_int 127.Fn m_fixhdr "struct mbuf *mbuf" 128.Ft void 129.Fn m_dup_pkthdr "struct mbuf *to" "struct mbuf *from" 130.Ft void 131.Fn m_move_pkthdr "struct mbuf *to" "struct mbuf *from" 132.Ft u_int 133.Fn m_length "struct mbuf *mbuf" "struct mbuf **last" 134.Ft struct mbuf * 135.Fn m_split "struct mbuf *mbuf" "int len" "int how" 136.Ft int 137.Fn m_apply "struct mbuf *mbuf" "int off" "int len" "int (*f)(void *arg, void *data, u_int len)" "void *arg" 138.Ft struct mbuf * 139.Fn m_getptr "struct mbuf *mbuf" "int loc" "int *off" 140.Ft struct mbuf * 141.Fn m_defrag "struct mbuf *m0" "int how" 142.Ft struct mbuf * 143.Fn m_unshare "struct mbuf *m0" "int how" 144.\" 145.Sh DESCRIPTION 146An 147.Vt mbuf 148is a basic unit of memory management in the kernel IPC subsystem. 149Network packets and socket buffers are stored in 150.Vt mbufs . 151A network packet may span multiple 152.Vt mbufs 153arranged into a 154.Vt mbuf chain 155(linked list), 156which allows adding or trimming 157network headers with little overhead. 158.Pp 159While a developer should not bother with 160.Vt mbuf 161internals without serious 162reason in order to avoid incompatibilities with future changes, it 163is useful to understand the general structure of an 164.Vt mbuf . 165.Pp 166An 167.Vt mbuf 168consists of a variable-sized header and a small internal 169buffer for data. 170The total size of an 171.Vt mbuf , 172.Dv MSIZE , 173is a constant defined in 174.In sys/param.h . 175The 176.Vt mbuf 177header includes: 178.Bl -tag -width "m_nextpkt" -offset indent 179.It Va m_next 180.Pq Vt struct mbuf * 181A pointer to the next 182.Vt mbuf 183in the 184.Vt mbuf chain . 185.It Va m_nextpkt 186.Pq Vt struct mbuf * 187A pointer to the next 188.Vt mbuf chain 189in the queue. 190.It Va m_data 191.Pq Vt caddr_t 192A pointer to data attached to this 193.Vt mbuf . 194.It Va m_len 195.Pq Vt int 196The length of the data. 197.It Va m_type 198.Pq Vt short 199The type of the data. 200.It Va m_flags 201.Pq Vt int 202The 203.Vt mbuf 204flags. 205.El 206.Pp 207The 208.Vt mbuf 209flag bits are defined as follows: 210.Bd -literal 211/* mbuf flags */ 212#define M_EXT 0x00000001 /* has associated external storage */ 213#define M_PKTHDR 0x00000002 /* start of record */ 214#define M_EOR 0x00000004 /* end of record */ 215#define M_RDONLY 0x00000008 /* associated data marked read-only */ 216#define M_PROTO1 0x00001000 /* protocol-specific */ 217#define M_PROTO2 0x00002000 /* protocol-specific */ 218#define M_PROTO3 0x00004000 /* protocol-specific */ 219#define M_PROTO4 0x00008000 /* protocol-specific */ 220#define M_PROTO5 0x00010000 /* protocol-specific */ 221#define M_PROTO6 0x00020000 /* protocol-specific */ 222#define M_PROTO7 0x00040000 /* protocol-specific */ 223#define M_PROTO8 0x00080000 /* protocol-specific */ 224#define M_PROTO9 0x00100000 /* protocol-specific */ 225#define M_PROTO10 0x00200000 /* protocol-specific */ 226#define M_PROTO11 0x00400000 /* protocol-specific */ 227#define M_PROTO12 0x00800000 /* protocol-specific */ 228 229/* mbuf pkthdr flags (also stored in m_flags) */ 230#define M_BCAST 0x00000010 /* send/received as link-level broadcast */ 231#define M_MCAST 0x00000020 /* send/received as link-level multicast */ 232.Ed 233.Pp 234The available 235.Vt mbuf 236types are defined as follows: 237.Bd -literal 238/* mbuf types */ 239#define MT_DATA 1 /* dynamic (data) allocation */ 240#define MT_HEADER MT_DATA /* packet header */ 241#define MT_SONAME 8 /* socket name */ 242#define MT_CONTROL 14 /* extra-data protocol message */ 243#define MT_OOBDATA 15 /* expedited data */ 244.Ed 245.Pp 246The available external buffer types are defined as follows: 247.Bd -literal 248/* external buffer types */ 249#define EXT_CLUSTER 1 /* mbuf cluster */ 250#define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */ 251#define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */ 252#define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 253#define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 254#define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 255#define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */ 256#define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */ 257#define EXT_MOD_TYPE 253 /* custom module's ext_buf type */ 258#define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */ 259#define EXT_EXTREF 255 /* has externally maintained ref_cnt ptr */ 260.Ed 261.Pp 262If the 263.Dv M_PKTHDR 264flag is set, a 265.Vt struct pkthdr Va m_pkthdr 266is added to the 267.Vt mbuf 268header. 269It contains a pointer to the interface 270the packet has been received from 271.Pq Vt struct ifnet Va *rcvif , 272and the total packet length 273.Pq Vt int Va len . 274Optionally, it may also contain an attached list of packet tags 275.Pq Vt "struct m_tag" . 276See 277.Xr mbuf_tags 9 278for details. 279Fields used in offloading checksum calculation to the hardware are kept in 280.Va m_pkthdr 281as well. 282See 283.Sx HARDWARE-ASSISTED CHECKSUM CALCULATION 284for details. 285.Pp 286If small enough, data is stored in the internal data buffer of an 287.Vt mbuf . 288If the data is sufficiently large, another 289.Vt mbuf 290may be added to the 291.Vt mbuf chain , 292or external storage may be associated with the 293.Vt mbuf . 294.Dv MHLEN 295bytes of data can fit into an 296.Vt mbuf 297with the 298.Dv M_PKTHDR 299flag set, 300.Dv MLEN 301bytes can otherwise. 302.Pp 303If external storage is being associated with an 304.Vt mbuf , 305the 306.Va m_ext 307header is added at the cost of losing the internal data buffer. 308It includes a pointer to external storage, the size of the storage, 309a pointer to a function used for freeing the storage, 310a pointer to an optional argument that can be passed to the function, 311and a pointer to a reference counter. 312An 313.Vt mbuf 314using external storage has the 315.Dv M_EXT 316flag set. 317.Pp 318The system supplies a macro for allocating the desired external storage 319buffer, 320.Dv MEXTADD . 321.Pp 322The allocation and management of the reference counter is handled by the 323subsystem. 324.Pp 325The system also supplies a default type of external storage buffer called an 326.Vt mbuf cluster . 327.Vt Mbuf clusters 328can be allocated and configured with the use of the 329.Dv MCLGET 330macro. 331Each 332.Vt mbuf cluster 333is 334.Dv MCLBYTES 335in size, where MCLBYTES is a machine-dependent constant. 336The system defines an advisory macro 337.Dv MINCLSIZE , 338which is the smallest amount of data to put into an 339.Vt mbuf cluster . 340It is equal to 341.Dv MHLEN 342plus one. 343It is typically preferable to store data into the data region of an 344.Vt mbuf , 345if size permits, as opposed to allocating a separate 346.Vt mbuf cluster 347to hold the same data. 348.\" 349.Ss Macros and Functions 350There are numerous predefined macros and functions that provide the 351developer with common utilities. 352.\" 353.Bl -ohang -offset indent 354.It Fn mtod mbuf type 355Convert an 356.Fa mbuf 357pointer to a data pointer. 358The macro expands to the data pointer cast to the specified 359.Fa type . 360.Sy Note : 361It is advisable to ensure that there is enough contiguous data in 362.Fa mbuf . 363See 364.Fn m_pullup 365for details. 366.It Fn MGET mbuf how type 367Allocate an 368.Vt mbuf 369and initialize it to contain internal data. 370.Fa mbuf 371will point to the allocated 372.Vt mbuf 373on success, or be set to 374.Dv NULL 375on failure. 376The 377.Fa how 378argument is to be set to 379.Dv M_WAITOK 380or 381.Dv M_NOWAIT . 382It specifies whether the caller is willing to block if necessary. 383A number of other functions and macros related to 384.Vt mbufs 385have the same argument because they may 386at some point need to allocate new 387.Vt mbufs . 388.It Fn MGETHDR mbuf how type 389Allocate an 390.Vt mbuf 391and initialize it to contain a packet header 392and internal data. 393See 394.Fn MGET 395for details. 396.It Fn MEXTADD mbuf buf size free opt_arg1 opt_arg2 flags type 397Associate externally managed data with 398.Fa mbuf . 399Any internal data contained in the mbuf will be discarded, and the 400.Dv M_EXT 401flag will be set. 402The 403.Fa buf 404and 405.Fa size 406arguments are the address and length, respectively, of the data. 407The 408.Fa free 409argument points to a function which will be called to free the data 410when the mbuf is freed; it is only used if 411.Fa type 412is 413.Dv EXT_EXTREF . 414The 415.Fa opt_arg1 416and 417.Fa opt_arg2 418arguments will be passed unmodified to 419.Fa free . 420The 421.Fa flags 422argument specifies additional 423.Vt mbuf 424flags; it is not necessary to specify 425.Dv M_EXT . 426Finally, the 427.Fa type 428argument specifies the type of external data, which controls how it 429will be disposed of when the 430.Vt mbuf 431is freed. 432In most cases, the correct value is 433.Dv EXT_EXTREF . 434.It Fn MCLGET mbuf how 435Allocate and attach an 436.Vt mbuf cluster 437to 438.Fa mbuf . 439If the macro fails, the 440.Dv M_EXT 441flag will not be set in 442.Fa mbuf . 443.It Fn M_ALIGN mbuf len 444Set the pointer 445.Fa mbuf->m_data 446to place an object of the size 447.Fa len 448at the end of the internal data area of 449.Fa mbuf , 450long word aligned. 451Applicable only if 452.Fa mbuf 453is newly allocated with 454.Fn MGET 455or 456.Fn m_get . 457.It Fn MH_ALIGN mbuf len 458Serves the same purpose as 459.Fn M_ALIGN 460does, but only for 461.Fa mbuf 462newly allocated with 463.Fn MGETHDR 464or 465.Fn m_gethdr , 466or initialized by 467.Fn m_dup_pkthdr 468or 469.Fn m_move_pkthdr . 470.It Fn m_align mbuf len 471Services the same purpose as 472.Fn M_ALIGN 473but handles any type of mbuf. 474.It Fn M_LEADINGSPACE mbuf 475Returns the number of bytes available before the beginning 476of data in 477.Fa mbuf . 478.It Fn M_TRAILINGSPACE mbuf 479Returns the number of bytes available after the end of data in 480.Fa mbuf . 481.It Fn M_PREPEND mbuf len how 482This macro operates on an 483.Vt mbuf chain . 484It is an optimized wrapper for 485.Fn m_prepend 486that can make use of possible empty space before data 487(e.g.\& left after trimming of a link-layer header). 488The new 489.Vt mbuf chain 490pointer or 491.Dv NULL 492is in 493.Fa mbuf 494after the call. 495.It Fn M_MOVE_PKTHDR to from 496Using this macro is equivalent to calling 497.Fn m_move_pkthdr to from . 498.It Fn M_WRITABLE mbuf 499This macro will evaluate true if 500.Fa mbuf 501is not marked 502.Dv M_RDONLY 503and if either 504.Fa mbuf 505does not contain external storage or, 506if it does, 507then if the reference count of the storage is not greater than 1. 508The 509.Dv M_RDONLY 510flag can be set in 511.Fa mbuf->m_flags . 512This can be achieved during setup of the external storage, 513by passing the 514.Dv M_RDONLY 515bit as a 516.Fa flags 517argument to the 518.Fn MEXTADD 519macro, or can be directly set in individual 520.Vt mbufs . 521.It Fn MCHTYPE mbuf type 522Change the type of 523.Fa mbuf 524to 525.Fa type . 526This is a relatively expensive operation and should be avoided. 527.El 528.Pp 529The functions are: 530.Bl -ohang -offset indent 531.It Fn m_get how type 532A function version of 533.Fn MGET 534for non-critical paths. 535.It Fn m_get2 size how type flags 536Allocate an 537.Vt mbuf 538with enough space to hold specified amount of data. 539.It Fn m_getm orig len how type 540Allocate 541.Fa len 542bytes worth of 543.Vt mbufs 544and 545.Vt mbuf clusters 546if necessary and append the resulting allocated 547.Vt mbuf chain 548to the 549.Vt mbuf chain 550.Fa orig , 551if it is 552.No non- Ns Dv NULL . 553If the allocation fails at any point, 554free whatever was allocated and return 555.Dv NULL . 556If 557.Fa orig 558is 559.No non- Ns Dv NULL , 560it will not be freed. 561It is possible to use 562.Fn m_getm 563to either append 564.Fa len 565bytes to an existing 566.Vt mbuf 567or 568.Vt mbuf chain 569(for example, one which may be sitting in a pre-allocated ring) 570or to simply perform an all-or-nothing 571.Vt mbuf 572and 573.Vt mbuf cluster 574allocation. 575.It Fn m_gethdr how type 576A function version of 577.Fn MGETHDR 578for non-critical paths. 579.It Fn m_getcl how type flags 580Fetch an 581.Vt mbuf 582with a 583.Vt mbuf cluster 584attached to it. 585If one of the allocations fails, the entire allocation fails. 586This routine is the preferred way of fetching both the 587.Vt mbuf 588and 589.Vt mbuf cluster 590together, as it avoids having to unlock/relock between allocations. 591Returns 592.Dv NULL 593on failure. 594.It Fn m_getjcl how type flags size 595This is like 596.Fn m_getcl 597but it the size of the cluster allocated will be large enough for 598.Fa size 599bytes. 600.It Fn m_getclr how type 601Allocate an 602.Vt mbuf 603and zero out the data region. 604.It Fn m_free mbuf 605Frees 606.Vt mbuf . 607Returns 608.Va m_next 609of the freed 610.Vt mbuf . 611.El 612.Pp 613The functions below operate on 614.Vt mbuf chains . 615.Bl -ohang -offset indent 616.It Fn m_freem mbuf 617Free an entire 618.Vt mbuf chain , 619including any external storage. 620.\" 621.It Fn m_adj mbuf len 622Trim 623.Fa len 624bytes from the head of an 625.Vt mbuf chain 626if 627.Fa len 628is positive, from the tail otherwise. 629.\" 630.It Fn m_append mbuf len cp 631Append 632.Vt len 633bytes of data 634.Vt cp 635to the 636.Vt mbuf chain . 637Extend the mbuf chain if the new data does not fit in 638existing space. 639.\" 640.It Fn m_prepend mbuf len how 641Allocate a new 642.Vt mbuf 643and prepend it to the 644.Vt mbuf chain , 645handle 646.Dv M_PKTHDR 647properly. 648.Sy Note : 649It does not allocate any 650.Vt mbuf clusters , 651so 652.Fa len 653must be less than 654.Dv MLEN 655or 656.Dv MHLEN , 657depending on the 658.Dv M_PKTHDR 659flag setting. 660.\" 661.It Fn m_copyup mbuf len dstoff 662Similar to 663.Fn m_pullup 664but copies 665.Fa len 666bytes of data into a new mbuf at 667.Fa dstoff 668bytes into the mbuf. 669The 670.Fa dstoff 671argument aligns the data and leaves room for a link layer header. 672Returns the new 673.Vt mbuf chain 674on success, 675and frees the 676.Vt mbuf chain 677and returns 678.Dv NULL 679on failure. 680.Sy Note : 681The function does not allocate 682.Vt mbuf clusters , 683so 684.Fa len + dstoff 685must be less than 686.Dv MHLEN . 687.\" 688.It Fn m_pullup mbuf len 689Arrange that the first 690.Fa len 691bytes of an 692.Vt mbuf chain 693are contiguous and lay in the data area of 694.Fa mbuf , 695so they are accessible with 696.Fn mtod mbuf type . 697It is important to remember that this may involve 698reallocating some mbufs and moving data so all pointers 699referencing data within the old mbuf chain 700must be recalculated or made invalid. 701Return the new 702.Vt mbuf chain 703on success, 704.Dv NULL 705on failure 706(the 707.Vt mbuf chain 708is freed in this case). 709.Sy Note : 710It does not allocate any 711.Vt mbuf clusters , 712so 713.Fa len 714must be less than or equal to 715.Dv MHLEN . 716.\" 717.It Fn m_pulldown mbuf offset len offsetp 718Arrange that 719.Fa len 720bytes between 721.Fa offset 722and 723.Fa offset + len 724in the 725.Vt mbuf chain 726are contiguous and lay in the data area of 727.Fa mbuf , 728so they are accessible with 729.Fn mtod mbuf type . 730.Fa len 731must be smaller than, or equal to, the size of an 732.Vt mbuf cluster . 733Return a pointer to an intermediate 734.Vt mbuf 735in the chain containing the requested region; 736the offset in the data region of the 737.Vt mbuf chain 738to the data contained in the returned mbuf is stored in 739.Fa *offsetp . 740If 741.Fa offsetp 742is NULL, the region may be accessed using 743.Fn mtod mbuf type . 744If 745.Fa offsetp 746is non-NULL, the region may be accessed using 747.Fn mtod mbuf uint8_t 748+ *offsetp. 749The region of the mbuf chain between its beginning and 750.Fa offset 751is not modified, therefore it is safe to hold pointers to data within 752this region before calling 753.Fn m_pulldown . 754.\" 755.It Fn m_copym mbuf offset len how 756Make a copy of an 757.Vt mbuf chain 758starting 759.Fa offset 760bytes from the beginning, continuing for 761.Fa len 762bytes. 763If 764.Fa len 765is 766.Dv M_COPYALL , 767copy to the end of the 768.Vt mbuf chain . 769.Sy Note : 770The copy is read-only, because the 771.Vt mbuf clusters 772are not copied, only their reference counts are incremented. 773.\" 774.It Fn m_copypacket mbuf how 775Copy an entire packet including header, which must be present. 776This is an optimized version of the common case 777.Fn m_copym mbuf 0 M_COPYALL how . 778.Sy Note : 779the copy is read-only, because the 780.Vt mbuf clusters 781are not copied, only their reference counts are incremented. 782.\" 783.It Fn m_dup mbuf how 784Copy a packet header 785.Vt mbuf chain 786into a completely new 787.Vt mbuf chain , 788including copying any 789.Vt mbuf clusters . 790Use this instead of 791.Fn m_copypacket 792when you need a writable copy of an 793.Vt mbuf chain . 794.\" 795.It Fn m_copydata mbuf offset len buf 796Copy data from an 797.Vt mbuf chain 798starting 799.Fa off 800bytes from the beginning, continuing for 801.Fa len 802bytes, into the indicated buffer 803.Fa buf . 804.\" 805.It Fn m_copyback mbuf offset len buf 806Copy 807.Fa len 808bytes from the buffer 809.Fa buf 810back into the indicated 811.Vt mbuf chain , 812starting at 813.Fa offset 814bytes from the beginning of the 815.Vt mbuf chain , 816extending the 817.Vt mbuf chain 818if necessary. 819.Sy Note : 820It does not allocate any 821.Vt mbuf clusters , 822just adds 823.Vt mbufs 824to the 825.Vt mbuf chain . 826It is safe to set 827.Fa offset 828beyond the current 829.Vt mbuf chain 830end: zeroed 831.Vt mbufs 832will be allocated to fill the space. 833.\" 834.It Fn m_length mbuf last 835Return the length of the 836.Vt mbuf chain , 837and optionally a pointer to the last 838.Vt mbuf . 839.\" 840.It Fn m_dup_pkthdr to from how 841Upon the function's completion, the 842.Vt mbuf 843.Fa to 844will contain an identical copy of 845.Fa from->m_pkthdr 846and the per-packet attributes found in the 847.Vt mbuf chain 848.Fa from . 849The 850.Vt mbuf 851.Fa from 852must have the flag 853.Dv M_PKTHDR 854initially set, and 855.Fa to 856must be empty on entry. 857.\" 858.It Fn m_move_pkthdr to from 859Move 860.Va m_pkthdr 861and the per-packet attributes from the 862.Vt mbuf chain 863.Fa from 864to the 865.Vt mbuf 866.Fa to . 867The 868.Vt mbuf 869.Fa from 870must have the flag 871.Dv M_PKTHDR 872initially set, and 873.Fa to 874must be empty on entry. 875Upon the function's completion, 876.Fa from 877will have the flag 878.Dv M_PKTHDR 879and the per-packet attributes cleared. 880.\" 881.It Fn m_fixhdr mbuf 882Set the packet-header length to the length of the 883.Vt mbuf chain . 884.\" 885.It Fn m_devget buf len offset ifp copy 886Copy data from a device local memory pointed to by 887.Fa buf 888to an 889.Vt mbuf chain . 890The copy is done using a specified copy routine 891.Fa copy , 892or 893.Fn bcopy 894if 895.Fa copy 896is 897.Dv NULL . 898.\" 899.It Fn m_cat m n 900Concatenate 901.Fa n 902to 903.Fa m . 904Both 905.Vt mbuf chains 906must be of the same type. 907.Fa N 908is still valid after the function returned. 909.Sy Note : 910It does not handle 911.Dv M_PKTHDR 912and friends. 913.\" 914.It Fn m_split mbuf len how 915Partition an 916.Vt mbuf chain 917in two pieces, returning the tail: 918all but the first 919.Fa len 920bytes. 921In case of failure, it returns 922.Dv NULL 923and attempts to restore the 924.Vt mbuf chain 925to its original state. 926.\" 927.It Fn m_apply mbuf off len f arg 928Apply a function to an 929.Vt mbuf chain , 930at offset 931.Fa off , 932for length 933.Fa len 934bytes. 935Typically used to avoid calls to 936.Fn m_pullup 937which would otherwise be unnecessary or undesirable. 938.Fa arg 939is a convenience argument which is passed to the callback function 940.Fa f . 941.Pp 942Each time 943.Fn f 944is called, it will be passed 945.Fa arg , 946a pointer to the 947.Fa data 948in the current mbuf, and the length 949.Fa len 950of the data in this mbuf to which the function should be applied. 951.Pp 952The function should return zero to indicate success; 953otherwise, if an error is indicated, then 954.Fn m_apply 955will return the error and stop iterating through the 956.Vt mbuf chain . 957.\" 958.It Fn m_getptr mbuf loc off 959Return a pointer to the mbuf containing the data located at 960.Fa loc 961bytes from the beginning of the 962.Vt mbuf chain . 963The corresponding offset into the mbuf will be stored in 964.Fa *off . 965.It Fn m_defrag m0 how 966Defragment an mbuf chain, returning the shortest possible 967chain of mbufs and clusters. 968If allocation fails and this can not be completed, 969.Dv NULL 970will be returned and the original chain will be unchanged. 971Upon success, the original chain will be freed and the new 972chain will be returned. 973.Fa how 974should be either 975.Dv M_WAITOK 976or 977.Dv M_NOWAIT , 978depending on the caller's preference. 979.Pp 980This function is especially useful in network drivers, where 981certain long mbuf chains must be shortened before being added 982to TX descriptor lists. 983.It Fn m_unshare m0 how 984Create a version of the specified mbuf chain whose 985contents can be safely modified without affecting other users. 986If allocation fails and this operation can not be completed, 987.Dv NULL 988will be returned. 989The original mbuf chain is always reclaimed and the reference 990count of any shared mbuf clusters is decremented. 991.Fa how 992should be either 993.Dv M_WAITOK 994or 995.Dv M_NOWAIT , 996depending on the caller's preference. 997As a side-effect of this process the returned 998mbuf chain may be compacted. 999.Pp 1000This function is especially useful in the transmit path of 1001network code, when data must be encrypted or otherwise 1002altered prior to transmission. 1003.El 1004.Sh HARDWARE-ASSISTED CHECKSUM CALCULATION 1005This section currently applies to TCP/IP only. 1006In order to save the host CPU resources, computing checksums is 1007offloaded to the network interface hardware if possible. 1008The 1009.Va m_pkthdr 1010member of the leading 1011.Vt mbuf 1012of a packet contains two fields used for that purpose, 1013.Vt int Va csum_flags 1014and 1015.Vt int Va csum_data . 1016The meaning of those fields depends on the direction a packet flows in, 1017and on whether the packet is fragmented. 1018Henceforth, 1019.Va csum_flags 1020or 1021.Va csum_data 1022of a packet 1023will denote the corresponding field of the 1024.Va m_pkthdr 1025member of the leading 1026.Vt mbuf 1027in the 1028.Vt mbuf chain 1029containing the packet. 1030.Pp 1031On output, checksum offloading is attempted after the outgoing 1032interface has been determined for a packet. 1033The interface-specific field 1034.Va ifnet.if_data.ifi_hwassist 1035(see 1036.Xr ifnet 9 ) 1037is consulted for the capabilities of the interface to assist in 1038computing checksums. 1039The 1040.Va csum_flags 1041field of the packet header is set to indicate which actions the interface 1042is supposed to perform on it. 1043The actions unsupported by the network interface are done in the 1044software prior to passing the packet down to the interface driver; 1045such actions will never be requested through 1046.Va csum_flags . 1047.Pp 1048The flags demanding a particular action from an interface are as follows: 1049.Bl -tag -width ".Dv CSUM_TCP" -offset indent 1050.It Dv CSUM_IP 1051The IP header checksum is to be computed and stored in the 1052corresponding field of the packet. 1053The hardware is expected to know the format of an IP header 1054to determine the offset of the IP checksum field. 1055.It Dv CSUM_TCP 1056The TCP checksum is to be computed. 1057(See below.) 1058.It Dv CSUM_UDP 1059The UDP checksum is to be computed. 1060(See below.) 1061.El 1062.Pp 1063Should a TCP or UDP checksum be offloaded to the hardware, 1064the field 1065.Va csum_data 1066will contain the byte offset of the checksum field relative to the 1067end of the IP header. 1068In this case, the checksum field will be initially 1069set by the TCP/IP module to the checksum of the pseudo header 1070defined by the TCP and UDP specifications. 1071.Pp 1072On input, an interface indicates the actions it has performed 1073on a packet by setting one or more of the following flags in 1074.Va csum_flags 1075associated with the packet: 1076.Bl -tag -width ".Dv CSUM_IP_CHECKED" -offset indent 1077.It Dv CSUM_IP_CHECKED 1078The IP header checksum has been computed. 1079.It Dv CSUM_IP_VALID 1080The IP header has a valid checksum. 1081This flag can appear only in combination with 1082.Dv CSUM_IP_CHECKED . 1083.It Dv CSUM_DATA_VALID 1084The checksum of the data portion of the IP packet has been computed 1085and stored in the field 1086.Va csum_data 1087in network byte order. 1088.It Dv CSUM_PSEUDO_HDR 1089Can be set only along with 1090.Dv CSUM_DATA_VALID 1091to indicate that the IP data checksum found in 1092.Va csum_data 1093allows for the pseudo header defined by the TCP and UDP specifications. 1094Otherwise the checksum of the pseudo header must be calculated by 1095the host CPU and added to 1096.Va csum_data 1097to obtain the final checksum to be used for TCP or UDP validation purposes. 1098.El 1099.Pp 1100If a particular network interface just indicates success or 1101failure of TCP or UDP checksum validation without returning 1102the exact value of the checksum to the host CPU, its driver can mark 1103.Dv CSUM_DATA_VALID 1104and 1105.Dv CSUM_PSEUDO_HDR 1106in 1107.Va csum_flags , 1108and set 1109.Va csum_data 1110to 1111.Li 0xFFFF 1112hexadecimal to indicate a valid checksum. 1113It is a peculiarity of the algorithm used that the Internet checksum 1114calculated over any valid packet will be 1115.Li 0xFFFF 1116as long as the original checksum field is included. 1117.Sh STRESS TESTING 1118When running a kernel compiled with the option 1119.Dv MBUF_STRESS_TEST , 1120the following 1121.Xr sysctl 8 Ns 1122-controlled options may be used to create 1123various failure/extreme cases for testing of network drivers 1124and other parts of the kernel that rely on 1125.Vt mbufs . 1126.Bl -tag -width ident 1127.It Va net.inet.ip.mbuf_frag_size 1128Causes 1129.Fn ip_output 1130to fragment outgoing 1131.Vt mbuf chains 1132into fragments of the specified size. 1133Setting this variable to 1 is an excellent way to 1134test the long 1135.Vt mbuf chain 1136handling ability of network drivers. 1137.It Va kern.ipc.m_defragrandomfailures 1138Causes the function 1139.Fn m_defrag 1140to randomly fail, returning 1141.Dv NULL . 1142Any piece of code which uses 1143.Fn m_defrag 1144should be tested with this feature. 1145.El 1146.Sh RETURN VALUES 1147See above. 1148.Sh SEE ALSO 1149.Xr ifnet 9 , 1150.Xr mbuf_tags 9 1151.Sh HISTORY 1152.\" Please correct me if I'm wrong 1153.Vt Mbufs 1154appeared in an early version of 1155.Bx . 1156Besides being used for network packets, they were used 1157to store various dynamic structures, such as routing table 1158entries, interface addresses, protocol control blocks, etc. 1159In more recent 1160.Fx 1161use of 1162.Vt mbufs 1163is almost entirely limited to packet storage, with 1164.Xr uma 9 1165zones being used directly to store other network-related memory. 1166.Pp 1167Historically, the 1168.Vt mbuf 1169allocator has been a special-purpose memory allocator able to run in 1170interrupt contexts and allocating from a special kernel address space map. 1171As of 1172.Fx 5.3 , 1173the 1174.Vt mbuf 1175allocator is a wrapper around 1176.Xr uma 9 , 1177allowing caching of 1178.Vt mbufs , 1179clusters, and 1180.Vt mbuf 1181+ cluster pairs in per-CPU caches, as well as bringing other benefits of 1182slab allocation. 1183.Sh AUTHORS 1184The original 1185.Nm 1186manual page was written by Yar Tikhiy. 1187The 1188.Xr uma 9 1189.Vt mbuf 1190allocator was written by Bosko Milekic. 1191