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 January 5, 2015 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.Ft int 44.Fn MCLGET "struct mbuf *mbuf" "int how" 45.Fo MEXTADD 46.Fa "struct mbuf *mbuf" 47.Fa "caddr_t buf" 48.Fa "u_int size" 49.Fa "void (*free)(void *opt_arg1, void *opt_arg2)" 50.Fa "void *opt_arg1" 51.Fa "void *opt_arg2" 52.Fa "short flags" 53.Fa "int type" 54.Fc 55.Fn MEXTFREE "struct mbuf *mbuf" 56.Fn MFREE "struct mbuf *mbuf" "struct mbuf *successor" 57.\" 58.Ss Mbuf utility macros 59.Fn mtod "struct mbuf *mbuf" "type" 60.Fn M_ALIGN "struct mbuf *mbuf" "u_int len" 61.Fn MH_ALIGN "struct mbuf *mbuf" "u_int len" 62.Ft int 63.Fn M_LEADINGSPACE "struct mbuf *mbuf" 64.Ft int 65.Fn M_TRAILINGSPACE "struct mbuf *mbuf" 66.Fn M_MOVE_PKTHDR "struct mbuf *to" "struct mbuf *from" 67.Fn M_PREPEND "struct mbuf *mbuf" "int len" "int how" 68.Fn MCHTYPE "struct mbuf *mbuf" "short type" 69.Ft int 70.Fn M_WRITABLE "struct mbuf *mbuf" 71.\" 72.Ss Mbuf allocation functions 73.Ft struct mbuf * 74.Fn m_get "int how" "short type" 75.Ft struct mbuf * 76.Fn m_get2 "int size" "int how" "short type" "int flags" 77.Ft struct mbuf * 78.Fn m_getm "struct mbuf *orig" "int len" "int how" "short type" 79.Ft struct mbuf * 80.Fn m_getjcl "int how" "short type" "int flags" "int size" 81.Ft struct mbuf * 82.Fn m_getcl "int how" "short type" "int flags" 83.Ft struct mbuf * 84.Fn m_getclr "int how" "short type" 85.Ft struct mbuf * 86.Fn m_gethdr "int how" "short type" 87.Ft struct mbuf * 88.Fn m_free "struct mbuf *mbuf" 89.Ft void 90.Fn m_freem "struct mbuf *mbuf" 91.\" 92.Ss Mbuf utility functions 93.Ft void 94.Fn m_adj "struct mbuf *mbuf" "int len" 95.Ft void 96.Fn m_align "struct mbuf *mbuf" "int len" 97.Ft int 98.Fn m_append "struct mbuf *mbuf" "int len" "c_caddr_t cp" 99.Ft struct mbuf * 100.Fn m_prepend "struct mbuf *mbuf" "int len" "int how" 101.Ft struct mbuf * 102.Fn m_copyup "struct mbuf *mbuf" "int len" "int dstoff" 103.Ft struct mbuf * 104.Fn m_pullup "struct mbuf *mbuf" "int len" 105.Ft struct mbuf * 106.Fn m_pulldown "struct mbuf *mbuf" "int offset" "int len" "int *offsetp" 107.Ft struct mbuf * 108.Fn m_copym "struct mbuf *mbuf" "int offset" "int len" "int how" 109.Ft struct mbuf * 110.Fn m_copypacket "struct mbuf *mbuf" "int how" 111.Ft struct mbuf * 112.Fn m_dup "struct mbuf *mbuf" "int how" 113.Ft void 114.Fn m_copydata "const struct mbuf *mbuf" "int offset" "int len" "caddr_t buf" 115.Ft void 116.Fn m_copyback "struct mbuf *mbuf" "int offset" "int len" "caddr_t buf" 117.Ft struct mbuf * 118.Fo m_devget 119.Fa "char *buf" 120.Fa "int len" 121.Fa "int offset" 122.Fa "struct ifnet *ifp" 123.Fa "void (*copy)(char *from, caddr_t to, u_int len)" 124.Fc 125.Ft void 126.Fn m_cat "struct mbuf *m" "struct mbuf *n" 127.Ft u_int 128.Fn m_fixhdr "struct mbuf *mbuf" 129.Ft void 130.Fn m_dup_pkthdr "struct mbuf *to" "struct mbuf *from" 131.Ft void 132.Fn m_move_pkthdr "struct mbuf *to" "struct mbuf *from" 133.Ft u_int 134.Fn m_length "struct mbuf *mbuf" "struct mbuf **last" 135.Ft struct mbuf * 136.Fn m_split "struct mbuf *mbuf" "int len" "int how" 137.Ft int 138.Fn m_apply "struct mbuf *mbuf" "int off" "int len" "int (*f)(void *arg, void *data, u_int len)" "void *arg" 139.Ft struct mbuf * 140.Fn m_getptr "struct mbuf *mbuf" "int loc" "int *off" 141.Ft struct mbuf * 142.Fn m_defrag "struct mbuf *m0" "int how" 143.Ft struct mbuf * 144.Fn m_unshare "struct mbuf *m0" "int how" 145.\" 146.Sh DESCRIPTION 147An 148.Vt mbuf 149is a basic unit of memory management in the kernel IPC subsystem. 150Network packets and socket buffers are stored in 151.Vt mbufs . 152A network packet may span multiple 153.Vt mbufs 154arranged into a 155.Vt mbuf chain 156(linked list), 157which allows adding or trimming 158network headers with little overhead. 159.Pp 160While a developer should not bother with 161.Vt mbuf 162internals without serious 163reason in order to avoid incompatibilities with future changes, it 164is useful to understand the general structure of an 165.Vt mbuf . 166.Pp 167An 168.Vt mbuf 169consists of a variable-sized header and a small internal 170buffer for data. 171The total size of an 172.Vt mbuf , 173.Dv MSIZE , 174is a constant defined in 175.In sys/param.h . 176The 177.Vt mbuf 178header includes: 179.Bl -tag -width "m_nextpkt" -offset indent 180.It Va m_next 181.Pq Vt struct mbuf * 182A pointer to the next 183.Vt mbuf 184in the 185.Vt mbuf chain . 186.It Va m_nextpkt 187.Pq Vt struct mbuf * 188A pointer to the next 189.Vt mbuf chain 190in the queue. 191.It Va m_data 192.Pq Vt caddr_t 193A pointer to data attached to this 194.Vt mbuf . 195.It Va m_len 196.Pq Vt int 197The length of the data. 198.It Va m_type 199.Pq Vt short 200The type of the data. 201.It Va m_flags 202.Pq Vt int 203The 204.Vt mbuf 205flags. 206.El 207.Pp 208The 209.Vt mbuf 210flag bits are defined as follows: 211.Bd -literal 212/* mbuf flags */ 213#define M_EXT 0x00000001 /* has associated external storage */ 214#define M_PKTHDR 0x00000002 /* start of record */ 215#define M_EOR 0x00000004 /* end of record */ 216#define M_RDONLY 0x00000008 /* associated data marked read-only */ 217#define M_PROTO1 0x00001000 /* protocol-specific */ 218#define M_PROTO2 0x00002000 /* protocol-specific */ 219#define M_PROTO3 0x00004000 /* protocol-specific */ 220#define M_PROTO4 0x00008000 /* protocol-specific */ 221#define M_PROTO5 0x00010000 /* protocol-specific */ 222#define M_PROTO6 0x00020000 /* protocol-specific */ 223#define M_PROTO7 0x00040000 /* protocol-specific */ 224#define M_PROTO8 0x00080000 /* protocol-specific */ 225#define M_PROTO9 0x00100000 /* protocol-specific */ 226#define M_PROTO10 0x00200000 /* protocol-specific */ 227#define M_PROTO11 0x00400000 /* protocol-specific */ 228#define M_PROTO12 0x00800000 /* protocol-specific */ 229 230/* mbuf pkthdr flags (also stored in m_flags) */ 231#define M_BCAST 0x00000010 /* send/received as link-level broadcast */ 232#define M_MCAST 0x00000020 /* send/received as link-level multicast */ 233.Ed 234.Pp 235The available 236.Vt mbuf 237types are defined as follows: 238.Bd -literal 239/* mbuf types */ 240#define MT_DATA 1 /* dynamic (data) allocation */ 241#define MT_HEADER MT_DATA /* packet header */ 242#define MT_SONAME 8 /* socket name */ 243#define MT_CONTROL 14 /* extra-data protocol message */ 244#define MT_OOBDATA 15 /* expedited data */ 245.Ed 246.Pp 247The available external buffer types are defined as follows: 248.Bd -literal 249/* external buffer types */ 250#define EXT_CLUSTER 1 /* mbuf cluster */ 251#define EXT_SFBUF 2 /* sendfile(2)'s sf_bufs */ 252#define EXT_JUMBOP 3 /* jumbo cluster 4096 bytes */ 253#define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */ 254#define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */ 255#define EXT_PACKET 6 /* mbuf+cluster from packet zone */ 256#define EXT_MBUF 7 /* external mbuf reference (M_IOVEC) */ 257#define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */ 258#define EXT_MOD_TYPE 253 /* custom module's ext_buf type */ 259#define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */ 260#define EXT_EXTREF 255 /* has externally maintained ref_cnt ptr */ 261.Ed 262.Pp 263If the 264.Dv M_PKTHDR 265flag is set, a 266.Vt struct pkthdr Va m_pkthdr 267is added to the 268.Vt mbuf 269header. 270It contains a pointer to the interface 271the packet has been received from 272.Pq Vt struct ifnet Va *rcvif , 273and the total packet length 274.Pq Vt int Va len . 275Optionally, it may also contain an attached list of packet tags 276.Pq Vt "struct m_tag" . 277See 278.Xr mbuf_tags 9 279for details. 280Fields used in offloading checksum calculation to the hardware are kept in 281.Va m_pkthdr 282as well. 283See 284.Sx HARDWARE-ASSISTED CHECKSUM CALCULATION 285for details. 286.Pp 287If small enough, data is stored in the internal data buffer of an 288.Vt mbuf . 289If the data is sufficiently large, another 290.Vt mbuf 291may be added to the 292.Vt mbuf chain , 293or external storage may be associated with the 294.Vt mbuf . 295.Dv MHLEN 296bytes of data can fit into an 297.Vt mbuf 298with the 299.Dv M_PKTHDR 300flag set, 301.Dv MLEN 302bytes can otherwise. 303.Pp 304If external storage is being associated with an 305.Vt mbuf , 306the 307.Va m_ext 308header is added at the cost of losing the internal data buffer. 309It includes a pointer to external storage, the size of the storage, 310a pointer to a function used for freeing the storage, 311a pointer to an optional argument that can be passed to the function, 312and a pointer to a reference counter. 313An 314.Vt mbuf 315using external storage has the 316.Dv M_EXT 317flag set. 318.Pp 319The system supplies a macro for allocating the desired external storage 320buffer, 321.Dv MEXTADD . 322.Pp 323The allocation and management of the reference counter is handled by the 324subsystem. 325.Pp 326The system also supplies a default type of external storage buffer called an 327.Vt mbuf cluster . 328.Vt Mbuf clusters 329can be allocated and configured with the use of the 330.Dv MCLGET 331macro. 332Each 333.Vt mbuf cluster 334is 335.Dv MCLBYTES 336in size, where MCLBYTES is a machine-dependent constant. 337The system defines an advisory macro 338.Dv MINCLSIZE , 339which is the smallest amount of data to put into an 340.Vt mbuf cluster . 341It is equal to 342.Dv MHLEN 343plus one. 344It is typically preferable to store data into the data region of an 345.Vt mbuf , 346if size permits, as opposed to allocating a separate 347.Vt mbuf cluster 348to hold the same data. 349.\" 350.Ss Macros and Functions 351There are numerous predefined macros and functions that provide the 352developer with common utilities. 353.\" 354.Bl -ohang -offset indent 355.It Fn mtod mbuf type 356Convert an 357.Fa mbuf 358pointer to a data pointer. 359The macro expands to the data pointer cast to the specified 360.Fa type . 361.Sy Note : 362It is advisable to ensure that there is enough contiguous data in 363.Fa mbuf . 364See 365.Fn m_pullup 366for details. 367.It Fn MGET mbuf how type 368Allocate an 369.Vt mbuf 370and initialize it to contain internal data. 371.Fa mbuf 372will point to the allocated 373.Vt mbuf 374on success, or be set to 375.Dv NULL 376on failure. 377The 378.Fa how 379argument is to be set to 380.Dv M_WAITOK 381or 382.Dv M_NOWAIT . 383It specifies whether the caller is willing to block if necessary. 384A number of other functions and macros related to 385.Vt mbufs 386have the same argument because they may 387at some point need to allocate new 388.Vt mbufs . 389.It Fn MGETHDR mbuf how type 390Allocate an 391.Vt mbuf 392and initialize it to contain a packet header 393and internal data. 394See 395.Fn MGET 396for details. 397.It Fn MEXTADD mbuf buf size free opt_arg1 opt_arg2 flags type 398Associate externally managed data with 399.Fa mbuf . 400Any internal data contained in the mbuf will be discarded, and the 401.Dv M_EXT 402flag will be set. 403The 404.Fa buf 405and 406.Fa size 407arguments are the address and length, respectively, of the data. 408The 409.Fa free 410argument points to a function which will be called to free the data 411when the mbuf is freed; it is only used if 412.Fa type 413is 414.Dv EXT_EXTREF . 415The 416.Fa opt_arg1 417and 418.Fa opt_arg2 419arguments will be passed unmodified to 420.Fa free . 421The 422.Fa flags 423argument specifies additional 424.Vt mbuf 425flags; it is not necessary to specify 426.Dv M_EXT . 427Finally, the 428.Fa type 429argument specifies the type of external data, which controls how it 430will be disposed of when the 431.Vt mbuf 432is freed. 433In most cases, the correct value is 434.Dv EXT_EXTREF . 435.It Fn MCLGET mbuf how 436Allocate and attach an 437.Vt mbuf cluster 438to 439.Fa mbuf . 440On success, a non-zero value returned; otherwise, 0. 441Historically, consumers would check for success by testing the 442.Dv M_EXT 443flag on the mbuf, but this is now discouraged to avoid unnecessary awareness 444of the implementation of external storage in protocol stacks and device 445drivers. 446.It Fn M_ALIGN mbuf len 447Set the pointer 448.Fa mbuf->m_data 449to place an object of the size 450.Fa len 451at the end of the internal data area of 452.Fa mbuf , 453long word aligned. 454Applicable only if 455.Fa mbuf 456is newly allocated with 457.Fn MGET 458or 459.Fn m_get . 460.It Fn MH_ALIGN mbuf len 461Serves the same purpose as 462.Fn M_ALIGN 463does, but only for 464.Fa mbuf 465newly allocated with 466.Fn MGETHDR 467or 468.Fn m_gethdr , 469or initialized by 470.Fn m_dup_pkthdr 471or 472.Fn m_move_pkthdr . 473.It Fn m_align mbuf len 474Services the same purpose as 475.Fn M_ALIGN 476but handles any type of mbuf. 477.It Fn M_LEADINGSPACE mbuf 478Returns the number of bytes available before the beginning 479of data in 480.Fa mbuf . 481.It Fn M_TRAILINGSPACE mbuf 482Returns the number of bytes available after the end of data in 483.Fa mbuf . 484.It Fn M_PREPEND mbuf len how 485This macro operates on an 486.Vt mbuf chain . 487It is an optimized wrapper for 488.Fn m_prepend 489that can make use of possible empty space before data 490(e.g.\& left after trimming of a link-layer header). 491The new 492.Vt mbuf chain 493pointer or 494.Dv NULL 495is in 496.Fa mbuf 497after the call. 498.It Fn M_MOVE_PKTHDR to from 499Using this macro is equivalent to calling 500.Fn m_move_pkthdr to from . 501.It Fn M_WRITABLE mbuf 502This macro will evaluate true if 503.Fa mbuf 504is not marked 505.Dv M_RDONLY 506and if either 507.Fa mbuf 508does not contain external storage or, 509if it does, 510then if the reference count of the storage is not greater than 1. 511The 512.Dv M_RDONLY 513flag can be set in 514.Fa mbuf->m_flags . 515This can be achieved during setup of the external storage, 516by passing the 517.Dv M_RDONLY 518bit as a 519.Fa flags 520argument to the 521.Fn MEXTADD 522macro, or can be directly set in individual 523.Vt mbufs . 524.It Fn MCHTYPE mbuf type 525Change the type of 526.Fa mbuf 527to 528.Fa type . 529This is a relatively expensive operation and should be avoided. 530.El 531.Pp 532The functions are: 533.Bl -ohang -offset indent 534.It Fn m_get how type 535A function version of 536.Fn MGET 537for non-critical paths. 538.It Fn m_get2 size how type flags 539Allocate an 540.Vt mbuf 541with enough space to hold specified amount of data. 542.It Fn m_getm orig len how type 543Allocate 544.Fa len 545bytes worth of 546.Vt mbufs 547and 548.Vt mbuf clusters 549if necessary and append the resulting allocated 550.Vt mbuf chain 551to the 552.Vt mbuf chain 553.Fa orig , 554if it is 555.No non- Ns Dv NULL . 556If the allocation fails at any point, 557free whatever was allocated and return 558.Dv NULL . 559If 560.Fa orig 561is 562.No non- Ns Dv NULL , 563it will not be freed. 564It is possible to use 565.Fn m_getm 566to either append 567.Fa len 568bytes to an existing 569.Vt mbuf 570or 571.Vt mbuf chain 572(for example, one which may be sitting in a pre-allocated ring) 573or to simply perform an all-or-nothing 574.Vt mbuf 575and 576.Vt mbuf cluster 577allocation. 578.It Fn m_gethdr how type 579A function version of 580.Fn MGETHDR 581for non-critical paths. 582.It Fn m_getcl how type flags 583Fetch an 584.Vt mbuf 585with a 586.Vt mbuf cluster 587attached to it. 588If one of the allocations fails, the entire allocation fails. 589This routine is the preferred way of fetching both the 590.Vt mbuf 591and 592.Vt mbuf cluster 593together, as it avoids having to unlock/relock between allocations. 594Returns 595.Dv NULL 596on failure. 597.It Fn m_getjcl how type flags size 598This is like 599.Fn m_getcl 600but it the size of the cluster allocated will be large enough for 601.Fa size 602bytes. 603.It Fn m_getclr how type 604Allocate an 605.Vt mbuf 606and zero out the data region. 607.It Fn m_free mbuf 608Frees 609.Vt mbuf . 610Returns 611.Va m_next 612of the freed 613.Vt mbuf . 614.El 615.Pp 616The functions below operate on 617.Vt mbuf chains . 618.Bl -ohang -offset indent 619.It Fn m_freem mbuf 620Free an entire 621.Vt mbuf chain , 622including any external storage. 623.\" 624.It Fn m_adj mbuf len 625Trim 626.Fa len 627bytes from the head of an 628.Vt mbuf chain 629if 630.Fa len 631is positive, from the tail otherwise. 632.\" 633.It Fn m_append mbuf len cp 634Append 635.Vt len 636bytes of data 637.Vt cp 638to the 639.Vt mbuf chain . 640Extend the mbuf chain if the new data does not fit in 641existing space. 642.\" 643.It Fn m_prepend mbuf len how 644Allocate a new 645.Vt mbuf 646and prepend it to the 647.Vt mbuf chain , 648handle 649.Dv M_PKTHDR 650properly. 651.Sy Note : 652It does not allocate any 653.Vt mbuf clusters , 654so 655.Fa len 656must be less than 657.Dv MLEN 658or 659.Dv MHLEN , 660depending on the 661.Dv M_PKTHDR 662flag setting. 663.\" 664.It Fn m_copyup mbuf len dstoff 665Similar to 666.Fn m_pullup 667but copies 668.Fa len 669bytes of data into a new mbuf at 670.Fa dstoff 671bytes into the mbuf. 672The 673.Fa dstoff 674argument aligns the data and leaves room for a link layer header. 675Returns the new 676.Vt mbuf chain 677on success, 678and frees the 679.Vt mbuf chain 680and returns 681.Dv NULL 682on failure. 683.Sy Note : 684The function does not allocate 685.Vt mbuf clusters , 686so 687.Fa len + dstoff 688must be less than 689.Dv MHLEN . 690.\" 691.It Fn m_pullup mbuf len 692Arrange that the first 693.Fa len 694bytes of an 695.Vt mbuf chain 696are contiguous and lay in the data area of 697.Fa mbuf , 698so they are accessible with 699.Fn mtod mbuf type . 700It is important to remember that this may involve 701reallocating some mbufs and moving data so all pointers 702referencing data within the old mbuf chain 703must be recalculated or made invalid. 704Return the new 705.Vt mbuf chain 706on success, 707.Dv NULL 708on failure 709(the 710.Vt mbuf chain 711is freed in this case). 712.Sy Note : 713It does not allocate any 714.Vt mbuf clusters , 715so 716.Fa len 717must be less than or equal to 718.Dv MHLEN . 719.\" 720.It Fn m_pulldown mbuf offset len offsetp 721Arrange that 722.Fa len 723bytes between 724.Fa offset 725and 726.Fa offset + len 727in the 728.Vt mbuf chain 729are contiguous and lay in the data area of 730.Fa mbuf , 731so they are accessible with 732.Fn mtod mbuf type . 733.Fa len 734must be smaller than, or equal to, the size of an 735.Vt mbuf cluster . 736Return a pointer to an intermediate 737.Vt mbuf 738in the chain containing the requested region; 739the offset in the data region of the 740.Vt mbuf chain 741to the data contained in the returned mbuf is stored in 742.Fa *offsetp . 743If 744.Fa offsetp 745is NULL, the region may be accessed using 746.Fn mtod mbuf type . 747If 748.Fa offsetp 749is non-NULL, the region may be accessed using 750.Fn mtod mbuf uint8_t 751+ *offsetp. 752The region of the mbuf chain between its beginning and 753.Fa offset 754is not modified, therefore it is safe to hold pointers to data within 755this region before calling 756.Fn m_pulldown . 757.\" 758.It Fn m_copym mbuf offset len how 759Make a copy of an 760.Vt mbuf chain 761starting 762.Fa offset 763bytes from the beginning, continuing for 764.Fa len 765bytes. 766If 767.Fa len 768is 769.Dv M_COPYALL , 770copy to the end of the 771.Vt mbuf chain . 772.Sy Note : 773The copy is read-only, because the 774.Vt mbuf clusters 775are not copied, only their reference counts are incremented. 776.\" 777.It Fn m_copypacket mbuf how 778Copy an entire packet including header, which must be present. 779This is an optimized version of the common case 780.Fn m_copym mbuf 0 M_COPYALL how . 781.Sy Note : 782the copy is read-only, because the 783.Vt mbuf clusters 784are not copied, only their reference counts are incremented. 785.\" 786.It Fn m_dup mbuf how 787Copy a packet header 788.Vt mbuf chain 789into a completely new 790.Vt mbuf chain , 791including copying any 792.Vt mbuf clusters . 793Use this instead of 794.Fn m_copypacket 795when you need a writable copy of an 796.Vt mbuf chain . 797.\" 798.It Fn m_copydata mbuf offset len buf 799Copy data from an 800.Vt mbuf chain 801starting 802.Fa off 803bytes from the beginning, continuing for 804.Fa len 805bytes, into the indicated buffer 806.Fa buf . 807.\" 808.It Fn m_copyback mbuf offset len buf 809Copy 810.Fa len 811bytes from the buffer 812.Fa buf 813back into the indicated 814.Vt mbuf chain , 815starting at 816.Fa offset 817bytes from the beginning of the 818.Vt mbuf chain , 819extending the 820.Vt mbuf chain 821if necessary. 822.Sy Note : 823It does not allocate any 824.Vt mbuf clusters , 825just adds 826.Vt mbufs 827to the 828.Vt mbuf chain . 829It is safe to set 830.Fa offset 831beyond the current 832.Vt mbuf chain 833end: zeroed 834.Vt mbufs 835will be allocated to fill the space. 836.\" 837.It Fn m_length mbuf last 838Return the length of the 839.Vt mbuf chain , 840and optionally a pointer to the last 841.Vt mbuf . 842.\" 843.It Fn m_dup_pkthdr to from how 844Upon the function's completion, the 845.Vt mbuf 846.Fa to 847will contain an identical copy of 848.Fa from->m_pkthdr 849and the per-packet attributes found in the 850.Vt mbuf chain 851.Fa from . 852The 853.Vt mbuf 854.Fa from 855must have the flag 856.Dv M_PKTHDR 857initially set, and 858.Fa to 859must be empty on entry. 860.\" 861.It Fn m_move_pkthdr to from 862Move 863.Va m_pkthdr 864and the per-packet attributes from the 865.Vt mbuf chain 866.Fa from 867to the 868.Vt mbuf 869.Fa to . 870The 871.Vt mbuf 872.Fa from 873must have the flag 874.Dv M_PKTHDR 875initially set, and 876.Fa to 877must be empty on entry. 878Upon the function's completion, 879.Fa from 880will have the flag 881.Dv M_PKTHDR 882and the per-packet attributes cleared. 883.\" 884.It Fn m_fixhdr mbuf 885Set the packet-header length to the length of the 886.Vt mbuf chain . 887.\" 888.It Fn m_devget buf len offset ifp copy 889Copy data from a device local memory pointed to by 890.Fa buf 891to an 892.Vt mbuf chain . 893The copy is done using a specified copy routine 894.Fa copy , 895or 896.Fn bcopy 897if 898.Fa copy 899is 900.Dv NULL . 901.\" 902.It Fn m_cat m n 903Concatenate 904.Fa n 905to 906.Fa m . 907Both 908.Vt mbuf chains 909must be of the same type. 910.Fa N 911is still valid after the function returned. 912.Sy Note : 913It does not handle 914.Dv M_PKTHDR 915and friends. 916.\" 917.It Fn m_split mbuf len how 918Partition an 919.Vt mbuf chain 920in two pieces, returning the tail: 921all but the first 922.Fa len 923bytes. 924In case of failure, it returns 925.Dv NULL 926and attempts to restore the 927.Vt mbuf chain 928to its original state. 929.\" 930.It Fn m_apply mbuf off len f arg 931Apply a function to an 932.Vt mbuf chain , 933at offset 934.Fa off , 935for length 936.Fa len 937bytes. 938Typically used to avoid calls to 939.Fn m_pullup 940which would otherwise be unnecessary or undesirable. 941.Fa arg 942is a convenience argument which is passed to the callback function 943.Fa f . 944.Pp 945Each time 946.Fn f 947is called, it will be passed 948.Fa arg , 949a pointer to the 950.Fa data 951in the current mbuf, and the length 952.Fa len 953of the data in this mbuf to which the function should be applied. 954.Pp 955The function should return zero to indicate success; 956otherwise, if an error is indicated, then 957.Fn m_apply 958will return the error and stop iterating through the 959.Vt mbuf chain . 960.\" 961.It Fn m_getptr mbuf loc off 962Return a pointer to the mbuf containing the data located at 963.Fa loc 964bytes from the beginning of the 965.Vt mbuf chain . 966The corresponding offset into the mbuf will be stored in 967.Fa *off . 968.It Fn m_defrag m0 how 969Defragment an mbuf chain, returning the shortest possible 970chain of mbufs and clusters. 971If allocation fails and this can not be completed, 972.Dv NULL 973will be returned and the original chain will be unchanged. 974Upon success, the original chain will be freed and the new 975chain will be returned. 976.Fa how 977should be either 978.Dv M_WAITOK 979or 980.Dv M_NOWAIT , 981depending on the caller's preference. 982.Pp 983This function is especially useful in network drivers, where 984certain long mbuf chains must be shortened before being added 985to TX descriptor lists. 986.It Fn m_unshare m0 how 987Create a version of the specified mbuf chain whose 988contents can be safely modified without affecting other users. 989If allocation fails and this operation can not be completed, 990.Dv NULL 991will be returned. 992The original mbuf chain is always reclaimed and the reference 993count of any shared mbuf clusters is decremented. 994.Fa how 995should be either 996.Dv M_WAITOK 997or 998.Dv M_NOWAIT , 999depending on the caller's preference. 1000As a side-effect of this process the returned 1001mbuf chain may be compacted. 1002.Pp 1003This function is especially useful in the transmit path of 1004network code, when data must be encrypted or otherwise 1005altered prior to transmission. 1006.El 1007.Sh HARDWARE-ASSISTED CHECKSUM CALCULATION 1008This section currently applies to TCP/IP only. 1009In order to save the host CPU resources, computing checksums is 1010offloaded to the network interface hardware if possible. 1011The 1012.Va m_pkthdr 1013member of the leading 1014.Vt mbuf 1015of a packet contains two fields used for that purpose, 1016.Vt int Va csum_flags 1017and 1018.Vt int Va csum_data . 1019The meaning of those fields depends on the direction a packet flows in, 1020and on whether the packet is fragmented. 1021Henceforth, 1022.Va csum_flags 1023or 1024.Va csum_data 1025of a packet 1026will denote the corresponding field of the 1027.Va m_pkthdr 1028member of the leading 1029.Vt mbuf 1030in the 1031.Vt mbuf chain 1032containing the packet. 1033.Pp 1034On output, checksum offloading is attempted after the outgoing 1035interface has been determined for a packet. 1036The interface-specific field 1037.Va ifnet.if_data.ifi_hwassist 1038(see 1039.Xr ifnet 9 ) 1040is consulted for the capabilities of the interface to assist in 1041computing checksums. 1042The 1043.Va csum_flags 1044field of the packet header is set to indicate which actions the interface 1045is supposed to perform on it. 1046The actions unsupported by the network interface are done in the 1047software prior to passing the packet down to the interface driver; 1048such actions will never be requested through 1049.Va csum_flags . 1050.Pp 1051The flags demanding a particular action from an interface are as follows: 1052.Bl -tag -width ".Dv CSUM_TCP" -offset indent 1053.It Dv CSUM_IP 1054The IP header checksum is to be computed and stored in the 1055corresponding field of the packet. 1056The hardware is expected to know the format of an IP header 1057to determine the offset of the IP checksum field. 1058.It Dv CSUM_TCP 1059The TCP checksum is to be computed. 1060(See below.) 1061.It Dv CSUM_UDP 1062The UDP checksum is to be computed. 1063(See below.) 1064.El 1065.Pp 1066Should a TCP or UDP checksum be offloaded to the hardware, 1067the field 1068.Va csum_data 1069will contain the byte offset of the checksum field relative to the 1070end of the IP header. 1071In this case, the checksum field will be initially 1072set by the TCP/IP module to the checksum of the pseudo header 1073defined by the TCP and UDP specifications. 1074.Pp 1075On input, an interface indicates the actions it has performed 1076on a packet by setting one or more of the following flags in 1077.Va csum_flags 1078associated with the packet: 1079.Bl -tag -width ".Dv CSUM_IP_CHECKED" -offset indent 1080.It Dv CSUM_IP_CHECKED 1081The IP header checksum has been computed. 1082.It Dv CSUM_IP_VALID 1083The IP header has a valid checksum. 1084This flag can appear only in combination with 1085.Dv CSUM_IP_CHECKED . 1086.It Dv CSUM_DATA_VALID 1087The checksum of the data portion of the IP packet has been computed 1088and stored in the field 1089.Va csum_data 1090in network byte order. 1091.It Dv CSUM_PSEUDO_HDR 1092Can be set only along with 1093.Dv CSUM_DATA_VALID 1094to indicate that the IP data checksum found in 1095.Va csum_data 1096allows for the pseudo header defined by the TCP and UDP specifications. 1097Otherwise the checksum of the pseudo header must be calculated by 1098the host CPU and added to 1099.Va csum_data 1100to obtain the final checksum to be used for TCP or UDP validation purposes. 1101.El 1102.Pp 1103If a particular network interface just indicates success or 1104failure of TCP or UDP checksum validation without returning 1105the exact value of the checksum to the host CPU, its driver can mark 1106.Dv CSUM_DATA_VALID 1107and 1108.Dv CSUM_PSEUDO_HDR 1109in 1110.Va csum_flags , 1111and set 1112.Va csum_data 1113to 1114.Li 0xFFFF 1115hexadecimal to indicate a valid checksum. 1116It is a peculiarity of the algorithm used that the Internet checksum 1117calculated over any valid packet will be 1118.Li 0xFFFF 1119as long as the original checksum field is included. 1120.Sh STRESS TESTING 1121When running a kernel compiled with the option 1122.Dv MBUF_STRESS_TEST , 1123the following 1124.Xr sysctl 8 Ns 1125-controlled options may be used to create 1126various failure/extreme cases for testing of network drivers 1127and other parts of the kernel that rely on 1128.Vt mbufs . 1129.Bl -tag -width ident 1130.It Va net.inet.ip.mbuf_frag_size 1131Causes 1132.Fn ip_output 1133to fragment outgoing 1134.Vt mbuf chains 1135into fragments of the specified size. 1136Setting this variable to 1 is an excellent way to 1137test the long 1138.Vt mbuf chain 1139handling ability of network drivers. 1140.It Va kern.ipc.m_defragrandomfailures 1141Causes the function 1142.Fn m_defrag 1143to randomly fail, returning 1144.Dv NULL . 1145Any piece of code which uses 1146.Fn m_defrag 1147should be tested with this feature. 1148.El 1149.Sh RETURN VALUES 1150See above. 1151.Sh SEE ALSO 1152.Xr ifnet 9 , 1153.Xr mbuf_tags 9 1154.Sh HISTORY 1155.\" Please correct me if I'm wrong 1156.Vt Mbufs 1157appeared in an early version of 1158.Bx . 1159Besides being used for network packets, they were used 1160to store various dynamic structures, such as routing table 1161entries, interface addresses, protocol control blocks, etc. 1162In more recent 1163.Fx 1164use of 1165.Vt mbufs 1166is almost entirely limited to packet storage, with 1167.Xr uma 9 1168zones being used directly to store other network-related memory. 1169.Pp 1170Historically, the 1171.Vt mbuf 1172allocator has been a special-purpose memory allocator able to run in 1173interrupt contexts and allocating from a special kernel address space map. 1174As of 1175.Fx 5.3 , 1176the 1177.Vt mbuf 1178allocator is a wrapper around 1179.Xr uma 9 , 1180allowing caching of 1181.Vt mbufs , 1182clusters, and 1183.Vt mbuf 1184+ cluster pairs in per-CPU caches, as well as bringing other benefits of 1185slab allocation. 1186.Sh AUTHORS 1187The original 1188.Nm 1189manual page was written by 1190.An Yar Tikhiy . 1191The 1192.Xr uma 9 1193.Vt mbuf 1194allocator was written by 1195.An Bosko Milekic . 1196