1.\" $NetBSD: mbuf.9,v 1.49 2009/03/09 17:28:12 joerg Exp $ 2.\" 3.\" Copyright (c) 1997 The NetBSD Foundation, Inc. 4.\" All rights reserved. 5.\" 6.\" This documentation is derived from text contributed to The NetBSD Foundation 7.\" by S.P.Zeidler (aka stargazer). 8.\" 9.\" Redistribution and use in source and binary forms, with or without 10.\" modification, are permitted provided that the following conditions 11.\" are met: 12.\" 1. Redistributions of source code must retain the above copyright 13.\" notice, this list of conditions and the following disclaimer. 14.\" 2. Redistributions in binary form must reproduce the above copyright 15.\" notice, this list of conditions and the following disclaimer in the 16.\" documentation and/or other materials provided with the distribution. 17.\" 18.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 19.\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 20.\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 21.\" PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 22.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 23.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 24.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 25.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 26.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 27.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 28.\" POSSIBILITY OF SUCH DAMAGE. 29.\" 30.Dd March 24, 2008 31.Dt MBUF 9 32.Os 33.Sh NAME 34.Nm mbuf , 35.Nm m_get , 36.Nm m_getclr , 37.Nm m_gethdr , 38.Nm m_devget , 39.Nm m_copym , 40.Nm m_copypacket , 41.Nm m_copydata , 42.Nm m_copyback , 43.Nm m_copyback_cow , 44.Nm m_cat , 45.Nm m_dup , 46.Nm m_makewritable , 47.Nm m_prepend , 48.Nm m_pulldown , 49.Nm m_pullup , 50.Nm m_split , 51.Nm m_adj , 52.Nm m_apply , 53.Nm m_free , 54.Nm m_freem , 55.Nm mtod , 56.Nm MGET , 57.Nm MGETHDR , 58.Nm MEXTMALLOC , 59.Nm MEXTADD , 60.Nm MCLGET , 61.Nm M_COPY_PKTHDR , 62.Nm M_MOVE_PKTHDR , 63.Nm M_ALIGN , 64.Nm MH_ALIGN , 65.Nm M_LEADINGSPACE , 66.Nm M_TRAILINGSPACE , 67.Nm M_PREPEND , 68.Nm MCHTYPE , 69.Nm MFREE 70.Nd "functions and macros for managing memory used by networking code" 71.Sh SYNOPSIS 72.In sys/mbuf.h 73.Ft struct mbuf * 74.Fn m_get "int nowait" "int type" 75.Ft struct mbuf * 76.Fn m_getclr "int nowait" "int type" 77.Ft struct mbuf * 78.Fn m_gethdr "int nowait" "int type" 79.Ft struct mbuf * 80.Fn m_devget "char *buf" "int totlen" "int off0" "struct ifnet *ifp" "void (*copy)(const void *, void *, size_t)" 81.Ft struct mbuf * 82.Fn m_copym "struct mbuf *m" "int off0" "int len" "int wait" 83.Ft struct mbuf * 84.Fn m_copypacket "struct mbuf *m" "int how" 85.Ft void 86.Fn m_copydata "struct mbuf *m" "int off" "int len" "void *cp" 87.Ft void 88.Fn m_copyback "struct mbuf *m0" "int off" "int len" "void *cp" 89.Ft struct mbuf * 90.Fn m_copyback_cow "struct mbuf *m0" "int off" "int len" "void *cp" "int how" 91.Ft int 92.Fn m_makewritable "struct mbuf **mp" "int off" "int len" "int how" 93.Ft void 94.Fn m_cat "struct mbuf *m" "struct mbuf *n" 95.Ft struct mbuf * 96.Fn m_dup "struct mbuf *m" "int off0" "int len" "int wait" 97.Ft struct mbuf * 98.Fn m_prepend "struct mbuf *m" "int len" "int how" 99.Ft struct mbuf * 100.Fn m_pulldown "struct mbuf *m" "int off" "int len" "int *offp" 101.Ft struct mbuf * 102.Fn m_pullup "struct mbuf *n" "int len" 103.Ft struct mbuf * 104.Fn m_split "struct mbuf *m0" "int len0" "int wait" 105.Ft void 106.Fn m_adj "struct mbuf *mp" "int req_len" 107.Ft int 108.Fn m_apply "struct mbuf *m" "int off" "int len" "int *f(void *, void *, unsigned int)" "void *arg" 109.Ft struct mbuf * 110.Fn m_free "struct mbuf *m" 111.Ft void 112.Fn m_freem "struct mbuf *m" 113.Ft datatype 114.Fn mtod "struct mbuf *m" "datatype" 115.Ft void 116.Fn MGET "struct mbuf *m" "int how" "int type" 117.Ft void 118.Fn MGETHDR "struct mbuf *m" "int how" "int type" 119.Ft void 120.Fn MEXTMALLOC "struct mbuf *m" "int len" "int how" 121.Ft void 122.Fn MEXTADD "struct mbuf *m" "void *buf" "int size" "int type" "void (*free)(struct mbuf *, void *, size_t, void *)" "void *arg" 123.Ft void 124.Fn MCLGET "struct mbuf *m" "int how" 125.Ft void 126.Fn M_COPY_PKTHDR "struct mbuf *to" "struct mbuf *from" 127.Ft void 128.Fn M_MOVE_PKTHDR "struct mbuf *to" "struct mbuf *from" 129.Ft void 130.Fn M_ALIGN "struct mbuf *m" "int len" 131.Ft void 132.Fn MH_ALIGN "struct mbuf *m" "int len" 133.Ft int 134.Fn M_LEADINGSPACE "struct mbuf *m" 135.Ft int 136.Fn M_TRAILINGSPACE "struct mbuf *m" 137.Ft void 138.Fn M_PREPEND "struct mbuf *m" "int plen" "int how" 139.Ft void 140.Fn MCHTYPE "struct mbuf *m" "int type" 141.Ft void 142.Fn MFREE "struct mbuf *m" "struct mbuf *n" 143.Sh DESCRIPTION 144The 145.Nm 146functions and macros provide an easy and consistent way to handle 147a networking stack's memory management needs. 148An 149.Nm 150consists of a header and a data area. 151It is of a fixed size, 152.Dv MSIZE 153.Pq defined in Aq Pa machine/param.h , 154which includes overhead. 155The header contains a pointer to the next 156.Nm 157in the 158.Sq "mbuf chain" , 159a pointer to the next 160.Sq "mbuf chain" , 161a pointer to the data area, the amount of data in this mbuf, its type 162and a 163.Dv flags 164field. 165.Pp 166The 167.Dv type 168variable can signify: 169.Bl -tag -compact -offset indent -width "XXXXXXXXXXX" 170.It Dv MT_FREE 171the mbuf should be on the ``free'' list 172.It Dv MT_DATA 173data was dynamically allocated 174.It Dv MT_HEADER 175data is a packet header 176.It Dv MT_SONAME 177data is a socket name 178.It Dv MT_SOOPTS 179data is socket options 180.It Dv MT_FTABLE 181data is the fragment reassembly header 182.It Dv MT_CONTROL 183mbuf contains ancillary \&(protocol control\&) data 184.It Dv MT_OOBDATA 185mbuf contains out-of-band data. 186.El 187.Pp 188The 189.Dv flags 190variable contains information describing the 191.Nm , 192notably: 193.Bl -tag -compact -offset indent -width "XXXXXXXXXXX" 194.It Dv M_EXT 195has external storage 196.It Dv M_PKTHDR 197is start of record 198.It Dv M_EOR 199is end of record 200.It Dv M_CLUSTER 201external storage is a cluster. 202.El 203.Pp 204If an 205.Nm 206designates the start of a record 207.Pq Dv M_PKTHDR , 208its 209.Dv flags 210field may contain additional information describing the content of 211the record: 212.Bl -tag -compact -offset indent -width "XXXXXXXXXXX" 213.It Dv M_BCAST 214sent/received as link-level broadcast 215.It Dv M_MCAST 216sent/received as link-level multicast 217.It Dv M_LINK0 , 218.It Dv M_LINK1 , 219.It Dv M_LINK2 220three link-level specific flags. 221.El 222.Pp 223An 224.Nm 225may add a single 226.Sq "mbuf cluster" 227of 228.Dv MCLBYTES 229bytes 230.Pq also defined in Aq Pa machine/param.h , 231which has no additional overhead 232and is used instead of the internal data area; this is done when at least 233.Dv MINCLSIZE 234bytes of data must be stored. 235.Pp 236When the 237.Dv M_EXT 238flag is raised for an mbuf, 239the external storage area could be shared among multiple mbufs. 240Be careful when you attempt to overwrite the data content of the mbuf. 241.Bl -tag -width compact 242.It Fn m_get "int nowait" "int type" 243Allocates an mbuf and initializes it to contain internal data. 244The 245.Fa nowait 246parameter is a choice of 247.Dv M_WAIT / M_DONTWAIT 248from caller. 249.Dv M_WAIT 250means the call cannot fail, but may take forever. 251The 252.Fa type 253parameter is an mbuf type. 254.It Fn m_getclr "int nowait" "int type" 255Allocates an mbuf and initializes it to contain internal data, then 256zeros the data area. 257The 258.Fa nowait 259parameter is a choice of 260.Dv M_WAIT / M_DONTWAIT 261from caller. 262The 263.Fa type 264parameter is an mbuf type. 265.It Fn m_gethdr "int nowait" "int type" 266Allocates an mbuf and initializes it to contain a packet header and internal 267data. 268The 269.Fa nowait 270parameter is a choice of 271.Dv M_WAIT / M_DONTWAIT 272from caller. 273The 274.Fa type 275parameter is an mbuf type. 276.It Fn m_devget "char *buf" "int totlen" "int off0" "struct ifnet *ifp" "void (*copy)(const void *, void *, size_t)" 277Copies 278.Fa len 279bytes from device local memory into mbufs using copy routine 280.Fa copy . 281If parameter 282.Fa off 283is non-zero, the packet is supposed to be trailer-encapsulated and 284.Fa off 285bytes plus the type and length fields will be skipped before copying. 286Returns the top of the mbuf chain it created. 287.It Fn m_copym "struct mbuf *m" "int off0" "int len" "int wait" 288Creates a copy of an mbuf chain starting 289.Fa off0 290bytes from the beginning, continuing for 291.Fa len 292bytes. 293If the 294.Fa len 295requested is 296.Dv M_COPYALL , 297the complete mbuf chain will be copied. 298The 299.Fa wait 300parameter is a choice of 301.Dv M_WAIT / M_DONTWAIT 302from caller. 303.It Fn m_copypacket "struct mbuf *m" "int how" 304Copies an entire packet, including header (which must be present). 305This function is an optimization of the common case 306.Li m_copym(m, 0, Dv M_COPYALL, Fa how ) . 307.It Fn m_copydata "struct mbuf *m" "int off" "int len" "void *cp" 308Copies 309.Fa len 310bytes data from mbuf chain 311.Fa m 312into the buffer 313.Fa cp , 314starting 315.Fa off 316bytes from the beginning. 317.It Fn m_copyback "struct mbuf *m0" "int off" "int len" "void *cp" 318Copies 319.Fa len 320bytes data from buffer 321.Fa cp 322back into the mbuf chain 323.Fa m0 , 324starting 325.Fa off 326bytes from the beginning of the chain, extending the mbuf chain if necessary. 327.Fn m_copyback 328can only fail when extending the chain. 329The caller should check for this kind of failure 330by checking the resulting length of the chain in that case. 331It is an error to use 332.Fn m_copyback 333on read-only mbufs. 334.It Fn m_copyback_cow "struct mbuf *m0" "int off" "int len" "void *cp" \ 335"int how" 336Copies 337.Fa len 338bytes data from buffer 339.Fa cp 340back into the mbuf chain 341.Fa m0 342as 343.Fn m_copyback 344does. 345Unlike 346.Fn m_copyback , 347it is safe to use 348.Fn m_copyback_cow 349on read-only mbufs. 350If needed, 351.Fn m_copyback_cow 352automatically allocates new mbufs and adjusts the chain. 353On success, it returns a pointer to the resulting mbuf chain, 354and frees the original mbuf 355.Fa m0 . 356Otherwise, it returns 357.Dv NULL . 358The 359.Fa how 360parameter is a choice of 361.Dv M_WAIT / M_DONTWAIT 362from the caller. 363Unlike 364.Fn m_copyback , 365extending the mbuf chain isn't supported. 366It is an error to attempt to extend the mbuf chain using 367.Fn m_copyback_cow . 368.It Fn m_makewritable "struct mbuf **mp" "int off" "int len" "int how" 369Rearranges an mbuf chain so that 370.Fa len 371bytes from offset 372.Fa off 373are writable. 374When it meets read-only mbufs, it allocates new mbufs, adjusts the chain as 375.Fn m_copyback_cow 376does, and copies the original content into them. 377.Fn m_makewritable 378does 379.Em not 380guarantee that all 381.Fa len 382bytes at 383.Fa off 384are consecutive. 385The 386.Fa how 387parameter is a choice of 388.Dv M_WAIT / M_DONTWAIT 389from the caller. 390.Fn m_makewritable 391preserves the contents of the mbuf chain even in the case of failure. 392It updates a pointer to the mbuf chain pointed to by 393.Fa mp . 394It returns 0 on success. 395Otherwise, it returns an error code, typically 396.Er ENOBUFS . 397.It Fn m_cat "struct mbuf *m" "struct mbuf *n" 398Concatenates mbuf chain 399.Fa n 400to 401.Fa m . 402Both chains must be of the same type; packet headers will 403.Em not 404be updated if present. 405.It Fn m_dup "struct mbuf *m" "int off0" "int len" "int wait" 406Similarly to 407.Fn m_copym , 408the function creates a copy of an mbuf chain starting 409.Fa off0 410bytes from the beginning, continuing for 411.Fa len 412bytes. 413While 414.Fn m_copym 415tries to share external storage for mbufs with 416.Dv M_EXT 417flag, 418.Fn m_dup 419will deep-copy the whole data content into new mbuf chain 420and avoids shared external storage. 421.It Fn m_prepend "struct mbuf *m" "int len" "int how" 422Lesser-used path for 423.Fn M_PREPEND : 424allocates new mbuf 425.Fa m 426of size 427.Fa len 428to prepend to the chain, copying junk along. 429The 430.Fa how 431parameter is a choice of 432.Dv M_WAIT / M_DONTWAIT 433from caller. 434.It Fn m_pulldown "struct mbuf *m" "int off" "int len" "int *offp" 435Rearranges an mbuf chain so that 436.Fa len 437bytes from offset 438.Fa off 439are contiguous and in the data area of an mbuf. 440The return value points to an mbuf in the middle of the mbuf chain 441.Fa m . 442If we call the return value 443.Fa n , 444the contiguous data region is available at 445.Li "mtod(n, void *) + *offp" , 446or 447.Li "mtod(n, void *)" 448if 449.Fa offp 450is 451.Dv NULL . 452The top of the mbuf chain 453.Fa m , 454and mbufs up to 455.Fa off , 456will not be modified. 457On successful return, it is guaranteed that the mbuf pointed to by 458.Fa n 459does not have a shared external storage, 460therefore it is safe to update the contiguous region. 461Returns 462.Dv NULL 463and frees the mbuf chain on failure. 464.Fa len 465must be smaller or equal than 466.Dv MCLBYTES . 467.It Fn m_pullup "struct mbuf *m" "int len" 468Rearranges an mbuf chain so that 469.Fa len 470bytes are contiguous 471and in the data area of an mbuf (so that 472.Fn mtod 473will work for a structure of size 474.Fa len ) . 475Returns the resulting 476mbuf chain on success, frees it and returns 477.Dv NULL 478on failure. 479If there is room, it will add up to 480.Dv max_protohdr 481- 482.Fa len 483extra bytes to the 484contiguous region to possibly avoid being called again. 485.Fa len 486must be smaller or equal than 487.Dv MHLEN . 488.It Fn m_split "struct mbuf *m0" "int len0" "int wait" 489Partitions an mbuf chain in two pieces, returning the tail, 490which is all but the first 491.Fa len0 492bytes. 493In case of failure, it returns 494.Dv NULL 495and attempts to 496restore the chain to its original state. 497.It Fn m_adj "struct mbuf *mp" "int req_len" 498Shaves off 499.Fa req_len 500bytes from head or tail of the (valid) data area. 501If 502.Fa req_len 503is greater than zero, front bytes are being shaved off, if it's smaller, 504from the back (and if it is zero, the mbuf will stay bearded). 505This function does not move data in any way, but is used to manipulate the 506data area pointer and data length variable of the mbuf in a non-clobbering 507way. 508.It Fn m_apply "struct mbuf *m" "int off" "int len" "int (*f)(void *, void *, unsigned int)" "void *arg" 509Apply function 510.Fa f 511to the data in an mbuf chain starting 512.Fa off 513bytes from the beginning, continuing for 514.Fa len 515bytes. 516Neither 517.Fa off 518nor 519.Fa len 520may be negative. 521.Fa arg 522will be supplied as first argument for 523.Fa f , 524the second argument will be the pointer to the data buffer of a 525packet (starting after 526.Fa off 527bytes in the stream), and the third argument is the amount 528of data in bytes in this call. 529If 530.Fa f 531returns something not equal to zero 532.Fn m_apply 533will bail out, returning the return code of 534.Fa f . 535Upon successful completion it will return zero. 536.It Fn m_free "struct mbuf *m" 537Frees mbuf 538.Fa m . 539.It Fn m_freem "struct mbuf *m" 540Frees the mbuf chain beginning with 541.Fa m . 542This function contains the elementary sanity check for a 543.Dv NULL 544pointer. 545.It Fn mtod "struct mbuf *m" "datatype" 546Returns a pointer to the data contained in the specified mbuf 547.Fa m , 548type-casted to the specified data type 549.Fa datatype . 550Implemented as a macro. 551.It Fn MGET "struct mbuf *m" "int how" "int type" 552Allocates mbuf 553.Fa m 554and initializes it to contain internal data. 555See 556.Fn m_get . 557Implemented as a macro. 558.It Fn MGETHDR "struct mbuf *m" "int how" "int type" 559Allocates mbuf 560.Fa m 561and initializes it to contain a packet header. 562See 563.Fn m_gethdr . 564Implemented as a macro. 565.It Fn MEXTMALLOC "struct mbuf *m" "int len" "int how" 566Allocates external storage of size 567.Fa len 568for mbuf 569.Fa m . 570The 571.Fa how 572parameter is a choice of 573.Dv M_WAIT / M_DONTWAIT 574from caller. 575The flag 576.Dv M_EXT 577is set upon success. 578Implemented as a macro. 579.It Fn MEXTADD "struct mbuf *m" "void *buf" "int size" "int type" "void (*free)(struct mbuf *, void *, size_t, void *)" "void *arg" 580Adds pre-allocated external storage 581.Fa buf 582to a normal mbuf 583.Fa m ; 584the parameters 585.Fa size , 586.Fa type , 587.Fa free 588and 589.Fa arg 590describe the external storage. 591.Fa size 592is the size of the storage, 593.Fa type 594describes its 595.Xr malloc 9 596type, 597.Fa free 598is a free routine (if not the usual one), and 599.Fa arg 600is a possible argument to the free routine. 601The flag 602.Dv M_EXT 603is set upon success. 604Implemented as a macro. 605If a free routine is specified, it will be called when the mbuf is freed. 606In the case of former, the first argument for a free routine is the mbuf 607.Fa m 608and the routine is expected to free it in addition to the external storage 609pointed by second argument. 610In the case of latter, the first argument for the routine is NULL. 611.It Fn MCLGET "struct mbuf *m" "int how" 612Allocates and adds an mbuf cluster to a normal mbuf 613.Fa m . 614The 615.Fa how 616parameter is a choice of 617.Dv M_WAIT / M_DONTWAIT 618from caller. 619The flag 620.Dv M_EXT 621is set upon success. 622Implemented as a macro. 623.It Fn M_COPY_PKTHDR "struct mbuf *to" "struct mbuf *from" 624Copies the mbuf pkthdr from mbuf 625.Fa from 626to mbuf 627.Fa to . 628.Fa from 629must have the type flag 630.Dv M_PKTHDR 631set, and 632.Fa to 633must be empty. 634Implemented as a macro. 635.It Fn M_MOVE_PKTHDR "struct mbuf *to" "struct mbuf *from" 636Moves the mbuf pkthdr from mbuf 637.Fa from 638to mbuf 639.Fa to . 640.Fa from 641must have the type flag 642.Dv M_PKTHDR 643set, and 644.Fa to 645must be empty. 646The flag 647.Dv M_PKTHDR 648in mbuf 649.Fa from 650will be cleared. 651.It Fn M_ALIGN "struct mbuf *m" "int len" 652Sets the data pointer of a newly allocated mbuf 653.Fa m 654to 655.Fa len 656bytes from the end of the mbuf data area, so that 657.Fa len 658bytes of data written to the mbuf 659.Fa m , 660starting at the data pointer, will be aligned to the end of the data area. 661Implemented as a macro. 662.It Fn MH_ALIGN "struct mbuf *m" "int len" 663Sets the data pointer of a newly allocated packetheader mbuf 664.Fa m 665to 666.Fa len 667bytes from the end of the mbuf data area, so that 668.Fa len 669bytes of data written to the mbuf 670.Fa m , 671starting at the data pointer, will be aligned to the end of the data area. 672Implemented as a macro. 673.It Fn M_LEADINGSPACE "struct mbuf *m" 674Returns the amount of space available before the current start of valid 675data in mbuf 676.Fa m . 677Returns 0 if the mbuf data part is shared across multiple mbufs 678.Pq i.e. not writable . 679Implemented as a macro. 680.It Fn M_TRAILINGSPACE "struct mbuf *m" 681Returns the amount of space available after the current end of valid 682data in mbuf 683.Fa m . 684Returns 0 if the mbuf data part is shared across multiple mbufs 685.Pq i.e. not writable . 686Implemented as a macro. 687.It Fn M_PREPEND "struct mbuf *m" "int plen" "int how" 688Prepends space of size 689.Fa plen 690to mbuf 691.Fa m . 692If a new mbuf must be allocated, 693.Fa how 694specifies whether to wait. 695If 696.Fa how 697is 698.Dv M_DONTWAIT 699and allocation fails, the original mbuf chain is freed and 700.Fa m 701is set to 702.Dv NULL . 703Implemented as a macro. 704.It Fn MCHTYPE "struct mbuf *m" "int type" 705Change mbuf 706.Fa m 707to new type 708.Fa type . 709Implemented as a macro. 710.It Fn MFREE "struct mbuf *m" "struct mbuf *n" 711Frees a single mbuf 712.Fa m 713and places the successor, if any, in mbuf 714.Fa n . 715Implemented as a macro. 716.El 717.Sh FILES 718The 719.Nm 720management functions are implemented within the file 721.Pa sys/kern/uipc_mbuf.c . 722Function prototypes, and the functions implemented as macros 723are located in 724.Pa sys/sys/mbuf.h . 725Both pathnames are relative to the root of the 726.Nx 727source tree, 728.Pa /usr/src . 729.Sh SEE ALSO 730.Pa /usr/share/doc/smm/18.net , 731.Xr netstat 1 , 732.Xr m_tag 9 , 733.Xr malloc 9 734.Rs 735.%A Jun-ichiro Hagino 736.%T "Mbuf issues in 4.4BSD IPv6/IPsec support (experiences from KAME IPv6/IPsec implementation)" 737.%B "Proceedings of the freenix track: 2000 USENIX annual technical conference" 738.%D June 2000 739.Re 740.Sh AUTHORS 741.An -nosplit 742The original mbuf data structures were designed by Rob Gurwitz 743when he did the initial TCP/IP implementation at BBN. 744.Pp 745Further extensions and enhancements were made by Bill Joy, Sam Leffler, 746and Mike Karels at CSRG. 747.Pp 748Current implementation of external storage by 749.An Matt Thomas 750.Aq matt@3am-software.com 751and 752.An Jason R. Thorpe 753.Aq thorpej@NetBSD.org . 754