The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/sys/mbuf.h

Version: -  FREEBSD  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-2  -  FREEBSD-11-1  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-4  -  FREEBSD-10-3  -  FREEBSD-10-2  -  FREEBSD-10-1  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-3  -  FREEBSD-9-2  -  FREEBSD-9-1  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-4  -  FREEBSD-8-3  -  FREEBSD-8-2  -  FREEBSD-8-1  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-4  -  FREEBSD-7-3  -  FREEBSD-7-2  -  FREEBSD-7-1  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-4  -  FREEBSD-6-3  -  FREEBSD-6-2  -  FREEBSD-6-1  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-5  -  FREEBSD-5-4  -  FREEBSD-5-3  -  FREEBSD-5-2  -  FREEBSD-5-1  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1982, 1986, 1988, 1993
    3  *      The Regents of the University of California.
    4  * All rights reserved.
    5  *
    6  * Redistribution and use in source and binary forms, with or without
    7  * modification, are permitted provided that the following conditions
    8  * are met:
    9  * 1. Redistributions of source code must retain the above copyright
   10  *    notice, this list of conditions and the following disclaimer.
   11  * 2. Redistributions in binary form must reproduce the above copyright
   12  *    notice, this list of conditions and the following disclaimer in the
   13  *    documentation and/or other materials provided with the distribution.
   14  * 3. Neither the name of the University nor the names of its contributors
   15  *    may be used to endorse or promote products derived from this software
   16  *    without specific prior written permission.
   17  *
   18  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   28  * SUCH DAMAGE.
   29  *
   30  *      @(#)mbuf.h      8.5 (Berkeley) 2/19/95
   31  * $FreeBSD: releng/7.4/sys/sys/mbuf.h 205430 2010-03-22 00:11:31Z tuexen $
   32  */
   33 
   34 #ifndef _SYS_MBUF_H_
   35 #define _SYS_MBUF_H_
   36 
   37 /* XXX: These includes suck. Sorry! */
   38 #include <sys/queue.h>
   39 #ifdef _KERNEL
   40 #include <sys/systm.h>
   41 #include <vm/uma.h>
   42 #ifdef WITNESS
   43 #include <sys/lock.h>
   44 #endif
   45 #endif
   46 
   47 /*
   48  * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
   49  * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
   50  * sys/param.h), which has no additional overhead and is used instead of the
   51  * internal data area; this is done when at least MINCLSIZE of data must be
   52  * stored.  Additionally, it is possible to allocate a separate buffer
   53  * externally and attach it to the mbuf in a way similar to that of mbuf
   54  * clusters.
   55  */
   56 #define MLEN            (MSIZE - sizeof(struct m_hdr))  /* normal data len */
   57 #define MHLEN           (MLEN - sizeof(struct pkthdr))  /* data len w/pkthdr */
   58 #define MINCLSIZE       (MHLEN + 1)     /* smallest amount to put in cluster */
   59 #define M_MAXCOMPRESS   (MHLEN / 2)     /* max amount to copy for compression */
   60 
   61 #ifdef _KERNEL
   62 /*-
   63  * Macros for type conversion:
   64  * mtod(m, t)   -- Convert mbuf pointer to data pointer of correct type.
   65  * dtom(x)      -- Convert data pointer within mbuf to mbuf pointer (XXX).
   66  */
   67 #define mtod(m, t)      ((t)((m)->m_data))
   68 #define dtom(x)         ((struct mbuf *)((intptr_t)(x) & ~(MSIZE-1)))
   69 
   70 /*
   71  * Argument structure passed to UMA routines during mbuf and packet
   72  * allocations.
   73  */
   74 struct mb_args {
   75         int     flags;  /* Flags for mbuf being allocated */
   76         short   type;   /* Type of mbuf being allocated */
   77 };
   78 #endif /* _KERNEL */
   79 
   80 #if defined(__LP64__)
   81 #define M_HDR_PAD    6
   82 #else
   83 #define M_HDR_PAD    2
   84 #endif
   85 
   86 /*
   87  * Header present at the beginning of every mbuf.
   88  */
   89 struct m_hdr {
   90         struct mbuf     *mh_next;       /* next buffer in chain */
   91         struct mbuf     *mh_nextpkt;    /* next chain in queue/record */
   92         caddr_t          mh_data;       /* location of data */
   93         int              mh_len;        /* amount of data in this mbuf */
   94         int              mh_flags;      /* flags; see below */
   95         short            mh_type;       /* type of data in this mbuf */
   96         uint8_t          pad[M_HDR_PAD];/* word align                  */
   97 };
   98 
   99 /*
  100  * Packet tag structure (see below for details).
  101  */
  102 struct m_tag {
  103         SLIST_ENTRY(m_tag)      m_tag_link;     /* List of packet tags */
  104         u_int16_t               m_tag_id;       /* Tag ID */
  105         u_int16_t               m_tag_len;      /* Length of data */
  106         u_int32_t               m_tag_cookie;   /* ABI/Module ID */
  107         void                    (*m_tag_free)(struct m_tag *);
  108 };
  109 
  110 /*
  111  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
  112  */
  113 struct pkthdr {
  114         struct ifnet    *rcvif;         /* rcv interface */
  115         /* variables for ip and tcp reassembly */
  116         void            *header;        /* pointer to packet header */
  117         int              len;           /* total packet length */
  118         /* variables for hardware checksum */
  119         int              csum_flags;    /* flags regarding checksum */
  120         int              csum_data;     /* data field used by csum routines */
  121         u_int16_t        tso_segsz;     /* TSO segment size */
  122         u_int16_t        ether_vtag;    /* Ethernet 802.1p+q vlan tag */
  123         SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
  124 };
  125 
  126 /*
  127  * Description of external storage mapped into mbuf; valid only if M_EXT is
  128  * set.
  129  */
  130 struct m_ext {
  131         caddr_t          ext_buf;       /* start of buffer */
  132         void            (*ext_free)     /* free routine if not the usual */
  133                             (void *, void *);
  134         void            *ext_args;      /* optional argument pointer */
  135         u_int            ext_size;      /* size of buffer, for ext_free */
  136         volatile u_int  *ref_cnt;       /* pointer to ref count info */
  137         int              ext_type;      /* type of external storage */
  138 };
  139 
  140 /*
  141  * The core of the mbuf object along with some shortcut defines for practical
  142  * purposes.
  143  */
  144 struct mbuf {
  145         struct m_hdr    m_hdr;
  146         union {
  147                 struct {
  148                         struct pkthdr   MH_pkthdr;      /* M_PKTHDR set */
  149                         union {
  150                                 struct m_ext    MH_ext; /* M_EXT set */
  151                                 char            MH_databuf[MHLEN];
  152                         } MH_dat;
  153                 } MH;
  154                 char    M_databuf[MLEN];                /* !M_PKTHDR, !M_EXT */
  155         } M_dat;
  156 };
  157 #define m_next          m_hdr.mh_next
  158 #define m_len           m_hdr.mh_len
  159 #define m_data          m_hdr.mh_data
  160 #define m_type          m_hdr.mh_type
  161 #define m_flags         m_hdr.mh_flags
  162 #define m_nextpkt       m_hdr.mh_nextpkt
  163 #define m_act           m_nextpkt
  164 #define m_pkthdr        M_dat.MH.MH_pkthdr
  165 #define m_ext           M_dat.MH.MH_dat.MH_ext
  166 #define m_pktdat        M_dat.MH.MH_dat.MH_databuf
  167 #define m_dat           M_dat.M_databuf
  168 
  169 /*
  170  * mbuf flags.
  171  */
  172 #define M_EXT           0x00000001 /* has associated external storage */
  173 #define M_PKTHDR        0x00000002 /* start of record */
  174 #define M_EOR           0x00000004 /* end of record */
  175 #define M_RDONLY        0x00000008 /* associated data is marked read-only */
  176 #define M_PROTO1        0x00000010 /* protocol-specific */
  177 #define M_PROTO2        0x00000020 /* protocol-specific */
  178 #define M_PROTO3        0x00000040 /* protocol-specific */
  179 #define M_PROTO4        0x00000080 /* protocol-specific */
  180 #define M_PROTO5        0x00000100 /* protocol-specific */
  181 #define M_BCAST         0x00000200 /* send/received as link-level broadcast */
  182 #define M_MCAST         0x00000400 /* send/received as link-level multicast */
  183 #define M_FRAG          0x00000800 /* packet is a fragment of a larger packet */
  184 #define M_FIRSTFRAG     0x00001000 /* packet is first fragment */
  185 #define M_LASTFRAG      0x00002000 /* packet is last fragment */
  186 #define M_SKIP_FIREWALL 0x00004000 /* skip firewall processing */
  187 #define M_FREELIST      0x00008000 /* mbuf is on the free list */
  188 #define M_VLANTAG       0x00010000 /* ether_vtag is valid */
  189 #define M_PROMISC       0x00020000 /* packet was not for us */
  190 #define M_NOFREE        0x00040000 /* do not free mbuf, embedded in cluster */
  191 #define M_PROTO6        0x00080000 /* protocol-specific */
  192 #define M_PROTO7        0x00100000 /* protocol-specific */
  193 #define M_PROTO8        0x00200000 /* protocol-specific */
  194 /*
  195  * For RELENG_{6,7} steal these flags for limited multiple routing table
  196  * support. In RELENG_8 and beyond, use just one flag and a tag.
  197  */
  198 #define M_FIB           0xF0000000 /* steal some bits to store fib number. */
  199 
  200 #define M_NOTIFICATION  M_PROTO5    /* SCTP notification */
  201 
  202 /*
  203  * Flags to purge when crossing layers.
  204  */
  205 #define M_PROTOFLAGS \
  206     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8)
  207 
  208 /*
  209  * Flags preserved when copying m_pkthdr.
  210  */
  211 #define M_COPYFLAGS \
  212     (M_PKTHDR|M_EOR|M_RDONLY|M_PROTOFLAGS|M_SKIP_FIREWALL|M_BCAST|M_MCAST|\
  213      M_FRAG|M_FIRSTFRAG|M_LASTFRAG|M_VLANTAG|M_PROMISC|M_FIB)
  214 
  215 /*
  216  * External buffer types: identify ext_buf type.
  217  */
  218 #define EXT_CLUSTER     1       /* mbuf cluster */
  219 #define EXT_SFBUF       2       /* sendfile(2)'s sf_bufs */
  220 #define EXT_JUMBOP      3       /* jumbo cluster 4096 bytes */
  221 #define EXT_JUMBO9      4       /* jumbo cluster 9216 bytes */
  222 #define EXT_JUMBO16     5       /* jumbo cluster 16184 bytes */
  223 #define EXT_PACKET      6       /* mbuf+cluster from packet zone */
  224 #define EXT_MBUF        7       /* external mbuf reference (M_IOVEC) */
  225 #define EXT_NET_DRV     100     /* custom ext_buf provided by net driver(s) */
  226 #define EXT_MOD_TYPE    200     /* custom module's ext_buf type */
  227 #define EXT_DISPOSABLE  300     /* can throw this buffer away w/page flipping */
  228 #define EXT_EXTREF      400     /* has externally maintained ref_cnt ptr */
  229 
  230 /*
  231  * Flags indicating hw checksum support and sw checksum requirements.  This
  232  * field can be directly tested against if_data.ifi_hwassist.
  233  */
  234 #define CSUM_IP                 0x0001          /* will csum IP */
  235 #define CSUM_TCP                0x0002          /* will csum TCP */
  236 #define CSUM_UDP                0x0004          /* will csum UDP */
  237 #define CSUM_IP_FRAGS           0x0008          /* will csum IP fragments */
  238 #define CSUM_FRAGMENT           0x0010          /* will do IP fragmentation */
  239 #define CSUM_TSO                0x0020          /* will do TSO */
  240 #define CSUM_SCTP               0x0040          /* will csum SCTP */
  241 
  242 #define CSUM_IP_CHECKED         0x0100          /* did csum IP */
  243 #define CSUM_IP_VALID           0x0200          /*   ... the csum is valid */
  244 #define CSUM_DATA_VALID         0x0400          /* csum_data field is valid */
  245 #define CSUM_PSEUDO_HDR         0x0800          /* csum_data has pseudo hdr */
  246 #define CSUM_SCTP_VALID         0x1000          /* SCTP checksum is valid */
  247 
  248 #define CSUM_DELAY_DATA         (CSUM_TCP | CSUM_UDP)
  249 #define CSUM_DELAY_IP           (CSUM_IP)       /* XXX add ipv6 here too? */
  250 
  251 /*
  252  * mbuf types.
  253  */
  254 #define MT_NOTMBUF      0       /* USED INTERNALLY ONLY! Object is not mbuf */
  255 #define MT_DATA         1       /* dynamic (data) allocation */
  256 #define MT_HEADER       MT_DATA /* packet header, use M_PKTHDR instead */
  257 #define MT_SONAME       8       /* socket name */
  258 #define MT_CONTROL      14      /* extra-data protocol message */
  259 #define MT_OOBDATA      15      /* expedited data  */
  260 #define MT_NTYPES       16      /* number of mbuf types for mbtypes[] */
  261 
  262 #define MT_NOINIT       255     /* Not a type but a flag to allocate
  263                                    a non-initialized mbuf */
  264 
  265 #define MB_NOTAGS       0x1UL   /* no tags attached to mbuf */
  266 
  267 /*
  268  * General mbuf allocator statistics structure.
  269  *
  270  * Many of these statistics are no longer used; we instead track many
  271  * allocator statistics through UMA's built in statistics mechanism.
  272  */
  273 struct mbstat {
  274         u_long  m_mbufs;        /* XXX */
  275         u_long  m_mclusts;      /* XXX */
  276 
  277         u_long  m_drain;        /* times drained protocols for space */
  278         u_long  m_mcfail;       /* XXX: times m_copym failed */
  279         u_long  m_mpfail;       /* XXX: times m_pullup failed */
  280         u_long  m_msize;        /* length of an mbuf */
  281         u_long  m_mclbytes;     /* length of an mbuf cluster */
  282         u_long  m_minclsize;    /* min length of data to allocate a cluster */
  283         u_long  m_mlen;         /* length of data in an mbuf */
  284         u_long  m_mhlen;        /* length of data in a header mbuf */
  285 
  286         /* Number of mbtypes (gives # elems in mbtypes[] array) */
  287         short   m_numtypes;
  288 
  289         /* XXX: Sendfile stats should eventually move to their own struct */
  290         u_long  sf_iocnt;       /* times sendfile had to do disk I/O */
  291         u_long  sf_allocfail;   /* times sfbuf allocation failed */
  292         u_long  sf_allocwait;   /* times sfbuf allocation had to wait */
  293 };
  294 
  295 /*
  296  * Flags specifying how an allocation should be made.
  297  *
  298  * The flag to use is as follows:
  299  * - M_DONTWAIT or M_NOWAIT from an interrupt handler to not block allocation.
  300  * - M_WAIT or M_WAITOK or M_TRYWAIT from wherever it is safe to block.
  301  *
  302  * M_DONTWAIT/M_NOWAIT means that we will not block the thread explicitly and
  303  * if we cannot allocate immediately we may return NULL, whereas
  304  * M_WAIT/M_WAITOK/M_TRYWAIT means that if we cannot allocate resources we
  305  * will block until they are available, and thus never return NULL.
  306  *
  307  * XXX Eventually just phase this out to use M_WAITOK/M_NOWAIT.
  308  */
  309 #define MBTOM(how)      (how)
  310 #define M_DONTWAIT      M_NOWAIT
  311 #define M_TRYWAIT       M_WAITOK
  312 #define M_WAIT          M_WAITOK
  313 
  314 /*
  315  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
  316  * !_KERNEL so that monitoring tools can look up the zones with
  317  * libmemstat(3).
  318  */
  319 #define MBUF_MEM_NAME           "mbuf"
  320 #define MBUF_CLUSTER_MEM_NAME   "mbuf_cluster"
  321 #define MBUF_PACKET_MEM_NAME    "mbuf_packet"
  322 #define MBUF_JUMBOP_MEM_NAME    "mbuf_jumbo_pagesize"
  323 #define MBUF_JUMBO9_MEM_NAME    "mbuf_jumbo_9k"
  324 #define MBUF_JUMBO16_MEM_NAME   "mbuf_jumbo_16k"
  325 #define MBUF_TAG_MEM_NAME       "mbuf_tag"
  326 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
  327 
  328 #ifdef _KERNEL
  329 
  330 #ifdef WITNESS
  331 #define MBUF_CHECKSLEEP(how) do {                                       \
  332         if (how == M_WAITOK)                                            \
  333                 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,         \
  334                     "Sleeping in \"%s\"", __func__);                    \
  335 } while (0)
  336 #else
  337 #define MBUF_CHECKSLEEP(how)
  338 #endif
  339 
  340 /*
  341  * Network buffer allocation API
  342  *
  343  * The rest of it is defined in kern/kern_mbuf.c
  344  */
  345 
  346 extern uma_zone_t       zone_mbuf;
  347 extern uma_zone_t       zone_clust;
  348 extern uma_zone_t       zone_pack;
  349 extern uma_zone_t       zone_jumbop;
  350 extern uma_zone_t       zone_jumbo9;
  351 extern uma_zone_t       zone_jumbo16;
  352 extern uma_zone_t       zone_ext_refcnt;
  353 
  354 static __inline struct mbuf     *m_getcl(int how, short type, int flags);
  355 static __inline struct mbuf     *m_get(int how, short type);
  356 static __inline struct mbuf     *m_gethdr(int how, short type);
  357 static __inline struct mbuf     *m_getjcl(int how, short type, int flags,
  358                                     int size);
  359 static __inline struct mbuf     *m_getclr(int how, short type); /* XXX */
  360 static __inline struct mbuf     *m_free(struct mbuf *m);
  361 static __inline void             m_clget(struct mbuf *m, int how);
  362 static __inline void            *m_cljget(struct mbuf *m, int how, int size);
  363 static __inline void             m_chtype(struct mbuf *m, short new_type);
  364 void                             mb_free_ext(struct mbuf *);
  365 static __inline struct mbuf     *m_last(struct mbuf *m);
  366 
  367 static __inline int
  368 m_gettype(int size)
  369 {
  370         int type;
  371         
  372         switch (size) {
  373         case MSIZE:
  374                 type = EXT_MBUF;
  375                 break;
  376         case MCLBYTES:
  377                 type = EXT_CLUSTER;
  378                 break;
  379 #if MJUMPAGESIZE != MCLBYTES
  380         case MJUMPAGESIZE:
  381                 type = EXT_JUMBOP;
  382                 break;
  383 #endif
  384         case MJUM9BYTES:
  385                 type = EXT_JUMBO9;
  386                 break;
  387         case MJUM16BYTES:
  388                 type = EXT_JUMBO16;
  389                 break;
  390         default:
  391                 panic("%s: m_getjcl: invalid cluster size", __func__);
  392         }
  393 
  394         return (type);
  395 }
  396 
  397 static __inline uma_zone_t
  398 m_getzone(int size)
  399 {
  400         uma_zone_t zone;
  401         
  402         switch (size) {
  403         case MSIZE:
  404                 zone = zone_mbuf;
  405                 break;
  406         case MCLBYTES:
  407                 zone = zone_clust;
  408                 break;
  409 #if MJUMPAGESIZE != MCLBYTES
  410         case MJUMPAGESIZE:
  411                 zone = zone_jumbop;
  412                 break;
  413 #endif
  414         case MJUM9BYTES:
  415                 zone = zone_jumbo9;
  416                 break;
  417         case MJUM16BYTES:
  418                 zone = zone_jumbo16;
  419                 break;
  420         default:
  421                 panic("%s: m_getjcl: invalid cluster type", __func__);
  422         }
  423 
  424         return (zone);
  425 }
  426 
  427 static __inline struct mbuf *
  428 m_get(int how, short type)
  429 {
  430         struct mb_args args;
  431 
  432         args.flags = 0;
  433         args.type = type;
  434         return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
  435 }
  436 
  437 /*
  438  * XXX This should be deprecated, very little use.
  439  */
  440 static __inline struct mbuf *
  441 m_getclr(int how, short type)
  442 {
  443         struct mbuf *m;
  444         struct mb_args args;
  445 
  446         args.flags = 0;
  447         args.type = type;
  448         m = uma_zalloc_arg(zone_mbuf, &args, how);
  449         if (m != NULL)
  450                 bzero(m->m_data, MLEN);
  451         return (m);
  452 }
  453 
  454 static __inline struct mbuf *
  455 m_gethdr(int how, short type)
  456 {
  457         struct mb_args args;
  458 
  459         args.flags = M_PKTHDR;
  460         args.type = type;
  461         return ((struct mbuf *)(uma_zalloc_arg(zone_mbuf, &args, how)));
  462 }
  463 
  464 static __inline struct mbuf *
  465 m_getcl(int how, short type, int flags)
  466 {
  467         struct mb_args args;
  468 
  469         args.flags = flags;
  470         args.type = type;
  471         return ((struct mbuf *)(uma_zalloc_arg(zone_pack, &args, how)));
  472 }
  473 
  474 /*
  475  * m_getjcl() returns an mbuf with a cluster of the specified size attached.
  476  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
  477  *
  478  * XXX: This is rather large, should be real function maybe.
  479  */
  480 static __inline struct mbuf *
  481 m_getjcl(int how, short type, int flags, int size)
  482 {
  483         struct mb_args args;
  484         struct mbuf *m, *n;
  485         uma_zone_t zone;
  486 
  487         args.flags = flags;
  488         args.type = type;
  489 
  490         m = uma_zalloc_arg(zone_mbuf, &args, how);
  491         if (m == NULL)
  492                 return (NULL);
  493 
  494         zone = m_getzone(size);
  495         n = uma_zalloc_arg(zone, m, how);
  496         if (n == NULL) {
  497                 uma_zfree(zone_mbuf, m);
  498                 return (NULL);
  499         }
  500         return (m);
  501 }
  502 
  503 static __inline void
  504 m_free_fast(struct mbuf *m)
  505 {
  506         KASSERT(SLIST_EMPTY(&m->m_pkthdr.tags), ("doing fast free of mbuf with tags"));
  507 
  508         uma_zfree_arg(zone_mbuf, m, (void *)MB_NOTAGS);
  509 }
  510 
  511 static __inline struct mbuf *
  512 m_free(struct mbuf *m)
  513 {
  514         struct mbuf *n = m->m_next;
  515 
  516         if (m->m_flags & M_EXT)
  517                 mb_free_ext(m);
  518         else if ((m->m_flags & M_NOFREE) == 0)
  519                 uma_zfree(zone_mbuf, m);
  520         return (n);
  521 }
  522 
  523 static __inline void
  524 m_clget(struct mbuf *m, int how)
  525 {
  526 
  527         if (m->m_flags & M_EXT)
  528                 printf("%s: %p mbuf already has cluster\n", __func__, m);
  529         m->m_ext.ext_buf = (char *)NULL;
  530         uma_zalloc_arg(zone_clust, m, how);
  531         /*
  532          * On a cluster allocation failure, drain the packet zone and retry,
  533          * we might be able to loosen a few clusters up on the drain.
  534          */
  535         if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
  536                 zone_drain(zone_pack);
  537                 uma_zalloc_arg(zone_clust, m, how);
  538         }
  539 }
  540 
  541 /*
  542  * m_cljget() is different from m_clget() as it can allocate clusters without
  543  * attaching them to an mbuf.  In that case the return value is the pointer
  544  * to the cluster of the requested size.  If an mbuf was specified, it gets
  545  * the cluster attached to it and the return value can be safely ignored.
  546  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
  547  */
  548 static __inline void *
  549 m_cljget(struct mbuf *m, int how, int size)
  550 {
  551         uma_zone_t zone;
  552 
  553         if (m && m->m_flags & M_EXT)
  554                 printf("%s: %p mbuf already has cluster\n", __func__, m);
  555         if (m != NULL)
  556                 m->m_ext.ext_buf = NULL;
  557 
  558         zone = m_getzone(size);
  559         return (uma_zalloc_arg(zone, m, how));
  560 }
  561 
  562 static __inline void
  563 m_cljset(struct mbuf *m, void *cl, int type)
  564 {
  565         uma_zone_t zone;
  566         int size;
  567         
  568         switch (type) {
  569         case EXT_CLUSTER:
  570                 size = MCLBYTES;
  571                 zone = zone_clust;
  572                 break;
  573 #if MJUMPAGESIZE != MCLBYTES
  574         case EXT_JUMBOP:
  575                 size = MJUMPAGESIZE;
  576                 zone = zone_jumbop;
  577                 break;
  578 #endif
  579         case EXT_JUMBO9:
  580                 size = MJUM9BYTES;
  581                 zone = zone_jumbo9;
  582                 break;
  583         case EXT_JUMBO16:
  584                 size = MJUM16BYTES;
  585                 zone = zone_jumbo16;
  586                 break;
  587         default:
  588                 panic("unknown cluster type");
  589                 break;
  590         }
  591 
  592         m->m_data = m->m_ext.ext_buf = cl;
  593         m->m_ext.ext_free = m->m_ext.ext_args = NULL;
  594         m->m_ext.ext_size = size;
  595         m->m_ext.ext_type = type;
  596         m->m_ext.ref_cnt = uma_find_refcnt(zone, cl);
  597         m->m_flags |= M_EXT;
  598 
  599 }
  600 
  601 static __inline void
  602 m_chtype(struct mbuf *m, short new_type)
  603 {
  604 
  605         m->m_type = new_type;
  606 }
  607 
  608 static __inline struct mbuf *
  609 m_last(struct mbuf *m)
  610 {
  611 
  612         while (m->m_next)
  613                 m = m->m_next;
  614         return (m);
  615 }
  616 
  617 /*
  618  * mbuf, cluster, and external object allocation macros (for compatibility
  619  * purposes).
  620  */
  621 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
  622 #define MGET(m, how, type)      ((m) = m_get((how), (type)))
  623 #define MGETHDR(m, how, type)   ((m) = m_gethdr((how), (type)))
  624 #define MCLGET(m, how)          m_clget((m), (how))
  625 #define MEXTADD(m, buf, size, free, args, flags, type)                  \
  626     m_extadd((m), (caddr_t)(buf), (size), (free), (args), (flags), (type))
  627 #define m_getm(m, len, how, type)                                       \
  628     m_getm2((m), (len), (how), (type), M_PKTHDR)
  629 
  630 /*
  631  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
  632  * be both the local data payload, or an external buffer area, depending on
  633  * whether M_EXT is set).
  634  */
  635 #define M_WRITABLE(m)   (!((m)->m_flags & M_RDONLY) &&                  \
  636                          (!(((m)->m_flags & M_EXT)) ||                  \
  637                          (*((m)->m_ext.ref_cnt) == 1)) )                \
  638 
  639 /* Check if the supplied mbuf has a packet header, or else panic. */
  640 #define M_ASSERTPKTHDR(m)                                               \
  641         KASSERT(m != NULL && m->m_flags & M_PKTHDR,                     \
  642             ("%s: no mbuf packet header!", __func__))
  643 
  644 /*
  645  * Ensure that the supplied mbuf is a valid, non-free mbuf.
  646  *
  647  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
  648  */
  649 #define M_ASSERTVALID(m)                                                \
  650         KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,                 \
  651             ("%s: attempted use of a free mbuf!", __func__))
  652 
  653 /*
  654  * Set the m_data pointer of a newly-allocated mbuf (m_get/MGET) to place an
  655  * object of the specified size at the end of the mbuf, longword aligned.
  656  */
  657 #define M_ALIGN(m, len) do {                                            \
  658         KASSERT(!((m)->m_flags & (M_PKTHDR|M_EXT)),                     \
  659                 ("%s: M_ALIGN not normal mbuf", __func__));             \
  660         KASSERT((m)->m_data == (m)->m_dat,                              \
  661                 ("%s: M_ALIGN not a virgin mbuf", __func__));           \
  662         (m)->m_data += (MLEN - (len)) & ~(sizeof(long) - 1);            \
  663 } while (0)
  664 
  665 /*
  666  * As above, for mbufs allocated with m_gethdr/MGETHDR or initialized by
  667  * M_DUP/MOVE_PKTHDR.
  668  */
  669 #define MH_ALIGN(m, len) do {                                           \
  670         KASSERT((m)->m_flags & M_PKTHDR && !((m)->m_flags & M_EXT),     \
  671                 ("%s: MH_ALIGN not PKTHDR mbuf", __func__));            \
  672         KASSERT((m)->m_data == (m)->m_pktdat,                           \
  673                 ("%s: MH_ALIGN not a virgin mbuf", __func__));          \
  674         (m)->m_data += (MHLEN - (len)) & ~(sizeof(long) - 1);           \
  675 } while (0)
  676 
  677 /*
  678  * Compute the amount of space available before the current start of data in
  679  * an mbuf.
  680  *
  681  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
  682  * of checking writability of the mbuf data area rests solely with the caller.
  683  */
  684 #define M_LEADINGSPACE(m)                                               \
  685         ((m)->m_flags & M_EXT ?                                         \
  686             (M_WRITABLE(m) ? (m)->m_data - (m)->m_ext.ext_buf : 0):     \
  687             (m)->m_flags & M_PKTHDR ? (m)->m_data - (m)->m_pktdat :     \
  688             (m)->m_data - (m)->m_dat)
  689 
  690 /*
  691  * Compute the amount of space available after the end of data in an mbuf.
  692  *
  693  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
  694  * of checking writability of the mbuf data area rests solely with the caller.
  695  */
  696 #define M_TRAILINGSPACE(m)                                              \
  697         ((m)->m_flags & M_EXT ?                                         \
  698             (M_WRITABLE(m) ? (m)->m_ext.ext_buf + (m)->m_ext.ext_size   \
  699                 - ((m)->m_data + (m)->m_len) : 0) :                     \
  700             &(m)->m_dat[MLEN] - ((m)->m_data + (m)->m_len))
  701 
  702 /*
  703  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
  704  * allocated, how specifies whether to wait.  If the allocation fails, the
  705  * original mbuf chain is freed and m is set to NULL.
  706  */
  707 #define M_PREPEND(m, plen, how) do {                                    \
  708         struct mbuf **_mmp = &(m);                                      \
  709         struct mbuf *_mm = *_mmp;                                       \
  710         int _mplen = (plen);                                            \
  711         int __mhow = (how);                                             \
  712                                                                         \
  713         MBUF_CHECKSLEEP(how);                                           \
  714         if (M_LEADINGSPACE(_mm) >= _mplen) {                            \
  715                 _mm->m_data -= _mplen;                                  \
  716                 _mm->m_len += _mplen;                                   \
  717         } else                                                          \
  718                 _mm = m_prepend(_mm, _mplen, __mhow);                   \
  719         if (_mm != NULL && _mm->m_flags & M_PKTHDR)                     \
  720                 _mm->m_pkthdr.len += _mplen;                            \
  721         *_mmp = _mm;                                                    \
  722 } while (0)
  723 
  724 /*
  725  * Change mbuf to new type.  This is a relatively expensive operation and
  726  * should be avoided.
  727  */
  728 #define MCHTYPE(m, t)   m_chtype((m), (t))
  729 
  730 /* Length to m_copy to copy all. */
  731 #define M_COPYALL       1000000000
  732 
  733 /* Compatibility with 4.3. */
  734 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT)
  735 
  736 extern int              max_datalen;    /* MHLEN - max_hdr */
  737 extern int              max_hdr;        /* Largest link + protocol header */
  738 extern int              max_linkhdr;    /* Largest link-level header */
  739 extern int              max_protohdr;   /* Largest protocol header */
  740 extern struct mbstat    mbstat;         /* General mbuf stats/infos */
  741 extern int              nmbclusters;    /* Maximum number of clusters */
  742 
  743 struct uio;
  744 
  745 void             m_adj(struct mbuf *, int);
  746 void             m_align(struct mbuf *, int);
  747 int              m_apply(struct mbuf *, int, int,
  748                     int (*)(void *, void *, u_int), void *);
  749 int              m_append(struct mbuf *, int, c_caddr_t);
  750 void             m_cat(struct mbuf *, struct mbuf *);
  751 void             m_extadd(struct mbuf *, caddr_t, u_int,
  752                     void (*)(void *, void *), void *, int, int);
  753 struct mbuf     *m_collapse(struct mbuf *, int, int);
  754 void             m_copyback(struct mbuf *, int, int, c_caddr_t);
  755 void             m_copydata(const struct mbuf *, int, int, caddr_t);
  756 struct mbuf     *m_copym(struct mbuf *, int, int, int);
  757 struct mbuf     *m_copymdata(struct mbuf *, struct mbuf *,
  758                     int, int, int, int);
  759 struct mbuf     *m_copypacket(struct mbuf *, int);
  760 void             m_copy_pkthdr(struct mbuf *, struct mbuf *);
  761 struct mbuf     *m_copyup(struct mbuf *n, int len, int dstoff);
  762 struct mbuf     *m_defrag(struct mbuf *, int);
  763 void             m_demote(struct mbuf *, int);
  764 struct mbuf     *m_devget(char *, int, int, struct ifnet *,
  765                     void (*)(char *, caddr_t, u_int));
  766 struct mbuf     *m_dup(struct mbuf *, int);
  767 int              m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
  768 u_int            m_fixhdr(struct mbuf *);
  769 struct mbuf     *m_fragment(struct mbuf *, int, int);
  770 void             m_freem(struct mbuf *);
  771 struct mbuf     *m_getm2(struct mbuf *, int, int, short, int);
  772 struct mbuf     *m_getptr(struct mbuf *, int, int *);
  773 u_int            m_length(struct mbuf *, struct mbuf **);
  774 void             m_move_pkthdr(struct mbuf *, struct mbuf *);
  775 struct mbuf     *m_prepend(struct mbuf *, int, int);
  776 void             m_print(const struct mbuf *, int);
  777 struct mbuf     *m_pulldown(struct mbuf *, int, int, int *);
  778 struct mbuf     *m_pullup(struct mbuf *, int);
  779 int             m_sanity(struct mbuf *, int);
  780 struct mbuf     *m_split(struct mbuf *, int, int);
  781 struct mbuf     *m_uiotombuf(struct uio *, int, int, int, int);
  782 struct mbuf     *m_unshare(struct mbuf *, int how);
  783 
  784 /*-
  785  * Network packets may have annotations attached by affixing a list of
  786  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
  787  * allocated semi-opaque data structures that have a fixed header
  788  * (struct m_tag) that specifies the size of the memory block and a
  789  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
  790  * unsigned value used to identify a module or ABI.  By convention this value
  791  * is chosen as the date+time that the module is created, expressed as the
  792  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
  793  * value is an ABI/module-specific value that identifies a particular
  794  * annotation and is private to the module.  For compatibility with systems
  795  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
  796  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
  797  * compatibility shim functions and several tag types are defined below.
  798  * Users that do not require compatibility should use a private cookie value
  799  * so that packet tag-related definitions can be maintained privately.
  800  *
  801  * Note that the packet tag returned by m_tag_alloc has the default memory
  802  * alignment implemented by malloc.  To reference private data one can use a
  803  * construct like:
  804  *
  805  *      struct m_tag *mtag = m_tag_alloc(...);
  806  *      struct foo *p = (struct foo *)(mtag+1);
  807  *
  808  * if the alignment of struct m_tag is sufficient for referencing members of
  809  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
  810  * private data structure to insure proper alignment; e.g.,
  811  *
  812  *      struct foo {
  813  *              struct m_tag    tag;
  814  *              ...
  815  *      };
  816  *      struct foo *p = (struct foo *) m_tag_alloc(...);
  817  *      struct m_tag *mtag = &p->tag;
  818  */
  819 
  820 /*
  821  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
  822  * tags are expected to ``vanish'' when they pass through a network
  823  * interface.  For most interfaces this happens normally as the tags are
  824  * reclaimed when the mbuf is free'd.  However in some special cases
  825  * reclaiming must be done manually.  An example is packets that pass through
  826  * the loopback interface.  Also, one must be careful to do this when
  827  * ``turning around'' packets (e.g., icmp_reflect).
  828  *
  829  * To mark a tag persistent bit-or this flag in when defining the tag id.
  830  * The tag will then be treated as described above.
  831  */
  832 #define MTAG_PERSISTENT                         0x800
  833 
  834 #define PACKET_TAG_NONE                         0  /* Nadda */
  835 
  836 /* Packet tags for use with PACKET_ABI_COMPAT. */
  837 #define PACKET_TAG_IPSEC_IN_DONE                1  /* IPsec applied, in */
  838 #define PACKET_TAG_IPSEC_OUT_DONE               2  /* IPsec applied, out */
  839 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE         3  /* NIC IPsec crypto done */
  840 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED      4  /* NIC IPsec crypto req'ed */
  841 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO     5  /* NIC notifies IPsec */
  842 #define PACKET_TAG_IPSEC_PENDING_TDB            6  /* Reminder to do IPsec */
  843 #define PACKET_TAG_BRIDGE                       7  /* Bridge processing done */
  844 #define PACKET_TAG_GIF                          8  /* GIF processing done */
  845 #define PACKET_TAG_GRE                          9  /* GRE processing done */
  846 #define PACKET_TAG_IN_PACKET_CHECKSUM           10 /* NIC checksumming done */
  847 #define PACKET_TAG_ENCAP                        11 /* Encap.  processing */
  848 #define PACKET_TAG_IPSEC_SOCKET                 12 /* IPSEC socket ref */
  849 #define PACKET_TAG_IPSEC_HISTORY                13 /* IPSEC history */
  850 #define PACKET_TAG_IPV6_INPUT                   14 /* IPV6 input processing */
  851 #define PACKET_TAG_DUMMYNET                     15 /* dummynet info */
  852 #define PACKET_TAG_DIVERT                       17 /* divert info */
  853 #define PACKET_TAG_IPFORWARD                    18 /* ipforward info */
  854 #define PACKET_TAG_MACLABEL     (19 | MTAG_PERSISTENT) /* MAC label */
  855 #define PACKET_TAG_PF                           21 /* PF + ALTQ information */
  856 #define PACKET_TAG_RTSOCKFAM                    25 /* rtsock sa family */
  857 #define PACKET_TAG_IPOPTIONS                    27 /* Saved IP options */
  858 #define PACKET_TAG_CARP                         28 /* CARP info */
  859 
  860 /* Specific cookies and tags. */
  861 
  862 /* Packet tag routines. */
  863 struct m_tag    *m_tag_alloc(u_int32_t, int, int, int);
  864 void             m_tag_delete(struct mbuf *, struct m_tag *);
  865 void             m_tag_delete_chain(struct mbuf *, struct m_tag *);
  866 void             m_tag_free_default(struct m_tag *);
  867 struct m_tag    *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
  868 struct m_tag    *m_tag_copy(struct m_tag *, int);
  869 int              m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
  870 void             m_tag_delete_nonpersistent(struct mbuf *);
  871 
  872 /*
  873  * Initialize the list of tags associated with an mbuf.
  874  */
  875 static __inline void
  876 m_tag_init(struct mbuf *m)
  877 {
  878 
  879         SLIST_INIT(&m->m_pkthdr.tags);
  880 }
  881 
  882 /*
  883  * Set up the contents of a tag.  Note that this does not fill in the free
  884  * method; the caller is expected to do that.
  885  *
  886  * XXX probably should be called m_tag_init, but that was already taken.
  887  */
  888 static __inline void
  889 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
  890 {
  891 
  892         t->m_tag_id = type;
  893         t->m_tag_len = len;
  894         t->m_tag_cookie = cookie;
  895 }
  896 
  897 /*
  898  * Reclaim resources associated with a tag.
  899  */
  900 static __inline void
  901 m_tag_free(struct m_tag *t)
  902 {
  903 
  904         (*t->m_tag_free)(t);
  905 }
  906 
  907 /*
  908  * Return the first tag associated with an mbuf.
  909  */
  910 static __inline struct m_tag *
  911 m_tag_first(struct mbuf *m)
  912 {
  913 
  914         return (SLIST_FIRST(&m->m_pkthdr.tags));
  915 }
  916 
  917 /*
  918  * Return the next tag in the list of tags associated with an mbuf.
  919  */
  920 static __inline struct m_tag *
  921 m_tag_next(struct mbuf *m, struct m_tag *t)
  922 {
  923 
  924         return (SLIST_NEXT(t, m_tag_link));
  925 }
  926 
  927 /*
  928  * Prepend a tag to the list of tags associated with an mbuf.
  929  */
  930 static __inline void
  931 m_tag_prepend(struct mbuf *m, struct m_tag *t)
  932 {
  933 
  934         SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
  935 }
  936 
  937 /*
  938  * Unlink a tag from the list of tags associated with an mbuf.
  939  */
  940 static __inline void
  941 m_tag_unlink(struct mbuf *m, struct m_tag *t)
  942 {
  943 
  944         SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
  945 }
  946 
  947 /* These are for OpenBSD compatibility. */
  948 #define MTAG_ABI_COMPAT         0               /* compatibility ABI */
  949 
  950 static __inline struct m_tag *
  951 m_tag_get(int type, int length, int wait)
  952 {
  953         return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
  954 }
  955 
  956 static __inline struct m_tag *
  957 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
  958 {
  959         return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
  960             m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
  961 }
  962 
  963 /* XXX temporary FIB methods probably eventually use tags.*/
  964 #define M_FIBSHIFT    28
  965 #define M_FIBMASK       0x0F
  966 
  967 /* get the fib from an mbuf and if it is not set, return the default */
  968 #define M_GETFIB(_m) \
  969     ((((_m)->m_flags & M_FIB) >> M_FIBSHIFT) & M_FIBMASK)
  970 
  971 #define M_SETFIB(_m, _fib) do {                                         \
  972         _m->m_flags &= ~M_FIB;                                          \
  973         _m->m_flags |= (((_fib) << M_FIBSHIFT) & M_FIB);  \
  974 } while (0) 
  975 
  976 #endif /* _KERNEL */
  977 
  978 #endif /* !_SYS_MBUF_H_ */

Cache object: 184d8e636f8c1125aa058a3a84f93c87


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.