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  -  FREEBSD10  -  FREEBSD9  -  FREEBSD92  -  FREEBSD91  -  FREEBSD90  -  FREEBSD8  -  FREEBSD82  -  FREEBSD81  -  FREEBSD80  -  FREEBSD7  -  FREEBSD74  -  FREEBSD73  -  FREEBSD72  -  FREEBSD71  -  FREEBSD70  -  FREEBSD6  -  FREEBSD64  -  FREEBSD63  -  FREEBSD62  -  FREEBSD61  -  FREEBSD60  -  FREEBSD5  -  FREEBSD55  -  FREEBSD54  -  FREEBSD53  -  FREEBSD52  -  FREEBSD51  -  FREEBSD50  -  FREEBSD4  -  FREEBSD3  -  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: head/sys/sys/mbuf.h 279213 2015-02-23 18:57:09Z glebius $
   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  * NB: These calculation do not take actual compiler-induced alignment and
   57  * padding inside the complete struct mbuf into account.  Appropriate
   58  * attention is required when changing members of struct mbuf.
   59  *
   60  * MLEN is data length in a normal mbuf.
   61  * MHLEN is data length in an mbuf with pktheader.
   62  * MINCLSIZE is a smallest amount of data that should be put into cluster.
   63  *
   64  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
   65  * they are sensible.
   66  */
   67 struct mbuf;
   68 #define MHSIZE          offsetof(struct mbuf, m_dat)
   69 #define MPKTHSIZE       offsetof(struct mbuf, m_pktdat)
   70 #define MLEN            ((int)(MSIZE - MHSIZE))
   71 #define MHLEN           ((int)(MSIZE - MPKTHSIZE))
   72 #define MINCLSIZE       (MHLEN + 1)
   73 
   74 #ifdef _KERNEL
   75 /*-
   76  * Macro for type conversion: convert mbuf pointer to data pointer of correct
   77  * type:
   78  *
   79  * mtod(m, t)   -- Convert mbuf pointer to data pointer of correct type.
   80  * mtodo(m, o) -- Same as above but with offset 'o' into data.
   81  */
   82 #define mtod(m, t)      ((t)((m)->m_data))
   83 #define mtodo(m, o)     ((void *)(((m)->m_data) + (o)))
   84 
   85 /*
   86  * Argument structure passed to UMA routines during mbuf and packet
   87  * allocations.
   88  */
   89 struct mb_args {
   90         int     flags;  /* Flags for mbuf being allocated */
   91         short   type;   /* Type of mbuf being allocated */
   92 };
   93 #endif /* _KERNEL */
   94 
   95 /*
   96  * Packet tag structure (see below for details).
   97  */
   98 struct m_tag {
   99         SLIST_ENTRY(m_tag)      m_tag_link;     /* List of packet tags */
  100         u_int16_t               m_tag_id;       /* Tag ID */
  101         u_int16_t               m_tag_len;      /* Length of data */
  102         u_int32_t               m_tag_cookie;   /* ABI/Module ID */
  103         void                    (*m_tag_free)(struct m_tag *);
  104 };
  105 
  106 /*
  107  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
  108  * Size ILP32: 48
  109  *       LP64: 56
  110  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
  111  * they are correct.
  112  */
  113 struct pkthdr {
  114         struct ifnet    *rcvif;         /* rcv interface */
  115         SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
  116         int32_t          len;           /* total packet length */
  117 
  118         /* Layer crossing persistent information. */
  119         uint32_t         flowid;        /* packet's 4-tuple system */
  120         uint64_t         csum_flags;    /* checksum and offload features */
  121         uint16_t         fibnum;        /* this packet should use this fib */
  122         uint8_t          cosqos;        /* class/quality of service */
  123         uint8_t          rsstype;       /* hash type */
  124         uint8_t          l2hlen;        /* layer 2 header length */
  125         uint8_t          l3hlen;        /* layer 3 header length */
  126         uint8_t          l4hlen;        /* layer 4 header length */
  127         uint8_t          l5hlen;        /* layer 5 header length */
  128         union {
  129                 uint8_t  eight[8];
  130                 uint16_t sixteen[4];
  131                 uint32_t thirtytwo[2];
  132                 uint64_t sixtyfour[1];
  133                 uintptr_t unintptr[1];
  134                 void    *ptr;
  135         } PH_per;
  136 
  137         /* Layer specific non-persistent local storage for reassembly, etc. */
  138         union {
  139                 uint8_t  eight[8];
  140                 uint16_t sixteen[4];
  141                 uint32_t thirtytwo[2];
  142                 uint64_t sixtyfour[1];
  143                 uintptr_t unintptr[1];
  144                 void    *ptr;
  145         } PH_loc;
  146 };
  147 #define ether_vtag      PH_per.sixteen[0]
  148 #define PH_vt           PH_per
  149 #define vt_nrecs        sixteen[0]
  150 #define tso_segsz       PH_per.sixteen[1]
  151 #define csum_phsum      PH_per.sixteen[2]
  152 #define csum_data       PH_per.thirtytwo[1]
  153 #define pkt_tcphdr      PH_loc.ptr
  154 
  155 /*
  156  * Description of external storage mapped into mbuf; valid only if M_EXT is
  157  * set.
  158  * Size ILP32: 28
  159  *       LP64: 48
  160  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
  161  * they are correct.
  162  */
  163 struct m_ext {
  164         volatile u_int  *ext_cnt;       /* pointer to ref count info */
  165         caddr_t          ext_buf;       /* start of buffer */
  166         uint32_t         ext_size;      /* size of buffer, for ext_free */
  167         uint32_t         ext_type:8,    /* type of external storage */
  168                          ext_flags:24;  /* external storage mbuf flags */
  169         void            (*ext_free)     /* free routine if not the usual */
  170                             (struct mbuf *, void *, void *);
  171         void            *ext_arg1;      /* optional argument pointer */
  172         void            *ext_arg2;      /* optional argument pointer */
  173 };
  174 
  175 /*
  176  * The core of the mbuf object along with some shortcut defines for practical
  177  * purposes.
  178  */
  179 struct mbuf {
  180         /*
  181          * Header present at the beginning of every mbuf.
  182          * Size ILP32: 24
  183          *      LP64: 32
  184          * Compile-time assertions in uipc_mbuf.c test these values to ensure
  185          * that they are correct.
  186          */
  187         union { /* next buffer in chain */
  188                 struct mbuf             *m_next;
  189                 SLIST_ENTRY(mbuf)       m_slist;
  190                 STAILQ_ENTRY(mbuf)      m_stailq;
  191         };
  192         union { /* next chain in queue/record */
  193                 struct mbuf             *m_nextpkt;
  194                 SLIST_ENTRY(mbuf)       m_slistpkt;
  195                 STAILQ_ENTRY(mbuf)      m_stailqpkt;
  196         };
  197         caddr_t          m_data;        /* location of data */
  198         int32_t          m_len;         /* amount of data in this mbuf */
  199         uint32_t         m_type:8,      /* type of data in this mbuf */
  200                          m_flags:24;    /* flags; see below */
  201 #if !defined(__LP64__)
  202         uint32_t         m_pad;         /* pad for 64bit alignment */
  203 #endif
  204 
  205         /*
  206          * A set of optional headers (packet header, external storage header)
  207          * and internal data storage.  Historically, these arrays were sized
  208          * to MHLEN (space left after a packet header) and MLEN (space left
  209          * after only a regular mbuf header); they are now variable size in
  210          * order to support future work on variable-size mbufs.
  211          */
  212         union {
  213                 struct {
  214                         struct pkthdr   m_pkthdr;       /* M_PKTHDR set */
  215                         union {
  216                                 struct m_ext    m_ext;  /* M_EXT set */
  217                                 char            m_pktdat[0];
  218                         };
  219                 };
  220                 char    m_dat[0];                       /* !M_PKTHDR, !M_EXT */
  221         };
  222 };
  223 
  224 /*
  225  * mbuf flags of global significance and layer crossing.
  226  * Those of only protocol/layer specific significance are to be mapped
  227  * to M_PROTO[1-12] and cleared at layer handoff boundaries.
  228  * NB: Limited to the lower 24 bits.
  229  */
  230 #define M_EXT           0x00000001 /* has associated external storage */
  231 #define M_PKTHDR        0x00000002 /* start of record */
  232 #define M_EOR           0x00000004 /* end of record */
  233 #define M_RDONLY        0x00000008 /* associated data is marked read-only */
  234 #define M_BCAST         0x00000010 /* send/received as link-level broadcast */
  235 #define M_MCAST         0x00000020 /* send/received as link-level multicast */
  236 #define M_PROMISC       0x00000040 /* packet was not for us */
  237 #define M_VLANTAG       0x00000080 /* ether_vtag is valid */
  238 #define M_UNUSED_8      0x00000100 /* --available-- */
  239 #define M_NOFREE        0x00000200 /* do not free mbuf, embedded in cluster */
  240 
  241 #define M_PROTO1        0x00001000 /* protocol-specific */
  242 #define M_PROTO2        0x00002000 /* protocol-specific */
  243 #define M_PROTO3        0x00004000 /* protocol-specific */
  244 #define M_PROTO4        0x00008000 /* protocol-specific */
  245 #define M_PROTO5        0x00010000 /* protocol-specific */
  246 #define M_PROTO6        0x00020000 /* protocol-specific */
  247 #define M_PROTO7        0x00040000 /* protocol-specific */
  248 #define M_PROTO8        0x00080000 /* protocol-specific */
  249 #define M_PROTO9        0x00100000 /* protocol-specific */
  250 #define M_PROTO10       0x00200000 /* protocol-specific */
  251 #define M_PROTO11       0x00400000 /* protocol-specific */
  252 #define M_PROTO12       0x00800000 /* protocol-specific */
  253 
  254 /*
  255  * Flags to purge when crossing layers.
  256  */
  257 #define M_PROTOFLAGS \
  258     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
  259      M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
  260 
  261 /*
  262  * Flags preserved when copying m_pkthdr.
  263  */
  264 #define M_COPYFLAGS \
  265     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG| \
  266      M_PROTOFLAGS)
  267 
  268 /*
  269  * Mbuf flag description for use with printf(9) %b identifier.
  270  */
  271 #define M_FLAG_BITS \
  272     "\2\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
  273     "\7M_PROMISC\10M_VLANTAG"
  274 #define M_FLAG_PROTOBITS \
  275     "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
  276     "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
  277     "\27M_PROTO11\30M_PROTO12"
  278 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
  279 
  280 /*
  281  * Network interface cards are able to hash protocol fields (such as IPv4
  282  * addresses and TCP port numbers) classify packets into flows.  These flows
  283  * can then be used to maintain ordering while delivering packets to the OS
  284  * via parallel input queues, as well as to provide a stateless affinity
  285  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
  286  * m_flag fields to indicate how the hash should be interpreted by the
  287  * network stack.
  288  *
  289  * Most NICs support RSS, which provides ordering and explicit affinity, and
  290  * use the hash m_flag bits to indicate what header fields were covered by
  291  * the hash.  M_HASHTYPE_OPAQUE can be set by non-RSS cards or configurations
  292  * that provide an opaque flow identifier, allowing for ordering and
  293  * distribution without explicit affinity.
  294  */
  295 /* Microsoft RSS standard hash types */
  296 #define M_HASHTYPE_NONE                 0
  297 #define M_HASHTYPE_RSS_IPV4             1       /* IPv4 2-tuple */
  298 #define M_HASHTYPE_RSS_TCP_IPV4         2       /* TCPv4 4-tuple */
  299 #define M_HASHTYPE_RSS_IPV6             3       /* IPv6 2-tuple */
  300 #define M_HASHTYPE_RSS_TCP_IPV6         4       /* TCPv6 4-tuple */
  301 #define M_HASHTYPE_RSS_IPV6_EX          5       /* IPv6 2-tuple + ext hdrs */
  302 #define M_HASHTYPE_RSS_TCP_IPV6_EX      6       /* TCPv6 4-tiple + ext hdrs */
  303 /* Non-standard RSS hash types */
  304 #define M_HASHTYPE_RSS_UDP_IPV4         7       /* IPv4 UDP 4-tuple */
  305 #define M_HASHTYPE_RSS_UDP_IPV4_EX      8       /* IPv4 UDP 4-tuple + ext hdrs */
  306 #define M_HASHTYPE_RSS_UDP_IPV6         9       /* IPv6 UDP 4-tuple */
  307 #define M_HASHTYPE_RSS_UDP_IPV6_EX      10      /* IPv6 UDP 4-tuple + ext hdrs */
  308 
  309 #define M_HASHTYPE_OPAQUE               255     /* ordering, not affinity */
  310 
  311 #define M_HASHTYPE_CLEAR(m)     ((m)->m_pkthdr.rsstype = 0)
  312 #define M_HASHTYPE_GET(m)       ((m)->m_pkthdr.rsstype)
  313 #define M_HASHTYPE_SET(m, v)    ((m)->m_pkthdr.rsstype = (v))
  314 #define M_HASHTYPE_TEST(m, v)   (M_HASHTYPE_GET(m) == (v))
  315 
  316 /*
  317  * COS/QOS class and quality of service tags.
  318  * It uses DSCP code points as base.
  319  */
  320 #define QOS_DSCP_CS0            0x00
  321 #define QOS_DSCP_DEF            QOS_DSCP_CS0
  322 #define QOS_DSCP_CS1            0x20
  323 #define QOS_DSCP_AF11           0x28
  324 #define QOS_DSCP_AF12           0x30
  325 #define QOS_DSCP_AF13           0x38
  326 #define QOS_DSCP_CS2            0x40
  327 #define QOS_DSCP_AF21           0x48
  328 #define QOS_DSCP_AF22           0x50
  329 #define QOS_DSCP_AF23           0x58
  330 #define QOS_DSCP_CS3            0x60
  331 #define QOS_DSCP_AF31           0x68
  332 #define QOS_DSCP_AF32           0x70
  333 #define QOS_DSCP_AF33           0x78
  334 #define QOS_DSCP_CS4            0x80
  335 #define QOS_DSCP_AF41           0x88
  336 #define QOS_DSCP_AF42           0x90
  337 #define QOS_DSCP_AF43           0x98
  338 #define QOS_DSCP_CS5            0xa0
  339 #define QOS_DSCP_EF             0xb8
  340 #define QOS_DSCP_CS6            0xc0
  341 #define QOS_DSCP_CS7            0xe0
  342 
  343 /*
  344  * External mbuf storage buffer types.
  345  */
  346 #define EXT_CLUSTER     1       /* mbuf cluster */
  347 #define EXT_SFBUF       2       /* sendfile(2)'s sf_bufs */
  348 #define EXT_JUMBOP      3       /* jumbo cluster 4096 bytes */
  349 #define EXT_JUMBO9      4       /* jumbo cluster 9216 bytes */
  350 #define EXT_JUMBO16     5       /* jumbo cluster 16184 bytes */
  351 #define EXT_PACKET      6       /* mbuf+cluster from packet zone */
  352 #define EXT_MBUF        7       /* external mbuf reference (M_IOVEC) */
  353 
  354 #define EXT_VENDOR1     224     /* for vendor-internal use */
  355 #define EXT_VENDOR2     225     /* for vendor-internal use */
  356 #define EXT_VENDOR3     226     /* for vendor-internal use */
  357 #define EXT_VENDOR4     227     /* for vendor-internal use */
  358 
  359 #define EXT_EXP1        244     /* for experimental use */
  360 #define EXT_EXP2        245     /* for experimental use */
  361 #define EXT_EXP3        246     /* for experimental use */
  362 #define EXT_EXP4        247     /* for experimental use */
  363 
  364 #define EXT_NET_DRV     252     /* custom ext_buf provided by net driver(s) */
  365 #define EXT_MOD_TYPE    253     /* custom module's ext_buf type */
  366 #define EXT_DISPOSABLE  254     /* can throw this buffer away w/page flipping */
  367 #define EXT_EXTREF      255     /* has externally maintained ext_cnt ptr */
  368 
  369 /*
  370  * Flags for external mbuf buffer types.
  371  * NB: limited to the lower 24 bits.
  372  */
  373 #define EXT_FLAG_EMBREF         0x000001        /* embedded ext_cnt, notyet */
  374 #define EXT_FLAG_EXTREF         0x000002        /* external ext_cnt, notyet */
  375 #define EXT_FLAG_NOFREE         0x000010        /* don't free mbuf to pool, notyet */
  376 
  377 #define EXT_FLAG_VENDOR1        0x010000        /* for vendor-internal use */
  378 #define EXT_FLAG_VENDOR2        0x020000        /* for vendor-internal use */
  379 #define EXT_FLAG_VENDOR3        0x040000        /* for vendor-internal use */
  380 #define EXT_FLAG_VENDOR4        0x080000        /* for vendor-internal use */
  381 
  382 #define EXT_FLAG_EXP1           0x100000        /* for experimental use */
  383 #define EXT_FLAG_EXP2           0x200000        /* for experimental use */
  384 #define EXT_FLAG_EXP3           0x400000        /* for experimental use */
  385 #define EXT_FLAG_EXP4           0x800000        /* for experimental use */
  386 
  387 /*
  388  * EXT flag description for use with printf(9) %b identifier.
  389  */
  390 #define EXT_FLAG_BITS \
  391     "\2\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
  392     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
  393     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
  394     "\30EXT_FLAG_EXP4"
  395 
  396 /*
  397  * External reference/free functions.
  398  */
  399 void sf_ext_ref(void *, void *);
  400 void sf_ext_free(void *, void *);
  401 
  402 /*
  403  * Flags indicating checksum, segmentation and other offload work to be
  404  * done, or already done, by hardware or lower layers.  It is split into
  405  * separate inbound and outbound flags.
  406  *
  407  * Outbound flags that are set by upper protocol layers requesting lower
  408  * layers, or ideally the hardware, to perform these offloading tasks.
  409  * For outbound packets this field and its flags can be directly tested
  410  * against ifnet if_hwassist.
  411  */
  412 #define CSUM_IP                 0x00000001      /* IP header checksum offload */
  413 #define CSUM_IP_UDP             0x00000002      /* UDP checksum offload */
  414 #define CSUM_IP_TCP             0x00000004      /* TCP checksum offload */
  415 #define CSUM_IP_SCTP            0x00000008      /* SCTP checksum offload */
  416 #define CSUM_IP_TSO             0x00000010      /* TCP segmentation offload */
  417 #define CSUM_IP_ISCSI           0x00000020      /* iSCSI checksum offload */
  418 
  419 #define CSUM_IP6_UDP            0x00000200      /* UDP checksum offload */
  420 #define CSUM_IP6_TCP            0x00000400      /* TCP checksum offload */
  421 #define CSUM_IP6_SCTP           0x00000800      /* SCTP checksum offload */
  422 #define CSUM_IP6_TSO            0x00001000      /* TCP segmentation offload */
  423 #define CSUM_IP6_ISCSI          0x00002000      /* iSCSI checksum offload */
  424 
  425 /* Inbound checksum support where the checksum was verified by hardware. */
  426 #define CSUM_L3_CALC            0x01000000      /* calculated layer 3 csum */
  427 #define CSUM_L3_VALID           0x02000000      /* checksum is correct */
  428 #define CSUM_L4_CALC            0x04000000      /* calculated layer 4 csum */
  429 #define CSUM_L4_VALID           0x08000000      /* checksum is correct */
  430 #define CSUM_L5_CALC            0x10000000      /* calculated layer 5 csum */
  431 #define CSUM_L5_VALID           0x20000000      /* checksum is correct */
  432 #define CSUM_COALESED           0x40000000      /* contains merged segments */
  433 
  434 /*
  435  * CSUM flag description for use with printf(9) %b identifier.
  436  */
  437 #define CSUM_BITS \
  438     "\2\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
  439     "\6CSUM_IP_ISCSI" \
  440     "\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP\15CSUM_IP6_TSO" \
  441     "\16CSUM_IP6_ISCSI" \
  442     "\31CSUM_L3_CALC\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID" \
  443     "\35CSUM_L5_CALC\36CSUM_L5_VALID\37CSUM_COALESED"
  444 
  445 /* CSUM flags compatibility mappings. */
  446 #define CSUM_IP_CHECKED         CSUM_L3_CALC
  447 #define CSUM_IP_VALID           CSUM_L3_VALID
  448 #define CSUM_DATA_VALID         CSUM_L4_VALID
  449 #define CSUM_PSEUDO_HDR         CSUM_L4_CALC
  450 #define CSUM_SCTP_VALID         CSUM_L4_VALID
  451 #define CSUM_DELAY_DATA         (CSUM_TCP|CSUM_UDP)
  452 #define CSUM_DELAY_IP           CSUM_IP         /* Only v4, no v6 IP hdr csum */
  453 #define CSUM_DELAY_DATA_IPV6    (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
  454 #define CSUM_DATA_VALID_IPV6    CSUM_DATA_VALID
  455 #define CSUM_TCP                CSUM_IP_TCP
  456 #define CSUM_UDP                CSUM_IP_UDP
  457 #define CSUM_SCTP               CSUM_IP_SCTP
  458 #define CSUM_TSO                (CSUM_IP_TSO|CSUM_IP6_TSO)
  459 #define CSUM_UDP_IPV6           CSUM_IP6_UDP
  460 #define CSUM_TCP_IPV6           CSUM_IP6_TCP
  461 #define CSUM_SCTP_IPV6          CSUM_IP6_SCTP
  462 
  463 /*
  464  * mbuf types describing the content of the mbuf (including external storage).
  465  */
  466 #define MT_NOTMBUF      0       /* USED INTERNALLY ONLY! Object is not mbuf */
  467 #define MT_DATA         1       /* dynamic (data) allocation */
  468 #define MT_HEADER       MT_DATA /* packet header, use M_PKTHDR instead */
  469 
  470 #define MT_VENDOR1      4       /* for vendor-internal use */
  471 #define MT_VENDOR2      5       /* for vendor-internal use */
  472 #define MT_VENDOR3      6       /* for vendor-internal use */
  473 #define MT_VENDOR4      7       /* for vendor-internal use */
  474 
  475 #define MT_SONAME       8       /* socket name */
  476 
  477 #define MT_EXP1         9       /* for experimental use */
  478 #define MT_EXP2         10      /* for experimental use */
  479 #define MT_EXP3         11      /* for experimental use */
  480 #define MT_EXP4         12      /* for experimental use */
  481 
  482 #define MT_CONTROL      14      /* extra-data protocol message */
  483 #define MT_OOBDATA      15      /* expedited data  */
  484 #define MT_NTYPES       16      /* number of mbuf types for mbtypes[] */
  485 
  486 #define MT_NOINIT       255     /* Not a type but a flag to allocate
  487                                    a non-initialized mbuf */
  488 
  489 /*
  490  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
  491  * !_KERNEL so that monitoring tools can look up the zones with
  492  * libmemstat(3).
  493  */
  494 #define MBUF_MEM_NAME           "mbuf"
  495 #define MBUF_CLUSTER_MEM_NAME   "mbuf_cluster"
  496 #define MBUF_PACKET_MEM_NAME    "mbuf_packet"
  497 #define MBUF_JUMBOP_MEM_NAME    "mbuf_jumbo_page"
  498 #define MBUF_JUMBO9_MEM_NAME    "mbuf_jumbo_9k"
  499 #define MBUF_JUMBO16_MEM_NAME   "mbuf_jumbo_16k"
  500 #define MBUF_TAG_MEM_NAME       "mbuf_tag"
  501 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
  502 
  503 #ifdef _KERNEL
  504 
  505 #ifdef WITNESS
  506 #define MBUF_CHECKSLEEP(how) do {                                       \
  507         if (how == M_WAITOK)                                            \
  508                 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,         \
  509                     "Sleeping in \"%s\"", __func__);                    \
  510 } while (0)
  511 #else
  512 #define MBUF_CHECKSLEEP(how)
  513 #endif
  514 
  515 /*
  516  * Network buffer allocation API
  517  *
  518  * The rest of it is defined in kern/kern_mbuf.c
  519  */
  520 extern uma_zone_t       zone_mbuf;
  521 extern uma_zone_t       zone_clust;
  522 extern uma_zone_t       zone_pack;
  523 extern uma_zone_t       zone_jumbop;
  524 extern uma_zone_t       zone_jumbo9;
  525 extern uma_zone_t       zone_jumbo16;
  526 extern uma_zone_t       zone_ext_refcnt;
  527 
  528 void             mb_free_ext(struct mbuf *);
  529 int              m_pkthdr_init(struct mbuf *, int);
  530 
  531 static __inline int
  532 m_gettype(int size)
  533 {
  534         int type;
  535 
  536         switch (size) {
  537         case MSIZE:
  538                 type = EXT_MBUF;
  539                 break;
  540         case MCLBYTES:
  541                 type = EXT_CLUSTER;
  542                 break;
  543 #if MJUMPAGESIZE != MCLBYTES
  544         case MJUMPAGESIZE:
  545                 type = EXT_JUMBOP;
  546                 break;
  547 #endif
  548         case MJUM9BYTES:
  549                 type = EXT_JUMBO9;
  550                 break;
  551         case MJUM16BYTES:
  552                 type = EXT_JUMBO16;
  553                 break;
  554         default:
  555                 panic("%s: invalid cluster size %d", __func__, size);
  556         }
  557 
  558         return (type);
  559 }
  560 
  561 /*
  562  * Associated an external reference counted buffer with an mbuf.
  563  */
  564 static __inline void
  565 m_extaddref(struct mbuf *m, caddr_t buf, u_int size, u_int *ref_cnt,
  566     void (*freef)(struct mbuf *, void *, void *), void *arg1, void *arg2)
  567 {
  568 
  569         KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
  570 
  571         atomic_add_int(ref_cnt, 1);
  572         m->m_flags |= M_EXT;
  573         m->m_ext.ext_buf = buf;
  574         m->m_ext.ext_cnt = ref_cnt;
  575         m->m_data = m->m_ext.ext_buf;
  576         m->m_ext.ext_size = size;
  577         m->m_ext.ext_free = freef;
  578         m->m_ext.ext_arg1 = arg1;
  579         m->m_ext.ext_arg2 = arg2;
  580         m->m_ext.ext_type = EXT_EXTREF;
  581 }
  582 
  583 static __inline uma_zone_t
  584 m_getzone(int size)
  585 {
  586         uma_zone_t zone;
  587 
  588         switch (size) {
  589         case MCLBYTES:
  590                 zone = zone_clust;
  591                 break;
  592 #if MJUMPAGESIZE != MCLBYTES
  593         case MJUMPAGESIZE:
  594                 zone = zone_jumbop;
  595                 break;
  596 #endif
  597         case MJUM9BYTES:
  598                 zone = zone_jumbo9;
  599                 break;
  600         case MJUM16BYTES:
  601                 zone = zone_jumbo16;
  602                 break;
  603         default:
  604                 panic("%s: invalid cluster size %d", __func__, size);
  605         }
  606 
  607         return (zone);
  608 }
  609 
  610 /*
  611  * Initialize an mbuf with linear storage.
  612  *
  613  * Inline because the consumer text overhead will be roughly the same to
  614  * initialize or call a function with this many parameters and M_PKTHDR
  615  * should go away with constant propagation for !MGETHDR.
  616  */
  617 static __inline int
  618 m_init(struct mbuf *m, uma_zone_t zone, int size, int how, short type,
  619     int flags)
  620 {
  621         int error;
  622 
  623         m->m_next = NULL;
  624         m->m_nextpkt = NULL;
  625         m->m_data = m->m_dat;
  626         m->m_len = 0;
  627         m->m_flags = flags;
  628         m->m_type = type;
  629         if (flags & M_PKTHDR) {
  630                 if ((error = m_pkthdr_init(m, how)) != 0)
  631                         return (error);
  632         }
  633 
  634         return (0);
  635 }
  636 
  637 static __inline struct mbuf *
  638 m_get(int how, short type)
  639 {
  640         struct mb_args args;
  641 
  642         args.flags = 0;
  643         args.type = type;
  644         return (uma_zalloc_arg(zone_mbuf, &args, how));
  645 }
  646 
  647 /*
  648  * XXX This should be deprecated, very little use.
  649  */
  650 static __inline struct mbuf *
  651 m_getclr(int how, short type)
  652 {
  653         struct mbuf *m;
  654         struct mb_args args;
  655 
  656         args.flags = 0;
  657         args.type = type;
  658         m = uma_zalloc_arg(zone_mbuf, &args, how);
  659         if (m != NULL)
  660                 bzero(m->m_data, MLEN);
  661         return (m);
  662 }
  663 
  664 static __inline struct mbuf *
  665 m_gethdr(int how, short type)
  666 {
  667         struct mb_args args;
  668 
  669         args.flags = M_PKTHDR;
  670         args.type = type;
  671         return (uma_zalloc_arg(zone_mbuf, &args, how));
  672 }
  673 
  674 static __inline struct mbuf *
  675 m_getcl(int how, short type, int flags)
  676 {
  677         struct mb_args args;
  678 
  679         args.flags = flags;
  680         args.type = type;
  681         return (uma_zalloc_arg(zone_pack, &args, how));
  682 }
  683 
  684 static __inline int
  685 m_clget(struct mbuf *m, int how)
  686 {
  687 
  688         if (m->m_flags & M_EXT)
  689                 printf("%s: %p mbuf already has external storage\n", __func__, m);
  690         m->m_ext.ext_buf = (char *)NULL;
  691         uma_zalloc_arg(zone_clust, m, how);
  692         /*
  693          * On a cluster allocation failure, drain the packet zone and retry,
  694          * we might be able to loosen a few clusters up on the drain.
  695          */
  696         if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
  697                 zone_drain(zone_pack);
  698                 uma_zalloc_arg(zone_clust, m, how);
  699         }
  700         return (m->m_flags & M_EXT);
  701 }
  702 
  703 /*
  704  * m_cljget() is different from m_clget() as it can allocate clusters without
  705  * attaching them to an mbuf.  In that case the return value is the pointer
  706  * to the cluster of the requested size.  If an mbuf was specified, it gets
  707  * the cluster attached to it and the return value can be safely ignored.
  708  * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
  709  */
  710 static __inline void *
  711 m_cljget(struct mbuf *m, int how, int size)
  712 {
  713         uma_zone_t zone;
  714 
  715         if (m && m->m_flags & M_EXT)
  716                 printf("%s: %p mbuf already has external storage\n", __func__, m);
  717         if (m != NULL)
  718                 m->m_ext.ext_buf = NULL;
  719 
  720         zone = m_getzone(size);
  721         return (uma_zalloc_arg(zone, m, how));
  722 }
  723 
  724 static __inline void
  725 m_cljset(struct mbuf *m, void *cl, int type)
  726 {
  727         uma_zone_t zone;
  728         int size;
  729 
  730         switch (type) {
  731         case EXT_CLUSTER:
  732                 size = MCLBYTES;
  733                 zone = zone_clust;
  734                 break;
  735 #if MJUMPAGESIZE != MCLBYTES
  736         case EXT_JUMBOP:
  737                 size = MJUMPAGESIZE;
  738                 zone = zone_jumbop;
  739                 break;
  740 #endif
  741         case EXT_JUMBO9:
  742                 size = MJUM9BYTES;
  743                 zone = zone_jumbo9;
  744                 break;
  745         case EXT_JUMBO16:
  746                 size = MJUM16BYTES;
  747                 zone = zone_jumbo16;
  748                 break;
  749         default:
  750                 panic("%s: unknown cluster type %d", __func__, type);
  751                 break;
  752         }
  753 
  754         m->m_data = m->m_ext.ext_buf = cl;
  755         m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
  756         m->m_ext.ext_size = size;
  757         m->m_ext.ext_type = type;
  758         m->m_ext.ext_flags = 0;
  759         m->m_ext.ext_cnt = uma_find_refcnt(zone, cl);
  760         m->m_flags |= M_EXT;
  761 
  762 }
  763 
  764 static __inline void
  765 m_chtype(struct mbuf *m, short new_type)
  766 {
  767 
  768         m->m_type = new_type;
  769 }
  770 
  771 static __inline void
  772 m_clrprotoflags(struct mbuf *m)
  773 {
  774 
  775         while (m) {
  776                 m->m_flags &= ~M_PROTOFLAGS;
  777                 m = m->m_next;
  778         }
  779 }
  780 
  781 static __inline struct mbuf *
  782 m_last(struct mbuf *m)
  783 {
  784 
  785         while (m->m_next)
  786                 m = m->m_next;
  787         return (m);
  788 }
  789 
  790 /*
  791  * mbuf, cluster, and external object allocation macros (for compatibility
  792  * purposes).
  793  */
  794 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
  795 #define MGET(m, how, type)      ((m) = m_get((how), (type)))
  796 #define MGETHDR(m, how, type)   ((m) = m_gethdr((how), (type)))
  797 #define MCLGET(m, how)          m_clget((m), (how))
  798 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type)            \
  799     (void )m_extadd((m), (caddr_t)(buf), (size), (free), (arg1), (arg2),\
  800     (flags), (type), M_NOWAIT)
  801 #define m_getm(m, len, how, type)                                       \
  802     m_getm2((m), (len), (how), (type), M_PKTHDR)
  803 
  804 /*
  805  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
  806  * be both the local data payload, or an external buffer area, depending on
  807  * whether M_EXT is set).
  808  */
  809 #define M_WRITABLE(m)   (!((m)->m_flags & M_RDONLY) &&                  \
  810                          (!(((m)->m_flags & M_EXT)) ||                  \
  811                          (*((m)->m_ext.ext_cnt) == 1)) )                \
  812 
  813 /* Check if the supplied mbuf has a packet header, or else panic. */
  814 #define M_ASSERTPKTHDR(m)                                               \
  815         KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,                 \
  816             ("%s: no mbuf packet header!", __func__))
  817 
  818 /*
  819  * Ensure that the supplied mbuf is a valid, non-free mbuf.
  820  *
  821  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
  822  */
  823 #define M_ASSERTVALID(m)                                                \
  824         KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,                 \
  825             ("%s: attempted use of a free mbuf!", __func__))
  826 
  827 /*
  828  * Return the address of the start of the buffer associated with an mbuf,
  829  * handling external storage, packet-header mbufs, and regular data mbufs.
  830  */
  831 #define M_START(m)                                                      \
  832         (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :                  \
  833          ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :                \
  834          &(m)->m_dat[0])
  835 
  836 /*
  837  * Return the size of the buffer associated with an mbuf, handling external
  838  * storage, packet-header mbufs, and regular data mbufs.
  839  */
  840 #define M_SIZE(m)                                                       \
  841         (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :                 \
  842          ((m)->m_flags & M_PKTHDR) ? MHLEN :                            \
  843          MLEN)
  844 
  845 /*
  846  * Set the m_data pointer of a newly allocated mbuf to place an object of the
  847  * specified size at the end of the mbuf, longword aligned.
  848  *
  849  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
  850  * separate macros, each asserting that it was called at the proper moment.
  851  * This required callers to themselves test the storage type and call the
  852  * right one.  Rather than require callers to be aware of those layout
  853  * decisions, we centralize here.
  854  */
  855 static __inline void
  856 m_align(struct mbuf *m, int len)
  857 {
  858 #ifdef INVARIANTS
  859         const char *msg = "%s: not a virgin mbuf";
  860 #endif
  861         int adjust;
  862 
  863         KASSERT(m->m_data == M_START(m), (msg, __func__));
  864 
  865         adjust = M_SIZE(m) - len;
  866         m->m_data += adjust &~ (sizeof(long)-1);
  867 }
  868 
  869 #define M_ALIGN(m, len)         m_align(m, len)
  870 #define MH_ALIGN(m, len)        m_align(m, len)
  871 #define MEXT_ALIGN(m, len)      m_align(m, len)
  872 
  873 /*
  874  * Compute the amount of space available before the current start of data in
  875  * an mbuf.
  876  *
  877  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
  878  * of checking writability of the mbuf data area rests solely with the caller.
  879  *
  880  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
  881  * for mbufs with external storage.  We now allow mbuf-embedded data to be
  882  * read-only as well.
  883  */
  884 #define M_LEADINGSPACE(m)                                               \
  885         (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
  886 
  887 /*
  888  * Compute the amount of space available after the end of data in an mbuf.
  889  *
  890  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
  891  * of checking writability of the mbuf data area rests solely with the caller.
  892  *
  893  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
  894  * for mbufs with external storage.  We now allow mbuf-embedded data to be
  895  * read-only as well.
  896  */
  897 #define M_TRAILINGSPACE(m)                                              \
  898         (M_WRITABLE(m) ?                                                \
  899             ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
  900 
  901 /*
  902  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
  903  * allocated, how specifies whether to wait.  If the allocation fails, the
  904  * original mbuf chain is freed and m is set to NULL.
  905  */
  906 #define M_PREPEND(m, plen, how) do {                                    \
  907         struct mbuf **_mmp = &(m);                                      \
  908         struct mbuf *_mm = *_mmp;                                       \
  909         int _mplen = (plen);                                            \
  910         int __mhow = (how);                                             \
  911                                                                         \
  912         MBUF_CHECKSLEEP(how);                                           \
  913         if (M_LEADINGSPACE(_mm) >= _mplen) {                            \
  914                 _mm->m_data -= _mplen;                                  \
  915                 _mm->m_len += _mplen;                                   \
  916         } else                                                          \
  917                 _mm = m_prepend(_mm, _mplen, __mhow);                   \
  918         if (_mm != NULL && _mm->m_flags & M_PKTHDR)                     \
  919                 _mm->m_pkthdr.len += _mplen;                            \
  920         *_mmp = _mm;                                                    \
  921 } while (0)
  922 
  923 /*
  924  * Change mbuf to new type.  This is a relatively expensive operation and
  925  * should be avoided.
  926  */
  927 #define MCHTYPE(m, t)   m_chtype((m), (t))
  928 
  929 /* Length to m_copy to copy all. */
  930 #define M_COPYALL       1000000000
  931 
  932 /* Compatibility with 4.3. */
  933 #define m_copy(m, o, l) m_copym((m), (o), (l), M_NOWAIT)
  934 
  935 extern int              max_datalen;    /* MHLEN - max_hdr */
  936 extern int              max_hdr;        /* Largest link + protocol header */
  937 extern int              max_linkhdr;    /* Largest link-level header */
  938 extern int              max_protohdr;   /* Largest protocol header */
  939 extern int              nmbclusters;    /* Maximum number of clusters */
  940 
  941 struct uio;
  942 
  943 void             m_adj(struct mbuf *, int);
  944 int              m_apply(struct mbuf *, int, int,
  945                     int (*)(void *, void *, u_int), void *);
  946 int              m_append(struct mbuf *, int, c_caddr_t);
  947 void             m_cat(struct mbuf *, struct mbuf *);
  948 void             m_catpkt(struct mbuf *, struct mbuf *);
  949 int              m_extadd(struct mbuf *, caddr_t, u_int,
  950                     void (*)(struct mbuf *, void *, void *), void *, void *,
  951                     int, int, int);
  952 struct mbuf     *m_collapse(struct mbuf *, int, int);
  953 void             m_copyback(struct mbuf *, int, int, c_caddr_t);
  954 void             m_copydata(const struct mbuf *, int, int, caddr_t);
  955 struct mbuf     *m_copym(struct mbuf *, int, int, int);
  956 struct mbuf     *m_copypacket(struct mbuf *, int);
  957 void             m_copy_pkthdr(struct mbuf *, struct mbuf *);
  958 struct mbuf     *m_copyup(struct mbuf *, int, int);
  959 struct mbuf     *m_defrag(struct mbuf *, int);
  960 void             m_demote(struct mbuf *, int, int);
  961 struct mbuf     *m_devget(char *, int, int, struct ifnet *,
  962                     void (*)(char *, caddr_t, u_int));
  963 struct mbuf     *m_dup(struct mbuf *, int);
  964 int              m_dup_pkthdr(struct mbuf *, struct mbuf *, int);
  965 u_int            m_fixhdr(struct mbuf *);
  966 struct mbuf     *m_fragment(struct mbuf *, int, int);
  967 void             m_freem(struct mbuf *);
  968 struct mbuf     *m_get2(int, int, short, int);
  969 struct mbuf     *m_getjcl(int, short, int, int);
  970 struct mbuf     *m_getm2(struct mbuf *, int, int, short, int);
  971 struct mbuf     *m_getptr(struct mbuf *, int, int *);
  972 u_int            m_length(struct mbuf *, struct mbuf **);
  973 int              m_mbuftouio(struct uio *, struct mbuf *, int);
  974 void             m_move_pkthdr(struct mbuf *, struct mbuf *);
  975 struct mbuf     *m_prepend(struct mbuf *, int, int);
  976 void             m_print(const struct mbuf *, int);
  977 struct mbuf     *m_pulldown(struct mbuf *, int, int, int *);
  978 struct mbuf     *m_pullup(struct mbuf *, int);
  979 int              m_sanity(struct mbuf *, int);
  980 struct mbuf     *m_split(struct mbuf *, int, int);
  981 struct mbuf     *m_uiotombuf(struct uio *, int, int, int, int);
  982 struct mbuf     *m_unshare(struct mbuf *, int);
  983 
  984 /*-
  985  * Network packets may have annotations attached by affixing a list of
  986  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
  987  * allocated semi-opaque data structures that have a fixed header
  988  * (struct m_tag) that specifies the size of the memory block and a
  989  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
  990  * unsigned value used to identify a module or ABI.  By convention this value
  991  * is chosen as the date+time that the module is created, expressed as the
  992  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
  993  * value is an ABI/module-specific value that identifies a particular
  994  * annotation and is private to the module.  For compatibility with systems
  995  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
  996  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
  997  * compatibility shim functions and several tag types are defined below.
  998  * Users that do not require compatibility should use a private cookie value
  999  * so that packet tag-related definitions can be maintained privately.
 1000  *
 1001  * Note that the packet tag returned by m_tag_alloc has the default memory
 1002  * alignment implemented by malloc.  To reference private data one can use a
 1003  * construct like:
 1004  *
 1005  *      struct m_tag *mtag = m_tag_alloc(...);
 1006  *      struct foo *p = (struct foo *)(mtag+1);
 1007  *
 1008  * if the alignment of struct m_tag is sufficient for referencing members of
 1009  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
 1010  * private data structure to insure proper alignment; e.g.,
 1011  *
 1012  *      struct foo {
 1013  *              struct m_tag    tag;
 1014  *              ...
 1015  *      };
 1016  *      struct foo *p = (struct foo *) m_tag_alloc(...);
 1017  *      struct m_tag *mtag = &p->tag;
 1018  */
 1019 
 1020 /*
 1021  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
 1022  * tags are expected to ``vanish'' when they pass through a network
 1023  * interface.  For most interfaces this happens normally as the tags are
 1024  * reclaimed when the mbuf is free'd.  However in some special cases
 1025  * reclaiming must be done manually.  An example is packets that pass through
 1026  * the loopback interface.  Also, one must be careful to do this when
 1027  * ``turning around'' packets (e.g., icmp_reflect).
 1028  *
 1029  * To mark a tag persistent bit-or this flag in when defining the tag id.
 1030  * The tag will then be treated as described above.
 1031  */
 1032 #define MTAG_PERSISTENT                         0x800
 1033 
 1034 #define PACKET_TAG_NONE                         0  /* Nadda */
 1035 
 1036 /* Packet tags for use with PACKET_ABI_COMPAT. */
 1037 #define PACKET_TAG_IPSEC_IN_DONE                1  /* IPsec applied, in */
 1038 #define PACKET_TAG_IPSEC_OUT_DONE               2  /* IPsec applied, out */
 1039 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE         3  /* NIC IPsec crypto done */
 1040 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED      4  /* NIC IPsec crypto req'ed */
 1041 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO     5  /* NIC notifies IPsec */
 1042 #define PACKET_TAG_IPSEC_PENDING_TDB            6  /* Reminder to do IPsec */
 1043 #define PACKET_TAG_BRIDGE                       7  /* Bridge processing done */
 1044 #define PACKET_TAG_GIF                          8  /* GIF processing done */
 1045 #define PACKET_TAG_GRE                          9  /* GRE processing done */
 1046 #define PACKET_TAG_IN_PACKET_CHECKSUM           10 /* NIC checksumming done */
 1047 #define PACKET_TAG_ENCAP                        11 /* Encap.  processing */
 1048 #define PACKET_TAG_IPSEC_SOCKET                 12 /* IPSEC socket ref */
 1049 #define PACKET_TAG_IPSEC_HISTORY                13 /* IPSEC history */
 1050 #define PACKET_TAG_IPV6_INPUT                   14 /* IPV6 input processing */
 1051 #define PACKET_TAG_DUMMYNET                     15 /* dummynet info */
 1052 #define PACKET_TAG_DIVERT                       17 /* divert info */
 1053 #define PACKET_TAG_IPFORWARD                    18 /* ipforward info */
 1054 #define PACKET_TAG_MACLABEL     (19 | MTAG_PERSISTENT) /* MAC label */
 1055 #define PACKET_TAG_PF           (21 | MTAG_PERSISTENT) /* PF/ALTQ information */
 1056 #define PACKET_TAG_RTSOCKFAM                    25 /* rtsock sa family */
 1057 #define PACKET_TAG_IPOPTIONS                    27 /* Saved IP options */
 1058 #define PACKET_TAG_CARP                         28 /* CARP info */
 1059 #define PACKET_TAG_IPSEC_NAT_T_PORTS            29 /* two uint16_t */
 1060 #define PACKET_TAG_ND_OUTGOING                  30 /* ND outgoing */
 1061 
 1062 /* Specific cookies and tags. */
 1063 
 1064 /* Packet tag routines. */
 1065 struct m_tag    *m_tag_alloc(u_int32_t, int, int, int);
 1066 void             m_tag_delete(struct mbuf *, struct m_tag *);
 1067 void             m_tag_delete_chain(struct mbuf *, struct m_tag *);
 1068 void             m_tag_free_default(struct m_tag *);
 1069 struct m_tag    *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
 1070 struct m_tag    *m_tag_copy(struct m_tag *, int);
 1071 int              m_tag_copy_chain(struct mbuf *, struct mbuf *, int);
 1072 void             m_tag_delete_nonpersistent(struct mbuf *);
 1073 
 1074 /*
 1075  * Initialize the list of tags associated with an mbuf.
 1076  */
 1077 static __inline void
 1078 m_tag_init(struct mbuf *m)
 1079 {
 1080 
 1081         SLIST_INIT(&m->m_pkthdr.tags);
 1082 }
 1083 
 1084 /*
 1085  * Set up the contents of a tag.  Note that this does not fill in the free
 1086  * method; the caller is expected to do that.
 1087  *
 1088  * XXX probably should be called m_tag_init, but that was already taken.
 1089  */
 1090 static __inline void
 1091 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
 1092 {
 1093 
 1094         t->m_tag_id = type;
 1095         t->m_tag_len = len;
 1096         t->m_tag_cookie = cookie;
 1097 }
 1098 
 1099 /*
 1100  * Reclaim resources associated with a tag.
 1101  */
 1102 static __inline void
 1103 m_tag_free(struct m_tag *t)
 1104 {
 1105 
 1106         (*t->m_tag_free)(t);
 1107 }
 1108 
 1109 /*
 1110  * Return the first tag associated with an mbuf.
 1111  */
 1112 static __inline struct m_tag *
 1113 m_tag_first(struct mbuf *m)
 1114 {
 1115 
 1116         return (SLIST_FIRST(&m->m_pkthdr.tags));
 1117 }
 1118 
 1119 /*
 1120  * Return the next tag in the list of tags associated with an mbuf.
 1121  */
 1122 static __inline struct m_tag *
 1123 m_tag_next(struct mbuf *m, struct m_tag *t)
 1124 {
 1125 
 1126         return (SLIST_NEXT(t, m_tag_link));
 1127 }
 1128 
 1129 /*
 1130  * Prepend a tag to the list of tags associated with an mbuf.
 1131  */
 1132 static __inline void
 1133 m_tag_prepend(struct mbuf *m, struct m_tag *t)
 1134 {
 1135 
 1136         SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
 1137 }
 1138 
 1139 /*
 1140  * Unlink a tag from the list of tags associated with an mbuf.
 1141  */
 1142 static __inline void
 1143 m_tag_unlink(struct mbuf *m, struct m_tag *t)
 1144 {
 1145 
 1146         SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
 1147 }
 1148 
 1149 /* These are for OpenBSD compatibility. */
 1150 #define MTAG_ABI_COMPAT         0               /* compatibility ABI */
 1151 
 1152 static __inline struct m_tag *
 1153 m_tag_get(int type, int length, int wait)
 1154 {
 1155         return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
 1156 }
 1157 
 1158 static __inline struct m_tag *
 1159 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
 1160 {
 1161         return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
 1162             m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
 1163 }
 1164 
 1165 static __inline struct mbuf *
 1166 m_free(struct mbuf *m)
 1167 {
 1168         struct mbuf *n = m->m_next;
 1169 
 1170         if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
 1171                 m_tag_delete_chain(m, NULL);
 1172         if (m->m_flags & M_EXT)
 1173                 mb_free_ext(m);
 1174         else if ((m->m_flags & M_NOFREE) == 0)
 1175                 uma_zfree(zone_mbuf, m);
 1176         return (n);
 1177 }
 1178 
 1179 static int inline
 1180 rt_m_getfib(struct mbuf *m)
 1181 {
 1182         KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
 1183         return (m->m_pkthdr.fibnum);
 1184 }
 1185 
 1186 #define M_GETFIB(_m)   rt_m_getfib(_m)
 1187 
 1188 #define M_SETFIB(_m, _fib) do {                                         \
 1189         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));  \
 1190         ((_m)->m_pkthdr.fibnum) = (_fib);                               \
 1191 } while (0)
 1192 
 1193 #ifdef MBUF_PROFILING
 1194  void m_profile(struct mbuf *m);
 1195  #define M_PROFILE(m) m_profile(m)
 1196 #else
 1197  #define M_PROFILE(m)
 1198 #endif
 1199 
 1200 struct mbufq {
 1201         STAILQ_HEAD(, mbuf)     mq_head;
 1202         int                     mq_len;
 1203         int                     mq_maxlen;
 1204 };
 1205 
 1206 static inline void
 1207 mbufq_init(struct mbufq *mq, int maxlen)
 1208 {
 1209 
 1210         STAILQ_INIT(&mq->mq_head);
 1211         mq->mq_maxlen = maxlen;
 1212         mq->mq_len = 0;
 1213 }
 1214 
 1215 static inline struct mbuf *
 1216 mbufq_flush(struct mbufq *mq)
 1217 {
 1218         struct mbuf *m;
 1219 
 1220         m = STAILQ_FIRST(&mq->mq_head);
 1221         STAILQ_INIT(&mq->mq_head);
 1222         mq->mq_len = 0;
 1223         return (m);
 1224 }
 1225 
 1226 static inline void
 1227 mbufq_drain(struct mbufq *mq)
 1228 {
 1229         struct mbuf *m, *n;
 1230 
 1231         n = mbufq_flush(mq);
 1232         while ((m = n) != NULL) {
 1233                 n = STAILQ_NEXT(m, m_stailqpkt);
 1234                 m_freem(m);
 1235         }
 1236 }
 1237 
 1238 static inline struct mbuf *
 1239 mbufq_first(const struct mbufq *mq)
 1240 {
 1241 
 1242         return (STAILQ_FIRST(&mq->mq_head));
 1243 }
 1244 
 1245 static inline struct mbuf *
 1246 mbufq_last(const struct mbufq *mq)
 1247 {
 1248 
 1249         return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
 1250 }
 1251 
 1252 static inline int
 1253 mbufq_full(const struct mbufq *mq)
 1254 {
 1255 
 1256         return (mq->mq_len >= mq->mq_maxlen);
 1257 }
 1258 
 1259 static inline int
 1260 mbufq_len(const struct mbufq *mq)
 1261 {
 1262 
 1263         return (mq->mq_len);
 1264 }
 1265 
 1266 static inline int
 1267 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
 1268 {
 1269 
 1270         if (mbufq_full(mq))
 1271                 return (ENOBUFS);
 1272         STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
 1273         mq->mq_len++;
 1274         return (0);
 1275 }
 1276 
 1277 static inline struct mbuf *
 1278 mbufq_dequeue(struct mbufq *mq)
 1279 {
 1280         struct mbuf *m;
 1281 
 1282         m = STAILQ_FIRST(&mq->mq_head);
 1283         if (m) {
 1284                 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
 1285                 m->m_nextpkt = NULL;
 1286                 mq->mq_len--;
 1287         }
 1288         return (m);
 1289 }
 1290 
 1291 static inline void
 1292 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
 1293 {
 1294 
 1295         STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
 1296         mq->mq_len++;
 1297 }
 1298 #endif /* _KERNEL */
 1299 #endif /* !_SYS_MBUF_H_ */

Cache object: 424d851cac9edbe3ecf1a3bde8739ce3


[ 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.