2 * Copyright (c) 1999-2020 Apple Inc. All rights reserved.
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
28 /* Copyright (c) 1998, 1999 Apple Computer, Inc. All Rights Reserved */
29 /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
31 * Mach Operating System
32 * Copyright (c) 1987 Carnegie-Mellon University
33 * All rights reserved. The CMU software License Agreement specifies
34 * the terms and conditions for use and redistribution.
37 * Copyright (c) 1994 NeXT Computer, Inc. All rights reserved.
39 * Copyright (c) 1982, 1986, 1988 Regents of the University of California.
40 * All rights reserved.
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. All advertising materials mentioning features or use of this software
51 * must display the following acknowledgement:
52 * This product includes software developed by the University of
53 * California, Berkeley and its contributors.
54 * 4. Neither the name of the University nor the names of its contributors
55 * may be used to endorse or promote products derived from this software
56 * without specific prior written permission.
58 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
59 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
60 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
61 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
62 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
63 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
64 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
65 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
66 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
67 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
70 * @(#)mbuf.h 8.3 (Berkeley) 1/21/94
73 * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
74 * support for mandatory and extensible security protections. This notice
75 * is included in support of clause 2.2 (b) of the Apple Public License,
82 #include <sys/appleapiopts.h>
83 #include <sys/cdefs.h>
84 #include <sys/_types/_u_int32_t.h> /* u_int32_t */
85 #include <sys/_types/_u_int64_t.h> /* u_int64_t */
86 #include <sys/_types/_u_short.h> /* u_short */
89 #include <sys/kpi_mbuf.h>
92 #ifdef XNU_KERNEL_PRIVATE
94 #include <sys/queue.h>
95 #include <machine/endian.h>
97 * Mbufs are of a single size, MSIZE (machine/param.h), which
98 * includes overhead. An mbuf may add a single "mbuf cluster" of size
99 * MCLBYTES/MBIGCLBYTES/M16KCLBYTES (also in machine/param.h), which has
100 * no additional overhead and is used instead of the internal data area;
101 * this is done when at least MINCLSIZE of data must be stored.
105 * The following _MLEN and _MHLEN macros are private to xnu. Private code
106 * that are outside of xnu must use the mbuf_get_{mlen,mhlen} routines since
107 * the sizes of the structures are dependent upon specific xnu configs.
109 #define _MLEN (MSIZE - sizeof(struct m_hdr)) /* normal data len */
110 #define _MHLEN (_MLEN - sizeof(struct pkthdr)) /* data len w/pkthdr */
112 #define NMBPGSHIFT (PAGE_SHIFT - MSIZESHIFT)
113 #define NMBPG (1 << NMBPGSHIFT) /* # of mbufs per page */
115 #define NCLPGSHIFT (PAGE_SHIFT - MCLSHIFT)
116 #define NCLPG (1 << NCLPGSHIFT) /* # of cl per page */
118 #define NBCLPGSHIFT (PAGE_SHIFT - MBIGCLSHIFT)
119 #define NBCLPG (1 << NBCLPGSHIFT) /* # of big cl per page */
121 #define NMBPCLSHIFT (MCLSHIFT - MSIZESHIFT)
122 #define NMBPCL (1 << NMBPCLSHIFT) /* # of mbufs per cl */
124 #define NCLPJCLSHIFT (M16KCLSHIFT - MCLSHIFT)
125 #define NCLPJCL (1 << NCLPJCLSHIFT) /* # of cl per jumbo cl */
127 #define NCLPBGSHIFT (MBIGCLSHIFT - MCLSHIFT)
128 #define NCLPBG (1 << NCLPBGSHIFT) /* # of cl per big cl */
130 #define NMBPBGSHIFT (MBIGCLSHIFT - MSIZESHIFT)
131 #define NMBPBG (1 << NMBPBGSHIFT) /* # of mbufs per big cl */
134 * Macros for type conversion
135 * mtod(m,t) - convert mbuf pointer to data pointer of correct type
136 * mtodo(m, o) -- Same as above but with offset 'o' into data.
137 * dtom(x) - convert data pointer within mbuf to mbuf pointer (XXX)
139 #define mtod(m, t) ((t)m_mtod(m))
140 #define mtodo(m, o) ((void *)(mtod(m, uint8_t *) + (o)))
141 #define dtom(x) m_dtom(x)
143 /* header at beginning of each mbuf: */
145 struct mbuf
*mh_next
; /* next buffer in chain */
146 struct mbuf
*mh_nextpkt
; /* next chain in queue/record */
147 caddr_t mh_data
; /* location of data */
148 int32_t mh_len
; /* amount of data in this mbuf */
149 u_int16_t mh_type
; /* type of data in this mbuf */
150 u_int16_t mh_flags
; /* flags; see below */
151 #if __arm__ && (__BIGGEST_ALIGNMENT__ > 4)
152 /* This is needed because of how _MLEN is defined and used. Ideally, _MLEN
153 * should be defined using the offsetof(struct mbuf, M_dat), since there is
154 * no guarantee that mbuf.M_dat will start where mbuf.m_hdr ends. The compiler
155 * may (and does in the armv7k case) insert padding between m_hdr and M_dat in
156 * mbuf. We cannot easily use offsetof, however, since _MLEN is referenced
157 * in the definition of mbuf.
159 } __attribute__((aligned(8)));
165 * Packet tag structure (see below for details).
168 uint64_t m_tag_cookie
; /* Error checking */
170 uint32_t pad
; /* For structure alignment */
171 #endif /* !__LP64__ */
172 SLIST_ENTRY(m_tag
) m_tag_link
; /* List of packet tags */
173 uint16_t m_tag_type
; /* Module specific type */
174 uint16_t m_tag_len
; /* Length of data */
175 uint32_t m_tag_id
; /* Module ID */
178 #define M_TAG_ALIGN(len) \
179 (P2ROUNDUP(len, sizeof (u_int64_t)) + sizeof (struct m_tag))
181 #define M_TAG_VALID_PATTERN 0xfeedfacefeedfaceULL
182 #define M_TAG_FREE_PATTERN 0xdeadbeefdeadbeefULL
185 * Packet tag header structure (at the top of mbuf). Pointers are
186 * 32-bit in ILP32; m_tag needs 64-bit alignment, hence padded.
190 u_int32_t pad
; /* For structure alignment */
191 #endif /* !__LP64__ */
192 u_int64_t refcnt
; /* Number of tags in this mbuf */
196 * Driver auxiliary metadata tag (KERNEL_TAG_TYPE_DRVAUX).
198 struct m_drvaux_tag
{
199 u_int32_t da_family
; /* IFNET_FAMILY values */
200 u_int32_t da_subfamily
; /* IFNET_SUBFAMILY values */
201 u_int32_t da_reserved
; /* for future */
202 u_int32_t da_length
; /* length of following data */
205 /* Values for pftag_flags (16-bit wide) */
206 #define PF_TAG_GENERATED 0x1 /* pkt generated by PF */
207 #define PF_TAG_FRAGCACHE 0x2
208 #define PF_TAG_TRANSLATE_LOCALHOST 0x4
210 #define PF_TAG_HDR_INET 0x8 /* hdr points to IPv4 */
211 #define PF_TAG_HDR_INET6 0x10 /* hdr points to IPv6 */
213 #define PF_TAG_REASSEMBLED 0x20 /* pkt reassembled by PF */
214 #define PF_TAG_REFRAGMENTED 0x40 /* pkt refragmented by PF */
219 u_int16_t pftag_flags
; /* PF_TAG flags */
220 u_int16_t pftag_rtableid
; /* alternate routing table id */
222 u_int16_t pftag_routed
;
224 void *pftag_hdr
; /* saved hdr pos in mbuf, for ECN */
228 /* System reserved PF tags */
229 #define PF_TAG_ID_SYSTEM_SERVICE 0xff00
230 #define PF_TAG_ID_STACK_DROP 0xff01
235 struct pf_fragment_tag
{
236 uint32_t ft_id
; /* fragment id */
237 uint16_t ft_hdrlen
; /* header length of reassembled pkt */
238 uint16_t ft_unfragpartlen
; /* length of the per-fragment headers */
239 uint16_t ft_extoff
; /* last extension header offset or 0 */
240 uint16_t ft_maxlen
; /* maximum fragment payload length */
249 uint32_t segsz
; /* segment size (actual MSS) */
250 uint32_t start_seq
; /* start seq of this packet */
255 uint8_t seg_cnt
; /* # of coalesced TCP pkts */
258 #define tso_segsz proto_mtag.__pr_u.tcp.tm_tcp.__offload.__tx.segsz
259 #define tx_start_seq proto_mtag.__pr_u.tcp.tm_tcp.__offload.__tx.start_seq
260 #define tx_tcp_pid proto_mtag.__pr_u.tcp.tm_tcp.__offload.__tx.pid
261 #define tx_tcp_e_pid proto_mtag.__pr_u.tcp.tm_tcp.__offload.__tx.e_pid
262 #define seg_cnt proto_mtag.__pr_u.tcp.tm_tcp.__offload.__rx.seg_cnt
268 struct mptcp_pktinfo
{
269 uint64_t mtpi_dsn
; /* MPTCP Data Sequence Number */
270 uint32_t mtpi_rel_seq
; /* Relative Seq Number */
271 uint16_t mtpi_length
; /* Length of mapping */
273 #define mp_dsn proto_mtag.__pr_u.tcp.tm_mptcp.mtpi_dsn
274 #define mp_rseq proto_mtag.__pr_u.tcp.tm_mptcp.mtpi_rel_seq
275 #define mp_rlen proto_mtag.__pr_u.tcp.tm_mptcp.mtpi_length
276 #define mp_csum proto_mtag.__pr_u.tcp.tm_mptcp.mtpi_csum
280 * TCP specific mbuf tag. Note that the current implementation uses
281 * MPTCP metadata strictly between MPTCP and the TCP subflow layers,
282 * hence tm_tcp and tm_mptcp are mutually exclusive. This also means
283 * that TCP messages functionality is currently incompatible with MPTCP.
287 struct tcp_pktinfo tm_tcp
; /* TCP and below */
288 struct mptcp_pktinfo tm_mptcp
; /* MPTCP-TCP only */
295 #define tx_udp_pid proto_mtag.__pr_u.udp._pid
296 #define tx_udp_e_pid proto_mtag.__pr_u.udp._e_pid
302 #define tx_rawip_pid proto_mtag.__pr_u.rawip._pid
303 #define tx_rawip_e_pid proto_mtag.__pr_u.rawip._e_pid
306 struct driver_mtag_
{
307 uintptr_t _drv_tx_compl_arg
;
308 uintptr_t _drv_tx_compl_data
;
309 kern_return_t _drv_tx_status
;
310 uint16_t _drv_flowid
;
311 #define drv_tx_compl_arg builtin_mtag._drv_mtag._drv_tx_compl_arg
312 #define drv_tx_compl_data builtin_mtag._drv_mtag._drv_tx_compl_data
313 #define drv_tx_status builtin_mtag._drv_mtag._drv_tx_status
314 #define drv_flowid builtin_mtag._drv_mtag._drv_flowid
318 * Protocol specific mbuf tag (at most one protocol metadata per mbuf).
320 * Care must be taken to ensure that they are mutually exclusive, e.g.
321 * IPsec policy ID implies no TCP segment offload (which is fine given
322 * that the former is used on the virtual ipsec interface that does
323 * not advertise the TSO capability.)
327 struct tcp_mtag tcp
; /* TCP specific */
328 struct udp_mtag udp
; /* UDP specific */
329 struct rawip_mtag rawip
; /* raw IPv4/IPv6 specific */
334 * NECP specific mbuf tag.
337 u_int32_t necp_policy_id
;
338 u_int32_t necp_skip_policy_id
;
339 u_int32_t necp_route_rule_id
;
340 u_int16_t necp_last_interface_index
;
341 u_int16_t necp_app_id
;
346 struct proto_mtag_ _proto_mtag
; /* built-in protocol-specific tag */
347 struct pf_mtag _pf_mtag
; /* built-in PF tag */
348 struct necp_mtag_ _necp_mtag
; /* built-in NECP tag */
350 struct driver_mtag_ _drv_mtag
;
351 #define necp_mtag builtin_mtag._net_mtag._necp_mtag
352 #define proto_mtag builtin_mtag._net_mtag._proto_mtag
353 #define driver_mtag builtin_mtag._drv_mtag
357 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR set.
360 struct ifnet
*rcvif
; /* rcv interface */
361 /* variables for ip and tcp reassembly */
362 void *pkt_hdr
; /* pointer to packet header */
363 int32_t len
; /* total packet length */
364 /* variables for hardware checksum */
365 /* Note: csum_flags is used for hardware checksum and VLAN */
366 u_int32_t csum_flags
; /* flags regarding checksum */
369 u_int16_t val
; /* checksum value */
370 u_int16_t start
; /* checksum start offset */
372 #define csum_rx_val _csum_rx.val
373 #define csum_rx_start _csum_rx.start
375 u_int16_t start
; /* checksum start offset */
376 u_int16_t stuff
; /* checksum stuff offset */
378 #define csum_tx_start _csum_tx.start
379 #define csum_tx_stuff _csum_tx.stuff
381 * Generic data field used by csum routines.
382 * It gets used differently in different contexts.
386 u_int16_t vlan_tag
; /* VLAN tag, host byte order */
388 * Packet classifier info
390 * PKTF_FLOW_ID set means valid flow ID. A non-zero flow ID value
391 * means the packet has been classified by one of the flow sources.
392 * It is also a prerequisite for flow control advisory, which is
393 * enabled by additionally setting PKTF_FLOW_ADV.
395 * The protocol value is a best-effort representation of the payload.
396 * It is opportunistically updated and used only for optimization.
397 * It is not a substitute for parsing the protocol header(s); use it
400 * If PKTF_IFAINFO is set, pkt_ifainfo contains one or both of the
401 * indices of interfaces which own the source and/or destination
402 * addresses of the packet. For the local/loopback case (PKTF_LOOP),
403 * both should be valid, and thus allows for the receiving end to
404 * quickly determine the actual interfaces used by the the addresses;
405 * they may not necessarily be the same or refer to the loopback
406 * interface. Otherwise, in the non-local/loopback case, the indices
407 * are opportunistically set, and because of that only one may be set
408 * (0 means the index has not been determined.) In addition, the
409 * interface address flags are also recorded. This allows us to avoid
410 * storing the corresponding {in,in6}_ifaddr in an mbuf tag. Ideally
411 * this would be a superset of {ia,ia6}_flags, but the namespaces are
412 * overlapping at present, so we'll need a new set of values in future
413 * to achieve this. For now, we will just rely on the address family
414 * related code paths examining this mbuf to interpret the flags.
416 u_int8_t pkt_proto
; /* IPPROTO value */
417 u_int8_t pkt_flowsrc
; /* FLOWSRC values */
418 u_int32_t pkt_flowid
; /* flow ID */
419 u_int32_t pkt_flags
; /* PKTF flags (see below) */
420 u_int32_t pkt_svc
; /* MBUF_SVC value */
422 u_int32_t pkt_compl_context
; /* Packet completion context */
426 u_int16_t src
; /* ifindex of src addr i/f */
427 u_int16_t src_flags
; /* src PKT_IFAIFF flags */
428 u_int16_t dst
; /* ifindex of dst addr i/f */
429 u_int16_t dst_flags
; /* dst PKT_IFAIFF flags */
431 #define src_ifindex _pkt_iaif.src
432 #define src_iff _pkt_iaif.src_flags
433 #define dst_ifindex _pkt_iaif.dst
434 #define dst_iff _pkt_iaif.dst_flags
435 u_int64_t pkt_ifainfo
; /* data field used by ifainfo */
437 u_int32_t if_data
; /* bytes in interface queue */
438 u_int32_t sndbuf_data
; /* bytes in socket buffer */
439 } _pkt_bsr
; /* Buffer status report used by cellular interface */
440 #define bufstatus_if _pkt_bsr.if_data
441 #define bufstatus_sndbuf _pkt_bsr.sndbuf_data
443 u_int64_t pkt_timestamp
; /* TX: enqueue time, RX: receive timestamp */
446 * Tags (external and built-in)
448 SLIST_HEAD(packet_tags
, m_tag
) tags
; /* list of external tags */
449 union builtin_mtag builtin_mtag
;
451 uint32_t comp_gencnt
;
454 * Module private scratch space (32-bit aligned), currently 16-bytes
455 * large. Anything stored here is not guaranteed to survive across
456 * modules. The AQM layer (outbound) uses all 16-bytes for both
457 * packet scheduling and flow advisory information.
461 u_int8_t __mpriv8
[16];
462 u_int16_t __mpriv16
[8];
466 u_int16_t __val16
[2];
470 u_int64_t __mpriv64
[2];
472 } pkt_mpriv
__attribute__((aligned(4)));
473 #define pkt_mpriv_hash pkt_mpriv.__mpriv_u.__mpriv32[0].__mpriv32_u.__val32
474 #define pkt_mpriv_flags pkt_mpriv.__mpriv_u.__mpriv32[1].__mpriv32_u.__val32
475 #define pkt_mpriv_srcid pkt_mpriv.__mpriv_u.__mpriv32[2].__mpriv32_u.__val32
476 #define pkt_mpriv_fidx pkt_mpriv.__mpriv_u.__mpriv32[3].__mpriv32_u.__val32
478 u_int32_t redzone
; /* red zone */
479 u_int32_t pkt_compl_callbacks
; /* Packet completion callbacks */
483 * Flow data source type. A data source module is responsible for generating
484 * a unique flow ID and associating it to each data flow as pkt_flowid.
485 * This is required for flow control/advisory, as it allows the output queue
486 * to identify the data source object and inform that it can resume its
487 * transmission (in the event it was flow controlled.)
489 #define FLOWSRC_INPCB 1 /* flow ID generated by INPCB */
490 #define FLOWSRC_IFNET 2 /* flow ID generated by interface */
491 #define FLOWSRC_PF 3 /* flow ID generated by PF */
492 #define FLOWSRC_CHANNEL 4 /* flow ID generated by channel */
495 * FLOWSRC_MPKL_INPUT is not a true flow data source and is used for
496 * multi-layer packet logging. We're usurping the pkt_flowsrc field because
497 * the mbuf packet header ran out of space and pkt_flowsrc is normally
498 * used on output so we assume we can safely overwrite the normal semantic of
500 * This value is meant to be used on incoming packet from a lower level protocol
501 * to pass information to some upper level protocol. When FLOWSRC_MPKL_INPUT
502 * is set, the following fields are used:
503 * - pkt_proto: the IP protocol ID of the lower level protocol
504 * - pkt_flowid: the identifier of the packet at the lower protocol.
505 * For example ESP would set pkt_proto to IPPROTO_ESP and pkt_flowid to the SPI.
509 * Packet flags. Unlike m_flags, all packet flags are copied along when
510 * copying m_pkthdr, i.e. no equivalent of M_COPYFLAGS here. These flags
511 * (and other classifier info) will be cleared during DLIL input.
513 * Some notes about M_LOOP and PKTF_LOOP:
515 * - M_LOOP flag is overloaded, and its use is discouraged. Historically,
516 * that flag was used by the KAME implementation for allowing certain
517 * certain exceptions to be made in the IP6_EXTHDR_CHECK() logic; this
518 * was originally meant to be set as the packet is looped back to the
519 * system, and in some circumstances temporarily set in ip6_output().
520 * Over time, this flag was used by the pre-output routines to indicate
521 * to the DLIL frameout and output routines, that the packet may be
522 * looped back to the system under the right conditions. In addition,
523 * this is an mbuf flag rather than an mbuf packet header flag.
525 * - PKTF_LOOP is an mbuf packet header flag, which is set if and only
526 * if the packet was looped back to the system. This flag should be
527 * used instead for newer code.
529 #define PKTF_FLOW_ID 0x1 /* pkt has valid flowid value */
530 #define PKTF_FLOW_ADV 0x2 /* pkt triggers local flow advisory */
531 #define PKTF_FLOW_LOCALSRC 0x4 /* pkt is locally originated */
532 #define PKTF_FLOW_RAWSOCK 0x8 /* pkt locally generated by raw sock */
533 #define PKTF_PRIO_PRIVILEGED 0x10 /* packet priority is privileged */
534 #define PKTF_PROXY_DST 0x20 /* processed but not locally destined */
535 #define PKTF_INET_RESOLVE 0x40 /* IPv4 resolver packet */
536 #define PKTF_INET6_RESOLVE 0x80 /* IPv6 resolver packet */
537 #define PKTF_RESOLVE_RTR 0x100 /* pkt is for resolving router */
538 #define PKTF_SKIP_PKTAP 0x200 /* pkt has already passed through pktap */
539 #define PKTF_MPTCP 0x800 /* TCP with MPTCP metadata */
540 #define PKTF_MPSO 0x1000 /* MPTCP socket meta data */
541 #define PKTF_LOOP 0x2000 /* loopbacked packet */
542 #define PKTF_IFAINFO 0x4000 /* pkt has valid interface addr info */
543 #define PKTF_SO_BACKGROUND 0x8000 /* data is from background source */
544 #define PKTF_FORWARDED 0x10000 /* pkt was forwarded from another i/f */
545 #define PKTF_PRIV_GUARDED 0x20000 /* pkt_mpriv area guard enabled */
546 #define PKTF_KEEPALIVE 0x40000 /* pkt is kernel-generated keepalive */
547 #define PKTF_SO_REALTIME 0x80000 /* data is realtime traffic */
548 #define PKTF_VALID_UNSENT_DATA 0x100000 /* unsent data is valid */
549 #define PKTF_TCP_REXMT 0x200000 /* packet is TCP retransmission */
550 #define PKTF_REASSEMBLED 0x400000 /* Packet was reassembled */
551 #define PKTF_TX_COMPL_TS_REQ 0x800000 /* tx completion timestamp requested */
552 #define PKTF_TS_VALID 0x1000000 /* pkt timestamp is valid */
553 #define PKTF_DRIVER_MTAG 0x2000000 /* driver mbuf tags fields inited */
554 #define PKTF_NEW_FLOW 0x4000000 /* Data from a new flow */
555 #define PKTF_START_SEQ 0x8000000 /* valid start sequence */
556 #define PKTF_LAST_PKT 0x10000000 /* last packet in the flow */
557 #define PKTF_MPTCP_REINJ 0x20000000 /* Packet has been reinjected for MPTCP */
558 #define PKTF_MPTCP_DFIN 0x40000000 /* Packet is a data-fin */
559 #define PKTF_HBH_CHKED 0x80000000 /* HBH option is checked */
561 /* flags related to flow control/advisory and identification */
562 #define PKTF_FLOW_MASK \
563 (PKTF_FLOW_ID | PKTF_FLOW_ADV | PKTF_FLOW_LOCALSRC | PKTF_FLOW_RAWSOCK)
566 * Description of external storage mapped into mbuf, valid only if M_EXT set.
568 typedef void (*m_ext_free_func_t
)(caddr_t
, u_int
, caddr_t
);
570 caddr_t ext_buf
; /* start of buffer */
571 m_ext_free_func_t ext_free
; /* free routine if not the usual */
572 u_int ext_size
; /* size of buffer, for ext_free */
573 caddr_t ext_arg
; /* additional ext_free argument */
585 /* define m_ext to a type since it gets redefined below */
586 typedef struct m_ext _m_ext_t
;
595 struct pkthdr MH_pkthdr
; /* M_PKTHDR set */
597 struct m_ext MH_ext
; /* M_EXT set */
598 char MH_databuf
[_MHLEN
];
601 char M_databuf
[_MLEN
]; /* !M_PKTHDR, !M_EXT */
605 #define m_next m_hdr.mh_next
606 #define m_len m_hdr.mh_len
607 #define m_data m_hdr.mh_data
608 #define m_type m_hdr.mh_type
609 #define m_flags m_hdr.mh_flags
610 #define m_nextpkt m_hdr.mh_nextpkt
611 #define m_act m_nextpkt
612 #define m_pkthdr M_dat.MH.MH_pkthdr
613 #define m_ext M_dat.MH.MH_dat.MH_ext
614 #define m_pktdat M_dat.MH.MH_dat.MH_databuf
615 #define m_dat M_dat.M_databuf
616 #define m_pktlen(_m) ((_m)->m_pkthdr.len)
617 #define m_pftag(_m) (&(_m)->m_pkthdr.builtin_mtag._net_mtag._pf_mtag)
619 /* mbuf flags (private) */
620 #define M_EXT 0x0001 /* has associated external storage */
621 #define M_PKTHDR 0x0002 /* start of record */
622 #define M_EOR 0x0004 /* end of record */
623 #define M_PROTO1 0x0008 /* protocol-specific */
624 #define M_PROTO2 0x0010 /* protocol-specific */
625 #define M_PROTO3 0x0020 /* protocol-specific */
626 #define M_LOOP 0x0040 /* packet is looped back (also see PKTF_LOOP) */
627 #define M_PROTO5 0x0080 /* protocol-specific */
629 /* mbuf pkthdr flags, also in m_flags (private) */
630 #define M_BCAST 0x0100 /* send/received as link-level broadcast */
631 #define M_MCAST 0x0200 /* send/received as link-level multicast */
632 #define M_FRAG 0x0400 /* packet is a fragment of a larger packet */
633 #define M_FIRSTFRAG 0x0800 /* packet is first fragment */
634 #define M_LASTFRAG 0x1000 /* packet is last fragment */
635 #define M_PROMISC 0x2000 /* packet is promiscuous (shouldn't go to stack) */
636 #define M_HASFCS 0x4000 /* packet has FCS */
637 #define M_TAGHDR 0x8000 /* m_tag hdr structure at top of mbuf data */
640 * Flags to purge when crossing layers.
642 #define M_PROTOFLAGS \
643 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO5)
645 /* flags copied when copying m_pkthdr */
646 #define M_COPYFLAGS \
647 (M_PKTHDR|M_EOR|M_PROTO1|M_PROTO2|M_PROTO3 | \
648 M_LOOP|M_PROTO5|M_BCAST|M_MCAST|M_FRAG | \
649 M_FIRSTFRAG|M_LASTFRAG|M_PROMISC|M_HASFCS)
651 /* flags indicating hw checksum support and sw checksum requirements */
652 #define CSUM_IP 0x0001 /* will csum IP */
653 #define CSUM_TCP 0x0002 /* will csum TCP */
654 #define CSUM_UDP 0x0004 /* will csum UDP */
655 #define CSUM_IP_FRAGS 0x0008 /* will csum IP fragments */
656 #define CSUM_FRAGMENT 0x0010 /* will do IP fragmentation */
657 #define CSUM_TCPIPV6 0x0020 /* will csum TCP for IPv6 */
658 #define CSUM_UDPIPV6 0x0040 /* will csum UDP for IPv6 */
659 #define CSUM_FRAGMENT_IPV6 0x0080 /* will do IPv6 fragmentation */
661 #define CSUM_IP_CHECKED 0x0100 /* did csum IP */
662 #define CSUM_IP_VALID 0x0200 /* ... the csum is valid */
663 #define CSUM_DATA_VALID 0x0400 /* csum_data field is valid */
664 #define CSUM_PSEUDO_HDR 0x0800 /* csum_data has pseudo hdr */
665 #define CSUM_PARTIAL 0x1000 /* simple Sum16 computation */
666 #define CSUM_ZERO_INVERT 0x2000 /* invert 0 to -0 (0xffff) */
668 #define CSUM_DELAY_DATA (CSUM_TCP | CSUM_UDP)
669 #define CSUM_DELAY_IP (CSUM_IP) /* IPv4 only: no IPv6 IP cksum */
670 #define CSUM_DELAY_IPV6_DATA (CSUM_TCPIPV6 | CSUM_UDPIPV6)
671 #define CSUM_DATA_IPV6_VALID CSUM_DATA_VALID /* csum_data field is valid */
673 #define CSUM_TX_FLAGS \
674 (CSUM_DELAY_IP | CSUM_DELAY_DATA | CSUM_DELAY_IPV6_DATA | \
675 CSUM_DATA_VALID | CSUM_PARTIAL | CSUM_ZERO_INVERT)
677 #define CSUM_RX_FULL_FLAGS \
678 (CSUM_IP_CHECKED | CSUM_IP_VALID | CSUM_PSEUDO_HDR | \
681 #define CSUM_RX_FLAGS \
682 (CSUM_RX_FULL_FLAGS | CSUM_PARTIAL)
687 * Note: see also IF_HWASSIST_CSUM defined in <net/if_var.h>
690 /* VLAN tag present */
691 #define CSUM_VLAN_TAG_VALID 0x00010000 /* vlan_tag field is valid */
693 /* checksum start adjustment has been done */
694 #define CSUM_ADJUST_DONE 0x00020000
696 /* TCP Segment Offloading requested on this mbuf */
697 #define CSUM_TSO_IPV4 0x00100000 /* This mbuf needs to be segmented by the NIC */
698 #define CSUM_TSO_IPV6 0x00200000 /* This mbuf needs to be segmented by the NIC */
700 #define TSO_IPV4_OK(_ifp, _m) \
701 (((_ifp)->if_hwassist & IFNET_TSO_IPV4) && \
702 ((_m)->m_pkthdr.csum_flags & CSUM_TSO_IPV4)) \
704 #define TSO_IPV4_NOTOK(_ifp, _m) \
705 (!((_ifp)->if_hwassist & IFNET_TSO_IPV4) && \
706 ((_m)->m_pkthdr.csum_flags & CSUM_TSO_IPV4)) \
708 #define TSO_IPV6_OK(_ifp, _m) \
709 (((_ifp)->if_hwassist & IFNET_TSO_IPV6) && \
710 ((_m)->m_pkthdr.csum_flags & CSUM_TSO_IPV6)) \
712 #define TSO_IPV6_NOTOK(_ifp, _m) \
713 (!((_ifp)->if_hwassist & IFNET_TSO_IPV6) && \
714 ((_m)->m_pkthdr.csum_flags & CSUM_TSO_IPV6)) \
716 #endif /* XNU_KERNEL_PRIVATE */
719 #define MT_FREE 0 /* should be on free list */
720 #define MT_DATA 1 /* dynamic (data) allocation */
721 #define MT_HEADER 2 /* packet header */
722 #define MT_SOCKET 3 /* socket structure */
723 #define MT_PCB 4 /* protocol control block */
724 #define MT_RTABLE 5 /* routing tables */
725 #define MT_HTABLE 6 /* IMP host tables */
726 #define MT_ATABLE 7 /* address resolution tables */
727 #define MT_SONAME 8 /* socket name */
728 #define MT_SOOPTS 10 /* socket options */
729 #define MT_FTABLE 11 /* fragment reassembly header */
730 #define MT_RIGHTS 12 /* access rights */
731 #define MT_IFADDR 13 /* interface address */
732 #define MT_CONTROL 14 /* extra-data protocol message */
733 #define MT_OOBDATA 15 /* expedited data */
734 #define MT_TAG 16 /* volatile metadata associated to pkts */
735 #define MT_MAX 32 /* enough? */
737 #ifdef XNU_KERNEL_PRIVATE
739 * mbuf allocation/deallocation macros:
741 * MGET(struct mbuf *m, int how, int type)
742 * allocates an mbuf and initializes it to contain internal data.
744 * MGETHDR(struct mbuf *m, int how, int type)
745 * allocates an mbuf and initializes it to contain a packet header
750 #define MCHECK(m) m_mcheck(m)
755 #define MGET(m, how, type) ((m) = m_get((how), (type)))
757 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
760 * Mbuf cluster macros.
761 * MCLALLOC(caddr_t p, int how) allocates an mbuf cluster.
762 * MCLGET adds such clusters to a normal mbuf;
763 * the flag M_EXT is set upon success.
764 * MCLFREE releases a reference to a cluster allocated by MCLALLOC,
765 * freeing the cluster if the reference count has reached 0.
767 * Normal mbuf clusters are normally treated as character arrays
768 * after allocation, but use the first word of the buffer as a free list
769 * pointer while on the free list.
772 union mcluster
*mcl_next
;
773 char mcl_buf
[MCLBYTES
];
776 #define MCLALLOC(p, how) ((p) = m_mclalloc(how))
778 #define MCLFREE(p) m_mclfree(p)
780 #define MCLGET(m, how) ((m) = m_mclget(m, how))
786 union mbigcluster
*mbc_next
;
787 char mbc_buf
[MBIGCLBYTES
];
794 union m16kcluster
*m16kcl_next
;
795 char m16kcl_buf
[M16KCLBYTES
];
798 #define MCLHASREFERENCE(m) m_mclhasreference(m)
801 * MFREE(struct mbuf *m, struct mbuf *n)
802 * Free a single mbuf and associated external storage.
803 * Place the successor, if any, in n.
806 #define MFREE(m, n) ((n) = m_free(m))
809 * Copy mbuf pkthdr from from to to.
810 * from must have M_PKTHDR set, and to must be empty.
811 * aux pointer will be moved to `to'.
813 #define M_COPY_PKTHDR(to, from) m_copy_pkthdr(to, from)
815 #define M_COPY_PFTAG(to, from) m_copy_pftag(to, from)
817 #define M_COPY_CLASSIFIER(to, from) m_copy_classifier(to, from)
820 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
821 * be both the local data payload, or an external buffer area, depending on
822 * whether M_EXT is set).
824 #define M_WRITABLE(m) (((m)->m_flags & M_EXT) == 0 || !MCLHASREFERENCE(m))
827 * These macros are mapped to the appropriate KPIs, so that private code
828 * can be simply recompiled in order to be forward-compatible with future
829 * changes toward the struture sizes.
831 #define MLEN mbuf_get_mlen() /* normal mbuf data len */
832 #define MHLEN mbuf_get_mhlen() /* data len in an mbuf w/pkthdr */
833 #define MINCLSIZE mbuf_get_minclsize() /* cluster usage threshold */
835 * Return the address of the start of the buffer associated with an mbuf,
836 * handling external storage, packet-header mbufs, and regular data mbufs.
839 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \
840 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \
844 * Return the size of the buffer associated with an mbuf, handling external
845 * storage, packet-header mbufs, and regular data mbufs.
848 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \
849 ((m)->m_flags & M_PKTHDR) ? MHLEN : \
852 #define M_ALIGN(m, len) m_align(m, len)
853 #define MH_ALIGN(m, len) m_align(m, len)
854 #define MEXT_ALIGN(m, len) m_align(m, len)
857 * Compute the amount of space available before the current start of data in
860 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
861 * of checking writability of the mbuf data area rests solely with the caller.
863 #define M_LEADINGSPACE(m) \
864 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
867 * Compute the amount of space available after the end of data in an mbuf.
869 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
870 * of checking writability of the mbuf data area rests solely with the caller.
872 #define M_TRAILINGSPACE(m) \
874 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
877 * Arrange to prepend space of size plen to mbuf m.
878 * If a new mbuf must be allocated, how specifies whether to wait.
879 * If how is M_DONTWAIT and allocation fails, the original mbuf chain
880 * is freed and m is set to NULL.
882 #define M_PREPEND(m, plen, how, align) \
883 ((m) = m_prepend_2((m), (plen), (how), (align)))
885 /* change mbuf to new type */
886 #define MCHTYPE(m, t) m_mchtype(m, t)
888 /* compatiblity with 4.3 */
889 #define m_copy(m, o, l) m_copym((m), (o), (l), M_DONTWAIT)
891 #define MBSHIFT 20 /* 1MB */
892 #define MBSIZE (1 << MBSHIFT)
893 #define GBSHIFT 30 /* 1GB */
894 #define GBSIZE (1 << GBSHIFT)
897 * M_STRUCT_GET ensures that intermediate protocol header (from "off" to
898 * "off+len") is located in single mbuf, on contiguous memory region.
899 * The pointer to the region will be returned to pointer variable "val",
902 * M_STRUCT_GET0 does the same, except that it aligns the structure at
903 * very top of mbuf. GET0 is likely to make memory copy than GET.
905 #define M_STRUCT_GET(val, typ, m, off, len) \
910 if ((m)->m_len >= (off) + (len)) { \
911 (val) = (typ)(mtod((m), caddr_t) + (off)); \
913 t = m_pulldown((m), (off), (len), &tmp); \
915 if (t->m_len < tmp + (len)) \
916 panic("m_pulldown malfunction"); \
917 (val) = (typ)(mtod(t, caddr_t) + tmp); \
925 #define M_STRUCT_GET0(val, typ, m, off, len) \
929 if ((off) == 0 && ((m)->m_len >= (len))) { \
930 (val) = (typ)(void *)mtod(m, caddr_t); \
932 t = m_pulldown((m), (off), (len), NULL); \
934 if (t->m_len < (len)) \
935 panic("m_pulldown malfunction"); \
936 (val) = (typ)(void *)mtod(t, caddr_t); \
944 #define MBUF_INPUT_CHECK(m, rcvif) \
946 if (!(m->m_flags & MBUF_PKTHDR) || \
948 m->m_len > ((njcl > 0) ? njclbytes : MBIGCLBYTES) || \
949 m->m_type == MT_FREE || \
950 ((m->m_flags & M_EXT) != 0 && m->m_ext.ext_buf == NULL)) { \
951 panic_plain("Failed mbuf validity check: mbuf %p len %d " \
952 "type %d flags 0x%x data %p rcvif %s ifflags 0x%x", \
953 m, m->m_len, m->m_type, m->m_flags, \
954 ((m->m_flags & M_EXT) ? m->m_ext.ext_buf : m->m_data), \
956 (rcvif->if_flags & 0xffff)); \
961 * Simple mbuf queueing system
963 * This is basically a SIMPLEQ adapted to mbuf use (i.e. using
964 * m_nextpkt instead of field.sqe_next).
966 * m_next is ignored, so queueing chains of mbufs is possible
968 #define MBUFQ_HEAD(name) \
970 struct mbuf *mq_first; /* first packet */ \
971 struct mbuf **mq_last; /* addr of last next packet */ \
974 #define MBUFQ_INIT(q) do { \
975 MBUFQ_FIRST(q) = NULL; \
976 (q)->mq_last = &MBUFQ_FIRST(q); \
979 #define MBUFQ_PREPEND(q, m) do { \
980 if ((MBUFQ_NEXT(m) = MBUFQ_FIRST(q)) == NULL) \
981 (q)->mq_last = &MBUFQ_NEXT(m); \
982 MBUFQ_FIRST(q) = (m); \
985 #define MBUFQ_ENQUEUE(q, m) do { \
986 MBUFQ_NEXT(m) = NULL; \
987 *(q)->mq_last = (m); \
988 (q)->mq_last = &MBUFQ_NEXT(m); \
991 #define MBUFQ_ENQUEUE_MULTI(q, m, n) do { \
992 MBUFQ_NEXT(n) = NULL; \
993 *(q)->mq_last = (m); \
994 (q)->mq_last = &MBUFQ_NEXT(n); \
997 #define MBUFQ_DEQUEUE(q, m) do { \
998 if (((m) = MBUFQ_FIRST(q)) != NULL) { \
999 if ((MBUFQ_FIRST(q) = MBUFQ_NEXT(m)) == NULL) \
1000 (q)->mq_last = &MBUFQ_FIRST(q); \
1002 MBUFQ_NEXT(m) = NULL; \
1006 #define MBUFQ_REMOVE(q, m) do { \
1007 if (MBUFQ_FIRST(q) == (m)) { \
1008 MBUFQ_DEQUEUE(q, m); \
1010 struct mbuf *_m = MBUFQ_FIRST(q); \
1011 while (MBUFQ_NEXT(_m) != (m)) \
1012 _m = MBUFQ_NEXT(_m); \
1013 if ((MBUFQ_NEXT(_m) = \
1014 MBUFQ_NEXT(MBUFQ_NEXT(_m))) == NULL) \
1015 (q)->mq_last = &MBUFQ_NEXT(_m); \
1019 #define MBUFQ_DRAIN(q) do { \
1020 struct mbuf *__m0; \
1021 while ((__m0 = MBUFQ_FIRST(q)) != NULL) { \
1022 MBUFQ_FIRST(q) = MBUFQ_NEXT(__m0); \
1023 MBUFQ_NEXT(__m0) = NULL; \
1026 (q)->mq_last = &MBUFQ_FIRST(q); \
1029 #define MBUFQ_FOREACH(m, q) \
1030 for ((m) = MBUFQ_FIRST(q); \
1032 (m) = MBUFQ_NEXT(m))
1034 #define MBUFQ_FOREACH_SAFE(m, q, tvar) \
1035 for ((m) = MBUFQ_FIRST(q); \
1036 (m) && ((tvar) = MBUFQ_NEXT(m), 1); \
1039 #define MBUFQ_EMPTY(q) ((q)->mq_first == NULL)
1040 #define MBUFQ_FIRST(q) ((q)->mq_first)
1041 #define MBUFQ_NEXT(m) ((m)->m_nextpkt)
1043 * mq_last is initialized to point to mq_first, so check if they're
1044 * equal and return NULL when the list is empty. Otherwise, we need
1045 * to subtract the offset of MBUQ_NEXT (i.e. m_nextpkt field) to get
1046 * to the base mbuf address to return to caller.
1048 #define MBUFQ_LAST(head) \
1049 (((head)->mq_last == &MBUFQ_FIRST(head)) ? NULL : \
1050 ((struct mbuf *)(void *)((char *)(head)->mq_last - \
1051 __builtin_offsetof(struct mbuf, m_nextpkt))))
1053 #define max_linkhdr (int)P2ROUNDUP(_max_linkhdr, sizeof (uint32_t))
1054 #define max_protohdr (int)P2ROUNDUP(_max_protohdr, sizeof (uint32_t))
1055 #endif /* XNU_KERNEL_PRIVATE */
1058 * Mbuf statistics (legacy).
1061 u_int32_t m_mbufs
; /* mbufs obtained from page pool */
1062 u_int32_t m_clusters
; /* clusters obtained from page pool */
1063 u_int32_t m_spare
; /* spare field */
1064 u_int32_t m_clfree
; /* free clusters */
1065 u_int32_t m_drops
; /* times failed to find space */
1066 u_int32_t m_wait
; /* times waited for space */
1067 u_int32_t m_drain
; /* times drained protocols for space */
1068 u_short m_mtypes
[256]; /* type specific mbuf allocations */
1069 u_int32_t m_mcfail
; /* times m_copym failed */
1070 u_int32_t m_mpfail
; /* times m_pullup failed */
1071 u_int32_t m_msize
; /* length of an mbuf */
1072 u_int32_t m_mclbytes
; /* length of an mbuf cluster */
1073 u_int32_t m_minclsize
; /* min length of data to allocate a cluster */
1074 u_int32_t m_mlen
; /* length of data in an mbuf */
1075 u_int32_t m_mhlen
; /* length of data in a header mbuf */
1076 u_int32_t m_bigclusters
; /* clusters obtained from page pool */
1077 u_int32_t m_bigclfree
; /* free clusters */
1078 u_int32_t m_bigmclbytes
; /* length of an mbuf cluster */
1081 /* Compatibillity with 10.3 */
1083 u_int32_t m_mbufs
; /* mbufs obtained from page pool */
1084 u_int32_t m_clusters
; /* clusters obtained from page pool */
1085 u_int32_t m_spare
; /* spare field */
1086 u_int32_t m_clfree
; /* free clusters */
1087 u_int32_t m_drops
; /* times failed to find space */
1088 u_int32_t m_wait
; /* times waited for space */
1089 u_int32_t m_drain
; /* times drained protocols for space */
1090 u_short m_mtypes
[256]; /* type specific mbuf allocations */
1091 u_int32_t m_mcfail
; /* times m_copym failed */
1092 u_int32_t m_mpfail
; /* times m_pullup failed */
1093 u_int32_t m_msize
; /* length of an mbuf */
1094 u_int32_t m_mclbytes
; /* length of an mbuf cluster */
1095 u_int32_t m_minclsize
; /* min length of data to allocate a cluster */
1096 u_int32_t m_mlen
; /* length of data in an mbuf */
1097 u_int32_t m_mhlen
; /* length of data in a header mbuf */
1101 * mbuf class statistics.
1103 #define MAX_MBUF_CNAME 15
1105 #if defined(XNU_KERNEL_PRIVATE)
1106 /* For backwards compatibility with 32-bit userland process */
1107 struct omb_class_stat
{
1108 char mbcl_cname
[MAX_MBUF_CNAME
+ 1]; /* class name */
1109 u_int32_t mbcl_size
; /* buffer size */
1110 u_int32_t mbcl_total
; /* # of buffers created */
1111 u_int32_t mbcl_active
; /* # of active buffers */
1112 u_int32_t mbcl_infree
; /* # of available buffers */
1113 u_int32_t mbcl_slab_cnt
; /* # of available slabs */
1114 u_int64_t mbcl_alloc_cnt
; /* # of times alloc is called */
1115 u_int64_t mbcl_free_cnt
; /* # of times free is called */
1116 u_int64_t mbcl_notified
; /* # of notified wakeups */
1117 u_int64_t mbcl_purge_cnt
; /* # of purges so far */
1118 u_int64_t mbcl_fail_cnt
; /* # of allocation failures */
1119 u_int32_t mbcl_ctotal
; /* total only for this class */
1120 u_int32_t mbcl_release_cnt
; /* amount of memory returned */
1122 * Cache layer statistics
1124 u_int32_t mbcl_mc_state
; /* cache state (see below) */
1125 u_int32_t mbcl_mc_cached
; /* # of cached buffers */
1126 u_int32_t mbcl_mc_waiter_cnt
; /* # waiters on the cache */
1127 u_int32_t mbcl_mc_wretry_cnt
; /* # of wait retries */
1128 u_int32_t mbcl_mc_nwretry_cnt
; /* # of no-wait retry attempts */
1129 u_int64_t mbcl_reserved
[4]; /* for future use */
1130 } __attribute__((__packed__
));
1131 #endif /* XNU_KERNEL_PRIVATE */
1133 typedef struct mb_class_stat
{
1134 char mbcl_cname
[MAX_MBUF_CNAME
+ 1]; /* class name */
1135 u_int32_t mbcl_size
; /* buffer size */
1136 u_int32_t mbcl_total
; /* # of buffers created */
1137 u_int32_t mbcl_active
; /* # of active buffers */
1138 u_int32_t mbcl_infree
; /* # of available buffers */
1139 u_int32_t mbcl_slab_cnt
; /* # of available slabs */
1140 #if defined(KERNEL) || defined(__LP64__)
1141 u_int32_t mbcl_pad
; /* padding */
1142 #endif /* KERNEL || __LP64__ */
1143 u_int64_t mbcl_alloc_cnt
; /* # of times alloc is called */
1144 u_int64_t mbcl_free_cnt
; /* # of times free is called */
1145 u_int64_t mbcl_notified
; /* # of notified wakeups */
1146 u_int64_t mbcl_purge_cnt
; /* # of purges so far */
1147 u_int64_t mbcl_fail_cnt
; /* # of allocation failures */
1148 u_int32_t mbcl_ctotal
; /* total only for this class */
1149 u_int32_t mbcl_release_cnt
; /* amount of memory returned */
1151 * Cache layer statistics
1153 u_int32_t mbcl_mc_state
; /* cache state (see below) */
1154 u_int32_t mbcl_mc_cached
; /* # of cached buffers */
1155 u_int32_t mbcl_mc_waiter_cnt
; /* # waiters on the cache */
1156 u_int32_t mbcl_mc_wretry_cnt
; /* # of wait retries */
1157 u_int32_t mbcl_mc_nwretry_cnt
; /* # of no-wait retry attempts */
1158 u_int32_t mbcl_peak_reported
; /* last usage peak reported */
1159 u_int32_t mbcl_reserved
[7]; /* for future use */
1162 #define MCS_DISABLED 0 /* cache is permanently disabled */
1163 #define MCS_ONLINE 1 /* cache is online */
1164 #define MCS_PURGING 2 /* cache is being purged */
1165 #define MCS_OFFLINE 3 /* cache is offline (resizing) */
1167 #if defined(XNU_KERNEL_PRIVATE)
1168 /* For backwards compatibility with 32-bit userland process */
1170 u_int32_t mbs_cnt
; /* number of classes */
1171 struct omb_class_stat mbs_class
[1]; /* class array */
1172 } __attribute__((__packed__
));
1173 #endif /* XNU_KERNEL_PRIVATE */
1175 typedef struct mb_stat
{
1176 u_int32_t mbs_cnt
; /* number of classes */
1177 #if defined(KERNEL) || defined(__LP64__)
1178 u_int32_t mbs_pad
; /* padding */
1179 #endif /* KERNEL || __LP64__ */
1180 mb_class_stat_t mbs_class
[1]; /* class array */
1184 #define MLEAK_STACK_DEPTH 16 /* Max PC stack depth */
1186 typedef struct mleak_trace_stat
{
1187 u_int64_t mltr_collisions
;
1188 u_int64_t mltr_hitcount
;
1189 u_int64_t mltr_allocs
;
1190 u_int64_t mltr_depth
;
1191 u_int64_t mltr_addr
[MLEAK_STACK_DEPTH
];
1192 } mleak_trace_stat_t
;
1194 typedef struct mleak_stat
{
1195 u_int32_t ml_isaddr64
; /* 64-bit KVA? */
1196 u_int32_t ml_cnt
; /* number of traces */
1197 mleak_trace_stat_t ml_trace
[1]; /* trace array */
1200 struct mleak_table
{
1201 u_int32_t mleak_capture
; /* sampling capture counter */
1202 u_int32_t mleak_sample_factor
; /* sample factor */
1204 /* Times two active records want to occupy the same spot */
1205 u_int64_t alloc_collisions
;
1206 u_int64_t trace_collisions
;
1208 /* Times new record lands on spot previously occupied by freed alloc */
1209 u_int64_t alloc_overwrites
;
1210 u_int64_t trace_overwrites
;
1212 /* Times a new alloc or trace is put into the hash table */
1213 u_int64_t alloc_recorded
;
1214 u_int64_t trace_recorded
;
1216 /* Total number of outstanding allocs */
1217 u_int64_t outstanding_allocs
;
1219 /* Times mleak_log returned false because couldn't acquire the lock */
1220 u_int64_t total_conflicts
;
1222 #endif /* PRIVATE */
1224 #ifdef KERNEL_PRIVATE
1228 * Exported (private)
1231 extern struct mbstat mbstat
; /* statistics */
1234 #endif /* KERNEL_PRIVATE */
1236 #ifdef XNU_KERNEL_PRIVATE
1240 * Not exported (xnu private)
1243 /* flags to m_get/MGET */
1244 /* Need to include malloc.h to get right options for malloc */
1245 #include <sys/malloc.h>
1249 /* length to m_copy to copy all */
1250 #define M_COPYALL 1000000000
1252 #define M_DONTWAIT M_NOWAIT
1253 #define M_WAIT M_WAITOK
1255 /* modes for m_copym and variants */
1256 #define M_COPYM_NOOP_HDR 0 /* don't copy/move pkthdr contents */
1257 #define M_COPYM_COPY_HDR 1 /* copy pkthdr from old to new */
1258 #define M_COPYM_MOVE_HDR 2 /* move pkthdr from old to new */
1259 #define M_COPYM_MUST_COPY_HDR 3 /* MUST copy pkthdr from old to new */
1260 #define M_COPYM_MUST_MOVE_HDR 4 /* MUST move pkthdr from old to new */
1262 extern void m_freem(struct mbuf
*) __XNU_INTERNAL(m_freem
);
1263 extern u_int64_t
mcl_to_paddr(char *);
1264 extern void m_adj(struct mbuf
*, int);
1265 extern void m_cat(struct mbuf
*, struct mbuf
*);
1266 extern void m_copydata(struct mbuf
*, int, int, void *);
1267 extern struct mbuf
*m_copym(struct mbuf
*, int, int, int);
1268 extern struct mbuf
*m_copym_mode(struct mbuf
*, int, int, int, uint32_t);
1269 extern struct mbuf
*m_get(int, int);
1270 extern struct mbuf
*m_gethdr(int, int);
1271 extern struct mbuf
*m_getpacket(void);
1272 extern struct mbuf
*m_getpackets(int, int, int);
1273 extern struct mbuf
*m_mclget(struct mbuf
*, int);
1274 extern void *m_mtod(struct mbuf
*);
1275 extern struct mbuf
*m_prepend_2(struct mbuf
*, int, int, int);
1276 extern struct mbuf
*m_pullup(struct mbuf
*, int);
1277 extern struct mbuf
*m_split(struct mbuf
*, int, int);
1278 extern void m_mclfree(caddr_t p
);
1281 * On platforms which require strict alignment (currently for anything but
1282 * i386 or x86_64), this macro checks whether the data pointer of an mbuf
1283 * is 32-bit aligned (this is the expected minimum alignment for protocol
1284 * headers), and assert otherwise.
1286 #if defined(__i386__) || defined(__x86_64__)
1287 #define MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(_m)
1288 #else /* !__i386__ && !__x86_64__ */
1289 #define MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(_m) do { \
1290 if (!IS_P2ALIGNED((_m)->m_data, sizeof (u_int32_t))) { \
1291 if (((_m)->m_flags & M_PKTHDR) && \
1292 (_m)->m_pkthdr.rcvif != NULL) { \
1293 panic_plain("\n%s: mbuf %p data ptr %p is not " \
1294 "32-bit aligned [%s: alignerrs=%lld]\n", \
1295 __func__, (_m), (_m)->m_data, \
1296 if_name((_m)->m_pkthdr.rcvif), \
1297 (_m)->m_pkthdr.rcvif->if_alignerrs); \
1299 panic_plain("\n%s: mbuf %p data ptr %p is not " \
1300 "32-bit aligned\n", \
1301 __func__, (_m), (_m)->m_data); \
1305 #endif /* !__i386__ && !__x86_64__ */
1307 /* Maximum number of MBUF_SC values (excluding MBUF_SC_UNSPEC) */
1308 #define MBUF_SC_MAX_CLASSES 10
1311 * These conversion macros rely on the corresponding MBUF_SC and
1312 * MBUF_TC values in order to establish the following mapping:
1314 * MBUF_SC_BK_SYS ] ==> MBUF_TC_BK
1317 * MBUF_SC_BE ] ==> MBUF_TC_BE
1321 * MBUF_SC_AV ] ==> MBUF_TC_VI
1326 * MBUF_SC_VO ] ==> MBUF_TC_VO
1329 * The values assigned to each service class allows for a fast mapping to
1330 * the corresponding MBUF_TC traffic class values, as well as to retrieve the
1331 * assigned index; therefore care must be taken when comparing against these
1332 * values. Use the corresponding class and index macros to retrieve the
1333 * corresponding portion, and never assume that a higher class corresponds
1334 * to a higher index.
1336 #define MBUF_SCVAL(x) ((x) & 0xffff)
1337 #define MBUF_SCIDX(x) ((((x) >> 16) & 0xff) >> 3)
1338 #define MBUF_SC2TC(_sc) (MBUF_SCVAL(_sc) >> 7)
1339 #define MBUF_TC2SCVAL(_tc) ((_tc) << 7)
1340 #define IS_MBUF_SC_BACKGROUND(_sc) (((_sc) == MBUF_SC_BK_SYS) || \
1341 ((_sc) == MBUF_SC_BK))
1342 #define IS_MBUF_SC_REALTIME(_sc) ((_sc) >= MBUF_SC_AV && (_sc) <= MBUF_SC_VO)
1343 #define IS_MBUF_SC_BESTEFFORT(_sc) ((_sc) == MBUF_SC_BE || \
1344 (_sc) == MBUF_SC_RD || (_sc) == MBUF_SC_OAM)
1346 #define SCIDX_BK_SYS MBUF_SCIDX(MBUF_SC_BK_SYS)
1347 #define SCIDX_BK MBUF_SCIDX(MBUF_SC_BK)
1348 #define SCIDX_BE MBUF_SCIDX(MBUF_SC_BE)
1349 #define SCIDX_RD MBUF_SCIDX(MBUF_SC_RD)
1350 #define SCIDX_OAM MBUF_SCIDX(MBUF_SC_OAM)
1351 #define SCIDX_AV MBUF_SCIDX(MBUF_SC_AV)
1352 #define SCIDX_RV MBUF_SCIDX(MBUF_SC_RV)
1353 #define SCIDX_VI MBUF_SCIDX(MBUF_SC_VI)
1354 #define SCIDX_SIG MBUF_SCIDX(MBUF_SC_SIG)
1355 #define SCIDX_VO MBUF_SCIDX(MBUF_SC_VO)
1356 #define SCIDX_CTL MBUF_SCIDX(MBUF_SC_CTL)
1358 #define SCVAL_BK_SYS MBUF_SCVAL(MBUF_SC_BK_SYS)
1359 #define SCVAL_BK MBUF_SCVAL(MBUF_SC_BK)
1360 #define SCVAL_BE MBUF_SCVAL(MBUF_SC_BE)
1361 #define SCVAL_RD MBUF_SCVAL(MBUF_SC_RD)
1362 #define SCVAL_OAM MBUF_SCVAL(MBUF_SC_OAM)
1363 #define SCVAL_AV MBUF_SCVAL(MBUF_SC_AV)
1364 #define SCVAL_RV MBUF_SCVAL(MBUF_SC_RV)
1365 #define SCVAL_VI MBUF_SCVAL(MBUF_SC_VI)
1366 #define SCVAL_SIG MBUF_SCVAL(MBUF_SC_SIG)
1367 #define SCVAL_VO MBUF_SCVAL(MBUF_SC_VO)
1368 #define SCVAL_CTL MBUF_SCVAL(MBUF_SC_CTL)
1370 #define MBUF_VALID_SC(c) \
1371 (c == MBUF_SC_BK_SYS || c == MBUF_SC_BK || c == MBUF_SC_BE || \
1372 c == MBUF_SC_RD || c == MBUF_SC_OAM || c == MBUF_SC_AV || \
1373 c == MBUF_SC_RV || c == MBUF_SC_VI || c == MBUF_SC_SIG || \
1374 c == MBUF_SC_VO || c == MBUF_SC_CTL)
1376 #define MBUF_VALID_SCIDX(c) \
1377 (c == SCIDX_BK_SYS || c == SCIDX_BK || c == SCIDX_BE || \
1378 c == SCIDX_RD || c == SCIDX_OAM || c == SCIDX_AV || \
1379 c == SCIDX_RV || c == SCIDX_VI || c == SCIDX_SIG || \
1380 c == SCIDX_VO || c == SCIDX_CTL)
1382 #define MBUF_VALID_SCVAL(c) \
1383 (c == SCVAL_BK_SYS || c == SCVAL_BK || c == SCVAL_BE || \
1384 c == SCVAL_RD || c == SCVAL_OAM || c == SCVAL_AV || \
1385 c == SCVAL_RV || c == SCVAL_VI || c == SCVAL_SIG || \
1386 c == SCVAL_VO || SCVAL_CTL)
1388 extern unsigned char *mbutl
; /* start VA of mbuf pool */
1389 extern unsigned char *embutl
; /* end VA of mbuf pool */
1390 extern unsigned int nmbclusters
; /* number of mapped clusters */
1391 extern int njcl
; /* # of jumbo clusters */
1392 extern int njclbytes
; /* size of a jumbo cluster */
1393 extern int max_hdr
; /* largest link+protocol header */
1394 extern int max_datalen
; /* MHLEN - max_hdr */
1396 /* Use max_linkhdr instead of _max_linkhdr */
1397 extern int _max_linkhdr
; /* largest link-level header */
1399 /* Use max_protohdr instead of _max_protohdr */
1400 extern int _max_protohdr
; /* largest protocol header */
1402 __private_extern__
unsigned int mbuf_default_ncl(uint64_t);
1403 __private_extern__
void mbinit(void);
1404 __private_extern__
struct mbuf
*m_clattach(struct mbuf
*, int, caddr_t
,
1405 void (*)(caddr_t
, u_int
, caddr_t
), u_int
, caddr_t
, int, int);
1406 __private_extern__ caddr_t
m_bigalloc(int);
1407 __private_extern__
void m_bigfree(caddr_t
, u_int
, caddr_t
);
1408 __private_extern__
struct mbuf
*m_mbigget(struct mbuf
*, int);
1409 __private_extern__ caddr_t
m_16kalloc(int);
1410 __private_extern__
void m_16kfree(caddr_t
, u_int
, caddr_t
);
1411 __private_extern__
struct mbuf
*m_m16kget(struct mbuf
*, int);
1412 __private_extern__
int m_reinit(struct mbuf
*, int);
1413 __private_extern__
struct mbuf
*m_free(struct mbuf
*) __XNU_INTERNAL(m_free
);
1414 __private_extern__
struct mbuf
*m_getclr(int, int);
1415 __private_extern__
struct mbuf
*m_getptr(struct mbuf
*, int, int *);
1416 __private_extern__
unsigned int m_length(struct mbuf
*);
1417 __private_extern__
unsigned int m_length2(struct mbuf
*, struct mbuf
**);
1418 __private_extern__
unsigned int m_fixhdr(struct mbuf
*);
1419 __private_extern__
struct mbuf
*m_defrag(struct mbuf
*, int);
1420 __private_extern__
struct mbuf
*m_defrag_offset(struct mbuf
*, u_int32_t
, int);
1421 __private_extern__
struct mbuf
*m_prepend(struct mbuf
*, int, int);
1422 __private_extern__
struct mbuf
*m_copyup(struct mbuf
*, int, int);
1423 __private_extern__
struct mbuf
*m_retry(int, int);
1424 __private_extern__
struct mbuf
*m_retryhdr(int, int);
1425 __private_extern__
int m_freem_list(struct mbuf
*);
1426 __private_extern__
int m_append(struct mbuf
*, int, caddr_t
);
1427 __private_extern__
struct mbuf
*m_last(struct mbuf
*);
1428 __private_extern__
struct mbuf
*m_devget(char *, int, int, struct ifnet
*,
1429 void (*)(const void *, void *, size_t));
1430 __private_extern__
struct mbuf
*m_pulldown(struct mbuf
*, int, int, int *);
1432 __private_extern__
struct mbuf
*m_getcl(int, int, int);
1433 __private_extern__ caddr_t
m_mclalloc(int);
1434 __private_extern__
int m_mclhasreference(struct mbuf
*);
1435 __private_extern__
void m_copy_pkthdr(struct mbuf
*, struct mbuf
*);
1436 __private_extern__
void m_copy_pftag(struct mbuf
*, struct mbuf
*);
1437 __private_extern__
void m_copy_classifier(struct mbuf
*, struct mbuf
*);
1439 __private_extern__
struct mbuf
*m_dtom(void *);
1440 __private_extern__
int m_mtocl(void *);
1441 __private_extern__
union mcluster
*m_cltom(int);
1443 __private_extern__
void m_align(struct mbuf
*, int);
1445 __private_extern__
struct mbuf
*m_normalize(struct mbuf
*m
);
1446 __private_extern__
void m_mchtype(struct mbuf
*m
, int t
);
1447 __private_extern__
void m_mcheck(struct mbuf
*);
1449 __private_extern__
void m_copyback(struct mbuf
*, int, int, const void *);
1450 __private_extern__
struct mbuf
*m_copyback_cow(struct mbuf
*, int, int,
1452 __private_extern__
int m_makewritable(struct mbuf
**, int, int, int);
1453 __private_extern__
struct mbuf
*m_dup(struct mbuf
*m
, int how
);
1454 __private_extern__
struct mbuf
*m_copym_with_hdrs(struct mbuf
*, int, int, int,
1455 struct mbuf
**, int *, uint32_t);
1456 __private_extern__
struct mbuf
*m_getpackethdrs(int, int);
1457 __private_extern__
struct mbuf
*m_getpacket_how(int);
1458 __private_extern__
struct mbuf
*m_getpackets_internal(unsigned int *, int,
1460 __private_extern__
struct mbuf
*m_allocpacket_internal(unsigned int *, size_t,
1461 unsigned int *, int, int, size_t);
1463 __private_extern__
int m_ext_set_prop(struct mbuf
*, uint32_t, uint32_t);
1464 __private_extern__
uint32_t m_ext_get_prop(struct mbuf
*);
1465 __private_extern__
int m_ext_paired_is_active(struct mbuf
*);
1466 __private_extern__
void m_ext_paired_activate(struct mbuf
*);
1468 __private_extern__
void mbuf_drain(boolean_t
);
1471 * Packets may have annotations attached by affixing a list of "packet
1472 * tags" to the pkthdr structure. Packet tags are dynamically allocated
1473 * semi-opaque data structures that have a fixed header (struct m_tag)
1474 * that specifies the size of the memory block and an <id,type> pair that
1475 * identifies it. The id identifies the module and the type identifies the
1476 * type of data for that module. The id of zero is reserved for the kernel.
1478 * Note that the packet tag returned by m_tag_allocate has the default
1479 * memory alignment implemented by malloc. To reference private data one
1480 * can use a construct like:
1482 * struct m_tag *mtag = m_tag_allocate(...);
1483 * struct foo *p = (struct foo *)(mtag+1);
1485 * if the alignment of struct m_tag is sufficient for referencing members
1486 * of struct foo. Otherwise it is necessary to embed struct m_tag within
1487 * the private data structure to insure proper alignment; e.g.
1493 * struct foo *p = (struct foo *) m_tag_allocate(...);
1494 * struct m_tag *mtag = &p->tag;
1497 #define KERNEL_MODULE_TAG_ID 0
1500 KERNEL_TAG_TYPE_NONE
= 0,
1501 KERNEL_TAG_TYPE_DUMMYNET
= 1,
1502 KERNEL_TAG_TYPE_DIVERT
= 2,
1503 KERNEL_TAG_TYPE_IPFORWARD
= 3,
1504 KERNEL_TAG_TYPE_IPFILT
= 4,
1505 KERNEL_TAG_TYPE_MACLABEL
= 5,
1506 KERNEL_TAG_TYPE_MAC_POLICY_LABEL
= 6,
1507 KERNEL_TAG_TYPE_ENCAP
= 8,
1508 KERNEL_TAG_TYPE_INET6
= 9,
1509 KERNEL_TAG_TYPE_IPSEC
= 10,
1510 KERNEL_TAG_TYPE_DRVAUX
= 11,
1511 KERNEL_TAG_TYPE_CFIL_UDP
= 13,
1512 KERNEL_TAG_TYPE_PF_REASS
= 14,
1515 /* Packet tag routines */
1516 __private_extern__
struct m_tag
*m_tag_alloc(u_int32_t
, u_int16_t
, int, int);
1517 __private_extern__
struct m_tag
*m_tag_create(u_int32_t
, u_int16_t
, int, int,
1519 __private_extern__
void m_tag_free(struct m_tag
*);
1520 __private_extern__
void m_tag_prepend(struct mbuf
*, struct m_tag
*);
1521 __private_extern__
void m_tag_unlink(struct mbuf
*, struct m_tag
*);
1522 __private_extern__
void m_tag_delete(struct mbuf
*, struct m_tag
*);
1523 __private_extern__
void m_tag_delete_chain(struct mbuf
*, struct m_tag
*);
1524 __private_extern__
struct m_tag
*m_tag_locate(struct mbuf
*, u_int32_t
,
1525 u_int16_t
, struct m_tag
*);
1526 __private_extern__
struct m_tag
*m_tag_copy(struct m_tag
*, int);
1527 __private_extern__
int m_tag_copy_chain(struct mbuf
*, struct mbuf
*, int);
1528 __private_extern__
void m_tag_init(struct mbuf
*, int);
1529 __private_extern__
struct m_tag
*m_tag_first(struct mbuf
*);
1530 __private_extern__
struct m_tag
*m_tag_next(struct mbuf
*, struct m_tag
*);
1532 __private_extern__
void m_scratch_init(struct mbuf
*);
1533 __private_extern__ u_int32_t
m_scratch_get(struct mbuf
*, u_int8_t
**);
1535 __private_extern__
void m_classifier_init(struct mbuf
*, uint32_t);
1537 __private_extern__
int m_set_service_class(struct mbuf
*, mbuf_svc_class_t
);
1538 __private_extern__ mbuf_svc_class_t
m_get_service_class(struct mbuf
*);
1539 __private_extern__ mbuf_svc_class_t
m_service_class_from_idx(u_int32_t
);
1540 __private_extern__ mbuf_svc_class_t
m_service_class_from_val(u_int32_t
);
1541 __private_extern__
int m_set_traffic_class(struct mbuf
*, mbuf_traffic_class_t
);
1542 __private_extern__ mbuf_traffic_class_t
m_get_traffic_class(struct mbuf
*);
1544 #define ADDCARRY(_x) do { \
1545 while (((_x) >> 16) != 0) \
1546 (_x) = ((_x) >> 16) + ((_x) & 0xffff); \
1549 __private_extern__ u_int16_t
m_adj_sum16(struct mbuf
*, u_int32_t
,
1550 u_int32_t
, u_int32_t
, u_int32_t
);
1551 __private_extern__ u_int16_t
m_sum16(struct mbuf
*, u_int32_t
, u_int32_t
);
1553 __private_extern__
void m_set_ext(struct mbuf
*, struct ext_ref
*,
1554 m_ext_free_func_t
, caddr_t
);
1555 __private_extern__
struct ext_ref
*m_get_rfa(struct mbuf
*);
1556 __private_extern__ m_ext_free_func_t
m_get_ext_free(struct mbuf
*);
1557 __private_extern__ caddr_t
m_get_ext_arg(struct mbuf
*);
1559 __private_extern__
void m_do_tx_compl_callback(struct mbuf
*, struct ifnet
*);
1560 __private_extern__ mbuf_tx_compl_func
m_get_tx_compl_callback(u_int32_t
);
1563 #endif /* XNU_KERNEL_PRIVATE */
1564 #endif /* !_SYS_MBUF_H_ */