1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 * @(#)mbuf.h 8.5 (Berkeley) 2/19/95
33 * $FreeBSD: stable/12/sys/sys/mbuf.h 371048 2021-11-19 06:40:25Z kp $
34 */
35
36 #ifndef _SYS_MBUF_H_
37 #define _SYS_MBUF_H_
38
39 /* XXX: These includes suck. Sorry! */
40 #include <sys/queue.h>
41 #ifdef _KERNEL
42 #include <sys/systm.h>
43 #include <vm/uma.h>
44 #ifdef WITNESS
45 #include <sys/lock.h>
46 #endif
47 #endif
48
49 #ifdef _KERNEL
50 #include <sys/sdt.h>
51
52 #define MBUF_PROBE1(probe, arg0) \
53 SDT_PROBE1(sdt, , , probe, arg0)
54 #define MBUF_PROBE2(probe, arg0, arg1) \
55 SDT_PROBE2(sdt, , , probe, arg0, arg1)
56 #define MBUF_PROBE3(probe, arg0, arg1, arg2) \
57 SDT_PROBE3(sdt, , , probe, arg0, arg1, arg2)
58 #define MBUF_PROBE4(probe, arg0, arg1, arg2, arg3) \
59 SDT_PROBE4(sdt, , , probe, arg0, arg1, arg2, arg3)
60 #define MBUF_PROBE5(probe, arg0, arg1, arg2, arg3, arg4) \
61 SDT_PROBE5(sdt, , , probe, arg0, arg1, arg2, arg3, arg4)
62
63 SDT_PROBE_DECLARE(sdt, , , m__init);
64 SDT_PROBE_DECLARE(sdt, , , m__gethdr);
65 SDT_PROBE_DECLARE(sdt, , , m__get);
66 SDT_PROBE_DECLARE(sdt, , , m__getcl);
67 SDT_PROBE_DECLARE(sdt, , , m__getjcl);
68 SDT_PROBE_DECLARE(sdt, , , m__clget);
69 SDT_PROBE_DECLARE(sdt, , , m__cljget);
70 SDT_PROBE_DECLARE(sdt, , , m__cljset);
71 SDT_PROBE_DECLARE(sdt, , , m__free);
72 SDT_PROBE_DECLARE(sdt, , , m__freem);
73
74 #endif /* _KERNEL */
75
76 /*
77 * Mbufs are of a single size, MSIZE (sys/param.h), which includes overhead.
78 * An mbuf may add a single "mbuf cluster" of size MCLBYTES (also in
79 * sys/param.h), which has no additional overhead and is used instead of the
80 * internal data area; this is done when at least MINCLSIZE of data must be
81 * stored. Additionally, it is possible to allocate a separate buffer
82 * externally and attach it to the mbuf in a way similar to that of mbuf
83 * clusters.
84 *
85 * NB: These calculation do not take actual compiler-induced alignment and
86 * padding inside the complete struct mbuf into account. Appropriate
87 * attention is required when changing members of struct mbuf.
88 *
89 * MLEN is data length in a normal mbuf.
90 * MHLEN is data length in an mbuf with pktheader.
91 * MINCLSIZE is a smallest amount of data that should be put into cluster.
92 *
93 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
94 * they are sensible.
95 */
96 struct mbuf;
97 #define MHSIZE offsetof(struct mbuf, m_dat)
98 #define MPKTHSIZE offsetof(struct mbuf, m_pktdat)
99 #define MLEN ((int)(MSIZE - MHSIZE))
100 #define MHLEN ((int)(MSIZE - MPKTHSIZE))
101 #define MINCLSIZE (MHLEN + 1)
102
103 #ifdef _KERNEL
104 /*-
105 * Macro for type conversion: convert mbuf pointer to data pointer of correct
106 * type:
107 *
108 * mtod(m, t) -- Convert mbuf pointer to data pointer of correct type.
109 * mtodo(m, o) -- Same as above but with offset 'o' into data.
110 */
111 #define mtod(m, t) ((t)((m)->m_data))
112 #define mtodo(m, o) ((void *)(((m)->m_data) + (o)))
113
114 /*
115 * Argument structure passed to UMA routines during mbuf and packet
116 * allocations.
117 */
118 struct mb_args {
119 int flags; /* Flags for mbuf being allocated */
120 short type; /* Type of mbuf being allocated */
121 };
122 #endif /* _KERNEL */
123
124 /*
125 * Packet tag structure (see below for details).
126 */
127 struct m_tag {
128 SLIST_ENTRY(m_tag) m_tag_link; /* List of packet tags */
129 u_int16_t m_tag_id; /* Tag ID */
130 u_int16_t m_tag_len; /* Length of data */
131 u_int32_t m_tag_cookie; /* ABI/Module ID */
132 void (*m_tag_free)(struct m_tag *);
133 };
134
135 /*
136 * Static network interface owned tag.
137 * Allocated through ifp->if_snd_tag_alloc().
138 */
139 struct m_snd_tag {
140 struct ifnet *ifp; /* network interface tag belongs to */
141 };
142
143 /*
144 * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
145 * Size ILP32: 48
146 * LP64: 56
147 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
148 * they are correct.
149 */
150 struct pkthdr {
151 union {
152 struct m_snd_tag *snd_tag; /* send tag, if any */
153 struct ifnet *rcvif; /* rcv interface */
154 };
155 SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
156 int32_t len; /* total packet length */
157
158 /* Layer crossing persistent information. */
159 uint32_t flowid; /* packet's 4-tuple system */
160 uint32_t csum_flags; /* checksum and offload features */
161 uint16_t fibnum; /* this packet should use this fib */
162 uint8_t cosqos; /* class/quality of service */
163 uint8_t rsstype; /* hash type */
164 union {
165 uint64_t rcv_tstmp; /* timestamp in ns */
166 struct {
167 uint8_t l2hlen; /* layer 2 hdr len */
168 uint8_t l3hlen; /* layer 3 hdr len */
169 uint8_t l4hlen; /* layer 4 hdr len */
170 uint8_t l5hlen; /* layer 5 hdr len */
171 uint8_t inner_l2hlen;
172 uint8_t inner_l3hlen;
173 uint8_t inner_l4hlen;
174 uint8_t inner_l5hlen;
175 };
176 };
177 union {
178 uint8_t eight[8];
179 uint16_t sixteen[4];
180 uint32_t thirtytwo[2];
181 uint64_t sixtyfour[1];
182 uintptr_t unintptr[1];
183 void *ptr;
184 } PH_per;
185
186 /* Layer specific non-persistent local storage for reassembly, etc. */
187 union {
188 uint8_t eight[8];
189 uint16_t sixteen[4];
190 uint32_t thirtytwo[2];
191 uint64_t sixtyfour[1];
192 uintptr_t unintptr[1];
193 void *ptr;
194 } PH_loc;
195 };
196 #define ether_vtag PH_per.sixteen[0]
197 #define PH_vt PH_per
198 #define vt_nrecs sixteen[0]
199 #define tso_segsz PH_per.sixteen[1]
200 #define lro_nsegs tso_segsz
201 #define csum_phsum PH_per.sixteen[2]
202 #define csum_data PH_per.thirtytwo[1]
203 #define pace_thoff PH_loc.sixteen[0]
204 #define pace_tlen PH_loc.sixteen[1]
205 #define pace_drphdrlen PH_loc.sixteen[2]
206 #define pace_tos PH_loc.eight[6]
207 #define pace_lock PH_loc.eight[7]
208
209 /*
210 * Description of external storage mapped into mbuf; valid only if M_EXT is
211 * set.
212 * Size ILP32: 28
213 * LP64: 48
214 * Compile-time assertions in uipc_mbuf.c test these values to ensure that
215 * they are correct.
216 */
217 typedef void m_ext_free_t(struct mbuf *);
218 struct m_ext {
219 union {
220 /*
221 * If EXT_FLAG_EMBREF is set, then we use refcount in the
222 * mbuf, the 'ext_count' member. Otherwise, we have a
223 * shadow copy and we use pointer 'ext_cnt'. The original
224 * mbuf is responsible to carry the pointer to free routine
225 * and its arguments. They aren't copied into shadows in
226 * mb_dupcl() to avoid dereferencing next cachelines.
227 */
228 volatile u_int ext_count;
229 volatile u_int *ext_cnt;
230 };
231 char *ext_buf; /* start of buffer */
232 uint32_t ext_size; /* size of buffer, for ext_free */
233 uint32_t ext_type:8, /* type of external storage */
234 ext_flags:24; /* external storage mbuf flags */
235 /*
236 * Fields below store the free context for the external storage.
237 * They are valid only in the refcount carrying mbuf, the one with
238 * EXT_FLAG_EMBREF flag, with exclusion for EXT_EXTREF type, where
239 * the free context is copied into all mbufs that use same external
240 * storage.
241 */
242 #define m_ext_copylen offsetof(struct m_ext, ext_free)
243 m_ext_free_t *ext_free; /* free routine if not the usual */
244 void *ext_arg1; /* optional argument pointer */
245 void *ext_arg2; /* optional argument pointer */
246 };
247
248 /*
249 * The core of the mbuf object along with some shortcut defines for practical
250 * purposes.
251 */
252 struct mbuf {
253 /*
254 * Header present at the beginning of every mbuf.
255 * Size ILP32: 24
256 * LP64: 32
257 * Compile-time assertions in uipc_mbuf.c test these values to ensure
258 * that they are correct.
259 */
260 union { /* next buffer in chain */
261 struct mbuf *m_next;
262 SLIST_ENTRY(mbuf) m_slist;
263 STAILQ_ENTRY(mbuf) m_stailq;
264 };
265 union { /* next chain in queue/record */
266 struct mbuf *m_nextpkt;
267 SLIST_ENTRY(mbuf) m_slistpkt;
268 STAILQ_ENTRY(mbuf) m_stailqpkt;
269 };
270 caddr_t m_data; /* location of data */
271 int32_t m_len; /* amount of data in this mbuf */
272 uint32_t m_type:8, /* type of data in this mbuf */
273 m_flags:24; /* flags; see below */
274 #if !defined(__LP64__)
275 uint32_t m_pad; /* pad for 64bit alignment */
276 #endif
277
278 /*
279 * A set of optional headers (packet header, external storage header)
280 * and internal data storage. Historically, these arrays were sized
281 * to MHLEN (space left after a packet header) and MLEN (space left
282 * after only a regular mbuf header); they are now variable size in
283 * order to support future work on variable-size mbufs.
284 */
285 union {
286 struct {
287 struct pkthdr m_pkthdr; /* M_PKTHDR set */
288 union {
289 struct m_ext m_ext; /* M_EXT set */
290 char m_pktdat[0];
291 };
292 };
293 char m_dat[0]; /* !M_PKTHDR, !M_EXT */
294 };
295 };
296
297 /*
298 * mbuf flags of global significance and layer crossing.
299 * Those of only protocol/layer specific significance are to be mapped
300 * to M_PROTO[1-12] and cleared at layer handoff boundaries.
301 * NB: Limited to the lower 24 bits.
302 */
303 #define M_EXT 0x00000001 /* has associated external storage */
304 #define M_PKTHDR 0x00000002 /* start of record */
305 #define M_EOR 0x00000004 /* end of record */
306 #define M_RDONLY 0x00000008 /* associated data is marked read-only */
307 #define M_BCAST 0x00000010 /* send/received as link-level broadcast */
308 #define M_MCAST 0x00000020 /* send/received as link-level multicast */
309 #define M_PROMISC 0x00000040 /* packet was not for us */
310 #define M_VLANTAG 0x00000080 /* ether_vtag is valid */
311 #define M_NOMAP 0x00000100 /* mbuf data is unmapped (soon from Drew) */
312 #define M_NOFREE 0x00000200 /* do not free mbuf, embedded in cluster */
313 #define M_TSTMP 0x00000400 /* rcv_tstmp field is valid */
314 #define M_TSTMP_HPREC 0x00000800 /* rcv_tstmp is high-prec, typically
315 hw-stamped on port (useful for IEEE 1588
316 and 802.1AS) */
317 #define M_TSTMP_LRO 0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
318
319 #define M_PROTO1 0x00001000 /* protocol-specific */
320 #define M_PROTO2 0x00002000 /* protocol-specific */
321 #define M_PROTO3 0x00004000 /* protocol-specific */
322 #define M_PROTO4 0x00008000 /* protocol-specific */
323 #define M_PROTO5 0x00010000 /* protocol-specific */
324 #define M_PROTO6 0x00020000 /* protocol-specific */
325 #define M_PROTO7 0x00040000 /* protocol-specific */
326 #define M_PROTO8 0x00080000 /* protocol-specific */
327 #define M_PROTO9 0x00100000 /* protocol-specific */
328 #define M_PROTO10 0x00200000 /* protocol-specific */
329 #define M_PROTO11 0x00400000 /* protocol-specific */
330 #define M_PROTO12 0x00800000 /* protocol-specific */
331
332 #define MB_DTOR_SKIP 0x1 /* don't pollute the cache by touching a freed mbuf */
333
334 /*
335 * Flags to purge when crossing layers.
336 */
337 #define M_PROTOFLAGS \
338 (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
339 M_PROTO9|M_PROTO10|M_PROTO11|M_PROTO12)
340
341 /*
342 * Flags preserved when copying m_pkthdr.
343 */
344 #define M_COPYFLAGS \
345 (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
346 M_TSTMP_HPREC|M_PROTOFLAGS)
347
348 /*
349 * Mbuf flag description for use with printf(9) %b identifier.
350 */
351 #define M_FLAG_BITS \
352 "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
353 "\7M_PROMISC\10M_VLANTAG\13M_TSTMP\14M_TSTMP_HPREC"
354 #define M_FLAG_PROTOBITS \
355 "\15M_PROTO1\16M_PROTO2\17M_PROTO3\20M_PROTO4\21M_PROTO5" \
356 "\22M_PROTO6\23M_PROTO7\24M_PROTO8\25M_PROTO9\26M_PROTO10" \
357 "\27M_PROTO11\30M_PROTO12"
358 #define M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
359
360 /*
361 * Network interface cards are able to hash protocol fields (such as IPv4
362 * addresses and TCP port numbers) classify packets into flows. These flows
363 * can then be used to maintain ordering while delivering packets to the OS
364 * via parallel input queues, as well as to provide a stateless affinity
365 * model. NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
366 * m_flag fields to indicate how the hash should be interpreted by the
367 * network stack.
368 *
369 * Most NICs support RSS, which provides ordering and explicit affinity, and
370 * use the hash m_flag bits to indicate what header fields were covered by
371 * the hash. M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
372 * RSS cards or configurations that provide an opaque flow identifier, allowing
373 * for ordering and distribution without explicit affinity. Additionally,
374 * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
375 * properties.
376 *
377 * The meaning of the IPV6_EX suffix:
378 * "o Home address from the home address option in the IPv6 destination
379 * options header. If the extension header is not present, use the Source
380 * IPv6 Address.
381 * o IPv6 address that is contained in the Routing-Header-Type-2 from the
382 * associated extension header. If the extension header is not present,
383 * use the Destination IPv6 Address."
384 * Quoted from:
385 * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
386 */
387 #define M_HASHTYPE_HASHPROP 0x80 /* has hash properties */
388 #define M_HASHTYPE_INNER 0x40 /* calculated from inner headers */
389 #define M_HASHTYPE_HASH(t) (M_HASHTYPE_HASHPROP | (t))
390 /* Microsoft RSS standard hash types */
391 #define M_HASHTYPE_NONE 0
392 #define M_HASHTYPE_RSS_IPV4 M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
393 #define M_HASHTYPE_RSS_TCP_IPV4 M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
394 #define M_HASHTYPE_RSS_IPV6 M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
395 #define M_HASHTYPE_RSS_TCP_IPV6 M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
396 #define M_HASHTYPE_RSS_IPV6_EX M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
397 * ext hdrs */
398 #define M_HASHTYPE_RSS_TCP_IPV6_EX M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
399 * ext hdrs */
400 #define M_HASHTYPE_RSS_UDP_IPV4 M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
401 #define M_HASHTYPE_RSS_UDP_IPV6 M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
402 #define M_HASHTYPE_RSS_UDP_IPV6_EX M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
403 * ext hdrs */
404
405 #define M_HASHTYPE_OPAQUE 0x3f /* ordering, not affinity */
406 #define M_HASHTYPE_OPAQUE_HASH M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
407 /* ordering+hash, not affinity*/
408
409 #define M_HASHTYPE_CLEAR(m) ((m)->m_pkthdr.rsstype = 0)
410 #define M_HASHTYPE_GET(m) ((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER)
411 #define M_HASHTYPE_SET(m, v) ((m)->m_pkthdr.rsstype = (v))
412 #define M_HASHTYPE_TEST(m, v) (M_HASHTYPE_GET(m) == (v))
413 #define M_HASHTYPE_ISHASH(m) \
414 (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0)
415 #define M_HASHTYPE_SETINNER(m) do { \
416 (m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER; \
417 } while (0)
418
419 /*
420 * COS/QOS class and quality of service tags.
421 * It uses DSCP code points as base.
422 */
423 #define QOS_DSCP_CS0 0x00
424 #define QOS_DSCP_DEF QOS_DSCP_CS0
425 #define QOS_DSCP_CS1 0x20
426 #define QOS_DSCP_AF11 0x28
427 #define QOS_DSCP_AF12 0x30
428 #define QOS_DSCP_AF13 0x38
429 #define QOS_DSCP_CS2 0x40
430 #define QOS_DSCP_AF21 0x48
431 #define QOS_DSCP_AF22 0x50
432 #define QOS_DSCP_AF23 0x58
433 #define QOS_DSCP_CS3 0x60
434 #define QOS_DSCP_AF31 0x68
435 #define QOS_DSCP_AF32 0x70
436 #define QOS_DSCP_AF33 0x78
437 #define QOS_DSCP_CS4 0x80
438 #define QOS_DSCP_AF41 0x88
439 #define QOS_DSCP_AF42 0x90
440 #define QOS_DSCP_AF43 0x98
441 #define QOS_DSCP_CS5 0xa0
442 #define QOS_DSCP_EF 0xb8
443 #define QOS_DSCP_CS6 0xc0
444 #define QOS_DSCP_CS7 0xe0
445
446 /*
447 * External mbuf storage buffer types.
448 */
449 #define EXT_CLUSTER 1 /* mbuf cluster */
450 #define EXT_SFBUF 2 /* sendfile(2)'s sf_buf */
451 #define EXT_JUMBOP 3 /* jumbo cluster page sized */
452 #define EXT_JUMBO9 4 /* jumbo cluster 9216 bytes */
453 #define EXT_JUMBO16 5 /* jumbo cluster 16184 bytes */
454 #define EXT_PACKET 6 /* mbuf+cluster from packet zone */
455 #define EXT_MBUF 7 /* external mbuf reference */
456
457 #define EXT_VENDOR1 224 /* for vendor-internal use */
458 #define EXT_VENDOR2 225 /* for vendor-internal use */
459 #define EXT_VENDOR3 226 /* for vendor-internal use */
460 #define EXT_VENDOR4 227 /* for vendor-internal use */
461
462 #define EXT_EXP1 244 /* for experimental use */
463 #define EXT_EXP2 245 /* for experimental use */
464 #define EXT_EXP3 246 /* for experimental use */
465 #define EXT_EXP4 247 /* for experimental use */
466
467 #define EXT_NET_DRV 252 /* custom ext_buf provided by net driver(s) */
468 #define EXT_MOD_TYPE 253 /* custom module's ext_buf type */
469 #define EXT_DISPOSABLE 254 /* can throw this buffer away w/page flipping */
470 #define EXT_EXTREF 255 /* has externally maintained ext_cnt ptr */
471
472 /*
473 * Flags for external mbuf buffer types.
474 * NB: limited to the lower 24 bits.
475 */
476 #define EXT_FLAG_EMBREF 0x000001 /* embedded ext_count */
477 #define EXT_FLAG_EXTREF 0x000002 /* external ext_cnt, notyet */
478
479 #define EXT_FLAG_NOFREE 0x000010 /* don't free mbuf to pool, notyet */
480
481 #define EXT_FLAG_VENDOR1 0x010000 /* These flags are vendor */
482 #define EXT_FLAG_VENDOR2 0x020000 /* or submodule specific, */
483 #define EXT_FLAG_VENDOR3 0x040000 /* not used by mbuf code. */
484 #define EXT_FLAG_VENDOR4 0x080000 /* Set/read by submodule. */
485
486 #define EXT_FLAG_EXP1 0x100000 /* for experimental use */
487 #define EXT_FLAG_EXP2 0x200000 /* for experimental use */
488 #define EXT_FLAG_EXP3 0x400000 /* for experimental use */
489 #define EXT_FLAG_EXP4 0x800000 /* for experimental use */
490
491 /*
492 * EXT flag description for use with printf(9) %b identifier.
493 */
494 #define EXT_FLAG_BITS \
495 "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
496 "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
497 "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
498 "\30EXT_FLAG_EXP4"
499
500 /*
501 * Flags indicating checksum, segmentation and other offload work to be
502 * done, or already done, by hardware or lower layers. It is split into
503 * separate inbound and outbound flags.
504 *
505 * Outbound flags that are set by upper protocol layers requesting lower
506 * layers, or ideally the hardware, to perform these offloading tasks.
507 * For outbound packets this field and its flags can be directly tested
508 * against ifnet if_hwassist. Note that the outbound and the inbound flags do
509 * not collide right now but they could be allowed to (as long as the flags are
510 * scrubbed appropriately when the direction of an mbuf changes). CSUM_BITS
511 * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
512 *
513 * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
514 * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
515 */
516 #define CSUM_IP 0x00000001 /* IP header checksum offload */
517 #define CSUM_IP_UDP 0x00000002 /* UDP checksum offload */
518 #define CSUM_IP_TCP 0x00000004 /* TCP checksum offload */
519 #define CSUM_IP_SCTP 0x00000008 /* SCTP checksum offload */
520 #define CSUM_IP_TSO 0x00000010 /* TCP segmentation offload */
521 #define CSUM_IP_ISCSI 0x00000020 /* iSCSI checksum offload */
522
523 #define CSUM_INNER_IP6_UDP 0x00000040
524 #define CSUM_INNER_IP6_TCP 0x00000080
525 #define CSUM_INNER_IP6_TSO 0x00000100
526 #define CSUM_IP6_UDP 0x00000200 /* UDP checksum offload */
527 #define CSUM_IP6_TCP 0x00000400 /* TCP checksum offload */
528 #define CSUM_IP6_SCTP 0x00000800 /* SCTP checksum offload */
529 #define CSUM_IP6_TSO 0x00001000 /* TCP segmentation offload */
530 #define CSUM_IP6_ISCSI 0x00002000 /* iSCSI checksum offload */
531
532 #define CSUM_INNER_IP 0x00004000
533 #define CSUM_INNER_IP_UDP 0x00008000
534 #define CSUM_INNER_IP_TCP 0x00010000
535 #define CSUM_INNER_IP_TSO 0x00020000
536
537 #define CSUM_ENCAP_VXLAN 0x00040000 /* VXLAN outer encapsulation */
538 #define CSUM_ENCAP_RSVD1 0x00080000
539
540 /* Inbound checksum support where the checksum was verified by hardware. */
541 #define CSUM_INNER_L3_CALC 0x00100000
542 #define CSUM_INNER_L3_VALID 0x00200000
543 #define CSUM_INNER_L4_CALC 0x00400000
544 #define CSUM_INNER_L4_VALID 0x00800000
545 #define CSUM_L3_CALC 0x01000000 /* calculated layer 3 csum */
546 #define CSUM_L3_VALID 0x02000000 /* checksum is correct */
547 #define CSUM_L4_CALC 0x04000000 /* calculated layer 4 csum */
548 #define CSUM_L4_VALID 0x08000000 /* checksum is correct */
549 #define CSUM_L5_CALC 0x10000000 /* calculated layer 5 csum */
550 #define CSUM_L5_VALID 0x20000000 /* checksum is correct */
551 #define CSUM_COALESCED 0x40000000 /* contains merged segments */
552
553 #define CSUM_SND_TAG 0x80000000 /* Packet header has send tag */
554
555 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
556 CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
557 CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
558 CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
559 CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
560 CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
561
562 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
563 CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
564 CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
565 CSUM_COALESCED)
566
567 /*
568 * CSUM flag description for use with printf(9) %b identifier.
569 */
570 #define CSUM_BITS \
571 "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
572 "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
573 "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
574 "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
575 "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
576 "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
577 "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
578 "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
579 "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
580
581 /* CSUM flags compatibility mappings. */
582 #define CSUM_IP_CHECKED CSUM_L3_CALC
583 #define CSUM_IP_VALID CSUM_L3_VALID
584 #define CSUM_DATA_VALID CSUM_L4_VALID
585 #define CSUM_PSEUDO_HDR CSUM_L4_CALC
586 #define CSUM_SCTP_VALID CSUM_L4_VALID
587 #define CSUM_DELAY_DATA (CSUM_TCP|CSUM_UDP)
588 #define CSUM_DELAY_IP CSUM_IP /* Only v4, no v6 IP hdr csum */
589 #define CSUM_DELAY_DATA_IPV6 (CSUM_TCP_IPV6|CSUM_UDP_IPV6)
590 #define CSUM_DATA_VALID_IPV6 CSUM_DATA_VALID
591 #define CSUM_TCP CSUM_IP_TCP
592 #define CSUM_UDP CSUM_IP_UDP
593 #define CSUM_SCTP CSUM_IP_SCTP
594 #define CSUM_TSO (CSUM_IP_TSO|CSUM_IP6_TSO)
595 #define CSUM_INNER_TSO (CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
596 #define CSUM_UDP_IPV6 CSUM_IP6_UDP
597 #define CSUM_TCP_IPV6 CSUM_IP6_TCP
598 #define CSUM_SCTP_IPV6 CSUM_IP6_SCTP
599
600 /*
601 * mbuf types describing the content of the mbuf (including external storage).
602 */
603 #define MT_NOTMBUF 0 /* USED INTERNALLY ONLY! Object is not mbuf */
604 #define MT_DATA 1 /* dynamic (data) allocation */
605 #define MT_HEADER MT_DATA /* packet header, use M_PKTHDR instead */
606
607 #define MT_VENDOR1 4 /* for vendor-internal use */
608 #define MT_VENDOR2 5 /* for vendor-internal use */
609 #define MT_VENDOR3 6 /* for vendor-internal use */
610 #define MT_VENDOR4 7 /* for vendor-internal use */
611
612 #define MT_SONAME 8 /* socket name */
613
614 #define MT_EXP1 9 /* for experimental use */
615 #define MT_EXP2 10 /* for experimental use */
616 #define MT_EXP3 11 /* for experimental use */
617 #define MT_EXP4 12 /* for experimental use */
618
619 #define MT_CONTROL 14 /* extra-data protocol message */
620 #define MT_EXTCONTROL 15 /* control message with externalized contents */
621 #define MT_OOBDATA 16 /* expedited data */
622
623 #define MT_NOINIT 255 /* Not a type but a flag to allocate
624 a non-initialized mbuf */
625
626 /*
627 * String names of mbuf-related UMA(9) and malloc(9) types. Exposed to
628 * !_KERNEL so that monitoring tools can look up the zones with
629 * libmemstat(3).
630 */
631 #define MBUF_MEM_NAME "mbuf"
632 #define MBUF_CLUSTER_MEM_NAME "mbuf_cluster"
633 #define MBUF_PACKET_MEM_NAME "mbuf_packet"
634 #define MBUF_JUMBOP_MEM_NAME "mbuf_jumbo_page"
635 #define MBUF_JUMBO9_MEM_NAME "mbuf_jumbo_9k"
636 #define MBUF_JUMBO16_MEM_NAME "mbuf_jumbo_16k"
637 #define MBUF_TAG_MEM_NAME "mbuf_tag"
638 #define MBUF_EXTREFCNT_MEM_NAME "mbuf_ext_refcnt"
639
640 #ifdef _KERNEL
641
642 #ifdef WITNESS
643 #define MBUF_CHECKSLEEP(how) do { \
644 if (how == M_WAITOK) \
645 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, \
646 "Sleeping in \"%s\"", __func__); \
647 } while (0)
648 #else
649 #define MBUF_CHECKSLEEP(how)
650 #endif
651
652 /*
653 * Network buffer allocation API
654 *
655 * The rest of it is defined in kern/kern_mbuf.c
656 */
657 extern uma_zone_t zone_mbuf;
658 extern uma_zone_t zone_clust;
659 extern uma_zone_t zone_pack;
660 extern uma_zone_t zone_jumbop;
661 extern uma_zone_t zone_jumbo9;
662 extern uma_zone_t zone_jumbo16;
663
664 void mb_dupcl(struct mbuf *, struct mbuf *);
665 void mb_free_ext(struct mbuf *);
666 void m_adj(struct mbuf *, int);
667 void m_adj_decap(struct mbuf *, int);
668 int m_apply(struct mbuf *, int, int,
669 int (*)(void *, void *, u_int), void *);
670 int m_append(struct mbuf *, int, c_caddr_t);
671 void m_cat(struct mbuf *, struct mbuf *);
672 void m_catpkt(struct mbuf *, struct mbuf *);
673 int m_clget(struct mbuf *m, int how);
674 void *m_cljget(struct mbuf *m, int how, int size);
675 struct mbuf *m_collapse(struct mbuf *, int, int);
676 void m_copyback(struct mbuf *, int, int, c_caddr_t);
677 void m_copydata(const struct mbuf *, int, int, caddr_t);
678 struct mbuf *m_copym(struct mbuf *, int, int, int);
679 struct mbuf *m_copypacket(struct mbuf *, int);
680 void m_copy_pkthdr(struct mbuf *, struct mbuf *);
681 struct mbuf *m_copyup(struct mbuf *, int, int);
682 struct mbuf *m_defrag(struct mbuf *, int);
683 void m_demote_pkthdr(struct mbuf *);
684 void m_demote(struct mbuf *, int, int);
685 struct mbuf *m_devget(char *, int, int, struct ifnet *,
686 void (*)(char *, caddr_t, u_int));
687 void m_dispose_extcontrolm(struct mbuf *m);
688 struct mbuf *m_dup(const struct mbuf *, int);
689 int m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
690 void m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
691 void *, void *, int, int);
692 u_int m_fixhdr(struct mbuf *);
693 struct mbuf *m_fragment(struct mbuf *, int, int);
694 void m_freem(struct mbuf *);
695 struct mbuf *m_get2(int, int, short, int);
696 struct mbuf *m_getjcl(int, short, int, int);
697 struct mbuf *m_getm2(struct mbuf *, int, int, short, int);
698 struct mbuf *m_getptr(struct mbuf *, int, int *);
699 u_int m_length(struct mbuf *, struct mbuf **);
700 int m_mbuftouio(struct uio *, const struct mbuf *, int);
701 void m_move_pkthdr(struct mbuf *, struct mbuf *);
702 int m_pkthdr_init(struct mbuf *, int);
703 struct mbuf *m_prepend(struct mbuf *, int, int);
704 void m_print(const struct mbuf *, int);
705 struct mbuf *m_pulldown(struct mbuf *, int, int, int *);
706 struct mbuf *m_pullup(struct mbuf *, int);
707 int m_sanity(struct mbuf *, int);
708 struct mbuf *m_split(struct mbuf *, int, int);
709 struct mbuf *m_uiotombuf(struct uio *, int, int, int, int);
710 struct mbuf *m_unshare(struct mbuf *, int);
711
712 static __inline int
m_gettype(int size)713 m_gettype(int size)
714 {
715 int type;
716
717 switch (size) {
718 case MSIZE:
719 type = EXT_MBUF;
720 break;
721 case MCLBYTES:
722 type = EXT_CLUSTER;
723 break;
724 #if MJUMPAGESIZE != MCLBYTES
725 case MJUMPAGESIZE:
726 type = EXT_JUMBOP;
727 break;
728 #endif
729 case MJUM9BYTES:
730 type = EXT_JUMBO9;
731 break;
732 case MJUM16BYTES:
733 type = EXT_JUMBO16;
734 break;
735 default:
736 panic("%s: invalid cluster size %d", __func__, size);
737 }
738
739 return (type);
740 }
741
742 /*
743 * Associated an external reference counted buffer with an mbuf.
744 */
745 static __inline void
m_extaddref(struct mbuf * m,char * buf,u_int size,u_int * ref_cnt,m_ext_free_t freef,void * arg1,void * arg2)746 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
747 m_ext_free_t freef, void *arg1, void *arg2)
748 {
749
750 KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
751
752 atomic_add_int(ref_cnt, 1);
753 m->m_flags |= M_EXT;
754 m->m_ext.ext_buf = buf;
755 m->m_ext.ext_cnt = ref_cnt;
756 m->m_data = m->m_ext.ext_buf;
757 m->m_ext.ext_size = size;
758 m->m_ext.ext_free = freef;
759 m->m_ext.ext_arg1 = arg1;
760 m->m_ext.ext_arg2 = arg2;
761 m->m_ext.ext_type = EXT_EXTREF;
762 m->m_ext.ext_flags = 0;
763 }
764
765 static __inline uma_zone_t
m_getzone(int size)766 m_getzone(int size)
767 {
768 uma_zone_t zone;
769
770 switch (size) {
771 case MCLBYTES:
772 zone = zone_clust;
773 break;
774 #if MJUMPAGESIZE != MCLBYTES
775 case MJUMPAGESIZE:
776 zone = zone_jumbop;
777 break;
778 #endif
779 case MJUM9BYTES:
780 zone = zone_jumbo9;
781 break;
782 case MJUM16BYTES:
783 zone = zone_jumbo16;
784 break;
785 default:
786 panic("%s: invalid cluster size %d", __func__, size);
787 }
788
789 return (zone);
790 }
791
792 /*
793 * Initialize an mbuf with linear storage.
794 *
795 * Inline because the consumer text overhead will be roughly the same to
796 * initialize or call a function with this many parameters and M_PKTHDR
797 * should go away with constant propagation for !MGETHDR.
798 */
799 static __inline int
m_init(struct mbuf * m,int how,short type,int flags)800 m_init(struct mbuf *m, int how, short type, int flags)
801 {
802 int error;
803
804 m->m_next = NULL;
805 m->m_nextpkt = NULL;
806 m->m_data = m->m_dat;
807 m->m_len = 0;
808 m->m_flags = flags;
809 m->m_type = type;
810 if (flags & M_PKTHDR)
811 error = m_pkthdr_init(m, how);
812 else
813 error = 0;
814
815 MBUF_PROBE5(m__init, m, how, type, flags, error);
816 return (error);
817 }
818
819 static __inline struct mbuf *
m_get(int how,short type)820 m_get(int how, short type)
821 {
822 struct mbuf *m;
823 struct mb_args args;
824
825 args.flags = 0;
826 args.type = type;
827 m = uma_zalloc_arg(zone_mbuf, &args, how);
828 MBUF_PROBE3(m__get, how, type, m);
829 return (m);
830 }
831
832 static __inline struct mbuf *
m_gethdr(int how,short type)833 m_gethdr(int how, short type)
834 {
835 struct mbuf *m;
836 struct mb_args args;
837
838 args.flags = M_PKTHDR;
839 args.type = type;
840 m = uma_zalloc_arg(zone_mbuf, &args, how);
841 MBUF_PROBE3(m__gethdr, how, type, m);
842 return (m);
843 }
844
845 static __inline struct mbuf *
m_getcl(int how,short type,int flags)846 m_getcl(int how, short type, int flags)
847 {
848 struct mbuf *m;
849 struct mb_args args;
850
851 args.flags = flags;
852 args.type = type;
853 m = uma_zalloc_arg(zone_pack, &args, how);
854 MBUF_PROBE4(m__getcl, how, type, flags, m);
855 return (m);
856 }
857
858 /*
859 * XXX: m_cljset() is a dangerous API. One must attach only a new,
860 * unreferenced cluster to an mbuf(9). It is not possible to assert
861 * that, so care can be taken only by users of the API.
862 */
863 static __inline void
m_cljset(struct mbuf * m,void * cl,int type)864 m_cljset(struct mbuf *m, void *cl, int type)
865 {
866 int size;
867
868 switch (type) {
869 case EXT_CLUSTER:
870 size = MCLBYTES;
871 break;
872 #if MJUMPAGESIZE != MCLBYTES
873 case EXT_JUMBOP:
874 size = MJUMPAGESIZE;
875 break;
876 #endif
877 case EXT_JUMBO9:
878 size = MJUM9BYTES;
879 break;
880 case EXT_JUMBO16:
881 size = MJUM16BYTES;
882 break;
883 default:
884 panic("%s: unknown cluster type %d", __func__, type);
885 break;
886 }
887
888 m->m_data = m->m_ext.ext_buf = cl;
889 m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
890 m->m_ext.ext_size = size;
891 m->m_ext.ext_type = type;
892 m->m_ext.ext_flags = EXT_FLAG_EMBREF;
893 m->m_ext.ext_count = 1;
894 m->m_flags |= M_EXT;
895 MBUF_PROBE3(m__cljset, m, cl, type);
896 }
897
898 static __inline void
m_chtype(struct mbuf * m,short new_type)899 m_chtype(struct mbuf *m, short new_type)
900 {
901
902 m->m_type = new_type;
903 }
904
905 static __inline void
m_clrprotoflags(struct mbuf * m)906 m_clrprotoflags(struct mbuf *m)
907 {
908
909 while (m) {
910 m->m_flags &= ~M_PROTOFLAGS;
911 m = m->m_next;
912 }
913 }
914
915 static __inline struct mbuf *
m_last(struct mbuf * m)916 m_last(struct mbuf *m)
917 {
918
919 while (m->m_next)
920 m = m->m_next;
921 return (m);
922 }
923
924 static inline u_int
m_extrefcnt(struct mbuf * m)925 m_extrefcnt(struct mbuf *m)
926 {
927
928 KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
929
930 return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
931 *m->m_ext.ext_cnt);
932 }
933
934 /*
935 * mbuf, cluster, and external object allocation macros (for compatibility
936 * purposes).
937 */
938 #define M_MOVE_PKTHDR(to, from) m_move_pkthdr((to), (from))
939 #define MGET(m, how, type) ((m) = m_get((how), (type)))
940 #define MGETHDR(m, how, type) ((m) = m_gethdr((how), (type)))
941 #define MCLGET(m, how) m_clget((m), (how))
942 #define MEXTADD(m, buf, size, free, arg1, arg2, flags, type) \
943 m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2), \
944 (flags), (type))
945 #define m_getm(m, len, how, type) \
946 m_getm2((m), (len), (how), (type), M_PKTHDR)
947
948 /*
949 * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
950 * be both the local data payload, or an external buffer area, depending on
951 * whether M_EXT is set).
952 */
953 #define M_WRITABLE(m) (!((m)->m_flags & M_RDONLY) && \
954 (!(((m)->m_flags & M_EXT)) || \
955 (m_extrefcnt(m) == 1)))
956
957 /* Check if the supplied mbuf has a packet header, or else panic. */
958 #define M_ASSERTPKTHDR(m) \
959 KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR, \
960 ("%s: no mbuf packet header!", __func__))
961
962 /*
963 * Ensure that the supplied mbuf is a valid, non-free mbuf.
964 *
965 * XXX: Broken at the moment. Need some UMA magic to make it work again.
966 */
967 #define M_ASSERTVALID(m) \
968 KASSERT((((struct mbuf *)m)->m_flags & 0) == 0, \
969 ("%s: attempted use of a free mbuf!", __func__))
970
971 /*
972 * Return the address of the start of the buffer associated with an mbuf,
973 * handling external storage, packet-header mbufs, and regular data mbufs.
974 */
975 #define M_START(m) \
976 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf : \
977 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] : \
978 &(m)->m_dat[0])
979
980 /*
981 * Return the size of the buffer associated with an mbuf, handling external
982 * storage, packet-header mbufs, and regular data mbufs.
983 */
984 #define M_SIZE(m) \
985 (((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size : \
986 ((m)->m_flags & M_PKTHDR) ? MHLEN : \
987 MLEN)
988
989 /*
990 * Set the m_data pointer of a newly allocated mbuf to place an object of the
991 * specified size at the end of the mbuf, longword aligned.
992 *
993 * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
994 * separate macros, each asserting that it was called at the proper moment.
995 * This required callers to themselves test the storage type and call the
996 * right one. Rather than require callers to be aware of those layout
997 * decisions, we centralize here.
998 */
999 static __inline void
m_align(struct mbuf * m,int len)1000 m_align(struct mbuf *m, int len)
1001 {
1002 #ifdef INVARIANTS
1003 const char *msg = "%s: not a virgin mbuf";
1004 #endif
1005 int adjust;
1006
1007 KASSERT(m->m_data == M_START(m), (msg, __func__));
1008
1009 adjust = M_SIZE(m) - len;
1010 m->m_data += adjust &~ (sizeof(long)-1);
1011 }
1012
1013 #define M_ALIGN(m, len) m_align(m, len)
1014 #define MH_ALIGN(m, len) m_align(m, len)
1015 #define MEXT_ALIGN(m, len) m_align(m, len)
1016
1017 /*
1018 * Compute the amount of space available before the current start of data in
1019 * an mbuf.
1020 *
1021 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1022 * of checking writability of the mbuf data area rests solely with the caller.
1023 *
1024 * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1025 * for mbufs with external storage. We now allow mbuf-embedded data to be
1026 * read-only as well.
1027 */
1028 #define M_LEADINGSPACE(m) \
1029 (M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1030
1031 /*
1032 * Compute the amount of space available after the end of data in an mbuf.
1033 *
1034 * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1035 * of checking writability of the mbuf data area rests solely with the caller.
1036 *
1037 * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1038 * for mbufs with external storage. We now allow mbuf-embedded data to be
1039 * read-only as well.
1040 */
1041 #define M_TRAILINGSPACE(m) \
1042 (M_WRITABLE(m) ? \
1043 ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1044
1045 /*
1046 * Arrange to prepend space of size plen to mbuf m. If a new mbuf must be
1047 * allocated, how specifies whether to wait. If the allocation fails, the
1048 * original mbuf chain is freed and m is set to NULL.
1049 */
1050 #define M_PREPEND(m, plen, how) do { \
1051 struct mbuf **_mmp = &(m); \
1052 struct mbuf *_mm = *_mmp; \
1053 int _mplen = (plen); \
1054 int __mhow = (how); \
1055 \
1056 MBUF_CHECKSLEEP(how); \
1057 if (M_LEADINGSPACE(_mm) >= _mplen) { \
1058 _mm->m_data -= _mplen; \
1059 _mm->m_len += _mplen; \
1060 } else \
1061 _mm = m_prepend(_mm, _mplen, __mhow); \
1062 if (_mm != NULL && _mm->m_flags & M_PKTHDR) \
1063 _mm->m_pkthdr.len += _mplen; \
1064 *_mmp = _mm; \
1065 } while (0)
1066
1067 /*
1068 * Change mbuf to new type. This is a relatively expensive operation and
1069 * should be avoided.
1070 */
1071 #define MCHTYPE(m, t) m_chtype((m), (t))
1072
1073 /* Length to m_copy to copy all. */
1074 #define M_COPYALL 1000000000
1075
1076 extern int max_datalen; /* MHLEN - max_hdr */
1077 extern int max_hdr; /* Largest link + protocol header */
1078 extern int max_linkhdr; /* Largest link-level header */
1079 extern int max_protohdr; /* Largest protocol header */
1080 extern int nmbclusters; /* Maximum number of clusters */
1081
1082 /*-
1083 * Network packets may have annotations attached by affixing a list of
1084 * "packet tags" to the pkthdr structure. Packet tags are dynamically
1085 * allocated semi-opaque data structures that have a fixed header
1086 * (struct m_tag) that specifies the size of the memory block and a
1087 * <cookie,type> pair that identifies it. The cookie is a 32-bit unique
1088 * unsigned value used to identify a module or ABI. By convention this value
1089 * is chosen as the date+time that the module is created, expressed as the
1090 * number of seconds since the epoch (e.g., using date -u +'%s'). The type
1091 * value is an ABI/module-specific value that identifies a particular
1092 * annotation and is private to the module. For compatibility with systems
1093 * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1094 * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1095 * compatibility shim functions and several tag types are defined below.
1096 * Users that do not require compatibility should use a private cookie value
1097 * so that packet tag-related definitions can be maintained privately.
1098 *
1099 * Note that the packet tag returned by m_tag_alloc has the default memory
1100 * alignment implemented by malloc. To reference private data one can use a
1101 * construct like:
1102 *
1103 * struct m_tag *mtag = m_tag_alloc(...);
1104 * struct foo *p = (struct foo *)(mtag+1);
1105 *
1106 * if the alignment of struct m_tag is sufficient for referencing members of
1107 * struct foo. Otherwise it is necessary to embed struct m_tag within the
1108 * private data structure to insure proper alignment; e.g.,
1109 *
1110 * struct foo {
1111 * struct m_tag tag;
1112 * ...
1113 * };
1114 * struct foo *p = (struct foo *) m_tag_alloc(...);
1115 * struct m_tag *mtag = &p->tag;
1116 */
1117
1118 /*
1119 * Persistent tags stay with an mbuf until the mbuf is reclaimed. Otherwise
1120 * tags are expected to ``vanish'' when they pass through a network
1121 * interface. For most interfaces this happens normally as the tags are
1122 * reclaimed when the mbuf is free'd. However in some special cases
1123 * reclaiming must be done manually. An example is packets that pass through
1124 * the loopback interface. Also, one must be careful to do this when
1125 * ``turning around'' packets (e.g., icmp_reflect).
1126 *
1127 * To mark a tag persistent bit-or this flag in when defining the tag id.
1128 * The tag will then be treated as described above.
1129 */
1130 #define MTAG_PERSISTENT 0x800
1131
1132 #define PACKET_TAG_NONE 0 /* Nadda */
1133
1134 /* Packet tags for use with PACKET_ABI_COMPAT. */
1135 #define PACKET_TAG_IPSEC_IN_DONE 1 /* IPsec applied, in */
1136 #define PACKET_TAG_IPSEC_OUT_DONE 2 /* IPsec applied, out */
1137 #define PACKET_TAG_IPSEC_IN_CRYPTO_DONE 3 /* NIC IPsec crypto done */
1138 #define PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED 4 /* NIC IPsec crypto req'ed */
1139 #define PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO 5 /* NIC notifies IPsec */
1140 #define PACKET_TAG_IPSEC_PENDING_TDB 6 /* Reminder to do IPsec */
1141 #define PACKET_TAG_BRIDGE 7 /* Bridge processing done */
1142 #define PACKET_TAG_GIF 8 /* GIF processing done */
1143 #define PACKET_TAG_GRE 9 /* GRE processing done */
1144 #define PACKET_TAG_IN_PACKET_CHECKSUM 10 /* NIC checksumming done */
1145 #define PACKET_TAG_ENCAP 11 /* Encap. processing */
1146 #define PACKET_TAG_IPSEC_SOCKET 12 /* IPSEC socket ref */
1147 #define PACKET_TAG_IPSEC_HISTORY 13 /* IPSEC history */
1148 #define PACKET_TAG_IPV6_INPUT 14 /* IPV6 input processing */
1149 #define PACKET_TAG_DUMMYNET 15 /* dummynet info */
1150 #define PACKET_TAG_DIVERT 17 /* divert info */
1151 #define PACKET_TAG_IPFORWARD 18 /* ipforward info */
1152 #define PACKET_TAG_MACLABEL (19 | MTAG_PERSISTENT) /* MAC label */
1153 #define PACKET_TAG_PF 21 /* PF/ALTQ information */
1154 #define PACKET_TAG_RTSOCKFAM 25 /* rtsock sa family */
1155 #define PACKET_TAG_IPOPTIONS 27 /* Saved IP options */
1156 #define PACKET_TAG_CARP 28 /* CARP info */
1157 #define PACKET_TAG_IPSEC_NAT_T_PORTS 29 /* two uint16_t */
1158 #define PACKET_TAG_ND_OUTGOING 30 /* ND outgoing */
1159
1160 /* Specific cookies and tags. */
1161
1162 /* Packet tag routines. */
1163 struct m_tag *m_tag_alloc(u_int32_t, int, int, int);
1164 void m_tag_delete(struct mbuf *, struct m_tag *);
1165 void m_tag_delete_chain(struct mbuf *, struct m_tag *);
1166 void m_tag_free_default(struct m_tag *);
1167 struct m_tag *m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1168 struct m_tag *m_tag_copy(struct m_tag *, int);
1169 int m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1170 void m_tag_delete_nonpersistent(struct mbuf *);
1171
1172 /*
1173 * Initialize the list of tags associated with an mbuf.
1174 */
1175 static __inline void
m_tag_init(struct mbuf * m)1176 m_tag_init(struct mbuf *m)
1177 {
1178
1179 SLIST_INIT(&m->m_pkthdr.tags);
1180 }
1181
1182 /*
1183 * Set up the contents of a tag. Note that this does not fill in the free
1184 * method; the caller is expected to do that.
1185 *
1186 * XXX probably should be called m_tag_init, but that was already taken.
1187 */
1188 static __inline void
m_tag_setup(struct m_tag * t,u_int32_t cookie,int type,int len)1189 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1190 {
1191
1192 t->m_tag_id = type;
1193 t->m_tag_len = len;
1194 t->m_tag_cookie = cookie;
1195 }
1196
1197 /*
1198 * Reclaim resources associated with a tag.
1199 */
1200 static __inline void
m_tag_free(struct m_tag * t)1201 m_tag_free(struct m_tag *t)
1202 {
1203
1204 (*t->m_tag_free)(t);
1205 }
1206
1207 /*
1208 * Return the first tag associated with an mbuf.
1209 */
1210 static __inline struct m_tag *
m_tag_first(struct mbuf * m)1211 m_tag_first(struct mbuf *m)
1212 {
1213
1214 return (SLIST_FIRST(&m->m_pkthdr.tags));
1215 }
1216
1217 /*
1218 * Return the next tag in the list of tags associated with an mbuf.
1219 */
1220 static __inline struct m_tag *
m_tag_next(struct mbuf * m __unused,struct m_tag * t)1221 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1222 {
1223
1224 return (SLIST_NEXT(t, m_tag_link));
1225 }
1226
1227 /*
1228 * Prepend a tag to the list of tags associated with an mbuf.
1229 */
1230 static __inline void
m_tag_prepend(struct mbuf * m,struct m_tag * t)1231 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1232 {
1233
1234 SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1235 }
1236
1237 /*
1238 * Unlink a tag from the list of tags associated with an mbuf.
1239 */
1240 static __inline void
m_tag_unlink(struct mbuf * m,struct m_tag * t)1241 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1242 {
1243
1244 SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1245 }
1246
1247 /* These are for OpenBSD compatibility. */
1248 #define MTAG_ABI_COMPAT 0 /* compatibility ABI */
1249
1250 static __inline struct m_tag *
m_tag_get(int type,int length,int wait)1251 m_tag_get(int type, int length, int wait)
1252 {
1253 return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1254 }
1255
1256 static __inline struct m_tag *
m_tag_find(struct mbuf * m,int type,struct m_tag * start)1257 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1258 {
1259 return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1260 m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1261 }
1262
1263 static __inline struct mbuf *
m_free(struct mbuf * m)1264 m_free(struct mbuf *m)
1265 {
1266 struct mbuf *n = m->m_next;
1267
1268 MBUF_PROBE1(m__free, m);
1269 if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1270 m_tag_delete_chain(m, NULL);
1271 if (m->m_flags & M_EXT)
1272 mb_free_ext(m);
1273 else if ((m->m_flags & M_NOFREE) == 0)
1274 uma_zfree(zone_mbuf, m);
1275 return (n);
1276 }
1277
1278 static __inline int
rt_m_getfib(struct mbuf * m)1279 rt_m_getfib(struct mbuf *m)
1280 {
1281 KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1282 return (m->m_pkthdr.fibnum);
1283 }
1284
1285 #define M_GETFIB(_m) rt_m_getfib(_m)
1286
1287 #define M_SETFIB(_m, _fib) do { \
1288 KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf.")); \
1289 ((_m)->m_pkthdr.fibnum) = (_fib); \
1290 } while (0)
1291
1292 /* flags passed as first argument for "m_xxx_tcpip_hash()" */
1293 #define MBUF_HASHFLAG_L2 (1 << 2)
1294 #define MBUF_HASHFLAG_L3 (1 << 3)
1295 #define MBUF_HASHFLAG_L4 (1 << 4)
1296
1297 /* mbuf hashing helper routines */
1298 uint32_t m_ether_tcpip_hash_init(void);
1299 uint32_t m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1300 uint32_t m_infiniband_tcpip_hash_init(void);
1301 uint32_t m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1302
1303 #ifdef MBUF_PROFILING
1304 void m_profile(struct mbuf *m);
1305 #define M_PROFILE(m) m_profile(m)
1306 #else
1307 #define M_PROFILE(m)
1308 #endif
1309
1310 struct mbufq {
1311 STAILQ_HEAD(, mbuf) mq_head;
1312 int mq_len;
1313 int mq_maxlen;
1314 };
1315
1316 static inline void
mbufq_init(struct mbufq * mq,int maxlen)1317 mbufq_init(struct mbufq *mq, int maxlen)
1318 {
1319
1320 STAILQ_INIT(&mq->mq_head);
1321 mq->mq_maxlen = maxlen;
1322 mq->mq_len = 0;
1323 }
1324
1325 static inline struct mbuf *
mbufq_flush(struct mbufq * mq)1326 mbufq_flush(struct mbufq *mq)
1327 {
1328 struct mbuf *m;
1329
1330 m = STAILQ_FIRST(&mq->mq_head);
1331 STAILQ_INIT(&mq->mq_head);
1332 mq->mq_len = 0;
1333 return (m);
1334 }
1335
1336 static inline void
mbufq_drain(struct mbufq * mq)1337 mbufq_drain(struct mbufq *mq)
1338 {
1339 struct mbuf *m, *n;
1340
1341 n = mbufq_flush(mq);
1342 while ((m = n) != NULL) {
1343 n = STAILQ_NEXT(m, m_stailqpkt);
1344 m_freem(m);
1345 }
1346 }
1347
1348 static inline struct mbuf *
mbufq_first(const struct mbufq * mq)1349 mbufq_first(const struct mbufq *mq)
1350 {
1351
1352 return (STAILQ_FIRST(&mq->mq_head));
1353 }
1354
1355 static inline struct mbuf *
mbufq_last(const struct mbufq * mq)1356 mbufq_last(const struct mbufq *mq)
1357 {
1358
1359 return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1360 }
1361
1362 static inline int
mbufq_full(const struct mbufq * mq)1363 mbufq_full(const struct mbufq *mq)
1364 {
1365
1366 return (mq->mq_len >= mq->mq_maxlen);
1367 }
1368
1369 static inline int
mbufq_len(const struct mbufq * mq)1370 mbufq_len(const struct mbufq *mq)
1371 {
1372
1373 return (mq->mq_len);
1374 }
1375
1376 static inline int
mbufq_enqueue(struct mbufq * mq,struct mbuf * m)1377 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1378 {
1379
1380 if (mbufq_full(mq))
1381 return (ENOBUFS);
1382 STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1383 mq->mq_len++;
1384 return (0);
1385 }
1386
1387 static inline struct mbuf *
mbufq_dequeue(struct mbufq * mq)1388 mbufq_dequeue(struct mbufq *mq)
1389 {
1390 struct mbuf *m;
1391
1392 m = STAILQ_FIRST(&mq->mq_head);
1393 if (m) {
1394 STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1395 m->m_nextpkt = NULL;
1396 mq->mq_len--;
1397 }
1398 return (m);
1399 }
1400
1401 static inline void
mbufq_prepend(struct mbufq * mq,struct mbuf * m)1402 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1403 {
1404
1405 STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1406 mq->mq_len++;
1407 }
1408
1409 /*
1410 * Note: this doesn't enforce the maximum list size for dst.
1411 */
1412 static inline void
mbufq_concat(struct mbufq * mq_dst,struct mbufq * mq_src)1413 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1414 {
1415
1416 mq_dst->mq_len += mq_src->mq_len;
1417 STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1418 mq_src->mq_len = 0;
1419 }
1420
1421 #ifdef _SYS_TIMESPEC_H_
1422 static inline void
mbuf_tstmp2timespec(struct mbuf * m,struct timespec * ts)1423 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1424 {
1425
1426 KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1427 KASSERT((m->m_flags & M_TSTMP) != 0, ("mbuf %p no M_TSTMP", m));
1428 ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1429 ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1430 }
1431 #endif
1432
1433 #ifdef NETDUMP
1434 /* Invoked from the netdump client code. */
1435 void netdump_mbuf_drain(void);
1436 void netdump_mbuf_dump(void);
1437 void netdump_mbuf_reinit(int nmbuf, int nclust, int clsize);
1438 #endif
1439
1440 #endif /* _KERNEL */
1441 #endif /* !_SYS_MBUF_H_ */
1442