xref: /freebsd-13-stable/sys/sys/mbuf.h (revision c62e10adba6391809ac20aaf942c1ab7ad2b255c)
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  */
34 
35 #ifndef _SYS_MBUF_H_
36 #define	_SYS_MBUF_H_
37 
38 /* XXX: These includes suck. Sorry! */
39 #include <sys/queue.h>
40 #ifdef _KERNEL
41 #include <sys/systm.h>
42 #include <sys/refcount.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 #define	M_NODOM		255
103 
104 #ifdef _KERNEL
105 /*-
106  * Macro for type conversion: convert mbuf pointer to data pointer of correct
107  * type:
108  *
109  * mtod(m, t)	-- Convert mbuf pointer to data pointer of correct type.
110  * mtodo(m, o) -- Same as above but with offset 'o' into data.
111  */
112 #define	mtod(m, t)	((t)((m)->m_data))
113 #define	mtodo(m, o)	((void *)(((m)->m_data) + (o)))
114 
115 /*
116  * Argument structure passed to UMA routines during mbuf and packet
117  * allocations.
118  */
119 struct mb_args {
120 	int	flags;	/* Flags for mbuf being allocated */
121 	short	type;	/* Type of mbuf being allocated */
122 };
123 #endif /* _KERNEL */
124 
125 /*
126  * Packet tag structure (see below for details).
127  */
128 struct m_tag {
129 	SLIST_ENTRY(m_tag)	m_tag_link;	/* List of packet tags */
130 	u_int16_t		m_tag_id;	/* Tag ID */
131 	u_int16_t		m_tag_len;	/* Length of data */
132 	u_int32_t		m_tag_cookie;	/* ABI/Module ID */
133 	void			(*m_tag_free)(struct m_tag *);
134 };
135 
136 /*
137  * Static network interface owned tag.
138  * Allocated through ifp->if_snd_tag_alloc().
139  */
140 struct m_snd_tag {
141 	struct ifnet *ifp;		/* network interface tag belongs to */
142 	volatile u_int refcount;
143 	u_int	type;			/* One of IF_SND_TAG_TYPE_*. */
144 };
145 
146 /*
147  * Record/packet header in first mbuf of chain; valid only if M_PKTHDR is set.
148  * Size ILP32: 48
149  *	 LP64: 56
150  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
151  * they are correct.
152  */
153 struct pkthdr {
154 	union {
155 		struct m_snd_tag *snd_tag;	/* send tag, if any */
156 		struct ifnet	*rcvif;		/* rcv interface */
157 	};
158 	SLIST_HEAD(packet_tags, m_tag) tags; /* list of packet tags */
159 	int32_t		 len;		/* total packet length */
160 
161 	/* Layer crossing persistent information. */
162 	uint32_t	 flowid;	/* packet's 4-tuple system */
163 	uint32_t	 csum_flags;	/* checksum and offload features */
164 	uint16_t	 fibnum;	/* this packet should use this fib */
165 	uint8_t		 numa_domain;	/* NUMA domain of recvd pkt */
166 	uint8_t		 rsstype;	/* hash type */
167 	union {
168 		uint64_t	rcv_tstmp;	/* timestamp in ns */
169 		struct {
170 			uint8_t		 l2hlen;	/* layer 2 hdr len */
171 			uint8_t		 l3hlen;	/* layer 3 hdr len */
172 			uint8_t		 l4hlen;	/* layer 4 hdr len */
173 			uint8_t		 l5hlen;	/* layer 5 hdr len */
174 			uint8_t		 inner_l2hlen;
175 			uint8_t		 inner_l3hlen;
176 			uint8_t		 inner_l4hlen;
177 			uint8_t		 inner_l5hlen;
178 		};
179 	};
180 	union {
181 		uint8_t  eight[8];
182 		uint16_t sixteen[4];
183 		uint32_t thirtytwo[2];
184 		uint64_t sixtyfour[1];
185 		uintptr_t unintptr[1];
186 		void	*ptr;
187 	} PH_per;
188 
189 	/* Layer specific non-persistent local storage for reassembly, etc. */
190 	union {
191 		uint8_t  eight[8];
192 		uint16_t sixteen[4];
193 		uint32_t thirtytwo[2];
194 		uint64_t sixtyfour[1];
195 		uintptr_t unintptr[1];
196 		void 	*ptr;
197 	} PH_loc;
198 };
199 #define	ether_vtag	PH_per.sixteen[0]
200 #define tcp_tun_port	PH_per.sixteen[0] /* outbound */
201 #define	PH_vt		PH_per
202 #define	vt_nrecs	sixteen[0]	  /* mld and v6-ND */
203 #define	tso_segsz	PH_per.sixteen[1] /* inbound after LRO */
204 #define	lro_nsegs	tso_segsz	  /* inbound after LRO */
205 #define	csum_data	PH_per.thirtytwo[1] /* inbound from hardware up */
206 #define	lro_tcp_d_len	PH_loc.sixteen[0] /* inbound during LRO (no reassembly) */
207 #define	lro_tcp_d_csum	PH_loc.sixteen[1] /* inbound during LRO (no reassembly) */
208 #define	lro_tcp_h_off	PH_loc.sixteen[2] /* inbound during LRO (no reassembly) */
209 #define	lro_etype	PH_loc.sixteen[3] /* inbound during LRO (no reassembly) */
210 /* Note PH_loc is used during IP reassembly (all 8 bytes as a ptr) */
211 
212 /*
213  * TLS records for TLS 1.0-1.2 can have the following header lengths:
214  * - 5 (AES-CBC with implicit IV)
215  * - 21 (AES-CBC with explicit IV)
216  * - 13 (AES-GCM with 8 byte explicit IV)
217  */
218 #define	MBUF_PEXT_HDR_LEN	23
219 
220 /*
221  * TLS records for TLS 1.0-1.2 can have the following maximum trailer
222  * lengths:
223  * - 16 (AES-GCM)
224  * - 36 (AES-CBC with SHA1 and up to 16 bytes of padding)
225  * - 48 (AES-CBC with SHA2-256 and up to 16 bytes of padding)
226  * - 64 (AES-CBC with SHA2-384 and up to 16 bytes of padding)
227  */
228 #define	MBUF_PEXT_TRAIL_LEN	64
229 
230 #if defined(__LP64__)
231 #define MBUF_PEXT_MAX_PGS (40 / sizeof(vm_paddr_t))
232 #else
233 #define MBUF_PEXT_MAX_PGS (72 / sizeof(vm_paddr_t))
234 #endif
235 
236 #define	MBUF_PEXT_MAX_BYTES						\
237     (MBUF_PEXT_MAX_PGS * PAGE_SIZE + MBUF_PEXT_HDR_LEN + MBUF_PEXT_TRAIL_LEN)
238 
239 struct ktls_session;
240 struct socket;
241 
242 /*
243  * Description of external storage mapped into mbuf; valid only if M_EXT is
244  * set.
245  * Size ILP32: 28
246  *	 LP64: 48
247  * Compile-time assertions in uipc_mbuf.c test these values to ensure that
248  * they are correct.
249  */
250 typedef	void m_ext_free_t(struct mbuf *);
251 struct m_ext {
252 	union {
253 		/*
254 		 * If EXT_FLAG_EMBREF is set, then we use refcount in the
255 		 * mbuf, the 'ext_count' member.  Otherwise, we have a
256 		 * shadow copy and we use pointer 'ext_cnt'.  The original
257 		 * mbuf is responsible to carry the pointer to free routine
258 		 * and its arguments.  They aren't copied into shadows in
259 		 * mb_dupcl() to avoid dereferencing next cachelines.
260 		 */
261 		volatile u_int	 ext_count;
262 		volatile u_int	*ext_cnt;
263 	};
264 	uint32_t	 ext_size;	/* size of buffer, for ext_free */
265 	uint32_t	 ext_type:8,	/* type of external storage */
266 			 ext_flags:24;	/* external storage mbuf flags */
267 	union {
268 		struct {
269 			/*
270 			 * Regular M_EXT mbuf:
271 			 * o ext_buf always points to the external buffer.
272 			 * o ext_free (below) and two optional arguments
273 			 *   ext_arg1 and ext_arg2 store the free context for
274 			 *   the external storage.  They are set only in the
275 			 *   refcount carrying mbuf, the one with
276 			 *   EXT_FLAG_EMBREF flag, with exclusion for
277 			 *   EXT_EXTREF type, where the free context is copied
278 			 *   into all mbufs that use same external storage.
279 			 */
280 			char 	*ext_buf;	/* start of buffer */
281 #define	m_ext_copylen	offsetof(struct m_ext, ext_arg2)
282 			void	*ext_arg2;
283 		};
284 		struct {
285 			/*
286 			 * Multi-page M_EXTPG mbuf:
287 			 * o extpg_pa - page vector.
288 			 * o extpg_trail and extpg_hdr - TLS trailer and
289 			 *   header.
290 			 * Uses ext_free and may also use ext_arg1.
291 			 */
292 			vm_paddr_t	extpg_pa[MBUF_PEXT_MAX_PGS];
293 			char		extpg_trail[MBUF_PEXT_TRAIL_LEN];
294 			char		extpg_hdr[MBUF_PEXT_HDR_LEN];
295 			/* Pretend these 3 fields are part of mbuf itself. */
296 #define	m_epg_pa	m_ext.extpg_pa
297 #define	m_epg_trail	m_ext.extpg_trail
298 #define	m_epg_hdr	m_ext.extpg_hdr
299 #define	m_epg_ext_copylen	offsetof(struct m_ext, ext_free)
300 		};
301 	};
302 	/*
303 	 * Free method and optional argument pointer, both
304 	 * used by M_EXT and M_EXTPG.
305 	 */
306 	m_ext_free_t	*ext_free;
307 	void		*ext_arg1;
308 };
309 
310 /*
311  * The core of the mbuf object along with some shortcut defines for practical
312  * purposes.
313  */
314 struct mbuf {
315 	/*
316 	 * Header present at the beginning of every mbuf.
317 	 * Size ILP32: 24
318 	 *      LP64: 32
319 	 * Compile-time assertions in uipc_mbuf.c test these values to ensure
320 	 * that they are correct.
321 	 */
322 	union {	/* next buffer in chain */
323 		struct mbuf		*m_next;
324 		SLIST_ENTRY(mbuf)	m_slist;
325 		STAILQ_ENTRY(mbuf)	m_stailq;
326 	};
327 	union {	/* next chain in queue/record */
328 		struct mbuf		*m_nextpkt;
329 		SLIST_ENTRY(mbuf)	m_slistpkt;
330 		STAILQ_ENTRY(mbuf)	m_stailqpkt;
331 	};
332 	caddr_t		 m_data;	/* location of data */
333 	int32_t		 m_len;		/* amount of data in this mbuf */
334 	uint32_t	 m_type:8,	/* type of data in this mbuf */
335 			 m_flags:24;	/* flags; see below */
336 #if !defined(__LP64__)
337 	uint32_t	 m_pad;		/* pad for 64bit alignment */
338 #endif
339 
340 	/*
341 	 * A set of optional headers (packet header, external storage header)
342 	 * and internal data storage.  Historically, these arrays were sized
343 	 * to MHLEN (space left after a packet header) and MLEN (space left
344 	 * after only a regular mbuf header); they are now variable size in
345 	 * order to support future work on variable-size mbufs.
346 	 */
347 	union {
348 		struct {
349 			union {
350 				/* M_PKTHDR set. */
351 				struct pkthdr	m_pkthdr;
352 
353 				/* M_EXTPG set.
354 				 * Multi-page M_EXTPG mbuf has its meta data
355 				 * split between the below anonymous structure
356 				 * and m_ext.  It carries vector of pages,
357 				 * optional header and trailer char vectors
358 				 * and pointers to socket/TLS data.
359 				 */
360 #define	m_epg_startcopy		m_epg_npgs
361 #define	m_epg_endcopy		m_epg_stailq
362 				struct {
363 					/* Overall count of pages and count of
364 					 * pages with I/O pending. */
365 					uint8_t	m_epg_npgs;
366 					uint8_t	m_epg_nrdy;
367 					/* TLS header and trailer lengths.
368 					 * The data itself resides in m_ext. */
369 					uint8_t	m_epg_hdrlen;
370 					uint8_t	m_epg_trllen;
371 					/* Offset into 1st page and length of
372 					 * data in the last page. */
373 					uint16_t m_epg_1st_off;
374 					uint16_t m_epg_last_len;
375 					uint8_t	m_epg_flags;
376 #define	EPG_FLAG_ANON	0x1	/* Data can be encrypted in place. */
377 #define	EPG_FLAG_2FREE	0x2	/* Scheduled for free. */
378 					uint8_t	m_epg_record_type;
379 					uint8_t	__spare[2];
380 					int	m_epg_enc_cnt;
381 					struct ktls_session *m_epg_tls;
382 					struct socket	*m_epg_so;
383 					uint64_t	m_epg_seqno;
384 					STAILQ_ENTRY(mbuf) m_epg_stailq;
385 				};
386 			};
387 			union {
388 				/* M_EXT or M_EXTPG set. */
389 				struct m_ext	m_ext;
390 				/* M_PKTHDR set, neither M_EXT nor M_EXTPG. */
391 				char		m_pktdat[0];
392 			};
393 		};
394 		char	m_dat[0];			/* !M_PKTHDR, !M_EXT */
395 	};
396 };
397 
398 #ifdef _KERNEL
399 static inline int
m_epg_pagelen(const struct mbuf * m,int pidx,int pgoff)400 m_epg_pagelen(const struct mbuf *m, int pidx, int pgoff)
401 {
402 
403 	KASSERT(pgoff == 0 || pidx == 0,
404 	    ("page %d with non-zero offset %d in %p", pidx, pgoff, m));
405 
406 	if (pidx == m->m_epg_npgs - 1) {
407 		return (m->m_epg_last_len);
408 	} else {
409 		return (PAGE_SIZE - pgoff);
410 	}
411 }
412 
413 #ifdef INVARIANTS
414 #define	MCHECK(ex, msg)	KASSERT((ex),				\
415 	    ("Multi page mbuf %p with " #msg " at %s:%d",	\
416 	    m, __FILE__, __LINE__))
417 /*
418  * NB: This expects a non-empty buffer (npgs > 0 and
419  * last_pg_len > 0).
420  */
421 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {				\
422 	MCHECK(m->m_epg_npgs > 0, "no valid pages");		\
423 	MCHECK(m->m_epg_npgs <= nitems(m->m_epg_pa),		\
424 	    "too many pages");						\
425 	MCHECK(m->m_epg_nrdy <= m->m_epg_npgs,			\
426 	    "too many ready pages");					\
427 	MCHECK(m->m_epg_1st_off < PAGE_SIZE,			\
428 		"too large page offset");				\
429 	MCHECK(m->m_epg_last_len > 0, "zero last page length");	\
430 	MCHECK(m->m_epg_last_len <= PAGE_SIZE,			\
431 	    "too large last page length");				\
432 	if (m->m_epg_npgs == 1)					\
433 		MCHECK(m->m_epg_1st_off +			\
434 		    m->m_epg_last_len <=	 PAGE_SIZE,		\
435 		    "single page too large");				\
436 	MCHECK(m->m_epg_hdrlen <= sizeof(m->m_epg_hdr),		\
437 	    "too large header length");					\
438 	MCHECK(m->m_epg_trllen <= sizeof(m->m_epg_trail),	\
439 	    "too large header length");					\
440 } while (0)
441 #else
442 #define	MBUF_EXT_PGS_ASSERT_SANITY(m)	do {} while (0)
443 #endif
444 #endif
445 
446 /*
447  * mbuf flags of global significance and layer crossing.
448  * Those of only protocol/layer specific significance are to be mapped
449  * to M_PROTO[1-11] and cleared at layer handoff boundaries.
450  * NB: Limited to the lower 24 bits.
451  */
452 #define	M_EXT		0x00000001 /* has associated external storage */
453 #define	M_PKTHDR	0x00000002 /* start of record */
454 #define	M_EOR		0x00000004 /* end of record */
455 #define	M_RDONLY	0x00000008 /* associated data is marked read-only */
456 #define	M_BCAST		0x00000010 /* send/received as link-level broadcast */
457 #define	M_MCAST		0x00000020 /* send/received as link-level multicast */
458 #define	M_PROMISC	0x00000040 /* packet was not for us */
459 #define	M_VLANTAG	0x00000080 /* ether_vtag is valid */
460 #define	M_EXTPG		0x00000100 /* has array of unmapped pages and TLS */
461 #define	M_NOFREE	0x00000200 /* do not free mbuf, embedded in cluster */
462 #define	M_TSTMP		0x00000400 /* rcv_tstmp field is valid */
463 #define	M_TSTMP_HPREC	0x00000800 /* rcv_tstmp is high-prec, typically
464 				      hw-stamped on port (useful for IEEE 1588
465 				      and 802.1AS) */
466 #define M_TSTMP_LRO	0x00001000 /* Time LRO pushed in pkt is valid in (PH_loc) */
467 
468 #define	M_PROTO1	0x00002000 /* protocol-specific */
469 #define	M_PROTO2	0x00004000 /* protocol-specific */
470 #define	M_PROTO3	0x00008000 /* protocol-specific */
471 #define	M_PROTO4	0x00010000 /* protocol-specific */
472 #define	M_PROTO5	0x00020000 /* protocol-specific */
473 #define	M_PROTO6	0x00040000 /* protocol-specific */
474 #define	M_PROTO7	0x00080000 /* protocol-specific */
475 #define	M_PROTO8	0x00100000 /* protocol-specific */
476 #define	M_PROTO9	0x00200000 /* protocol-specific */
477 #define	M_PROTO10	0x00400000 /* protocol-specific */
478 #define	M_PROTO11	0x00800000 /* protocol-specific */
479 
480 #define MB_DTOR_SKIP	0x1	/* don't pollute the cache by touching a freed mbuf */
481 
482 /*
483  * Flags to purge when crossing layers.
484  */
485 #define	M_PROTOFLAGS \
486     (M_PROTO1|M_PROTO2|M_PROTO3|M_PROTO4|M_PROTO5|M_PROTO6|M_PROTO7|M_PROTO8|\
487      M_PROTO9|M_PROTO10|M_PROTO11)
488 
489 /*
490  * Flags preserved when copying m_pkthdr.
491  */
492 #define M_COPYFLAGS \
493     (M_PKTHDR|M_EOR|M_RDONLY|M_BCAST|M_MCAST|M_PROMISC|M_VLANTAG|M_TSTMP| \
494      M_TSTMP_HPREC|M_TSTMP_LRO|M_PROTOFLAGS)
495 
496 /*
497  * Flags preserved during demote.
498  */
499 #define	M_DEMOTEFLAGS \
500     (M_EXT | M_RDONLY | M_NOFREE | M_EXTPG)
501 
502 /*
503  * Mbuf flag description for use with printf(9) %b identifier.
504  */
505 #define	M_FLAG_BITS \
506     "\20\1M_EXT\2M_PKTHDR\3M_EOR\4M_RDONLY\5M_BCAST\6M_MCAST" \
507     "\7M_PROMISC\10M_VLANTAG\11M_EXTPG\12M_NOFREE\13M_TSTMP\14M_TSTMP_HPREC\15M_TSTMP_LRO"
508 #define	M_FLAG_PROTOBITS \
509     "\16M_PROTO1\17M_PROTO2\20M_PROTO3\21M_PROTO4" \
510     "\22M_PROTO5\23M_PROTO6\24M_PROTO7\25M_PROTO8\26M_PROTO9" \
511     "\27M_PROTO10\28M_PROTO11"
512 #define	M_FLAG_PRINTF (M_FLAG_BITS M_FLAG_PROTOBITS)
513 
514 /*
515  * Network interface cards are able to hash protocol fields (such as IPv4
516  * addresses and TCP port numbers) classify packets into flows.  These flows
517  * can then be used to maintain ordering while delivering packets to the OS
518  * via parallel input queues, as well as to provide a stateless affinity
519  * model.  NIC drivers can pass up the hash via m->m_pkthdr.flowid, and set
520  * m_flag fields to indicate how the hash should be interpreted by the
521  * network stack.
522  *
523  * Most NICs support RSS, which provides ordering and explicit affinity, and
524  * use the hash m_flag bits to indicate what header fields were covered by
525  * the hash.  M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH can be set by non-
526  * RSS cards or configurations that provide an opaque flow identifier, allowing
527  * for ordering and distribution without explicit affinity.  Additionally,
528  * M_HASHTYPE_OPAQUE_HASH indicates that the flow identifier has hash
529  * properties.
530  *
531  * The meaning of the IPV6_EX suffix:
532  * "o  Home address from the home address option in the IPv6 destination
533  *     options header.  If the extension header is not present, use the Source
534  *     IPv6 Address.
535  *  o  IPv6 address that is contained in the Routing-Header-Type-2 from the
536  *     associated extension header.  If the extension header is not present,
537  *     use the Destination IPv6 Address."
538  * Quoted from:
539  * https://docs.microsoft.com/en-us/windows-hardware/drivers/network/rss-hashing-types#ndishashipv6ex
540  */
541 #define	M_HASHTYPE_HASHPROP		0x80	/* has hash properties */
542 #define	M_HASHTYPE_INNER		0x40	/* calculated from inner headers */
543 #define	M_HASHTYPE_HASH(t)		(M_HASHTYPE_HASHPROP | (t))
544 /* Microsoft RSS standard hash types */
545 #define	M_HASHTYPE_NONE			0
546 #define	M_HASHTYPE_RSS_IPV4		M_HASHTYPE_HASH(1) /* IPv4 2-tuple */
547 #define	M_HASHTYPE_RSS_TCP_IPV4		M_HASHTYPE_HASH(2) /* TCPv4 4-tuple */
548 #define	M_HASHTYPE_RSS_IPV6		M_HASHTYPE_HASH(3) /* IPv6 2-tuple */
549 #define	M_HASHTYPE_RSS_TCP_IPV6		M_HASHTYPE_HASH(4) /* TCPv6 4-tuple */
550 #define	M_HASHTYPE_RSS_IPV6_EX		M_HASHTYPE_HASH(5) /* IPv6 2-tuple +
551 							    * ext hdrs */
552 #define	M_HASHTYPE_RSS_TCP_IPV6_EX	M_HASHTYPE_HASH(6) /* TCPv6 4-tuple +
553 							    * ext hdrs */
554 #define	M_HASHTYPE_RSS_UDP_IPV4		M_HASHTYPE_HASH(7) /* IPv4 UDP 4-tuple*/
555 #define	M_HASHTYPE_RSS_UDP_IPV6		M_HASHTYPE_HASH(9) /* IPv6 UDP 4-tuple*/
556 #define	M_HASHTYPE_RSS_UDP_IPV6_EX	M_HASHTYPE_HASH(10)/* IPv6 UDP 4-tuple +
557 							    * ext hdrs */
558 
559 #define	M_HASHTYPE_OPAQUE		0x3f	/* ordering, not affinity */
560 #define	M_HASHTYPE_OPAQUE_HASH		M_HASHTYPE_HASH(M_HASHTYPE_OPAQUE)
561 						/* ordering+hash, not affinity*/
562 
563 #define	M_HASHTYPE_CLEAR(m)	((m)->m_pkthdr.rsstype = 0)
564 #define	M_HASHTYPE_GET(m)	((m)->m_pkthdr.rsstype & ~M_HASHTYPE_INNER)
565 #define	M_HASHTYPE_SET(m, v)	((m)->m_pkthdr.rsstype = (v))
566 #define	M_HASHTYPE_TEST(m, v)	(M_HASHTYPE_GET(m) == (v))
567 #define	M_HASHTYPE_ISHASH(m)	\
568     (((m)->m_pkthdr.rsstype & M_HASHTYPE_HASHPROP) != 0)
569 #define	M_HASHTYPE_SETINNER(m)	do {			\
570 	(m)->m_pkthdr.rsstype |= M_HASHTYPE_INNER;	\
571     } while (0)
572 
573 /*
574  * External mbuf storage buffer types.
575  */
576 #define	EXT_CLUSTER	1	/* mbuf cluster */
577 #define	EXT_SFBUF	2	/* sendfile(2)'s sf_buf */
578 #define	EXT_JUMBOP	3	/* jumbo cluster page sized */
579 #define	EXT_JUMBO9	4	/* jumbo cluster 9216 bytes */
580 #define	EXT_JUMBO16	5	/* jumbo cluster 16184 bytes */
581 #define	EXT_PACKET	6	/* mbuf+cluster from packet zone */
582 #define	EXT_MBUF	7	/* external mbuf reference */
583 #define	EXT_RXRING	8	/* data in NIC receive ring */
584 
585 #define	EXT_VENDOR1	224	/* for vendor-internal use */
586 #define	EXT_VENDOR2	225	/* for vendor-internal use */
587 #define	EXT_VENDOR3	226	/* for vendor-internal use */
588 #define	EXT_VENDOR4	227	/* for vendor-internal use */
589 
590 #define	EXT_EXP1	244	/* for experimental use */
591 #define	EXT_EXP2	245	/* for experimental use */
592 #define	EXT_EXP3	246	/* for experimental use */
593 #define	EXT_EXP4	247	/* for experimental use */
594 
595 #define	EXT_NET_DRV	252	/* custom ext_buf provided by net driver(s) */
596 #define	EXT_MOD_TYPE	253	/* custom module's ext_buf type */
597 #define	EXT_DISPOSABLE	254	/* can throw this buffer away w/page flipping */
598 #define	EXT_EXTREF	255	/* has externally maintained ext_cnt ptr */
599 
600 /*
601  * Flags for external mbuf buffer types.
602  * NB: limited to the lower 24 bits.
603  */
604 #define	EXT_FLAG_EMBREF		0x000001	/* embedded ext_count */
605 #define	EXT_FLAG_EXTREF		0x000002	/* external ext_cnt, notyet */
606 
607 #define	EXT_FLAG_NOFREE		0x000010	/* don't free mbuf to pool, notyet */
608 
609 #define	EXT_FLAG_VENDOR1	0x010000	/* These flags are vendor */
610 #define	EXT_FLAG_VENDOR2	0x020000	/* or submodule specific, */
611 #define	EXT_FLAG_VENDOR3	0x040000	/* not used by mbuf code. */
612 #define	EXT_FLAG_VENDOR4	0x080000	/* Set/read by submodule. */
613 
614 #define	EXT_FLAG_EXP1		0x100000	/* for experimental use */
615 #define	EXT_FLAG_EXP2		0x200000	/* for experimental use */
616 #define	EXT_FLAG_EXP3		0x400000	/* for experimental use */
617 #define	EXT_FLAG_EXP4		0x800000	/* for experimental use */
618 
619 /*
620  * EXT flag description for use with printf(9) %b identifier.
621  */
622 #define	EXT_FLAG_BITS \
623     "\20\1EXT_FLAG_EMBREF\2EXT_FLAG_EXTREF\5EXT_FLAG_NOFREE" \
624     "\21EXT_FLAG_VENDOR1\22EXT_FLAG_VENDOR2\23EXT_FLAG_VENDOR3" \
625     "\24EXT_FLAG_VENDOR4\25EXT_FLAG_EXP1\26EXT_FLAG_EXP2\27EXT_FLAG_EXP3" \
626     "\30EXT_FLAG_EXP4"
627 
628 /*
629  * Flags indicating checksum, segmentation and other offload work to be
630  * done, or already done, by hardware or lower layers.  It is split into
631  * separate inbound and outbound flags.
632  *
633  * Outbound flags that are set by upper protocol layers requesting lower
634  * layers, or ideally the hardware, to perform these offloading tasks.
635  * For outbound packets this field and its flags can be directly tested
636  * against ifnet if_hwassist.  Note that the outbound and the inbound flags do
637  * not collide right now but they could be allowed to (as long as the flags are
638  * scrubbed appropriately when the direction of an mbuf changes).  CSUM_BITS
639  * would also have to split into CSUM_BITS_TX and CSUM_BITS_RX.
640  *
641  * CSUM_INNER_<x> is the same as CSUM_<x> but it applies to the inner frame.
642  * The CSUM_ENCAP_<x> bits identify the outer encapsulation.
643  */
644 #define	CSUM_IP			0x00000001	/* IP header checksum offload */
645 #define	CSUM_IP_UDP		0x00000002	/* UDP checksum offload */
646 #define	CSUM_IP_TCP		0x00000004	/* TCP checksum offload */
647 #define	CSUM_IP_SCTP		0x00000008	/* SCTP checksum offload */
648 #define	CSUM_IP_TSO		0x00000010	/* TCP segmentation offload */
649 #define	CSUM_IP_ISCSI		0x00000020	/* iSCSI checksum offload */
650 
651 #define	CSUM_INNER_IP6_UDP	0x00000040
652 #define	CSUM_INNER_IP6_TCP	0x00000080
653 #define	CSUM_INNER_IP6_TSO	0x00000100
654 #define	CSUM_IP6_UDP		0x00000200	/* UDP checksum offload */
655 #define	CSUM_IP6_TCP		0x00000400	/* TCP checksum offload */
656 #define	CSUM_IP6_SCTP		0x00000800	/* SCTP checksum offload */
657 #define	CSUM_IP6_TSO		0x00001000	/* TCP segmentation offload */
658 #define	CSUM_IP6_ISCSI		0x00002000	/* iSCSI checksum offload */
659 
660 #define	CSUM_INNER_IP		0x00004000
661 #define	CSUM_INNER_IP_UDP	0x00008000
662 #define	CSUM_INNER_IP_TCP	0x00010000
663 #define	CSUM_INNER_IP_TSO	0x00020000
664 
665 #define	CSUM_ENCAP_VXLAN	0x00040000	/* VXLAN outer encapsulation */
666 #define	CSUM_ENCAP_RSVD1	0x00080000
667 
668 /* Inbound checksum support where the checksum was verified by hardware. */
669 #define	CSUM_INNER_L3_CALC	0x00100000
670 #define	CSUM_INNER_L3_VALID	0x00200000
671 #define	CSUM_INNER_L4_CALC	0x00400000
672 #define	CSUM_INNER_L4_VALID	0x00800000
673 #define	CSUM_L3_CALC		0x01000000	/* calculated layer 3 csum */
674 #define	CSUM_L3_VALID		0x02000000	/* checksum is correct */
675 #define	CSUM_L4_CALC		0x04000000	/* calculated layer 4 csum */
676 #define	CSUM_L4_VALID		0x08000000	/* checksum is correct */
677 #define	CSUM_L5_CALC		0x10000000	/* calculated layer 5 csum */
678 #define	CSUM_L5_VALID		0x20000000	/* checksum is correct */
679 #define	CSUM_COALESCED		0x40000000	/* contains merged segments */
680 
681 #define	CSUM_SND_TAG		0x80000000	/* Packet header has send tag */
682 
683 #define CSUM_FLAGS_TX (CSUM_IP | CSUM_IP_UDP | CSUM_IP_TCP | CSUM_IP_SCTP | \
684     CSUM_IP_TSO | CSUM_IP_ISCSI | CSUM_INNER_IP6_UDP | CSUM_INNER_IP6_TCP | \
685     CSUM_INNER_IP6_TSO | CSUM_IP6_UDP | CSUM_IP6_TCP | CSUM_IP6_SCTP | \
686     CSUM_IP6_TSO | CSUM_IP6_ISCSI | CSUM_INNER_IP | CSUM_INNER_IP_UDP | \
687     CSUM_INNER_IP_TCP | CSUM_INNER_IP_TSO | CSUM_ENCAP_VXLAN | \
688     CSUM_ENCAP_RSVD1 | CSUM_SND_TAG)
689 
690 #define CSUM_FLAGS_RX (CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID | \
691     CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID | CSUM_L3_CALC | CSUM_L3_VALID | \
692     CSUM_L4_CALC | CSUM_L4_VALID | CSUM_L5_CALC | CSUM_L5_VALID | \
693     CSUM_COALESCED)
694 
695 /*
696  * CSUM flag description for use with printf(9) %b identifier.
697  */
698 #define	CSUM_BITS \
699     "\20\1CSUM_IP\2CSUM_IP_UDP\3CSUM_IP_TCP\4CSUM_IP_SCTP\5CSUM_IP_TSO" \
700     "\6CSUM_IP_ISCSI\7CSUM_INNER_IP6_UDP\10CSUM_INNER_IP6_TCP" \
701     "\11CSUM_INNER_IP6_TSO\12CSUM_IP6_UDP\13CSUM_IP6_TCP\14CSUM_IP6_SCTP" \
702     "\15CSUM_IP6_TSO\16CSUM_IP6_ISCSI\17CSUM_INNER_IP\20CSUM_INNER_IP_UDP" \
703     "\21CSUM_INNER_IP_TCP\22CSUM_INNER_IP_TSO\23CSUM_ENCAP_VXLAN" \
704     "\24CSUM_ENCAP_RSVD1\25CSUM_INNER_L3_CALC\26CSUM_INNER_L3_VALID" \
705     "\27CSUM_INNER_L4_CALC\30CSUM_INNER_L4_VALID\31CSUM_L3_CALC" \
706     "\32CSUM_L3_VALID\33CSUM_L4_CALC\34CSUM_L4_VALID\35CSUM_L5_CALC" \
707     "\36CSUM_L5_VALID\37CSUM_COALESCED\40CSUM_SND_TAG"
708 
709 /* CSUM flags compatibility mappings. */
710 #define	CSUM_IP_CHECKED		CSUM_L3_CALC
711 #define	CSUM_IP_VALID		CSUM_L3_VALID
712 #define	CSUM_DATA_VALID		CSUM_L4_VALID
713 #define	CSUM_PSEUDO_HDR		CSUM_L4_CALC
714 #define	CSUM_SCTP_VALID		CSUM_L4_VALID
715 #define	CSUM_DELAY_DATA		(CSUM_TCP|CSUM_UDP)
716 #define	CSUM_DELAY_IP		CSUM_IP		/* Only v4, no v6 IP hdr csum */
717 #define	CSUM_DELAY_DATA_IPV6	(CSUM_TCP_IPV6|CSUM_UDP_IPV6)
718 #define	CSUM_DATA_VALID_IPV6	CSUM_DATA_VALID
719 #define	CSUM_TCP		CSUM_IP_TCP
720 #define	CSUM_UDP		CSUM_IP_UDP
721 #define	CSUM_SCTP		CSUM_IP_SCTP
722 #define	CSUM_TSO		(CSUM_IP_TSO|CSUM_IP6_TSO)
723 #define	CSUM_INNER_TSO		(CSUM_INNER_IP_TSO|CSUM_INNER_IP6_TSO)
724 #define	CSUM_UDP_IPV6		CSUM_IP6_UDP
725 #define	CSUM_TCP_IPV6		CSUM_IP6_TCP
726 #define	CSUM_SCTP_IPV6		CSUM_IP6_SCTP
727 #define	CSUM_TLS_MASK		(CSUM_L5_CALC|CSUM_L5_VALID)
728 #define	CSUM_TLS_DECRYPTED	CSUM_L5_CALC
729 
730 /*
731  * mbuf types describing the content of the mbuf (including external storage).
732  */
733 #define	MT_NOTMBUF	0	/* USED INTERNALLY ONLY! Object is not mbuf */
734 #define	MT_DATA		1	/* dynamic (data) allocation */
735 #define	MT_HEADER	MT_DATA	/* packet header, use M_PKTHDR instead */
736 
737 #define	MT_VENDOR1	4	/* for vendor-internal use */
738 #define	MT_VENDOR2	5	/* for vendor-internal use */
739 #define	MT_VENDOR3	6	/* for vendor-internal use */
740 #define	MT_VENDOR4	7	/* for vendor-internal use */
741 
742 #define	MT_SONAME	8	/* socket name */
743 
744 #define	MT_EXP1		9	/* for experimental use */
745 #define	MT_EXP2		10	/* for experimental use */
746 #define	MT_EXP3		11	/* for experimental use */
747 #define	MT_EXP4		12	/* for experimental use */
748 
749 #define	MT_CONTROL	14	/* extra-data protocol message */
750 #define	MT_EXTCONTROL	15	/* control message with externalized contents */
751 #define	MT_OOBDATA	16	/* expedited data  */
752 
753 #define	MT_NOINIT	255	/* Not a type but a flag to allocate
754 				   a non-initialized mbuf */
755 
756 /*
757  * String names of mbuf-related UMA(9) and malloc(9) types.  Exposed to
758  * !_KERNEL so that monitoring tools can look up the zones with
759  * libmemstat(3).
760  */
761 #define	MBUF_MEM_NAME		"mbuf"
762 #define	MBUF_CLUSTER_MEM_NAME	"mbuf_cluster"
763 #define	MBUF_PACKET_MEM_NAME	"mbuf_packet"
764 #define	MBUF_JUMBOP_MEM_NAME	"mbuf_jumbo_page"
765 #define	MBUF_JUMBO9_MEM_NAME	"mbuf_jumbo_9k"
766 #define	MBUF_JUMBO16_MEM_NAME	"mbuf_jumbo_16k"
767 #define	MBUF_TAG_MEM_NAME	"mbuf_tag"
768 #define	MBUF_EXTREFCNT_MEM_NAME	"mbuf_ext_refcnt"
769 #define	MBUF_EXTPGS_MEM_NAME	"mbuf_extpgs"
770 
771 #ifdef _KERNEL
772 union if_snd_tag_alloc_params;
773 
774 #ifdef WITNESS
775 #define	MBUF_CHECKSLEEP(how) do {					\
776 	if (how == M_WAITOK)						\
777 		WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,		\
778 		    "Sleeping in \"%s\"", __func__);			\
779 } while (0)
780 #else
781 #define	MBUF_CHECKSLEEP(how) do {} while (0)
782 #endif
783 
784 /*
785  * Network buffer allocation API
786  *
787  * The rest of it is defined in kern/kern_mbuf.c
788  */
789 extern uma_zone_t	zone_mbuf;
790 extern uma_zone_t	zone_clust;
791 extern uma_zone_t	zone_pack;
792 extern uma_zone_t	zone_jumbop;
793 extern uma_zone_t	zone_jumbo9;
794 extern uma_zone_t	zone_jumbo16;
795 extern uma_zone_t	zone_extpgs;
796 
797 void		 mb_dupcl(struct mbuf *, struct mbuf *);
798 void		 mb_free_ext(struct mbuf *);
799 void		 mb_free_extpg(struct mbuf *);
800 void		 mb_free_mext_pgs(struct mbuf *);
801 struct mbuf	*mb_alloc_ext_pgs(int, m_ext_free_t);
802 struct mbuf	*mb_alloc_ext_plus_pages(int, int);
803 struct mbuf	*mb_mapped_to_unmapped(struct mbuf *, int, int, int,
804 		    struct mbuf **);
805 int		 mb_unmapped_compress(struct mbuf *m);
806 struct mbuf 	*mb_unmapped_to_ext(struct mbuf *m);
807 void		 mb_free_notready(struct mbuf *m, int count);
808 void		 m_adj(struct mbuf *, int);
809 void		 m_adj_decap(struct mbuf *, int);
810 int		 m_apply(struct mbuf *, int, int,
811 		    int (*)(void *, void *, u_int), void *);
812 int		 m_append(struct mbuf *, int, c_caddr_t);
813 void		 m_cat(struct mbuf *, struct mbuf *);
814 void		 m_catpkt(struct mbuf *, struct mbuf *);
815 int		 m_clget(struct mbuf *m, int how);
816 void 		*m_cljget(struct mbuf *m, int how, int size);
817 struct mbuf	*m_collapse(struct mbuf *, int, int);
818 void		 m_copyback(struct mbuf *, int, int, c_caddr_t);
819 void		 m_copydata(const struct mbuf *, int, int, caddr_t);
820 struct mbuf	*m_copym(struct mbuf *, int, int, int);
821 struct mbuf	*m_copypacket(struct mbuf *, int);
822 void		 m_copy_pkthdr(struct mbuf *, struct mbuf *);
823 struct mbuf	*m_copyup(struct mbuf *, int, int);
824 struct mbuf	*m_defrag(struct mbuf *, int);
825 void		 m_demote_pkthdr(struct mbuf *);
826 void		 m_demote(struct mbuf *, int, int);
827 struct mbuf	*m_devget(char *, int, int, struct ifnet *,
828 		    void (*)(char *, caddr_t, u_int));
829 void		 m_dispose_extcontrolm(struct mbuf *m);
830 struct mbuf	*m_dup(const struct mbuf *, int);
831 int		 m_dup_pkthdr(struct mbuf *, const struct mbuf *, int);
832 void		 m_extadd(struct mbuf *, char *, u_int, m_ext_free_t,
833 		    void *, void *, int, int);
834 u_int		 m_fixhdr(struct mbuf *);
835 struct mbuf	*m_fragment(struct mbuf *, int, int);
836 void		 m_freem(struct mbuf *);
837 void		 m_free_raw(struct mbuf *);
838 struct mbuf	*m_get2(int, int, short, int);
839 struct mbuf	*m_getjcl(int, short, int, int);
840 struct mbuf	*m_getm2(struct mbuf *, int, int, short, int);
841 struct mbuf	*m_getptr(struct mbuf *, int, int *);
842 u_int		 m_length(struct mbuf *, struct mbuf **);
843 int		 m_mbuftouio(struct uio *, const struct mbuf *, int);
844 void		 m_move_pkthdr(struct mbuf *, struct mbuf *);
845 int		 m_pkthdr_init(struct mbuf *, int);
846 struct mbuf	*m_prepend(struct mbuf *, int, int);
847 void		 m_print(const struct mbuf *, int);
848 struct mbuf	*m_pulldown(struct mbuf *, int, int, int *);
849 struct mbuf	*m_pullup(struct mbuf *, int);
850 int		 m_sanity(struct mbuf *, int);
851 struct mbuf	*m_split(struct mbuf *, int, int);
852 struct mbuf	*m_uiotombuf(struct uio *, int, int, int, int);
853 int		 m_unmapped_uiomove(const struct mbuf *, int, struct uio *,
854 		    int);
855 struct mbuf	*m_unshare(struct mbuf *, int);
856 int		 m_snd_tag_alloc(struct ifnet *,
857 		    union if_snd_tag_alloc_params *, struct m_snd_tag **);
858 void		 m_snd_tag_init(struct m_snd_tag *, struct ifnet *, u_int);
859 void		 m_snd_tag_destroy(struct m_snd_tag *);
860 
861 static __inline int
m_gettype(int size)862 m_gettype(int size)
863 {
864 	int type;
865 
866 	switch (size) {
867 	case MSIZE:
868 		type = EXT_MBUF;
869 		break;
870 	case MCLBYTES:
871 		type = EXT_CLUSTER;
872 		break;
873 #if MJUMPAGESIZE != MCLBYTES
874 	case MJUMPAGESIZE:
875 		type = EXT_JUMBOP;
876 		break;
877 #endif
878 	case MJUM9BYTES:
879 		type = EXT_JUMBO9;
880 		break;
881 	case MJUM16BYTES:
882 		type = EXT_JUMBO16;
883 		break;
884 	default:
885 		panic("%s: invalid cluster size %d", __func__, size);
886 	}
887 
888 	return (type);
889 }
890 
891 /*
892  * Associated an external reference counted buffer with an mbuf.
893  */
894 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)895 m_extaddref(struct mbuf *m, char *buf, u_int size, u_int *ref_cnt,
896     m_ext_free_t freef, void *arg1, void *arg2)
897 {
898 
899 	KASSERT(ref_cnt != NULL, ("%s: ref_cnt not provided", __func__));
900 
901 	atomic_add_int(ref_cnt, 1);
902 	m->m_flags |= M_EXT;
903 	m->m_ext.ext_buf = buf;
904 	m->m_ext.ext_cnt = ref_cnt;
905 	m->m_data = m->m_ext.ext_buf;
906 	m->m_ext.ext_size = size;
907 	m->m_ext.ext_free = freef;
908 	m->m_ext.ext_arg1 = arg1;
909 	m->m_ext.ext_arg2 = arg2;
910 	m->m_ext.ext_type = EXT_EXTREF;
911 	m->m_ext.ext_flags = 0;
912 }
913 
914 static __inline uma_zone_t
m_getzone(int size)915 m_getzone(int size)
916 {
917 	uma_zone_t zone;
918 
919 	switch (size) {
920 	case MCLBYTES:
921 		zone = zone_clust;
922 		break;
923 #if MJUMPAGESIZE != MCLBYTES
924 	case MJUMPAGESIZE:
925 		zone = zone_jumbop;
926 		break;
927 #endif
928 	case MJUM9BYTES:
929 		zone = zone_jumbo9;
930 		break;
931 	case MJUM16BYTES:
932 		zone = zone_jumbo16;
933 		break;
934 	default:
935 		panic("%s: invalid cluster size %d", __func__, size);
936 	}
937 
938 	return (zone);
939 }
940 
941 /*
942  * Initialize an mbuf with linear storage.
943  *
944  * Inline because the consumer text overhead will be roughly the same to
945  * initialize or call a function with this many parameters and M_PKTHDR
946  * should go away with constant propagation for !MGETHDR.
947  */
948 static __inline int
m_init(struct mbuf * m,int how,short type,int flags)949 m_init(struct mbuf *m, int how, short type, int flags)
950 {
951 	int error;
952 
953 	m->m_next = NULL;
954 	m->m_nextpkt = NULL;
955 	m->m_data = m->m_dat;
956 	m->m_len = 0;
957 	m->m_flags = flags;
958 	m->m_type = type;
959 	if (flags & M_PKTHDR)
960 		error = m_pkthdr_init(m, how);
961 	else
962 		error = 0;
963 
964 	MBUF_PROBE5(m__init, m, how, type, flags, error);
965 	return (error);
966 }
967 
968 static __inline struct mbuf *
m_get(int how,short type)969 m_get(int how, short type)
970 {
971 	struct mbuf *m;
972 	struct mb_args args;
973 
974 	args.flags = 0;
975 	args.type = type;
976 	m = uma_zalloc_arg(zone_mbuf, &args, how);
977 	MBUF_PROBE3(m__get, how, type, m);
978 	return (m);
979 }
980 
981 static __inline struct mbuf *
m_gethdr(int how,short type)982 m_gethdr(int how, short type)
983 {
984 	struct mbuf *m;
985 	struct mb_args args;
986 
987 	args.flags = M_PKTHDR;
988 	args.type = type;
989 	m = uma_zalloc_arg(zone_mbuf, &args, how);
990 	MBUF_PROBE3(m__gethdr, how, type, m);
991 	return (m);
992 }
993 
994 static __inline struct mbuf *
m_getcl(int how,short type,int flags)995 m_getcl(int how, short type, int flags)
996 {
997 	struct mbuf *m;
998 	struct mb_args args;
999 
1000 	args.flags = flags;
1001 	args.type = type;
1002 	m = uma_zalloc_arg(zone_pack, &args, how);
1003 	MBUF_PROBE4(m__getcl, how, type, flags, m);
1004 	return (m);
1005 }
1006 
1007 /*
1008  * XXX: m_cljset() is a dangerous API.  One must attach only a new,
1009  * unreferenced cluster to an mbuf(9).  It is not possible to assert
1010  * that, so care can be taken only by users of the API.
1011  */
1012 static __inline void
m_cljset(struct mbuf * m,void * cl,int type)1013 m_cljset(struct mbuf *m, void *cl, int type)
1014 {
1015 	int size;
1016 
1017 	switch (type) {
1018 	case EXT_CLUSTER:
1019 		size = MCLBYTES;
1020 		break;
1021 #if MJUMPAGESIZE != MCLBYTES
1022 	case EXT_JUMBOP:
1023 		size = MJUMPAGESIZE;
1024 		break;
1025 #endif
1026 	case EXT_JUMBO9:
1027 		size = MJUM9BYTES;
1028 		break;
1029 	case EXT_JUMBO16:
1030 		size = MJUM16BYTES;
1031 		break;
1032 	default:
1033 		panic("%s: unknown cluster type %d", __func__, type);
1034 		break;
1035 	}
1036 
1037 	m->m_data = m->m_ext.ext_buf = cl;
1038 	m->m_ext.ext_free = m->m_ext.ext_arg1 = m->m_ext.ext_arg2 = NULL;
1039 	m->m_ext.ext_size = size;
1040 	m->m_ext.ext_type = type;
1041 	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
1042 	m->m_ext.ext_count = 1;
1043 	m->m_flags |= M_EXT;
1044 	MBUF_PROBE3(m__cljset, m, cl, type);
1045 }
1046 
1047 static __inline void
m_chtype(struct mbuf * m,short new_type)1048 m_chtype(struct mbuf *m, short new_type)
1049 {
1050 
1051 	m->m_type = new_type;
1052 }
1053 
1054 static __inline void
m_clrprotoflags(struct mbuf * m)1055 m_clrprotoflags(struct mbuf *m)
1056 {
1057 
1058 	while (m) {
1059 		m->m_flags &= ~M_PROTOFLAGS;
1060 		m = m->m_next;
1061 	}
1062 }
1063 
1064 static __inline struct mbuf *
m_last(struct mbuf * m)1065 m_last(struct mbuf *m)
1066 {
1067 
1068 	while (m->m_next)
1069 		m = m->m_next;
1070 	return (m);
1071 }
1072 
1073 static inline u_int
m_extrefcnt(struct mbuf * m)1074 m_extrefcnt(struct mbuf *m)
1075 {
1076 
1077 	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT missing", __func__));
1078 
1079 	return ((m->m_ext.ext_flags & EXT_FLAG_EMBREF) ? m->m_ext.ext_count :
1080 	    *m->m_ext.ext_cnt);
1081 }
1082 
1083 /*
1084  * mbuf, cluster, and external object allocation macros (for compatibility
1085  * purposes).
1086  */
1087 #define	M_MOVE_PKTHDR(to, from)	m_move_pkthdr((to), (from))
1088 #define	MGET(m, how, type)	((m) = m_get((how), (type)))
1089 #define	MGETHDR(m, how, type)	((m) = m_gethdr((how), (type)))
1090 #define	MCLGET(m, how)		m_clget((m), (how))
1091 #define	MEXTADD(m, buf, size, free, arg1, arg2, flags, type)		\
1092     m_extadd((m), (char *)(buf), (size), (free), (arg1), (arg2),	\
1093     (flags), (type))
1094 #define	m_getm(m, len, how, type)					\
1095     m_getm2((m), (len), (how), (type), M_PKTHDR)
1096 
1097 /*
1098  * Evaluate TRUE if it's safe to write to the mbuf m's data region (this can
1099  * be both the local data payload, or an external buffer area, depending on
1100  * whether M_EXT is set).
1101  */
1102 #define	M_WRITABLE(m)	(((m)->m_flags & (M_RDONLY | M_EXTPG)) == 0 &&	\
1103 			 (!(((m)->m_flags & M_EXT)) ||			\
1104 			 (m_extrefcnt(m) == 1)))
1105 
1106 /* Check if the supplied mbuf has a packet header, or else panic. */
1107 #define	M_ASSERTPKTHDR(m)						\
1108 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR,			\
1109 	    ("%s: no mbuf packet header!", __func__))
1110 
1111 /* Check if the supplied mbuf has no send tag, or else panic. */
1112 #define	M_ASSERT_NO_SND_TAG(m)						\
1113 	KASSERT((m) != NULL && (m)->m_flags & M_PKTHDR &&		\
1114 	       ((m)->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0,		\
1115 	    ("%s: receive mbuf has send tag!", __func__))
1116 
1117 /* Check if mbuf is multipage. */
1118 #define M_ASSERTEXTPG(m)						\
1119 	KASSERT(((m)->m_flags & (M_EXTPG|M_PKTHDR)) == M_EXTPG,		\
1120 	    ("%s: m %p is not multipage!", __func__, m))
1121 
1122 /*
1123  * Ensure that the supplied mbuf is a valid, non-free mbuf.
1124  *
1125  * XXX: Broken at the moment.  Need some UMA magic to make it work again.
1126  */
1127 #define	M_ASSERTVALID(m)						\
1128 	KASSERT((((struct mbuf *)m)->m_flags & 0) == 0,			\
1129 	    ("%s: attempted use of a free mbuf!", __func__))
1130 
1131 /* Check whether any mbuf in the chain is unmapped. */
1132 #ifdef INVARIANTS
1133 #define	M_ASSERTMAPPED(m) do {						\
1134 	for (struct mbuf *__m = (m); __m != NULL; __m = __m->m_next)	\
1135 		KASSERT((__m->m_flags & M_EXTPG) == 0,			\
1136 		    ("%s: chain %p contains an unmapped mbuf", __func__, (m)));\
1137 } while (0)
1138 #else
1139 #define	M_ASSERTMAPPED(m) do {} while (0)
1140 #endif
1141 
1142 /*
1143  * Return the address of the start of the buffer associated with an mbuf,
1144  * handling external storage, packet-header mbufs, and regular data mbufs.
1145  */
1146 #define	M_START(m)							\
1147 	(((m)->m_flags & M_EXTPG) ? NULL :				\
1148 	 ((m)->m_flags & M_EXT) ? (m)->m_ext.ext_buf :			\
1149 	 ((m)->m_flags & M_PKTHDR) ? &(m)->m_pktdat[0] :		\
1150 	 &(m)->m_dat[0])
1151 
1152 /*
1153  * Return the size of the buffer associated with an mbuf, handling external
1154  * storage, packet-header mbufs, and regular data mbufs.
1155  */
1156 #define	M_SIZE(m)							\
1157 	(((m)->m_flags & M_EXT) ? (m)->m_ext.ext_size :			\
1158 	 ((m)->m_flags & M_PKTHDR) ? MHLEN :				\
1159 	 MLEN)
1160 
1161 /*
1162  * Set the m_data pointer of a newly allocated mbuf to place an object of the
1163  * specified size at the end of the mbuf, longword aligned.
1164  *
1165  * NB: Historically, we had M_ALIGN(), MH_ALIGN(), and MEXT_ALIGN() as
1166  * separate macros, each asserting that it was called at the proper moment.
1167  * This required callers to themselves test the storage type and call the
1168  * right one.  Rather than require callers to be aware of those layout
1169  * decisions, we centralize here.
1170  */
1171 static __inline void
m_align(struct mbuf * m,int len)1172 m_align(struct mbuf *m, int len)
1173 {
1174 #ifdef INVARIANTS
1175 	const char *msg = "%s: not a virgin mbuf";
1176 #endif
1177 	int adjust;
1178 
1179 	KASSERT(m->m_data == M_START(m), (msg, __func__));
1180 
1181 	adjust = M_SIZE(m) - len;
1182 	m->m_data += adjust &~ (sizeof(long)-1);
1183 }
1184 
1185 #define	M_ALIGN(m, len)		m_align(m, len)
1186 #define	MH_ALIGN(m, len)	m_align(m, len)
1187 #define	MEXT_ALIGN(m, len)	m_align(m, len)
1188 
1189 /*
1190  * Compute the amount of space available before the current start of data in
1191  * an mbuf.
1192  *
1193  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1194  * of checking writability of the mbuf data area rests solely with the caller.
1195  *
1196  * NB: In previous versions, M_LEADINGSPACE() would only check M_WRITABLE()
1197  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1198  * read-only as well.
1199  */
1200 #define	M_LEADINGSPACE(m)						\
1201 	(M_WRITABLE(m) ? ((m)->m_data - M_START(m)) : 0)
1202 
1203 /*
1204  * Compute the amount of space available after the end of data in an mbuf.
1205  *
1206  * The M_WRITABLE() is a temporary, conservative safety measure: the burden
1207  * of checking writability of the mbuf data area rests solely with the caller.
1208  *
1209  * NB: In previous versions, M_TRAILINGSPACE() would only check M_WRITABLE()
1210  * for mbufs with external storage.  We now allow mbuf-embedded data to be
1211  * read-only as well.
1212  */
1213 #define	M_TRAILINGSPACE(m)						\
1214 	(M_WRITABLE(m) ?						\
1215 	    ((M_START(m) + M_SIZE(m)) - ((m)->m_data + (m)->m_len)) : 0)
1216 
1217 /*
1218  * Arrange to prepend space of size plen to mbuf m.  If a new mbuf must be
1219  * allocated, how specifies whether to wait.  If the allocation fails, the
1220  * original mbuf chain is freed and m is set to NULL.
1221  */
1222 #define	M_PREPEND(m, plen, how) do {					\
1223 	struct mbuf **_mmp = &(m);					\
1224 	struct mbuf *_mm = *_mmp;					\
1225 	int _mplen = (plen);						\
1226 	int __mhow = (how);						\
1227 									\
1228 	MBUF_CHECKSLEEP(how);						\
1229 	if (M_LEADINGSPACE(_mm) >= _mplen) {				\
1230 		_mm->m_data -= _mplen;					\
1231 		_mm->m_len += _mplen;					\
1232 	} else								\
1233 		_mm = m_prepend(_mm, _mplen, __mhow);			\
1234 	if (_mm != NULL && _mm->m_flags & M_PKTHDR)			\
1235 		_mm->m_pkthdr.len += _mplen;				\
1236 	*_mmp = _mm;							\
1237 } while (0)
1238 
1239 /*
1240  * Change mbuf to new type.  This is a relatively expensive operation and
1241  * should be avoided.
1242  */
1243 #define	MCHTYPE(m, t)	m_chtype((m), (t))
1244 
1245 /* Return the rcvif of a packet header. */
1246 static __inline struct ifnet *
m_rcvif(struct mbuf * m)1247 m_rcvif(struct mbuf *m)
1248 {
1249 
1250 	M_ASSERTPKTHDR(m);
1251 	if (m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1252 		return (NULL);
1253 	return (m->m_pkthdr.rcvif);
1254 }
1255 
1256 /* Length to m_copy to copy all. */
1257 #define	M_COPYALL	1000000000
1258 
1259 extern int		max_datalen;	/* MHLEN - max_hdr */
1260 extern int		max_hdr;	/* Largest link + protocol header */
1261 extern int		max_linkhdr;	/* Largest link-level header */
1262 extern int		max_protohdr;	/* Largest protocol header */
1263 extern int		nmbclusters;	/* Maximum number of clusters */
1264 extern bool		mb_use_ext_pgs;	/* Use ext_pgs for sendfile */
1265 
1266 /*-
1267  * Network packets may have annotations attached by affixing a list of
1268  * "packet tags" to the pkthdr structure.  Packet tags are dynamically
1269  * allocated semi-opaque data structures that have a fixed header
1270  * (struct m_tag) that specifies the size of the memory block and a
1271  * <cookie,type> pair that identifies it.  The cookie is a 32-bit unique
1272  * unsigned value used to identify a module or ABI.  By convention this value
1273  * is chosen as the date+time that the module is created, expressed as the
1274  * number of seconds since the epoch (e.g., using date -u +'%s').  The type
1275  * value is an ABI/module-specific value that identifies a particular
1276  * annotation and is private to the module.  For compatibility with systems
1277  * like OpenBSD that define packet tags w/o an ABI/module cookie, the value
1278  * PACKET_ABI_COMPAT is used to implement m_tag_get and m_tag_find
1279  * compatibility shim functions and several tag types are defined below.
1280  * Users that do not require compatibility should use a private cookie value
1281  * so that packet tag-related definitions can be maintained privately.
1282  *
1283  * Note that the packet tag returned by m_tag_alloc has the default memory
1284  * alignment implemented by malloc.  To reference private data one can use a
1285  * construct like:
1286  *
1287  *	struct m_tag *mtag = m_tag_alloc(...);
1288  *	struct foo *p = (struct foo *)(mtag+1);
1289  *
1290  * if the alignment of struct m_tag is sufficient for referencing members of
1291  * struct foo.  Otherwise it is necessary to embed struct m_tag within the
1292  * private data structure to insure proper alignment; e.g.,
1293  *
1294  *	struct foo {
1295  *		struct m_tag	tag;
1296  *		...
1297  *	};
1298  *	struct foo *p = (struct foo *) m_tag_alloc(...);
1299  *	struct m_tag *mtag = &p->tag;
1300  */
1301 
1302 /*
1303  * Persistent tags stay with an mbuf until the mbuf is reclaimed.  Otherwise
1304  * tags are expected to ``vanish'' when they pass through a network
1305  * interface.  For most interfaces this happens normally as the tags are
1306  * reclaimed when the mbuf is free'd.  However in some special cases
1307  * reclaiming must be done manually.  An example is packets that pass through
1308  * the loopback interface.  Also, one must be careful to do this when
1309  * ``turning around'' packets (e.g., icmp_reflect).
1310  *
1311  * To mark a tag persistent bit-or this flag in when defining the tag id.
1312  * The tag will then be treated as described above.
1313  */
1314 #define	MTAG_PERSISTENT				0x800
1315 
1316 #define	PACKET_TAG_NONE				0  /* Nadda */
1317 
1318 /* Packet tags for use with PACKET_ABI_COMPAT. */
1319 #define	PACKET_TAG_IPSEC_IN_DONE		1  /* IPsec applied, in */
1320 #define	PACKET_TAG_IPSEC_OUT_DONE		2  /* IPsec applied, out */
1321 #define	PACKET_TAG_IPSEC_IN_CRYPTO_DONE		3  /* NIC IPsec crypto done */
1322 #define	PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED	4  /* NIC IPsec crypto req'ed */
1323 #define	PACKET_TAG_IPSEC_IN_COULD_DO_CRYPTO	5  /* NIC notifies IPsec */
1324 #define	PACKET_TAG_IPSEC_PENDING_TDB		6  /* Reminder to do IPsec */
1325 #define	PACKET_TAG_BRIDGE			7  /* Bridge processing done */
1326 #define	PACKET_TAG_GIF				8  /* GIF processing done */
1327 #define	PACKET_TAG_GRE				9  /* GRE processing done */
1328 #define	PACKET_TAG_IN_PACKET_CHECKSUM		10 /* NIC checksumming done */
1329 #define	PACKET_TAG_ENCAP			11 /* Encap.  processing */
1330 #define	PACKET_TAG_IPSEC_SOCKET			12 /* IPSEC socket ref */
1331 #define	PACKET_TAG_IPSEC_HISTORY		13 /* IPSEC history */
1332 #define	PACKET_TAG_IPV6_INPUT			14 /* IPV6 input processing */
1333 #define	PACKET_TAG_DUMMYNET			15 /* dummynet info */
1334 #define	PACKET_TAG_DIVERT			17 /* divert info */
1335 #define	PACKET_TAG_IPFORWARD			18 /* ipforward info */
1336 #define	PACKET_TAG_MACLABEL	(19 | MTAG_PERSISTENT) /* MAC label */
1337 #define	PACKET_TAG_PF				21 /* PF/ALTQ information */
1338 #define	PACKET_TAG_RTSOCKFAM			25 /* rtsock sa family */
1339 #define	PACKET_TAG_IPOPTIONS			27 /* Saved IP options */
1340 #define	PACKET_TAG_CARP				28 /* CARP info */
1341 #define	PACKET_TAG_IPSEC_NAT_T_PORTS		29 /* two uint16_t */
1342 #define	PACKET_TAG_ND_OUTGOING			30 /* ND outgoing */
1343 
1344 /* Specific cookies and tags. */
1345 
1346 /* Packet tag routines. */
1347 struct m_tag	*m_tag_alloc(u_int32_t, int, int, int);
1348 void		 m_tag_delete(struct mbuf *, struct m_tag *);
1349 void		 m_tag_delete_chain(struct mbuf *, struct m_tag *);
1350 void		 m_tag_free_default(struct m_tag *);
1351 struct m_tag	*m_tag_locate(struct mbuf *, u_int32_t, int, struct m_tag *);
1352 struct m_tag	*m_tag_copy(struct m_tag *, int);
1353 int		 m_tag_copy_chain(struct mbuf *, const struct mbuf *, int);
1354 void		 m_tag_delete_nonpersistent(struct mbuf *);
1355 
1356 /*
1357  * Initialize the list of tags associated with an mbuf.
1358  */
1359 static __inline void
m_tag_init(struct mbuf * m)1360 m_tag_init(struct mbuf *m)
1361 {
1362 
1363 	SLIST_INIT(&m->m_pkthdr.tags);
1364 }
1365 
1366 /*
1367  * Set up the contents of a tag.  Note that this does not fill in the free
1368  * method; the caller is expected to do that.
1369  *
1370  * XXX probably should be called m_tag_init, but that was already taken.
1371  */
1372 static __inline void
m_tag_setup(struct m_tag * t,u_int32_t cookie,int type,int len)1373 m_tag_setup(struct m_tag *t, u_int32_t cookie, int type, int len)
1374 {
1375 
1376 	t->m_tag_id = type;
1377 	t->m_tag_len = len;
1378 	t->m_tag_cookie = cookie;
1379 }
1380 
1381 /*
1382  * Reclaim resources associated with a tag.
1383  */
1384 static __inline void
m_tag_free(struct m_tag * t)1385 m_tag_free(struct m_tag *t)
1386 {
1387 
1388 	(*t->m_tag_free)(t);
1389 }
1390 
1391 /*
1392  * Return the first tag associated with an mbuf.
1393  */
1394 static __inline struct m_tag *
m_tag_first(struct mbuf * m)1395 m_tag_first(struct mbuf *m)
1396 {
1397 
1398 	return (SLIST_FIRST(&m->m_pkthdr.tags));
1399 }
1400 
1401 /*
1402  * Return the next tag in the list of tags associated with an mbuf.
1403  */
1404 static __inline struct m_tag *
m_tag_next(struct mbuf * m __unused,struct m_tag * t)1405 m_tag_next(struct mbuf *m __unused, struct m_tag *t)
1406 {
1407 
1408 	return (SLIST_NEXT(t, m_tag_link));
1409 }
1410 
1411 /*
1412  * Prepend a tag to the list of tags associated with an mbuf.
1413  */
1414 static __inline void
m_tag_prepend(struct mbuf * m,struct m_tag * t)1415 m_tag_prepend(struct mbuf *m, struct m_tag *t)
1416 {
1417 
1418 	SLIST_INSERT_HEAD(&m->m_pkthdr.tags, t, m_tag_link);
1419 }
1420 
1421 /*
1422  * Unlink a tag from the list of tags associated with an mbuf.
1423  */
1424 static __inline void
m_tag_unlink(struct mbuf * m,struct m_tag * t)1425 m_tag_unlink(struct mbuf *m, struct m_tag *t)
1426 {
1427 
1428 	SLIST_REMOVE(&m->m_pkthdr.tags, t, m_tag, m_tag_link);
1429 }
1430 
1431 /* These are for OpenBSD compatibility. */
1432 #define	MTAG_ABI_COMPAT		0		/* compatibility ABI */
1433 
1434 static __inline struct m_tag *
m_tag_get(int type,int length,int wait)1435 m_tag_get(int type, int length, int wait)
1436 {
1437 	return (m_tag_alloc(MTAG_ABI_COMPAT, type, length, wait));
1438 }
1439 
1440 static __inline struct m_tag *
m_tag_find(struct mbuf * m,int type,struct m_tag * start)1441 m_tag_find(struct mbuf *m, int type, struct m_tag *start)
1442 {
1443 	return (SLIST_EMPTY(&m->m_pkthdr.tags) ? (struct m_tag *)NULL :
1444 	    m_tag_locate(m, MTAG_ABI_COMPAT, type, start));
1445 }
1446 
1447 static inline struct m_snd_tag *
m_snd_tag_ref(struct m_snd_tag * mst)1448 m_snd_tag_ref(struct m_snd_tag *mst)
1449 {
1450 
1451 	refcount_acquire(&mst->refcount);
1452 	return (mst);
1453 }
1454 
1455 static inline void
m_snd_tag_rele(struct m_snd_tag * mst)1456 m_snd_tag_rele(struct m_snd_tag *mst)
1457 {
1458 
1459 	if (refcount_release(&mst->refcount))
1460 		m_snd_tag_destroy(mst);
1461 }
1462 
1463 static __inline struct mbuf *
m_free(struct mbuf * m)1464 m_free(struct mbuf *m)
1465 {
1466 	struct mbuf *n = m->m_next;
1467 
1468 	MBUF_PROBE1(m__free, m);
1469 	if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE))
1470 		m_tag_delete_chain(m, NULL);
1471 	if (m->m_flags & M_PKTHDR && m->m_pkthdr.csum_flags & CSUM_SND_TAG)
1472 		m_snd_tag_rele(m->m_pkthdr.snd_tag);
1473 	if (m->m_flags & M_EXTPG)
1474 		mb_free_extpg(m);
1475 	else if (m->m_flags & M_EXT)
1476 		mb_free_ext(m);
1477 	else if ((m->m_flags & M_NOFREE) == 0)
1478 		uma_zfree(zone_mbuf, m);
1479 	return (n);
1480 }
1481 
1482 static __inline int
rt_m_getfib(struct mbuf * m)1483 rt_m_getfib(struct mbuf *m)
1484 {
1485 	KASSERT(m->m_flags & M_PKTHDR , ("Attempt to get FIB from non header mbuf."));
1486 	return (m->m_pkthdr.fibnum);
1487 }
1488 
1489 #define M_GETFIB(_m)   rt_m_getfib(_m)
1490 
1491 #define M_SETFIB(_m, _fib) do {						\
1492         KASSERT((_m)->m_flags & M_PKTHDR, ("Attempt to set FIB on non header mbuf."));	\
1493 	((_m)->m_pkthdr.fibnum) = (_fib);				\
1494 } while (0)
1495 
1496 /* flags passed as first argument for "m_xxx_tcpip_hash()" */
1497 #define	MBUF_HASHFLAG_L2	(1 << 2)
1498 #define	MBUF_HASHFLAG_L3	(1 << 3)
1499 #define	MBUF_HASHFLAG_L4	(1 << 4)
1500 
1501 /* mbuf hashing helper routines */
1502 uint32_t	m_ether_tcpip_hash_init(void);
1503 uint32_t	m_ether_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1504 uint32_t	m_infiniband_tcpip_hash_init(void);
1505 uint32_t	m_infiniband_tcpip_hash(const uint32_t, const struct mbuf *, uint32_t);
1506 
1507 #ifdef MBUF_PROFILING
1508  void m_profile(struct mbuf *m);
1509  #define M_PROFILE(m) m_profile(m)
1510 #else
1511  #define M_PROFILE(m)
1512 #endif
1513 
1514 struct mbufq {
1515 	STAILQ_HEAD(, mbuf)	mq_head;
1516 	int			mq_len;
1517 	int			mq_maxlen;
1518 };
1519 
1520 static inline void
mbufq_init(struct mbufq * mq,int maxlen)1521 mbufq_init(struct mbufq *mq, int maxlen)
1522 {
1523 
1524 	STAILQ_INIT(&mq->mq_head);
1525 	mq->mq_maxlen = maxlen;
1526 	mq->mq_len = 0;
1527 }
1528 
1529 static inline struct mbuf *
mbufq_flush(struct mbufq * mq)1530 mbufq_flush(struct mbufq *mq)
1531 {
1532 	struct mbuf *m;
1533 
1534 	m = STAILQ_FIRST(&mq->mq_head);
1535 	STAILQ_INIT(&mq->mq_head);
1536 	mq->mq_len = 0;
1537 	return (m);
1538 }
1539 
1540 static inline void
mbufq_drain(struct mbufq * mq)1541 mbufq_drain(struct mbufq *mq)
1542 {
1543 	struct mbuf *m, *n;
1544 
1545 	n = mbufq_flush(mq);
1546 	while ((m = n) != NULL) {
1547 		n = STAILQ_NEXT(m, m_stailqpkt);
1548 		m_freem(m);
1549 	}
1550 }
1551 
1552 static inline struct mbuf *
mbufq_first(const struct mbufq * mq)1553 mbufq_first(const struct mbufq *mq)
1554 {
1555 
1556 	return (STAILQ_FIRST(&mq->mq_head));
1557 }
1558 
1559 static inline struct mbuf *
mbufq_last(const struct mbufq * mq)1560 mbufq_last(const struct mbufq *mq)
1561 {
1562 
1563 	return (STAILQ_LAST(&mq->mq_head, mbuf, m_stailqpkt));
1564 }
1565 
1566 static inline bool
mbufq_empty(const struct mbufq * mq)1567 mbufq_empty(const struct mbufq *mq)
1568 {
1569 	return (mq->mq_len == 0);
1570 }
1571 
1572 static inline int
mbufq_full(const struct mbufq * mq)1573 mbufq_full(const struct mbufq *mq)
1574 {
1575 
1576 	return (mq->mq_maxlen > 0 && mq->mq_len >= mq->mq_maxlen);
1577 }
1578 
1579 static inline int
mbufq_len(const struct mbufq * mq)1580 mbufq_len(const struct mbufq *mq)
1581 {
1582 
1583 	return (mq->mq_len);
1584 }
1585 
1586 static inline int
mbufq_enqueue(struct mbufq * mq,struct mbuf * m)1587 mbufq_enqueue(struct mbufq *mq, struct mbuf *m)
1588 {
1589 
1590 	if (mbufq_full(mq))
1591 		return (ENOBUFS);
1592 	STAILQ_INSERT_TAIL(&mq->mq_head, m, m_stailqpkt);
1593 	mq->mq_len++;
1594 	return (0);
1595 }
1596 
1597 static inline struct mbuf *
mbufq_dequeue(struct mbufq * mq)1598 mbufq_dequeue(struct mbufq *mq)
1599 {
1600 	struct mbuf *m;
1601 
1602 	m = STAILQ_FIRST(&mq->mq_head);
1603 	if (m) {
1604 		STAILQ_REMOVE_HEAD(&mq->mq_head, m_stailqpkt);
1605 		m->m_nextpkt = NULL;
1606 		mq->mq_len--;
1607 	}
1608 	return (m);
1609 }
1610 
1611 static inline void
mbufq_prepend(struct mbufq * mq,struct mbuf * m)1612 mbufq_prepend(struct mbufq *mq, struct mbuf *m)
1613 {
1614 
1615 	STAILQ_INSERT_HEAD(&mq->mq_head, m, m_stailqpkt);
1616 	mq->mq_len++;
1617 }
1618 
1619 /*
1620  * Note: this doesn't enforce the maximum list size for dst.
1621  */
1622 static inline void
mbufq_concat(struct mbufq * mq_dst,struct mbufq * mq_src)1623 mbufq_concat(struct mbufq *mq_dst, struct mbufq *mq_src)
1624 {
1625 
1626 	mq_dst->mq_len += mq_src->mq_len;
1627 	STAILQ_CONCAT(&mq_dst->mq_head, &mq_src->mq_head);
1628 	mq_src->mq_len = 0;
1629 }
1630 
1631 #ifdef _SYS_TIMESPEC_H_
1632 static inline void
mbuf_tstmp2timespec(struct mbuf * m,struct timespec * ts)1633 mbuf_tstmp2timespec(struct mbuf *m, struct timespec *ts)
1634 {
1635 
1636 	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1637 	KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0,
1638 	    ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1639 	ts->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1640 	ts->tv_nsec = m->m_pkthdr.rcv_tstmp % 1000000000;
1641 }
1642 #endif
1643 
1644 static inline void
mbuf_tstmp2timeval(struct mbuf * m,struct timeval * tv)1645 mbuf_tstmp2timeval(struct mbuf *m, struct timeval *tv)
1646 {
1647 
1648 	KASSERT((m->m_flags & M_PKTHDR) != 0, ("mbuf %p no M_PKTHDR", m));
1649 	KASSERT((m->m_flags & (M_TSTMP|M_TSTMP_LRO)) != 0,
1650 	    ("mbuf %p no M_TSTMP or M_TSTMP_LRO", m));
1651 	tv->tv_sec = m->m_pkthdr.rcv_tstmp / 1000000000;
1652 	tv->tv_usec = (m->m_pkthdr.rcv_tstmp % 1000000000) / 1000;
1653 }
1654 
1655 #ifdef DEBUGNET
1656 /* Invoked from the debugnet client code. */
1657 void	debugnet_mbuf_drain(void);
1658 void	debugnet_mbuf_start(void);
1659 void	debugnet_mbuf_finish(void);
1660 void	debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize);
1661 #endif
1662 
1663 static inline bool
mbuf_has_tls_session(struct mbuf * m)1664 mbuf_has_tls_session(struct mbuf *m)
1665 {
1666 
1667 	if (m->m_flags & M_EXTPG) {
1668 		if (m->m_epg_tls != NULL) {
1669 			return (true);
1670 		}
1671 	}
1672 	return (false);
1673 }
1674 
1675 #endif /* _KERNEL */
1676 #endif /* !_SYS_MBUF_H_ */
1677