1/*-
2 * Copyright (c) 2007-2009 Robert N. M. Watson
3 * Copyright (c) 2010-2011 Juniper Networks, Inc.
4 * All rights reserved.
5 *
6 * This software was developed by Robert N. M. Watson under contract
7 * to Juniper Networks, Inc.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 *    notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in the
16 *    documentation and/or other materials provided with the distribution.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * SUCH DAMAGE.
29 *
30 * $FreeBSD$
31 */
32
33#ifndef _NET_NETISR_H_
34#define _NET_NETISR_H_
35
36/*
37 * The netisr (network interrupt service routine) provides a deferred
38 * execution evironment in which (generally inbound) network processing can
39 * take place.  Protocols register handlers which will be executed directly,
40 * or via deferred dispatch, depending on the circumstances.
41 *
42 * Historically, this was implemented by the BSD software ISR facility; it is
43 * now implemented via a software ithread (SWI).
44 */
45
46/*
47 * Protocol numbers, which are encoded in monitoring applications and kernel
48 * modules.  Internally, these are used in bit shift operations so must have
49 * a value 0 < proto < 32; we currently further limit at compile-time to 16
50 * for array-sizing purposes.
51 */
52#define	NETISR_IP	1
53#define	NETISR_IGMP	2		/* IGMPv3 output queue */
54#define	NETISR_ROUTE	3		/* routing socket */
55#define	NETISR_AARP	4		/* Appletalk ARP */
56#define	NETISR_ATALK2	5		/* Appletalk phase 2 */
57#define	NETISR_ATALK1	6		/* Appletalk phase 1 */
58#define	NETISR_ARP	7		/* same as AF_LINK */
59#define	NETISR_IPX	8		/* same as AF_IPX */
60#define	NETISR_ETHER	9		/* ethernet input */
61#define	NETISR_IPV6	10
62#define	NETISR_NATM	11
63#define	NETISR_EPAIR	12		/* if_epair(4) */
64
65/*
66 * Protocol ordering and affinity policy constants.  See the detailed
67 * discussion of policies later in the file.
68 */
69#define	NETISR_POLICY_SOURCE	1	/* Maintain source ordering. */
70#define	NETISR_POLICY_FLOW	2	/* Maintain flow ordering. */
71#define	NETISR_POLICY_CPU	3	/* Protocol determines CPU placement. */
72
73/*
74 * Protocol dispatch policy constants; selects whether and when direct
75 * dispatch is permitted.
76 */
77#define	NETISR_DISPATCH_DEFAULT		0	/* Use global default. */
78#define	NETISR_DISPATCH_DEFERRED	1	/* Always defer dispatch. */
79#define	NETISR_DISPATCH_HYBRID		2	/* Allow hybrid dispatch. */
80#define	NETISR_DISPATCH_DIRECT		3	/* Always direct dispatch. */
81
82/*
83 * Monitoring data structures, exported by sysctl(2).
84 *
85 * Three sysctls are defined.  First, a per-protocol structure exported by
86 * net.isr.proto.
87 */
88#define	NETISR_NAMEMAXLEN	32
89struct sysctl_netisr_proto {
90	u_int	snp_version;			/* Length of struct. */
91	char	snp_name[NETISR_NAMEMAXLEN];	/* nh_name */
92	u_int	snp_proto;			/* nh_proto */
93	u_int	snp_qlimit;			/* nh_qlimit */
94	u_int	snp_policy;			/* nh_policy */
95	u_int	snp_flags;			/* Various flags. */
96	u_int	snp_dispatch;			/* Dispatch policy. */
97	u_int	_snp_ispare[6];
98};
99
100/*
101 * Flags for sysctl_netisr_proto.snp_flags.
102 */
103#define	NETISR_SNP_FLAGS_M2FLOW		0x00000001	/* nh_m2flow */
104#define	NETISR_SNP_FLAGS_M2CPUID	0x00000002	/* nh_m2cpuid */
105#define	NETISR_SNP_FLAGS_DRAINEDCPU	0x00000004	/* nh_drainedcpu */
106
107/*
108 * Next, a structure per-workstream, with per-protocol data, exported as
109 * net.isr.workstream.
110 */
111struct sysctl_netisr_workstream {
112	u_int	snws_version;			/* Length of struct. */
113	u_int	snws_flags;			/* Various flags. */
114	u_int	snws_wsid;			/* Workstream ID. */
115	u_int	snws_cpu;			/* nws_cpu */
116	u_int	_snws_ispare[12];
117};
118
119/*
120 * Flags for sysctl_netisr_workstream.snws_flags
121 */
122#define	NETISR_SNWS_FLAGS_INTR		0x00000001	/* nws_intr_event */
123
124/*
125 * Finally, a per-workstream-per-protocol structure, exported as
126 * net.isr.work.
127 */
128struct sysctl_netisr_work {
129	u_int	snw_version;			/* Length of struct. */
130	u_int	snw_wsid;			/* Workstream ID. */
131	u_int	snw_proto;			/* Protocol number. */
132	u_int	snw_len;			/* nw_len */
133	u_int	snw_watermark;			/* nw_watermark */
134	u_int	_snw_ispare[3];
135
136	uint64_t	snw_dispatched;		/* nw_dispatched */
137	uint64_t	snw_hybrid_dispatched;	/* nw_hybrid_dispatched */
138	uint64_t	snw_qdrops;		/* nw_qdrops */
139	uint64_t	snw_queued;		/* nw_queued */
140	uint64_t	snw_handled;		/* nw_handled */
141
142	uint64_t	_snw_llspare[7];
143};
144
145#ifdef _KERNEL
146
147/*-
148 * Protocols express ordering constraints and affinity preferences by
149 * implementing one or neither of nh_m2flow and nh_m2cpuid, which are used by
150 * netisr to determine which per-CPU workstream to assign mbufs to.
151 *
152 * The following policies may be used by protocols:
153 *
154 * NETISR_POLICY_SOURCE - netisr should maintain source ordering without
155 *                        advice from the protocol.  netisr will ignore any
156 *                        flow IDs present on the mbuf for the purposes of
157 *                        work placement.
158 *
159 * NETISR_POLICY_FLOW - netisr should maintain flow ordering as defined by
160 *                      the mbuf header flow ID field.  If the protocol
161 *                      implements nh_m2flow, then netisr will query the
162 *                      protocol in the event that the mbuf doesn't have a
163 *                      flow ID, falling back on source ordering.
164 *
165 * NETISR_POLICY_CPU - netisr will delegate all work placement decisions to
166 *                     the protocol, querying nh_m2cpuid for each packet.
167 *
168 * Protocols might make decisions about work placement based on an existing
169 * calculated flow ID on the mbuf, such as one provided in hardware, the
170 * receive interface pointed to by the mbuf (if any), the optional source
171 * identifier passed at some dispatch points, or even parse packet headers to
172 * calculate a flow.  Both protocol handlers may return a new mbuf pointer
173 * for the chain, or NULL if the packet proves invalid or m_pullup() fails.
174 *
175 * XXXRW: If we eventually support dynamic reconfiguration, there should be
176 * protocol handlers to notify them of CPU configuration changes so that they
177 * can rebalance work.
178 */
179struct mbuf;
180typedef void		 netisr_handler_t(struct mbuf *m);
181typedef struct mbuf	*netisr_m2cpuid_t(struct mbuf *m, uintptr_t source,
182			 u_int *cpuid);
183typedef	struct mbuf	*netisr_m2flow_t(struct mbuf *m, uintptr_t source);
184typedef void		 netisr_drainedcpu_t(u_int cpuid);
185
186#define	NETISR_CPUID_NONE	((u_int)-1)	/* No affinity returned. */
187
188/*
189 * Data structure describing a protocol handler.
190 */
191struct netisr_handler {
192	const char	*nh_name;	/* Character string protocol name. */
193	netisr_handler_t *nh_handler;	/* Protocol handler. */
194	netisr_m2flow_t	*nh_m2flow;	/* Query flow for untagged packet. */
195	netisr_m2cpuid_t *nh_m2cpuid;	/* Query CPU to process mbuf on. */
196	netisr_drainedcpu_t *nh_drainedcpu; /* Callback when drained a queue. */
197	u_int		 nh_proto;	/* Integer protocol ID. */
198	u_int		 nh_qlimit;	/* Maximum per-CPU queue depth. */
199	u_int		 nh_policy;	/* Work placement policy. */
200	u_int		 nh_dispatch;	/* Dispatch policy. */
201	u_int		 nh_ispare[4];	/* For future use. */
202	void		*nh_pspare[4];	/* For future use. */
203};
204
205/*
206 * Register, unregister, and other netisr handler management functions.
207 */
208void	netisr_clearqdrops(const struct netisr_handler *nhp);
209void	netisr_getqdrops(const struct netisr_handler *nhp,
210	    u_int64_t *qdropsp);
211void	netisr_getqlimit(const struct netisr_handler *nhp, u_int *qlimitp);
212void	netisr_register(const struct netisr_handler *nhp);
213int	netisr_setqlimit(const struct netisr_handler *nhp, u_int qlimit);
214void	netisr_unregister(const struct netisr_handler *nhp);
215
216/*
217 * Process a packet destined for a protocol, and attempt direct dispatch.
218 * Supplemental source ordering information can be passed using the _src
219 * variant.
220 */
221int	netisr_dispatch(u_int proto, struct mbuf *m);
222int	netisr_dispatch_src(u_int proto, uintptr_t source, struct mbuf *m);
223int	netisr_queue(u_int proto, struct mbuf *m);
224int	netisr_queue_src(u_int proto, uintptr_t source, struct mbuf *m);
225
226/*
227 * Provide a default implementation of "map an ID to a CPU ID".
228 */
229u_int	netisr_default_flow2cpu(u_int flowid);
230
231/*
232 * Utility routines to return the number of CPUs participting in netisr, and
233 * to return a mapping from a number to a CPU ID that can be used with the
234 * scheduler.
235 */
236u_int	netisr_get_cpucount(void);
237u_int	netisr_get_cpuid(u_int cpunumber);
238
239/*
240 * Interfaces between DEVICE_POLLING and netisr.
241 */
242void	netisr_sched_poll(void);
243void	netisr_poll(void);
244void	netisr_pollmore(void);
245
246#endif /* !_KERNEL */
247#endif /* !_NET_NETISR_H_ */
248