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_ARP 4 /* same as AF_LINK */ 56 #define NETISR_ETHER 5 /* ethernet input */ 57 #define NETISR_IPV6 6 58 #define NETISR_EPAIR 8 /* if_epair(4) */ 59 #define NETISR_IP_DIRECT 9 /* direct-dispatch IPv4 */ 60 #define NETISR_IPV6_DIRECT 10 /* direct-dispatch IPv6 */ 61 62 /* 63 * Protocol ordering and affinity policy constants. See the detailed 64 * discussion of policies later in the file. 65 */ 66 #define NETISR_POLICY_SOURCE 1 /* Maintain source ordering. */ 67 #define NETISR_POLICY_FLOW 2 /* Maintain flow ordering. */ 68 #define NETISR_POLICY_CPU 3 /* Protocol determines CPU placement. */ 69 70 /* 71 * Protocol dispatch policy constants; selects whether and when direct 72 * dispatch is permitted. 73 */ 74 #define NETISR_DISPATCH_DEFAULT 0 /* Use global default. */ 75 #define NETISR_DISPATCH_DEFERRED 1 /* Always defer dispatch. */ 76 #define NETISR_DISPATCH_HYBRID 2 /* Allow hybrid dispatch. */ 77 #define NETISR_DISPATCH_DIRECT 3 /* Always direct dispatch. */ 78 79 /* 80 * Monitoring data structures, exported by sysctl(2). 81 * 82 * Three sysctls are defined. First, a per-protocol structure exported by 83 * net.isr.proto. 84 */ 85 #define NETISR_NAMEMAXLEN 32 86 struct sysctl_netisr_proto { 87 u_int snp_version; /* Length of struct. */ 88 char snp_name[NETISR_NAMEMAXLEN]; /* nh_name */ 89 u_int snp_proto; /* nh_proto */ 90 u_int snp_qlimit; /* nh_qlimit */ 91 u_int snp_policy; /* nh_policy */ 92 u_int snp_flags; /* Various flags. */ 93 u_int snp_dispatch; /* Dispatch policy. */ 94 u_int _snp_ispare[6]; 95 }; 96 97 /* 98 * Flags for sysctl_netisr_proto.snp_flags. 99 */ 100 #define NETISR_SNP_FLAGS_M2FLOW 0x00000001 /* nh_m2flow */ 101 #define NETISR_SNP_FLAGS_M2CPUID 0x00000002 /* nh_m2cpuid */ 102 #define NETISR_SNP_FLAGS_DRAINEDCPU 0x00000004 /* nh_drainedcpu */ 103 104 /* 105 * Next, a structure per-workstream, with per-protocol data, exported as 106 * net.isr.workstream. 107 */ 108 struct sysctl_netisr_workstream { 109 u_int snws_version; /* Length of struct. */ 110 u_int snws_flags; /* Various flags. */ 111 u_int snws_wsid; /* Workstream ID. */ 112 u_int snws_cpu; /* nws_cpu */ 113 u_int _snws_ispare[12]; 114 }; 115 116 /* 117 * Flags for sysctl_netisr_workstream.snws_flags 118 */ 119 #define NETISR_SNWS_FLAGS_INTR 0x00000001 /* nws_intr_event */ 120 121 /* 122 * Finally, a per-workstream-per-protocol structure, exported as 123 * net.isr.work. 124 */ 125 struct sysctl_netisr_work { 126 u_int snw_version; /* Length of struct. */ 127 u_int snw_wsid; /* Workstream ID. */ 128 u_int snw_proto; /* Protocol number. */ 129 u_int snw_len; /* nw_len */ 130 u_int snw_watermark; /* nw_watermark */ 131 u_int _snw_ispare[3]; 132 133 uint64_t snw_dispatched; /* nw_dispatched */ 134 uint64_t snw_hybrid_dispatched; /* nw_hybrid_dispatched */ 135 uint64_t snw_qdrops; /* nw_qdrops */ 136 uint64_t snw_queued; /* nw_queued */ 137 uint64_t snw_handled; /* nw_handled */ 138 139 uint64_t _snw_llspare[7]; 140 }; 141 142 #ifdef _KERNEL 143 144 /*- 145 * Protocols express ordering constraints and affinity preferences by 146 * implementing one or neither of nh_m2flow and nh_m2cpuid, which are used by 147 * netisr to determine which per-CPU workstream to assign mbufs to. 148 * 149 * The following policies may be used by protocols: 150 * 151 * NETISR_POLICY_SOURCE - netisr should maintain source ordering without 152 * advice from the protocol. netisr will ignore any 153 * flow IDs present on the mbuf for the purposes of 154 * work placement. 155 * 156 * NETISR_POLICY_FLOW - netisr should maintain flow ordering as defined by 157 * the mbuf header flow ID field. If the protocol 158 * implements nh_m2flow, then netisr will query the 159 * protocol in the event that the mbuf doesn't have a 160 * flow ID, falling back on source ordering. 161 * 162 * NETISR_POLICY_CPU - netisr will delegate all work placement decisions to 163 * the protocol, querying nh_m2cpuid for each packet. 164 * 165 * Protocols might make decisions about work placement based on an existing 166 * calculated flow ID on the mbuf, such as one provided in hardware, the 167 * receive interface pointed to by the mbuf (if any), the optional source 168 * identifier passed at some dispatch points, or even parse packet headers to 169 * calculate a flow. Both protocol handlers may return a new mbuf pointer 170 * for the chain, or NULL if the packet proves invalid or m_pullup() fails. 171 * 172 * XXXRW: If we eventually support dynamic reconfiguration, there should be 173 * protocol handlers to notify them of CPU configuration changes so that they 174 * can rebalance work. 175 */ 176 struct mbuf; 177 typedef void netisr_handler_t(struct mbuf *m); 178 typedef struct mbuf *netisr_m2cpuid_t(struct mbuf *m, uintptr_t source, 179 u_int *cpuid); 180 typedef struct mbuf *netisr_m2flow_t(struct mbuf *m, uintptr_t source); 181 typedef void netisr_drainedcpu_t(u_int cpuid); 182 183 #define NETISR_CPUID_NONE ((u_int)-1) /* No affinity returned. */ 184 185 /* 186 * Data structure describing a protocol handler. 187 */ 188 struct netisr_handler { 189 const char *nh_name; /* Character string protocol name. */ 190 netisr_handler_t *nh_handler; /* Protocol handler. */ 191 netisr_m2flow_t *nh_m2flow; /* Query flow for untagged packet. */ 192 netisr_m2cpuid_t *nh_m2cpuid; /* Query CPU to process mbuf on. */ 193 netisr_drainedcpu_t *nh_drainedcpu; /* Callback when drained a queue. */ 194 u_int nh_proto; /* Integer protocol ID. */ 195 u_int nh_qlimit; /* Maximum per-CPU queue depth. */ 196 u_int nh_policy; /* Work placement policy. */ 197 u_int nh_dispatch; /* Dispatch policy. */ 198 u_int nh_ispare[4]; /* For future use. */ 199 void *nh_pspare[4]; /* For future use. */ 200 }; 201 202 /* 203 * Register, unregister, and other netisr handler management functions. 204 */ 205 void netisr_clearqdrops(const struct netisr_handler *nhp); 206 void netisr_getqdrops(const struct netisr_handler *nhp, 207 u_int64_t *qdropsp); 208 void netisr_getqlimit(const struct netisr_handler *nhp, u_int *qlimitp); 209 void netisr_register(const struct netisr_handler *nhp); 210 int netisr_setqlimit(const struct netisr_handler *nhp, u_int qlimit); 211 void netisr_unregister(const struct netisr_handler *nhp); 212 #ifdef VIMAGE 213 void netisr_register_vnet(const struct netisr_handler *nhp); 214 void netisr_unregister_vnet(const struct netisr_handler *nhp); 215 #endif 216 217 /* 218 * Process a packet destined for a protocol, and attempt direct dispatch. 219 * Supplemental source ordering information can be passed using the _src 220 * variant. 221 */ 222 int netisr_dispatch(u_int proto, struct mbuf *m); 223 int netisr_dispatch_src(u_int proto, uintptr_t source, struct mbuf *m); 224 int netisr_queue(u_int proto, struct mbuf *m); 225 int netisr_queue_src(u_int proto, uintptr_t source, struct mbuf *m); 226 227 /* 228 * Provide a default implementation of "map an ID to a CPU ID". 229 */ 230 u_int netisr_default_flow2cpu(u_int flowid); 231 232 /* 233 * Utility routines to return the number of CPUs participting in netisr, and 234 * to return a mapping from a number to a CPU ID that can be used with the 235 * scheduler. 236 */ 237 u_int netisr_get_cpucount(void); 238 u_int netisr_get_cpuid(u_int cpunumber); 239 240 /* 241 * Interfaces between DEVICE_POLLING and netisr. 242 */ 243 void netisr_sched_poll(void); 244 void netisr_poll(void); 245 void netisr_pollmore(void); 246 247 #endif /* !_KERNEL */ 248 #endif /* !_NET_NETISR_H_ */ 249