1/* SPDX-License-Identifier: GPL-2.0 */ 2#ifndef __NET_PKT_SCHED_H 3#define __NET_PKT_SCHED_H 4 5#include <linux/jiffies.h> 6#include <linux/ktime.h> 7#include <linux/if_vlan.h> 8#include <linux/netdevice.h> 9#include <net/sch_generic.h> 10#include <net/net_namespace.h> 11#include <uapi/linux/pkt_sched.h> 12 13#define DEFAULT_TX_QUEUE_LEN 1000 14#define STAB_SIZE_LOG_MAX 30 15 16struct qdisc_walker { 17 int stop; 18 int skip; 19 int count; 20 int (*fn)(struct Qdisc *, unsigned long cl, struct qdisc_walker *); 21}; 22 23static inline void *qdisc_priv(struct Qdisc *q) 24{ 25 return &q->privdata; 26} 27 28static inline struct Qdisc *qdisc_from_priv(void *priv) 29{ 30 return container_of(priv, struct Qdisc, privdata); 31} 32 33/* 34 Timer resolution MUST BE < 10% of min_schedulable_packet_size/bandwidth 35 36 Normal IP packet size ~ 512byte, hence: 37 38 0.5Kbyte/1Mbyte/sec = 0.5msec, so that we need 50usec timer for 39 10Mbit ethernet. 40 41 10msec resolution -> <50Kbit/sec. 42 43 The result: [34]86 is not good choice for QoS router :-( 44 45 The things are not so bad, because we may use artificial 46 clock evaluated by integration of network data flow 47 in the most critical places. 48 */ 49 50typedef u64 psched_time_t; 51typedef long psched_tdiff_t; 52 53/* Avoid doing 64 bit divide */ 54#define PSCHED_SHIFT 6 55#define PSCHED_TICKS2NS(x) ((s64)(x) << PSCHED_SHIFT) 56#define PSCHED_NS2TICKS(x) ((x) >> PSCHED_SHIFT) 57 58#define PSCHED_TICKS_PER_SEC PSCHED_NS2TICKS(NSEC_PER_SEC) 59#define PSCHED_PASTPERFECT 0 60 61static inline psched_time_t psched_get_time(void) 62{ 63 return PSCHED_NS2TICKS(ktime_get_ns()); 64} 65 66struct qdisc_watchdog { 67 struct hrtimer timer; 68 struct Qdisc *qdisc; 69}; 70 71void qdisc_watchdog_init_clockid(struct qdisc_watchdog *wd, struct Qdisc *qdisc, 72 clockid_t clockid); 73void qdisc_watchdog_init(struct qdisc_watchdog *wd, struct Qdisc *qdisc); 74 75void qdisc_watchdog_schedule_range_ns(struct qdisc_watchdog *wd, u64 expires, 76 u64 delta_ns); 77 78static inline void qdisc_watchdog_schedule_ns(struct qdisc_watchdog *wd, 79 u64 expires) 80{ 81 return qdisc_watchdog_schedule_range_ns(wd, expires, 0ULL); 82} 83 84static inline void qdisc_watchdog_schedule(struct qdisc_watchdog *wd, 85 psched_time_t expires) 86{ 87 qdisc_watchdog_schedule_ns(wd, PSCHED_TICKS2NS(expires)); 88} 89 90void qdisc_watchdog_cancel(struct qdisc_watchdog *wd); 91 92extern struct Qdisc_ops pfifo_qdisc_ops; 93extern struct Qdisc_ops bfifo_qdisc_ops; 94extern struct Qdisc_ops pfifo_head_drop_qdisc_ops; 95 96int fifo_set_limit(struct Qdisc *q, unsigned int limit); 97struct Qdisc *fifo_create_dflt(struct Qdisc *sch, struct Qdisc_ops *ops, 98 unsigned int limit, 99 struct netlink_ext_ack *extack); 100 101int register_qdisc(struct Qdisc_ops *qops); 102void unregister_qdisc(struct Qdisc_ops *qops); 103void qdisc_get_default(char *id, size_t len); 104int qdisc_set_default(const char *id); 105 106void qdisc_hash_add(struct Qdisc *q, bool invisible); 107void qdisc_hash_del(struct Qdisc *q); 108struct Qdisc *qdisc_lookup(struct net_device *dev, u32 handle); 109struct Qdisc *qdisc_lookup_rcu(struct net_device *dev, u32 handle); 110struct qdisc_rate_table *qdisc_get_rtab(struct tc_ratespec *r, 111 struct nlattr *tab, 112 struct netlink_ext_ack *extack); 113void qdisc_put_rtab(struct qdisc_rate_table *tab); 114void qdisc_put_stab(struct qdisc_size_table *tab); 115void qdisc_warn_nonwc(const char *txt, struct Qdisc *qdisc); 116bool sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q, 117 struct net_device *dev, struct netdev_queue *txq, 118 spinlock_t *root_lock, bool validate); 119 120void __qdisc_run(struct Qdisc *q); 121 122static inline void qdisc_run(struct Qdisc *q) 123{ 124 if (qdisc_run_begin(q)) { 125 __qdisc_run(q); 126 qdisc_run_end(q); 127 } 128} 129 130extern const struct nla_policy rtm_tca_policy[TCA_MAX + 1]; 131 132/* Calculate maximal size of packet seen by hard_start_xmit 133 routine of this device. 134 */ 135static inline unsigned int psched_mtu(const struct net_device *dev) 136{ 137 return READ_ONCE(dev->mtu) + dev->hard_header_len; 138} 139 140static inline struct net *qdisc_net(struct Qdisc *q) 141{ 142 return dev_net(q->dev_queue->dev); 143} 144 145struct tc_query_caps_base { 146 enum tc_setup_type type; 147 void *caps; 148}; 149 150struct tc_cbs_qopt_offload { 151 u8 enable; 152 s32 queue; 153 s32 hicredit; 154 s32 locredit; 155 s32 idleslope; 156 s32 sendslope; 157}; 158 159struct tc_etf_qopt_offload { 160 u8 enable; 161 s32 queue; 162}; 163 164struct tc_mqprio_caps { 165 bool validate_queue_counts:1; 166}; 167 168struct tc_mqprio_qopt_offload { 169 /* struct tc_mqprio_qopt must always be the first element */ 170 struct tc_mqprio_qopt qopt; 171 struct netlink_ext_ack *extack; 172 u16 mode; 173 u16 shaper; 174 u32 flags; 175 u64 min_rate[TC_QOPT_MAX_QUEUE]; 176 u64 max_rate[TC_QOPT_MAX_QUEUE]; 177 unsigned long preemptible_tcs; 178}; 179 180struct tc_taprio_caps { 181 bool supports_queue_max_sdu:1; 182 bool gate_mask_per_txq:1; 183 /* Device expects lower TXQ numbers to have higher priority over higher 184 * TXQs, regardless of their TC mapping. DO NOT USE FOR NEW DRIVERS, 185 * INSTEAD ENFORCE A PROPER TC:TXQ MAPPING COMING FROM USER SPACE. 186 */ 187 bool broken_mqprio:1; 188}; 189 190enum tc_taprio_qopt_cmd { 191 TAPRIO_CMD_REPLACE, 192 TAPRIO_CMD_DESTROY, 193 TAPRIO_CMD_STATS, 194 TAPRIO_CMD_QUEUE_STATS, 195}; 196 197/** 198 * struct tc_taprio_qopt_stats - IEEE 802.1Qbv statistics 199 * @window_drops: Frames that were dropped because they were too large to be 200 * transmitted in any of the allotted time windows (open gates) for their 201 * traffic class. 202 * @tx_overruns: Frames still being transmitted by the MAC after the 203 * transmission gate associated with their traffic class has closed. 204 * Equivalent to `12.29.1.1.2 TransmissionOverrun` from 802.1Q-2018. 205 */ 206struct tc_taprio_qopt_stats { 207 u64 window_drops; 208 u64 tx_overruns; 209}; 210 211struct tc_taprio_qopt_queue_stats { 212 int queue; 213 struct tc_taprio_qopt_stats stats; 214}; 215 216struct tc_taprio_sched_entry { 217 u8 command; /* TC_TAPRIO_CMD_* */ 218 219 /* The gate_mask in the offloading side refers to traffic classes */ 220 u32 gate_mask; 221 u32 interval; 222}; 223 224struct tc_taprio_qopt_offload { 225 enum tc_taprio_qopt_cmd cmd; 226 227 union { 228 /* TAPRIO_CMD_STATS */ 229 struct tc_taprio_qopt_stats stats; 230 /* TAPRIO_CMD_QUEUE_STATS */ 231 struct tc_taprio_qopt_queue_stats queue_stats; 232 /* TAPRIO_CMD_REPLACE */ 233 struct { 234 struct tc_mqprio_qopt_offload mqprio; 235 struct netlink_ext_ack *extack; 236 ktime_t base_time; 237 u64 cycle_time; 238 u64 cycle_time_extension; 239 u32 max_sdu[TC_MAX_QUEUE]; 240 241 size_t num_entries; 242 struct tc_taprio_sched_entry entries[]; 243 }; 244 }; 245}; 246 247#if IS_ENABLED(CONFIG_NET_SCH_TAPRIO) 248 249/* Reference counting */ 250struct tc_taprio_qopt_offload *taprio_offload_get(struct tc_taprio_qopt_offload 251 *offload); 252void taprio_offload_free(struct tc_taprio_qopt_offload *offload); 253 254#else 255 256/* Reference counting */ 257static inline struct tc_taprio_qopt_offload * 258taprio_offload_get(struct tc_taprio_qopt_offload *offload) 259{ 260 return NULL; 261} 262 263static inline void taprio_offload_free(struct tc_taprio_qopt_offload *offload) 264{ 265} 266 267#endif 268 269/* Ensure skb_mstamp_ns, which might have been populated with the txtime, is 270 * not mistaken for a software timestamp, because this will otherwise prevent 271 * the dispatch of hardware timestamps to the socket. 272 */ 273static inline void skb_txtime_consumed(struct sk_buff *skb) 274{ 275 skb->tstamp = ktime_set(0, 0); 276} 277 278struct tc_skb_cb { 279 struct qdisc_skb_cb qdisc_cb; 280 281 u16 mru; 282 u8 post_ct:1; 283 u8 post_ct_snat:1; 284 u8 post_ct_dnat:1; 285 u16 zone; /* Only valid if post_ct = true */ 286}; 287 288static inline struct tc_skb_cb *tc_skb_cb(const struct sk_buff *skb) 289{ 290 struct tc_skb_cb *cb = (struct tc_skb_cb *)skb->cb; 291 292 BUILD_BUG_ON(sizeof(*cb) > sizeof_field(struct sk_buff, cb)); 293 return cb; 294} 295 296static inline bool tc_qdisc_stats_dump(struct Qdisc *sch, 297 unsigned long cl, 298 struct qdisc_walker *arg) 299{ 300 if (arg->count >= arg->skip && arg->fn(sch, cl, arg) < 0) { 301 arg->stop = 1; 302 return false; 303 } 304 305 arg->count++; 306 return true; 307} 308 309#endif 310