Michael S. Tsirkin | 2e0ab8c | 2016-06-13 23:54:31 +0300 | [diff] [blame^] | 1 | /* |
| 2 | * Definitions for the 'struct ptr_ring' datastructure. |
| 3 | * |
| 4 | * Author: |
| 5 | * Michael S. Tsirkin <mst@redhat.com> |
| 6 | * |
| 7 | * Copyright (C) 2016 Red Hat, Inc. |
| 8 | * |
| 9 | * This program is free software; you can redistribute it and/or modify it |
| 10 | * under the terms of the GNU General Public License as published by the |
| 11 | * Free Software Foundation; either version 2 of the License, or (at your |
| 12 | * option) any later version. |
| 13 | * |
| 14 | * This is a limited-size FIFO maintaining pointers in FIFO order, with |
| 15 | * one CPU producing entries and another consuming entries from a FIFO. |
| 16 | * |
| 17 | * This implementation tries to minimize cache-contention when there is a |
| 18 | * single producer and a single consumer CPU. |
| 19 | */ |
| 20 | |
| 21 | #ifndef _LINUX_PTR_RING_H |
| 22 | #define _LINUX_PTR_RING_H 1 |
| 23 | |
| 24 | #ifdef __KERNEL__ |
| 25 | #include <linux/spinlock.h> |
| 26 | #include <linux/cache.h> |
| 27 | #include <linux/types.h> |
| 28 | #include <linux/compiler.h> |
| 29 | #include <linux/cache.h> |
| 30 | #include <linux/slab.h> |
| 31 | #include <asm/errno.h> |
| 32 | #endif |
| 33 | |
| 34 | struct ptr_ring { |
| 35 | int producer ____cacheline_aligned_in_smp; |
| 36 | spinlock_t producer_lock; |
| 37 | int consumer ____cacheline_aligned_in_smp; |
| 38 | spinlock_t consumer_lock; |
| 39 | /* Shared consumer/producer data */ |
| 40 | /* Read-only by both the producer and the consumer */ |
| 41 | int size ____cacheline_aligned_in_smp; /* max entries in queue */ |
| 42 | void **queue; |
| 43 | }; |
| 44 | |
| 45 | /* Note: callers invoking this in a loop must use a compiler barrier, |
| 46 | * for example cpu_relax(). |
| 47 | * Callers don't need to take producer lock - if they don't |
| 48 | * the next call to __ptr_ring_produce may fail. |
| 49 | */ |
| 50 | static inline bool __ptr_ring_full(struct ptr_ring *r) |
| 51 | { |
| 52 | return r->queue[r->producer]; |
| 53 | } |
| 54 | |
| 55 | static inline bool ptr_ring_full(struct ptr_ring *r) |
| 56 | { |
| 57 | barrier(); |
| 58 | return __ptr_ring_full(r); |
| 59 | } |
| 60 | |
| 61 | /* Note: callers invoking this in a loop must use a compiler barrier, |
| 62 | * for example cpu_relax(). |
| 63 | */ |
| 64 | static inline int __ptr_ring_produce(struct ptr_ring *r, void *ptr) |
| 65 | { |
| 66 | if (__ptr_ring_full(r)) |
| 67 | return -ENOSPC; |
| 68 | |
| 69 | r->queue[r->producer++] = ptr; |
| 70 | if (unlikely(r->producer >= r->size)) |
| 71 | r->producer = 0; |
| 72 | return 0; |
| 73 | } |
| 74 | |
| 75 | static inline int ptr_ring_produce(struct ptr_ring *r, void *ptr) |
| 76 | { |
| 77 | int ret; |
| 78 | |
| 79 | spin_lock(&r->producer_lock); |
| 80 | ret = __ptr_ring_produce(r, ptr); |
| 81 | spin_unlock(&r->producer_lock); |
| 82 | |
| 83 | return ret; |
| 84 | } |
| 85 | |
| 86 | static inline int ptr_ring_produce_irq(struct ptr_ring *r, void *ptr) |
| 87 | { |
| 88 | int ret; |
| 89 | |
| 90 | spin_lock_irq(&r->producer_lock); |
| 91 | ret = __ptr_ring_produce(r, ptr); |
| 92 | spin_unlock_irq(&r->producer_lock); |
| 93 | |
| 94 | return ret; |
| 95 | } |
| 96 | |
| 97 | static inline int ptr_ring_produce_any(struct ptr_ring *r, void *ptr) |
| 98 | { |
| 99 | unsigned long flags; |
| 100 | int ret; |
| 101 | |
| 102 | spin_lock_irqsave(&r->producer_lock, flags); |
| 103 | ret = __ptr_ring_produce(r, ptr); |
| 104 | spin_unlock_irqrestore(&r->producer_lock, flags); |
| 105 | |
| 106 | return ret; |
| 107 | } |
| 108 | |
| 109 | static inline int ptr_ring_produce_bh(struct ptr_ring *r, void *ptr) |
| 110 | { |
| 111 | int ret; |
| 112 | |
| 113 | spin_lock_bh(&r->producer_lock); |
| 114 | ret = __ptr_ring_produce(r, ptr); |
| 115 | spin_unlock_bh(&r->producer_lock); |
| 116 | |
| 117 | return ret; |
| 118 | } |
| 119 | |
| 120 | /* Note: callers invoking this in a loop must use a compiler barrier, |
| 121 | * for example cpu_relax(). Callers must take consumer_lock |
| 122 | * if they dereference the pointer - see e.g. PTR_RING_PEEK_CALL. |
| 123 | * There's no need for a lock if pointer is merely tested - see e.g. |
| 124 | * ptr_ring_empty. |
| 125 | */ |
| 126 | static inline void *__ptr_ring_peek(struct ptr_ring *r) |
| 127 | { |
| 128 | return r->queue[r->consumer]; |
| 129 | } |
| 130 | |
| 131 | static inline bool ptr_ring_empty(struct ptr_ring *r) |
| 132 | { |
| 133 | barrier(); |
| 134 | return !__ptr_ring_peek(r); |
| 135 | } |
| 136 | |
| 137 | /* Must only be called after __ptr_ring_peek returned !NULL */ |
| 138 | static inline void __ptr_ring_discard_one(struct ptr_ring *r) |
| 139 | { |
| 140 | r->queue[r->consumer++] = NULL; |
| 141 | if (unlikely(r->consumer >= r->size)) |
| 142 | r->consumer = 0; |
| 143 | } |
| 144 | |
| 145 | static inline void *__ptr_ring_consume(struct ptr_ring *r) |
| 146 | { |
| 147 | void *ptr; |
| 148 | |
| 149 | ptr = __ptr_ring_peek(r); |
| 150 | if (ptr) |
| 151 | __ptr_ring_discard_one(r); |
| 152 | |
| 153 | return ptr; |
| 154 | } |
| 155 | |
| 156 | static inline void *ptr_ring_consume(struct ptr_ring *r) |
| 157 | { |
| 158 | void *ptr; |
| 159 | |
| 160 | spin_lock(&r->consumer_lock); |
| 161 | ptr = __ptr_ring_consume(r); |
| 162 | spin_unlock(&r->consumer_lock); |
| 163 | |
| 164 | return ptr; |
| 165 | } |
| 166 | |
| 167 | static inline void *ptr_ring_consume_irq(struct ptr_ring *r) |
| 168 | { |
| 169 | void *ptr; |
| 170 | |
| 171 | spin_lock_irq(&r->consumer_lock); |
| 172 | ptr = __ptr_ring_consume(r); |
| 173 | spin_unlock_irq(&r->consumer_lock); |
| 174 | |
| 175 | return ptr; |
| 176 | } |
| 177 | |
| 178 | static inline void *ptr_ring_consume_any(struct ptr_ring *r) |
| 179 | { |
| 180 | unsigned long flags; |
| 181 | void *ptr; |
| 182 | |
| 183 | spin_lock_irqsave(&r->consumer_lock, flags); |
| 184 | ptr = __ptr_ring_consume(r); |
| 185 | spin_unlock_irqrestore(&r->consumer_lock, flags); |
| 186 | |
| 187 | return ptr; |
| 188 | } |
| 189 | |
| 190 | static inline void *ptr_ring_consume_bh(struct ptr_ring *r) |
| 191 | { |
| 192 | void *ptr; |
| 193 | |
| 194 | spin_lock_bh(&r->consumer_lock); |
| 195 | ptr = __ptr_ring_consume(r); |
| 196 | spin_unlock_bh(&r->consumer_lock); |
| 197 | |
| 198 | return ptr; |
| 199 | } |
| 200 | |
| 201 | /* Cast to structure type and call a function without discarding from FIFO. |
| 202 | * Function must return a value. |
| 203 | * Callers must take consumer_lock. |
| 204 | */ |
| 205 | #define __PTR_RING_PEEK_CALL(r, f) ((f)(__ptr_ring_peek(r))) |
| 206 | |
| 207 | #define PTR_RING_PEEK_CALL(r, f) ({ \ |
| 208 | typeof((f)(NULL)) __PTR_RING_PEEK_CALL_v; \ |
| 209 | \ |
| 210 | spin_lock(&(r)->consumer_lock); \ |
| 211 | __PTR_RING_PEEK_CALL_v = __PTR_RING_PEEK_CALL(r, f); \ |
| 212 | spin_unlock(&(r)->consumer_lock); \ |
| 213 | __PTR_RING_PEEK_CALL_v; \ |
| 214 | }) |
| 215 | |
| 216 | #define PTR_RING_PEEK_CALL_IRQ(r, f) ({ \ |
| 217 | typeof((f)(NULL)) __PTR_RING_PEEK_CALL_v; \ |
| 218 | \ |
| 219 | spin_lock_irq(&(r)->consumer_lock); \ |
| 220 | __PTR_RING_PEEK_CALL_v = __PTR_RING_PEEK_CALL(r, f); \ |
| 221 | spin_unlock_irq(&(r)->consumer_lock); \ |
| 222 | __PTR_RING_PEEK_CALL_v; \ |
| 223 | }) |
| 224 | |
| 225 | #define PTR_RING_PEEK_CALL_BH(r, f) ({ \ |
| 226 | typeof((f)(NULL)) __PTR_RING_PEEK_CALL_v; \ |
| 227 | \ |
| 228 | spin_lock_bh(&(r)->consumer_lock); \ |
| 229 | __PTR_RING_PEEK_CALL_v = __PTR_RING_PEEK_CALL(r, f); \ |
| 230 | spin_unlock_bh(&(r)->consumer_lock); \ |
| 231 | __PTR_RING_PEEK_CALL_v; \ |
| 232 | }) |
| 233 | |
| 234 | #define PTR_RING_PEEK_CALL_ANY(r, f) ({ \ |
| 235 | typeof((f)(NULL)) __PTR_RING_PEEK_CALL_v; \ |
| 236 | unsigned long __PTR_RING_PEEK_CALL_f;\ |
| 237 | \ |
| 238 | spin_lock_irqsave(&(r)->consumer_lock, __PTR_RING_PEEK_CALL_f); \ |
| 239 | __PTR_RING_PEEK_CALL_v = __PTR_RING_PEEK_CALL(r, f); \ |
| 240 | spin_unlock_irqrestore(&(r)->consumer_lock, __PTR_RING_PEEK_CALL_f); \ |
| 241 | __PTR_RING_PEEK_CALL_v; \ |
| 242 | }) |
| 243 | |
| 244 | static inline int ptr_ring_init(struct ptr_ring *r, int size, gfp_t gfp) |
| 245 | { |
| 246 | r->queue = kzalloc(ALIGN(size * sizeof *(r->queue), SMP_CACHE_BYTES), |
| 247 | gfp); |
| 248 | if (!r->queue) |
| 249 | return -ENOMEM; |
| 250 | |
| 251 | r->size = size; |
| 252 | r->producer = r->consumer = 0; |
| 253 | spin_lock_init(&r->producer_lock); |
| 254 | spin_lock_init(&r->consumer_lock); |
| 255 | |
| 256 | return 0; |
| 257 | } |
| 258 | |
| 259 | static inline void ptr_ring_cleanup(struct ptr_ring *r) |
| 260 | { |
| 261 | kfree(r->queue); |
| 262 | } |
| 263 | |
| 264 | #endif /* _LINUX_PTR_RING_H */ |