xref: /openbmc/linux/net/xdp/xsk_queue.h (revision 26ba4e57)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* XDP user-space ring structure
3  * Copyright(c) 2018 Intel Corporation.
4  */
5 
6 #ifndef _LINUX_XSK_QUEUE_H
7 #define _LINUX_XSK_QUEUE_H
8 
9 #include <linux/types.h>
10 #include <linux/if_xdp.h>
11 #include <net/xdp_sock.h>
12 
13 #define RX_BATCH_SIZE 16
14 #define LAZY_UPDATE_THRESHOLD 128
15 
16 struct xdp_ring {
17 	u32 producer ____cacheline_aligned_in_smp;
18 	u32 consumer ____cacheline_aligned_in_smp;
19 	u32 flags;
20 };
21 
22 /* Used for the RX and TX queues for packets */
23 struct xdp_rxtx_ring {
24 	struct xdp_ring ptrs;
25 	struct xdp_desc desc[0] ____cacheline_aligned_in_smp;
26 };
27 
28 /* Used for the fill and completion queues for buffers */
29 struct xdp_umem_ring {
30 	struct xdp_ring ptrs;
31 	u64 desc[0] ____cacheline_aligned_in_smp;
32 };
33 
34 struct xsk_queue {
35 	u64 chunk_mask;
36 	u64 size;
37 	u32 ring_mask;
38 	u32 nentries;
39 	u32 prod_head;
40 	u32 prod_tail;
41 	u32 cons_head;
42 	u32 cons_tail;
43 	struct xdp_ring *ring;
44 	u64 invalid_descs;
45 };
46 
47 /* The structure of the shared state of the rings are the same as the
48  * ring buffer in kernel/events/ring_buffer.c. For the Rx and completion
49  * ring, the kernel is the producer and user space is the consumer. For
50  * the Tx and fill rings, the kernel is the consumer and user space is
51  * the producer.
52  *
53  * producer                         consumer
54  *
55  * if (LOAD ->consumer) {           LOAD ->producer
56  *                    (A)           smp_rmb()       (C)
57  *    STORE $data                   LOAD $data
58  *    smp_wmb()       (B)           smp_mb()        (D)
59  *    STORE ->producer              STORE ->consumer
60  * }
61  *
62  * (A) pairs with (D), and (B) pairs with (C).
63  *
64  * Starting with (B), it protects the data from being written after
65  * the producer pointer. If this barrier was missing, the consumer
66  * could observe the producer pointer being set and thus load the data
67  * before the producer has written the new data. The consumer would in
68  * this case load the old data.
69  *
70  * (C) protects the consumer from speculatively loading the data before
71  * the producer pointer actually has been read. If we do not have this
72  * barrier, some architectures could load old data as speculative loads
73  * are not discarded as the CPU does not know there is a dependency
74  * between ->producer and data.
75  *
76  * (A) is a control dependency that separates the load of ->consumer
77  * from the stores of $data. In case ->consumer indicates there is no
78  * room in the buffer to store $data we do not. So no barrier is needed.
79  *
80  * (D) protects the load of the data to be observed to happen after the
81  * store of the consumer pointer. If we did not have this memory
82  * barrier, the producer could observe the consumer pointer being set
83  * and overwrite the data with a new value before the consumer got the
84  * chance to read the old value. The consumer would thus miss reading
85  * the old entry and very likely read the new entry twice, once right
86  * now and again after circling through the ring.
87  */
88 
89 /* Common functions operating for both RXTX and umem queues */
90 
91 static inline u64 xskq_nb_invalid_descs(struct xsk_queue *q)
92 {
93 	return q ? q->invalid_descs : 0;
94 }
95 
96 static inline u32 xskq_nb_avail(struct xsk_queue *q, u32 dcnt)
97 {
98 	u32 entries = q->prod_tail - q->cons_tail;
99 
100 	if (entries == 0) {
101 		/* Refresh the local pointer */
102 		q->prod_tail = READ_ONCE(q->ring->producer);
103 		entries = q->prod_tail - q->cons_tail;
104 	}
105 
106 	return (entries > dcnt) ? dcnt : entries;
107 }
108 
109 static inline u32 xskq_nb_free(struct xsk_queue *q, u32 producer, u32 dcnt)
110 {
111 	u32 free_entries = q->nentries - (producer - q->cons_tail);
112 
113 	if (free_entries >= dcnt)
114 		return free_entries;
115 
116 	/* Refresh the local tail pointer */
117 	q->cons_tail = READ_ONCE(q->ring->consumer);
118 	return q->nentries - (producer - q->cons_tail);
119 }
120 
121 static inline bool xskq_has_addrs(struct xsk_queue *q, u32 cnt)
122 {
123 	u32 entries = q->prod_tail - q->cons_tail;
124 
125 	if (entries >= cnt)
126 		return true;
127 
128 	/* Refresh the local pointer. */
129 	q->prod_tail = READ_ONCE(q->ring->producer);
130 	entries = q->prod_tail - q->cons_tail;
131 
132 	return entries >= cnt;
133 }
134 
135 /* UMEM queue */
136 
137 static inline bool xskq_crosses_non_contig_pg(struct xdp_umem *umem, u64 addr,
138 					      u64 length)
139 {
140 	bool cross_pg = (addr & (PAGE_SIZE - 1)) + length > PAGE_SIZE;
141 	bool next_pg_contig =
142 		(unsigned long)umem->pages[(addr >> PAGE_SHIFT)].addr &
143 			XSK_NEXT_PG_CONTIG_MASK;
144 
145 	return cross_pg && !next_pg_contig;
146 }
147 
148 static inline bool xskq_is_valid_addr(struct xsk_queue *q, u64 addr)
149 {
150 	if (addr >= q->size) {
151 		q->invalid_descs++;
152 		return false;
153 	}
154 
155 	return true;
156 }
157 
158 static inline bool xskq_is_valid_addr_unaligned(struct xsk_queue *q, u64 addr,
159 						u64 length,
160 						struct xdp_umem *umem)
161 {
162 	u64 base_addr = xsk_umem_extract_addr(addr);
163 
164 	addr = xsk_umem_add_offset_to_addr(addr);
165 	if (base_addr >= q->size || addr >= q->size ||
166 	    xskq_crosses_non_contig_pg(umem, addr, length)) {
167 		q->invalid_descs++;
168 		return false;
169 	}
170 
171 	return true;
172 }
173 
174 static inline u64 *xskq_validate_addr(struct xsk_queue *q, u64 *addr,
175 				      struct xdp_umem *umem)
176 {
177 	while (q->cons_tail != q->cons_head) {
178 		struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
179 		unsigned int idx = q->cons_tail & q->ring_mask;
180 
181 		*addr = READ_ONCE(ring->desc[idx]) & q->chunk_mask;
182 
183 		if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG) {
184 			if (xskq_is_valid_addr_unaligned(q, *addr,
185 							 umem->chunk_size_nohr,
186 							 umem))
187 				return addr;
188 			goto out;
189 		}
190 
191 		if (xskq_is_valid_addr(q, *addr))
192 			return addr;
193 
194 out:
195 		q->cons_tail++;
196 	}
197 
198 	return NULL;
199 }
200 
201 static inline u64 *xskq_peek_addr(struct xsk_queue *q, u64 *addr,
202 				  struct xdp_umem *umem)
203 {
204 	if (q->cons_tail == q->cons_head) {
205 		smp_mb(); /* D, matches A */
206 		WRITE_ONCE(q->ring->consumer, q->cons_tail);
207 		q->cons_head = q->cons_tail + xskq_nb_avail(q, RX_BATCH_SIZE);
208 
209 		/* Order consumer and data */
210 		smp_rmb();
211 	}
212 
213 	return xskq_validate_addr(q, addr, umem);
214 }
215 
216 static inline void xskq_discard_addr(struct xsk_queue *q)
217 {
218 	q->cons_tail++;
219 }
220 
221 static inline int xskq_produce_addr(struct xsk_queue *q, u64 addr)
222 {
223 	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
224 
225 	if (xskq_nb_free(q, q->prod_tail, 1) == 0)
226 		return -ENOSPC;
227 
228 	/* A, matches D */
229 	ring->desc[q->prod_tail++ & q->ring_mask] = addr;
230 
231 	/* Order producer and data */
232 	smp_wmb(); /* B, matches C */
233 
234 	WRITE_ONCE(q->ring->producer, q->prod_tail);
235 	return 0;
236 }
237 
238 static inline int xskq_produce_addr_lazy(struct xsk_queue *q, u64 addr)
239 {
240 	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
241 
242 	if (xskq_nb_free(q, q->prod_head, LAZY_UPDATE_THRESHOLD) == 0)
243 		return -ENOSPC;
244 
245 	/* A, matches D */
246 	ring->desc[q->prod_head++ & q->ring_mask] = addr;
247 	return 0;
248 }
249 
250 static inline void xskq_produce_flush_addr_n(struct xsk_queue *q,
251 					     u32 nb_entries)
252 {
253 	/* Order producer and data */
254 	smp_wmb(); /* B, matches C */
255 
256 	q->prod_tail += nb_entries;
257 	WRITE_ONCE(q->ring->producer, q->prod_tail);
258 }
259 
260 static inline int xskq_reserve_addr(struct xsk_queue *q)
261 {
262 	if (xskq_nb_free(q, q->prod_head, 1) == 0)
263 		return -ENOSPC;
264 
265 	/* A, matches D */
266 	q->prod_head++;
267 	return 0;
268 }
269 
270 /* Rx/Tx queue */
271 
272 static inline bool xskq_is_valid_desc(struct xsk_queue *q, struct xdp_desc *d,
273 				      struct xdp_umem *umem)
274 {
275 	if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG) {
276 		if (!xskq_is_valid_addr_unaligned(q, d->addr, d->len, umem))
277 			return false;
278 
279 		if (d->len > umem->chunk_size_nohr || d->options) {
280 			q->invalid_descs++;
281 			return false;
282 		}
283 
284 		return true;
285 	}
286 
287 	if (!xskq_is_valid_addr(q, d->addr))
288 		return false;
289 
290 	if (((d->addr + d->len) & q->chunk_mask) != (d->addr & q->chunk_mask) ||
291 	    d->options) {
292 		q->invalid_descs++;
293 		return false;
294 	}
295 
296 	return true;
297 }
298 
299 static inline struct xdp_desc *xskq_validate_desc(struct xsk_queue *q,
300 						  struct xdp_desc *desc,
301 						  struct xdp_umem *umem)
302 {
303 	while (q->cons_tail != q->cons_head) {
304 		struct xdp_rxtx_ring *ring = (struct xdp_rxtx_ring *)q->ring;
305 		unsigned int idx = q->cons_tail & q->ring_mask;
306 
307 		*desc = READ_ONCE(ring->desc[idx]);
308 		if (xskq_is_valid_desc(q, desc, umem))
309 			return desc;
310 
311 		q->cons_tail++;
312 	}
313 
314 	return NULL;
315 }
316 
317 static inline struct xdp_desc *xskq_peek_desc(struct xsk_queue *q,
318 					      struct xdp_desc *desc,
319 					      struct xdp_umem *umem)
320 {
321 	if (q->cons_tail == q->cons_head) {
322 		smp_mb(); /* D, matches A */
323 		WRITE_ONCE(q->ring->consumer, q->cons_tail);
324 		q->cons_head = q->cons_tail + xskq_nb_avail(q, RX_BATCH_SIZE);
325 
326 		/* Order consumer and data */
327 		smp_rmb(); /* C, matches B */
328 	}
329 
330 	return xskq_validate_desc(q, desc, umem);
331 }
332 
333 static inline void xskq_discard_desc(struct xsk_queue *q)
334 {
335 	q->cons_tail++;
336 }
337 
338 static inline int xskq_produce_batch_desc(struct xsk_queue *q,
339 					  u64 addr, u32 len)
340 {
341 	struct xdp_rxtx_ring *ring = (struct xdp_rxtx_ring *)q->ring;
342 	unsigned int idx;
343 
344 	if (xskq_nb_free(q, q->prod_head, 1) == 0)
345 		return -ENOSPC;
346 
347 	/* A, matches D */
348 	idx = (q->prod_head++) & q->ring_mask;
349 	ring->desc[idx].addr = addr;
350 	ring->desc[idx].len = len;
351 
352 	return 0;
353 }
354 
355 static inline void xskq_produce_flush_desc(struct xsk_queue *q)
356 {
357 	/* Order producer and data */
358 	smp_wmb(); /* B, matches C */
359 
360 	q->prod_tail = q->prod_head;
361 	WRITE_ONCE(q->ring->producer, q->prod_tail);
362 }
363 
364 static inline bool xskq_full_desc(struct xsk_queue *q)
365 {
366 	return xskq_nb_avail(q, q->nentries) == q->nentries;
367 }
368 
369 static inline bool xskq_empty_desc(struct xsk_queue *q)
370 {
371 	return xskq_nb_free(q, q->prod_tail, q->nentries) == q->nentries;
372 }
373 
374 void xskq_set_umem(struct xsk_queue *q, u64 size, u64 chunk_mask);
375 struct xsk_queue *xskq_create(u32 nentries, bool umem_queue);
376 void xskq_destroy(struct xsk_queue *q_ops);
377 
378 /* Executed by the core when the entire UMEM gets freed */
379 void xsk_reuseq_destroy(struct xdp_umem *umem);
380 
381 #endif /* _LINUX_XSK_QUEUE_H */
382