1/*-
2 * Copyright (c) 2011, Bryan Venteicher <bryanv@FreeBSD.org>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice unmodified, this list of conditions, and the following
10 *    disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 *    notice, this list of conditions and the following disclaimer in the
13 *    documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25 *
26 * $FreeBSD$
27 */
28
29#ifndef _IF_VTNETVAR_H
30#define _IF_VTNETVAR_H
31
32struct vtnet_softc;
33
34struct vtnet_statistics {
35	uint64_t	mbuf_alloc_failed;
36
37	uint64_t	rx_frame_too_large;
38	uint64_t	rx_enq_replacement_failed;
39	uint64_t	rx_mergeable_failed;
40	uint64_t	rx_csum_bad_ethtype;
41	uint64_t	rx_csum_bad_ipproto;
42	uint64_t	rx_csum_bad_offset;
43	uint64_t	rx_csum_bad_proto;
44	uint64_t	tx_csum_bad_ethtype;
45	uint64_t	tx_tso_bad_ethtype;
46	uint64_t	tx_tso_not_tcp;
47	uint64_t	tx_defragged;
48	uint64_t	tx_defrag_failed;
49
50	/*
51	 * These are accumulated from each Rx/Tx queue.
52	 */
53	uint64_t	rx_csum_failed;
54	uint64_t	rx_csum_offloaded;
55	uint64_t	rx_task_rescheduled;
56	uint64_t	tx_csum_offloaded;
57	uint64_t	tx_tso_offloaded;
58	uint64_t	tx_task_rescheduled;
59};
60
61struct vtnet_rxq_stats {
62	uint64_t	vrxs_ipackets;	/* if_ipackets */
63	uint64_t	vrxs_ibytes;	/* if_ibytes */
64	uint64_t	vrxs_iqdrops;	/* if_iqdrops */
65	uint64_t	vrxs_ierrors;	/* if_ierrors */
66	uint64_t	vrxs_csum;
67	uint64_t	vrxs_csum_failed;
68	uint64_t	vrxs_rescheduled;
69};
70
71struct vtnet_rxq {
72	struct mtx		 vtnrx_mtx;
73	struct vtnet_softc	*vtnrx_sc;
74	struct virtqueue	*vtnrx_vq;
75	struct sglist		*vtnrx_sg;
76	int			 vtnrx_id;
77	struct vtnet_rxq_stats	 vtnrx_stats;
78	struct taskqueue	*vtnrx_tq;
79	struct task		 vtnrx_intrtask;
80	char			 vtnrx_name[16];
81} __aligned(CACHE_LINE_SIZE);
82
83#define VTNET_RXQ_LOCK(_rxq)	mtx_lock(&(_rxq)->vtnrx_mtx)
84#define VTNET_RXQ_UNLOCK(_rxq)	mtx_unlock(&(_rxq)->vtnrx_mtx)
85#define VTNET_RXQ_LOCK_ASSERT(_rxq)		\
86    mtx_assert(&(_rxq)->vtnrx_mtx, MA_OWNED)
87#define VTNET_RXQ_LOCK_ASSERT_NOTOWNED(_rxq)	\
88    mtx_assert(&(_rxq)->vtnrx_mtx, MA_NOTOWNED)
89
90struct vtnet_txq_stats {
91	uint64_t vtxs_opackets;	/* if_opackets */
92	uint64_t vtxs_obytes;	/* if_obytes */
93	uint64_t vtxs_omcasts;	/* if_omcasts */
94	uint64_t vtxs_csum;
95	uint64_t vtxs_tso;
96	uint64_t vtxs_rescheduled;
97};
98
99struct vtnet_txq {
100	struct mtx		 vtntx_mtx;
101	struct vtnet_softc	*vtntx_sc;
102	struct virtqueue	*vtntx_vq;
103	struct sglist		*vtntx_sg;
104#ifndef VTNET_LEGACY_TX
105	struct buf_ring		*vtntx_br;
106#endif
107	int			 vtntx_id;
108	int			 vtntx_watchdog;
109	struct vtnet_txq_stats	 vtntx_stats;
110	struct taskqueue	*vtntx_tq;
111	struct task		 vtntx_intrtask;
112#ifndef VTNET_LEGACY_TX
113	struct task		 vtntx_defrtask;
114#endif
115	char			 vtntx_name[16];
116} __aligned(CACHE_LINE_SIZE);
117
118#define VTNET_TXQ_LOCK(_txq)	mtx_lock(&(_txq)->vtntx_mtx)
119#define VTNET_TXQ_TRYLOCK(_txq)	mtx_trylock(&(_txq)->vtntx_mtx)
120#define VTNET_TXQ_UNLOCK(_txq)	mtx_unlock(&(_txq)->vtntx_mtx)
121#define VTNET_TXQ_LOCK_ASSERT(_txq)		\
122    mtx_assert(&(_txq)->vtntx_mtx, MA_OWNED)
123#define VTNET_TXQ_LOCK_ASSERT_NOTOWNED(_txq)	\
124    mtx_assert(&(_txq)->vtntx_mtx, MA_NOTOWNED)
125
126struct vtnet_softc {
127	device_t		 vtnet_dev;
128	struct ifnet		*vtnet_ifp;
129	struct vtnet_rxq	*vtnet_rxqs;
130	struct vtnet_txq	*vtnet_txqs;
131
132	uint32_t		 vtnet_flags;
133#define VTNET_FLAG_SUSPENDED	 0x0001
134#define VTNET_FLAG_MAC		 0x0002
135#define VTNET_FLAG_CTRL_VQ	 0x0004
136#define VTNET_FLAG_CTRL_RX	 0x0008
137#define VTNET_FLAG_CTRL_MAC	 0x0010
138#define VTNET_FLAG_VLAN_FILTER	 0x0020
139#define VTNET_FLAG_TSO_ECN	 0x0040
140#define VTNET_FLAG_MRG_RXBUFS	 0x0080
141#define VTNET_FLAG_LRO_NOMRG	 0x0100
142#define VTNET_FLAG_MULTIQ	 0x0200
143#define VTNET_FLAG_INDIRECT	 0x0400
144#define VTNET_FLAG_EVENT_IDX	 0x0800
145
146	int			 vtnet_link_active;
147	int			 vtnet_hdr_size;
148	int			 vtnet_rx_process_limit;
149	int			 vtnet_rx_nsegs;
150	int			 vtnet_rx_nmbufs;
151	int			 vtnet_rx_clsize;
152	int			 vtnet_rx_new_clsize;
153	int			 vtnet_tx_intr_thresh;
154	int			 vtnet_tx_nsegs;
155	int			 vtnet_if_flags;
156	int			 vtnet_act_vq_pairs;
157	int			 vtnet_max_vq_pairs;
158	int			 vtnet_requested_vq_pairs;
159
160	struct virtqueue	*vtnet_ctrl_vq;
161	struct vtnet_mac_filter	*vtnet_mac_filter;
162	uint32_t		*vtnet_vlan_filter;
163
164	uint64_t		 vtnet_features;
165	struct vtnet_statistics	 vtnet_stats;
166	struct callout		 vtnet_tick_ch;
167	struct ifmedia		 vtnet_media;
168	eventhandler_tag	 vtnet_vlan_attach;
169	eventhandler_tag	 vtnet_vlan_detach;
170
171	struct mtx		 vtnet_mtx;
172	char			 vtnet_mtx_name[16];
173	char			 vtnet_hwaddr[ETHER_ADDR_LEN];
174};
175
176/*
177 * Maximum number of queue pairs we will autoconfigure to.
178 */
179#define VTNET_MAX_QUEUE_PAIRS	8
180
181/*
182 * Additional completed entries can appear in a virtqueue before we can
183 * reenable interrupts. Number of times to retry before scheduling the
184 * taskqueue to process the completed entries.
185 */
186#define VTNET_INTR_DISABLE_RETRIES	4
187
188/*
189 * Similarly, additional completed entries can appear in a virtqueue
190 * between when lasted checked and before notifying the host. Number
191 * of times to retry before scheduling the taskqueue to process the
192 * queue.
193 */
194#define VTNET_NOTIFY_RETRIES		4
195
196/*
197 * Fake the media type. The host does not provide us with any real media
198 * information.
199 */
200#define VTNET_MEDIATYPE		 (IFM_ETHER | IFM_10G_T | IFM_FDX)
201
202/*
203 * Number of words to allocate for the VLAN shadow table. There is one
204 * bit for each VLAN.
205 */
206#define VTNET_VLAN_FILTER_NWORDS	(4096 / 32)
207
208/*
209 * When mergeable buffers are not negotiated, the vtnet_rx_header structure
210 * below is placed at the beginning of the mbuf data. Use 4 bytes of pad to
211 * both keep the VirtIO header and the data non-contiguous and to keep the
212 * frame's payload 4 byte aligned.
213 *
214 * When mergeable buffers are negotiated, the host puts the VirtIO header in
215 * the beginning of the first mbuf's data.
216 */
217#define VTNET_RX_HEADER_PAD	4
218struct vtnet_rx_header {
219	struct virtio_net_hdr	vrh_hdr;
220	char			vrh_pad[VTNET_RX_HEADER_PAD];
221} __packed;
222
223/*
224 * For each outgoing frame, the vtnet_tx_header below is allocated from
225 * the vtnet_tx_header_zone.
226 */
227struct vtnet_tx_header {
228	union {
229		struct virtio_net_hdr		hdr;
230		struct virtio_net_hdr_mrg_rxbuf	mhdr;
231	} vth_uhdr;
232
233	struct mbuf *vth_mbuf;
234};
235
236/*
237 * The VirtIO specification does not place a limit on the number of MAC
238 * addresses the guest driver may request to be filtered. In practice,
239 * the host is constrained by available resources. To simplify this driver,
240 * impose a reasonably high limit of MAC addresses we will filter before
241 * falling back to promiscuous or all-multicast modes.
242 */
243#define VTNET_MAX_MAC_ENTRIES	128
244
245struct vtnet_mac_table {
246	uint32_t	nentries;
247	uint8_t		macs[VTNET_MAX_MAC_ENTRIES][ETHER_ADDR_LEN];
248} __packed;
249
250struct vtnet_mac_filter {
251	struct vtnet_mac_table	vmf_unicast;
252	uint32_t		vmf_pad; /* Make tables non-contiguous. */
253	struct vtnet_mac_table	vmf_multicast;
254};
255
256/*
257 * The MAC filter table is malloc(9)'d when needed. Ensure it will
258 * always fit in one segment.
259 */
260CTASSERT(sizeof(struct vtnet_mac_filter) <= PAGE_SIZE);
261
262#define VTNET_TX_TIMEOUT	5
263#define VTNET_CSUM_OFFLOAD	(CSUM_TCP | CSUM_UDP | CSUM_SCTP)
264#define VTNET_CSUM_OFFLOAD_IPV6	(CSUM_TCP_IPV6 | CSUM_UDP_IPV6 | CSUM_SCTP_IPV6)
265
266#define VTNET_CSUM_ALL_OFFLOAD	\
267    (VTNET_CSUM_OFFLOAD | VTNET_CSUM_OFFLOAD_IPV6 | CSUM_TSO)
268
269/* Features desired/implemented by this driver. */
270#define VTNET_FEATURES \
271    (VIRTIO_NET_F_MAC			| \
272     VIRTIO_NET_F_STATUS		| \
273     VIRTIO_NET_F_CTRL_VQ		| \
274     VIRTIO_NET_F_CTRL_RX		| \
275     VIRTIO_NET_F_CTRL_MAC_ADDR		| \
276     VIRTIO_NET_F_CTRL_VLAN		| \
277     VIRTIO_NET_F_CSUM			| \
278     VIRTIO_NET_F_GSO			| \
279     VIRTIO_NET_F_HOST_TSO4		| \
280     VIRTIO_NET_F_HOST_TSO6		| \
281     VIRTIO_NET_F_HOST_ECN		| \
282     VIRTIO_NET_F_GUEST_CSUM		| \
283     VIRTIO_NET_F_GUEST_TSO4		| \
284     VIRTIO_NET_F_GUEST_TSO6		| \
285     VIRTIO_NET_F_GUEST_ECN		| \
286     VIRTIO_NET_F_MRG_RXBUF		| \
287     VIRTIO_NET_F_MQ			| \
288     VIRTIO_RING_F_EVENT_IDX		| \
289     VIRTIO_RING_F_INDIRECT_DESC)
290
291/*
292 * The VIRTIO_NET_F_HOST_TSO[46] features permit us to send the host
293 * frames larger than 1514 bytes.
294 */
295#define VTNET_TSO_FEATURES (VIRTIO_NET_F_GSO | VIRTIO_NET_F_HOST_TSO4 | \
296    VIRTIO_NET_F_HOST_TSO6 | VIRTIO_NET_F_HOST_ECN)
297
298/*
299 * The VIRTIO_NET_F_GUEST_TSO[46] features permit the host to send us
300 * frames larger than 1514 bytes. We do not yet support software LRO
301 * via tcp_lro_rx().
302 */
303#define VTNET_LRO_FEATURES (VIRTIO_NET_F_GUEST_TSO4 | \
304    VIRTIO_NET_F_GUEST_TSO6 | VIRTIO_NET_F_GUEST_ECN)
305
306#define VTNET_MAX_MTU		65536
307#define VTNET_MAX_RX_SIZE	65550
308
309/*
310 * Used to preallocate the Vq indirect descriptors. The first segment
311 * is reserved for the header, except for mergeable buffers since the
312 * header is placed inline with the data.
313 */
314#define VTNET_MRG_RX_SEGS	1
315#define VTNET_MIN_RX_SEGS	2
316#define VTNET_MAX_RX_SEGS	34
317#define VTNET_MIN_TX_SEGS	4
318#define VTNET_MAX_TX_SEGS	64
319
320/*
321 * Assert we can receive and transmit the maximum with regular
322 * size clusters.
323 */
324CTASSERT(((VTNET_MAX_RX_SEGS - 1) * MCLBYTES) >= VTNET_MAX_RX_SIZE);
325CTASSERT(((VTNET_MAX_TX_SEGS - 1) * MCLBYTES) >= VTNET_MAX_MTU);
326
327/*
328 * Number of slots in the Tx bufrings. This value matches most other
329 * multiqueue drivers.
330 */
331#define VTNET_DEFAULT_BUFRING_SIZE	4096
332
333/*
334 * Determine how many mbufs are in each receive buffer. For LRO without
335 * mergeable buffers, we must allocate an mbuf chain large enough to
336 * hold both the vtnet_rx_header and the maximum receivable data.
337 */
338#define VTNET_NEEDED_RX_MBUFS(_sc, _clsize)				\
339	((_sc)->vtnet_flags & VTNET_FLAG_LRO_NOMRG) == 0 ? 1 :		\
340	    howmany(sizeof(struct vtnet_rx_header) + VTNET_MAX_RX_SIZE,	\
341	        (_clsize))
342
343#define VTNET_CORE_MTX(_sc)		&(_sc)->vtnet_mtx
344#define VTNET_CORE_LOCK(_sc)		mtx_lock(VTNET_CORE_MTX((_sc)))
345#define VTNET_CORE_UNLOCK(_sc)		mtx_unlock(VTNET_CORE_MTX((_sc)))
346#define VTNET_CORE_LOCK_DESTROY(_sc)	mtx_destroy(VTNET_CORE_MTX((_sc)))
347#define VTNET_CORE_LOCK_ASSERT(_sc)		\
348    mtx_assert(VTNET_CORE_MTX((_sc)), MA_OWNED)
349#define VTNET_CORE_LOCK_ASSERT_NOTOWNED(_sc)	\
350    mtx_assert(VTNET_CORE_MTX((_sc)), MA_NOTOWNED)
351
352#define VTNET_CORE_LOCK_INIT(_sc) do {					\
353    snprintf((_sc)->vtnet_mtx_name, sizeof((_sc)->vtnet_mtx_name),	\
354        "%s", device_get_nameunit((_sc)->vtnet_dev));			\
355    mtx_init(VTNET_CORE_MTX((_sc)), (_sc)->vtnet_mtx_name,		\
356        "VTNET Core Lock", MTX_DEF);					\
357} while (0)
358
359#endif /* _IF_VTNETVAR_H */
360