xref: /freebsd/sys/dev/virtio/network/if_vtnetvar.h (revision 5dd76dd0cc19450133aa379ce0ce4a68ae07fb39)
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 
32 struct vtnet_softc;
33 
34 struct 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 
48 	/*
49 	 * These are accumulated from each Rx/Tx queue.
50 	 */
51 	uint64_t	rx_csum_failed;
52 	uint64_t	rx_csum_offloaded;
53 	uint64_t	rx_task_rescheduled;
54 	uint64_t	tx_csum_offloaded;
55 	uint64_t	tx_tso_offloaded;
56 	uint64_t	tx_task_rescheduled;
57 };
58 
59 struct vtnet_rxq_stats {
60 	uint64_t	vrxs_ipackets;	/* if_ipackets */
61 	uint64_t	vrxs_ibytes;	/* if_ibytes */
62 	uint64_t	vrxs_iqdrops;	/* if_iqdrops */
63 	uint64_t	vrxs_ierrors;	/* if_ierrors */
64 	uint64_t	vrxs_csum;
65 	uint64_t	vrxs_csum_failed;
66 	uint64_t	vrxs_rescheduled;
67 };
68 
69 struct vtnet_rxq {
70 	struct mtx		 vtnrx_mtx;
71 	struct vtnet_softc	*vtnrx_sc;
72 	struct virtqueue	*vtnrx_vq;
73 	int			 vtnrx_id;
74 	int			 vtnrx_process_limit;
75 	struct vtnet_rxq_stats	 vtnrx_stats;
76 	struct taskqueue	*vtnrx_tq;
77 	struct task		 vtnrx_intrtask;
78 	char			 vtnrx_name[16];
79 } __aligned(CACHE_LINE_SIZE);
80 
81 #define VTNET_RXQ_LOCK(_rxq)	mtx_lock(&(_rxq)->vtnrx_mtx)
82 #define VTNET_RXQ_UNLOCK(_rxq)	mtx_unlock(&(_rxq)->vtnrx_mtx)
83 #define VTNET_RXQ_LOCK_ASSERT(_rxq)		\
84     mtx_assert(&(_rxq)->vtnrx_mtx, MA_OWNED)
85 #define VTNET_RXQ_LOCK_ASSERT_NOTOWNED(_rxq)	\
86     mtx_assert(&(_rxq)->vtnrx_mtx, MA_NOTOWNED)
87 
88 struct vtnet_txq_stats {
89 	uint64_t vtxs_opackets;	/* if_opackets */
90 	uint64_t vtxs_obytes;	/* if_obytes */
91 	uint64_t vtxs_omcasts;	/* if_omcasts */
92 	uint64_t vtxs_csum;
93 	uint64_t vtxs_tso;
94 	uint64_t vtxs_collapsed;
95 	uint64_t vtxs_rescheduled;
96 };
97 
98 struct vtnet_txq {
99 	struct mtx		 vtntx_mtx;
100 	struct vtnet_softc	*vtntx_sc;
101 	struct virtqueue	*vtntx_vq;
102 #ifndef VTNET_LEGACY_TX
103 	struct buf_ring		*vtntx_br;
104 #endif
105 	int			 vtntx_id;
106 	int			 vtntx_watchdog;
107 	struct vtnet_txq_stats	 vtntx_stats;
108 	struct taskqueue	*vtntx_tq;
109 	struct task		 vtntx_intrtask;
110 #ifndef VTNET_LEGACY_TX
111 	struct task		 vtntx_defrtask;
112 #endif
113 	char			 vtntx_name[16];
114 } __aligned(CACHE_LINE_SIZE);
115 
116 #define VTNET_TXQ_LOCK(_txq)	mtx_lock(&(_txq)->vtntx_mtx)
117 #define VTNET_TXQ_TRYLOCK(_txq)	mtx_trylock(&(_txq)->vtntx_mtx)
118 #define VTNET_TXQ_UNLOCK(_txq)	mtx_unlock(&(_txq)->vtntx_mtx)
119 #define VTNET_TXQ_LOCK_ASSERT(_txq)		\
120     mtx_assert(&(_txq)->vtntx_mtx, MA_OWNED)
121 #define VTNET_TXQ_LOCK_ASSERT_NOTOWNED(_txq)	\
122     mtx_assert(&(_txq)->vtntx_mtx, MA_NOTOWNED)
123 
124 struct vtnet_softc {
125 	device_t		 vtnet_dev;
126 	struct ifnet		*vtnet_ifp;
127 	struct vtnet_rxq	*vtnet_rxqs;
128 	struct vtnet_txq	*vtnet_txqs;
129 
130 	uint32_t		 vtnet_flags;
131 #define VTNET_FLAG_SUSPENDED	 0x0001
132 #define VTNET_FLAG_MAC		 0x0002
133 #define VTNET_FLAG_CTRL_VQ	 0x0004
134 #define VTNET_FLAG_CTRL_RX	 0x0008
135 #define VTNET_FLAG_CTRL_MAC	 0x0010
136 #define VTNET_FLAG_VLAN_FILTER	 0x0020
137 #define VTNET_FLAG_TSO_ECN	 0x0040
138 #define VTNET_FLAG_MRG_RXBUFS	 0x0080
139 #define VTNET_FLAG_LRO_NOMRG	 0x0100
140 #define VTNET_FLAG_MULTIQ	 0x0200
141 #define VTNET_FLAG_EVENT_IDX	 0x0400
142 
143 	int			 vtnet_link_active;
144 	int			 vtnet_hdr_size;
145 	int			 vtnet_rx_process_limit;
146 	int			 vtnet_rx_nmbufs;
147 	int			 vtnet_rx_clsize;
148 	int			 vtnet_rx_new_clsize;
149 	int			 vtnet_if_flags;
150 	int			 vtnet_act_vq_pairs;
151 	int			 vtnet_max_vq_pairs;
152 
153 	struct virtqueue	*vtnet_ctrl_vq;
154 	struct vtnet_mac_filter	*vtnet_mac_filter;
155 	uint32_t		*vtnet_vlan_filter;
156 
157 	uint64_t		 vtnet_features;
158 	struct vtnet_statistics	 vtnet_stats;
159 	struct callout		 vtnet_tick_ch;
160 	struct ifmedia		 vtnet_media;
161 	eventhandler_tag	 vtnet_vlan_attach;
162 	eventhandler_tag	 vtnet_vlan_detach;
163 
164 	struct mtx		 vtnet_mtx;
165 	char			 vtnet_mtx_name[16];
166 	char			 vtnet_hwaddr[ETHER_ADDR_LEN];
167 };
168 
169 /*
170  * Maximum number of queue pairs we will autoconfigure to.
171  */
172 #define VTNET_MAX_QUEUE_PAIRS	8
173 
174 /*
175  * Additional completed entries can appear in a virtqueue before we can
176  * reenable interrupts. Number of times to retry before scheduling the
177  * taskqueue to process the completed entries.
178  */
179 #define VTNET_INTR_DISABLE_RETRIES	4
180 
181 /*
182  * Fake the media type. The host does not provide us with any real media
183  * information.
184  */
185 #define VTNET_MEDIATYPE		 (IFM_ETHER | IFM_10G_T | IFM_FDX)
186 
187 /*
188  * Number of words to allocate for the VLAN shadow table. There is one
189  * bit for each VLAN.
190  */
191 #define VTNET_VLAN_FILTER_NWORDS	(4096 / 32)
192 
193 /*
194  * When mergeable buffers are not negotiated, the vtnet_rx_header structure
195  * below is placed at the beginning of the mbuf data. Use 4 bytes of pad to
196  * both keep the VirtIO header and the data non-contiguous and to keep the
197  * frame's payload 4 byte aligned.
198  *
199  * When mergeable buffers are negotiated, the host puts the VirtIO header in
200  * the beginning of the first mbuf's data.
201  */
202 #define VTNET_RX_HEADER_PAD	4
203 struct vtnet_rx_header {
204 	struct virtio_net_hdr	vrh_hdr;
205 	char			vrh_pad[VTNET_RX_HEADER_PAD];
206 } __packed;
207 
208 /*
209  * For each outgoing frame, the vtnet_tx_header below is allocated from
210  * the vtnet_tx_header_zone.
211  */
212 struct vtnet_tx_header {
213 	union {
214 		struct virtio_net_hdr		hdr;
215 		struct virtio_net_hdr_mrg_rxbuf	mhdr;
216 	} vth_uhdr;
217 
218 	struct mbuf *vth_mbuf;
219 };
220 
221 /*
222  * The VirtIO specification does not place a limit on the number of MAC
223  * addresses the guest driver may request to be filtered. In practice,
224  * the host is constrained by available resources. To simplify this driver,
225  * impose a reasonably high limit of MAC addresses we will filter before
226  * falling back to promiscuous or all-multicast modes.
227  */
228 #define VTNET_MAX_MAC_ENTRIES	128
229 
230 struct vtnet_mac_table {
231 	uint32_t	nentries;
232 	uint8_t		macs[VTNET_MAX_MAC_ENTRIES][ETHER_ADDR_LEN];
233 } __packed;
234 
235 struct vtnet_mac_filter {
236 	struct vtnet_mac_table	vmf_unicast;
237 	uint32_t		vmf_pad; /* Make tables non-contiguous. */
238 	struct vtnet_mac_table	vmf_multicast;
239 };
240 
241 /*
242  * The MAC filter table is malloc(9)'d when needed. Ensure it will
243  * always fit in one segment.
244  */
245 CTASSERT(sizeof(struct vtnet_mac_filter) <= PAGE_SIZE);
246 
247 #define VTNET_TX_TIMEOUT	5
248 #define VTNET_CSUM_OFFLOAD	(CSUM_TCP | CSUM_UDP | CSUM_SCTP)
249 #define VTNET_CSUM_OFFLOAD_IPV6	(CSUM_TCP_IPV6 | CSUM_UDP_IPV6 | CSUM_SCTP_IPV6)
250 
251 #define VTNET_CSUM_ALL_OFFLOAD	\
252     (VTNET_CSUM_OFFLOAD | VTNET_CSUM_OFFLOAD_IPV6 | CSUM_TSO)
253 
254 /* Features desired/implemented by this driver. */
255 #define VTNET_FEATURES \
256     (VIRTIO_NET_F_MAC			| \
257      VIRTIO_NET_F_STATUS		| \
258      VIRTIO_NET_F_CTRL_VQ		| \
259      VIRTIO_NET_F_CTRL_RX		| \
260      VIRTIO_NET_F_CTRL_MAC_ADDR		| \
261      VIRTIO_NET_F_CTRL_VLAN		| \
262      VIRTIO_NET_F_CSUM			| \
263      VIRTIO_NET_F_GSO			| \
264      VIRTIO_NET_F_HOST_TSO4		| \
265      VIRTIO_NET_F_HOST_TSO6		| \
266      VIRTIO_NET_F_HOST_ECN		| \
267      VIRTIO_NET_F_GUEST_CSUM		| \
268      VIRTIO_NET_F_GUEST_TSO4		| \
269      VIRTIO_NET_F_GUEST_TSO6		| \
270      VIRTIO_NET_F_GUEST_ECN		| \
271      VIRTIO_NET_F_MRG_RXBUF		| \
272      VIRTIO_NET_F_MQ			| \
273      VIRTIO_RING_F_EVENT_IDX		| \
274      VIRTIO_RING_F_INDIRECT_DESC)
275 
276 /*
277  * The VIRTIO_NET_F_HOST_TSO[46] features permit us to send the host
278  * frames larger than 1514 bytes.
279  */
280 #define VTNET_TSO_FEATURES (VIRTIO_NET_F_GSO | VIRTIO_NET_F_HOST_TSO4 | \
281     VIRTIO_NET_F_HOST_TSO6 | VIRTIO_NET_F_HOST_ECN)
282 
283 /*
284  * The VIRTIO_NET_F_GUEST_TSO[46] features permit the host to send us
285  * frames larger than 1514 bytes. We do not yet support software LRO
286  * via tcp_lro_rx().
287  */
288 #define VTNET_LRO_FEATURES (VIRTIO_NET_F_GUEST_TSO4 | \
289     VIRTIO_NET_F_GUEST_TSO6 | VIRTIO_NET_F_GUEST_ECN)
290 
291 #define VTNET_MAX_MTU		65536
292 #define VTNET_MAX_RX_SIZE	65550
293 
294 /*
295  * Used to preallocate the Vq indirect descriptors. The first segment
296  * is reserved for the header.
297  */
298 #define VTNET_MIN_RX_SEGS	2
299 #define VTNET_MAX_RX_SEGS	34
300 #define VTNET_MAX_TX_SEGS	34
301 
302 /*
303  * Assert we can receive and transmit the maximum with regular
304  * size clusters.
305  */
306 CTASSERT(((VTNET_MAX_RX_SEGS - 1) * MCLBYTES) >= VTNET_MAX_RX_SIZE);
307 CTASSERT(((VTNET_MAX_TX_SEGS - 1) * MCLBYTES) >= VTNET_MAX_MTU);
308 
309 /*
310  * Number of slots in the Tx bufrings. This value matches most other
311  * multiqueue drivers.
312  */
313 #define VTNET_DEFAULT_BUFRING_SIZE	4096
314 
315 /*
316  * Determine how many mbufs are in each receive buffer. For LRO without
317  * mergeable descriptors, we must allocate an mbuf chain large enough to
318  * hold both the vtnet_rx_header and the maximum receivable data.
319  */
320 #define VTNET_NEEDED_RX_MBUFS(_sc, _clsize)				\
321 	((_sc)->vtnet_flags & VTNET_FLAG_LRO_NOMRG) == 0 ? 1 :		\
322 	    howmany(sizeof(struct vtnet_rx_header) + VTNET_MAX_RX_SIZE,	\
323 	        (_clsize))
324 
325 #define VTNET_CORE_MTX(_sc)		&(_sc)->vtnet_mtx
326 #define VTNET_CORE_LOCK(_sc)		mtx_lock(VTNET_CORE_MTX((_sc)))
327 #define VTNET_CORE_UNLOCK(_sc)		mtx_unlock(VTNET_CORE_MTX((_sc)))
328 #define VTNET_CORE_LOCK_DESTROY(_sc)	mtx_destroy(VTNET_CORE_MTX((_sc)))
329 #define VTNET_CORE_LOCK_ASSERT(_sc)		\
330     mtx_assert(VTNET_CORE_MTX((_sc)), MA_OWNED)
331 #define VTNET_CORE_LOCK_ASSERT_NOTOWNED(_sc)	\
332     mtx_assert(VTNET_CORE_MTX((_sc)), MA_NOTOWNED)
333 
334 #define VTNET_CORE_LOCK_INIT(_sc) do {					\
335     snprintf((_sc)->vtnet_mtx_name, sizeof((_sc)->vtnet_mtx_name),	\
336         "%s", device_get_nameunit((_sc)->vtnet_dev));			\
337     mtx_init(VTNET_CORE_MTX((_sc)), (_sc)->vtnet_mtx_name,		\
338         "VTNET Core Lock", MTX_DEF);					\
339 } while (0)
340 
341 #endif /* _IF_VTNETVAR_H */
342