xref: /linux/drivers/infiniband/ulp/opa_vnic/opa_vnic_netdev.c (revision 4f2c0a4acffbec01079c28f839422e64ddeff004)
1 /*
2  * Copyright(c) 2017 Intel Corporation.
3  *
4  * This file is provided under a dual BSD/GPLv2 license.  When using or
5  * redistributing this file, you may do so under either license.
6  *
7  * GPL LICENSE SUMMARY
8  *
9  * This program is free software; you can redistribute it and/or modify
10  * it under the terms of version 2 of the GNU General Public License as
11  * published by the Free Software Foundation.
12  *
13  * This program is distributed in the hope that it will be useful, but
14  * WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
16  * General Public License for more details.
17  *
18  * BSD LICENSE
19  *
20  * Redistribution and use in source and binary forms, with or without
21  * modification, are permitted provided that the following conditions
22  * are met:
23  *
24  *  - Redistributions of source code must retain the above copyright
25  *    notice, this list of conditions and the following disclaimer.
26  *  - Redistributions in binary form must reproduce the above copyright
27  *    notice, this list of conditions and the following disclaimer in
28  *    the documentation and/or other materials provided with the
29  *    distribution.
30  *  - Neither the name of Intel Corporation nor the names of its
31  *    contributors may be used to endorse or promote products derived
32  *    from this software without specific prior written permission.
33  *
34  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
35  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
36  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
37  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
38  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
39  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
40  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
41  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
42  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
43  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
44  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
45  *
46  */
47 
48 /*
49  * This file contains OPA Virtual Network Interface Controller (VNIC) driver
50  * netdev functionality.
51  */
52 
53 #include <linux/if_vlan.h>
54 #include <linux/crc32.h>
55 
56 #include "opa_vnic_internal.h"
57 
58 #define OPA_TX_TIMEOUT_MS 1000
59 
60 #define OPA_VNIC_SKB_HEADROOM  \
61 			ALIGN((OPA_VNIC_HDR_LEN + OPA_VNIC_SKB_MDATA_LEN), 8)
62 
63 /* This function is overloaded for opa_vnic specific implementation */
opa_vnic_get_stats64(struct net_device * netdev,struct rtnl_link_stats64 * stats)64 static void opa_vnic_get_stats64(struct net_device *netdev,
65 				 struct rtnl_link_stats64 *stats)
66 {
67 	struct opa_vnic_adapter *adapter = opa_vnic_priv(netdev);
68 	struct opa_vnic_stats vstats;
69 
70 	memset(&vstats, 0, sizeof(vstats));
71 	spin_lock(&adapter->stats_lock);
72 	adapter->rn_ops->ndo_get_stats64(netdev, &vstats.netstats);
73 	spin_unlock(&adapter->stats_lock);
74 	memcpy(stats, &vstats.netstats, sizeof(*stats));
75 }
76 
77 /* opa_netdev_start_xmit - transmit function */
opa_netdev_start_xmit(struct sk_buff * skb,struct net_device * netdev)78 static netdev_tx_t opa_netdev_start_xmit(struct sk_buff *skb,
79 					 struct net_device *netdev)
80 {
81 	struct opa_vnic_adapter *adapter = opa_vnic_priv(netdev);
82 
83 	v_dbg("xmit: queue %d skb len %d\n", skb->queue_mapping, skb->len);
84 	/* pad to ensure mininum ethernet packet length */
85 	if (unlikely(skb->len < ETH_ZLEN)) {
86 		if (skb_padto(skb, ETH_ZLEN))
87 			return NETDEV_TX_OK;
88 
89 		skb_put(skb, ETH_ZLEN - skb->len);
90 	}
91 
92 	opa_vnic_encap_skb(adapter, skb);
93 	return adapter->rn_ops->ndo_start_xmit(skb, netdev);
94 }
95 
opa_vnic_select_queue(struct net_device * netdev,struct sk_buff * skb,struct net_device * sb_dev)96 static u16 opa_vnic_select_queue(struct net_device *netdev, struct sk_buff *skb,
97 				 struct net_device *sb_dev)
98 {
99 	struct opa_vnic_adapter *adapter = opa_vnic_priv(netdev);
100 	struct opa_vnic_skb_mdata *mdata;
101 	int rc;
102 
103 	/* pass entropy and vl as metadata in skb */
104 	mdata = skb_push(skb, sizeof(*mdata));
105 	mdata->entropy = opa_vnic_calc_entropy(skb);
106 	mdata->vl = opa_vnic_get_vl(adapter, skb);
107 	rc = adapter->rn_ops->ndo_select_queue(netdev, skb, sb_dev);
108 	skb_pull(skb, sizeof(*mdata));
109 	return rc;
110 }
111 
opa_vnic_update_state(struct opa_vnic_adapter * adapter,bool up)112 static void opa_vnic_update_state(struct opa_vnic_adapter *adapter, bool up)
113 {
114 	struct __opa_veswport_info *info = &adapter->info;
115 
116 	mutex_lock(&adapter->lock);
117 	/* Operational state can only be DROP_ALL or FORWARDING */
118 	if ((info->vport.config_state == OPA_VNIC_STATE_FORWARDING) && up) {
119 		info->vport.oper_state = OPA_VNIC_STATE_FORWARDING;
120 		info->vport.eth_link_status = OPA_VNIC_ETH_LINK_UP;
121 	} else {
122 		info->vport.oper_state = OPA_VNIC_STATE_DROP_ALL;
123 		info->vport.eth_link_status = OPA_VNIC_ETH_LINK_DOWN;
124 	}
125 
126 	if (info->vport.config_state == OPA_VNIC_STATE_FORWARDING)
127 		netif_dormant_off(adapter->netdev);
128 	else
129 		netif_dormant_on(adapter->netdev);
130 	mutex_unlock(&adapter->lock);
131 }
132 
133 /* opa_vnic_process_vema_config - process vema configuration updates */
opa_vnic_process_vema_config(struct opa_vnic_adapter * adapter)134 void opa_vnic_process_vema_config(struct opa_vnic_adapter *adapter)
135 {
136 	struct __opa_veswport_info *info = &adapter->info;
137 	struct rdma_netdev *rn = netdev_priv(adapter->netdev);
138 	u8 port_num[OPA_VESW_MAX_NUM_DEF_PORT] = { 0 };
139 	struct net_device *netdev = adapter->netdev;
140 	u8 i, port_count = 0;
141 	u16 port_mask;
142 
143 	/* If the base_mac_addr is changed, update the interface mac address */
144 	if (memcmp(info->vport.base_mac_addr, adapter->vema_mac_addr,
145 		   ARRAY_SIZE(info->vport.base_mac_addr))) {
146 		struct sockaddr saddr;
147 
148 		memcpy(saddr.sa_data, info->vport.base_mac_addr,
149 		       ARRAY_SIZE(info->vport.base_mac_addr));
150 		mutex_lock(&adapter->lock);
151 		eth_commit_mac_addr_change(netdev, &saddr);
152 		memcpy(adapter->vema_mac_addr,
153 		       info->vport.base_mac_addr, ETH_ALEN);
154 		mutex_unlock(&adapter->lock);
155 	}
156 
157 	rn->set_id(netdev, info->vesw.vesw_id);
158 
159 	/* Handle MTU limit change */
160 	rtnl_lock();
161 	netdev->max_mtu = max_t(unsigned int, info->vesw.eth_mtu,
162 				netdev->min_mtu);
163 	if (netdev->mtu > netdev->max_mtu)
164 		dev_set_mtu(netdev, netdev->max_mtu);
165 	rtnl_unlock();
166 
167 	/* Update flow to default port redirection table */
168 	port_mask = info->vesw.def_port_mask;
169 	for (i = 0; i < OPA_VESW_MAX_NUM_DEF_PORT; i++) {
170 		if (port_mask & 1)
171 			port_num[port_count++] = i;
172 		port_mask >>= 1;
173 	}
174 
175 	/*
176 	 * Build the flow table. Flow table is required when destination LID
177 	 * is not available. Up to OPA_VNIC_FLOW_TBL_SIZE flows supported.
178 	 * Each flow need a default port number to get its dlid from the
179 	 * u_ucast_dlid array.
180 	 */
181 	for (i = 0; i < OPA_VNIC_FLOW_TBL_SIZE; i++)
182 		adapter->flow_tbl[i] = port_count ? port_num[i % port_count] :
183 						    OPA_VNIC_INVALID_PORT;
184 
185 	/* update state */
186 	opa_vnic_update_state(adapter, !!(netdev->flags & IFF_UP));
187 }
188 
189 /*
190  * Set the power on default values in adapter's vema interface structure.
191  */
opa_vnic_set_pod_values(struct opa_vnic_adapter * adapter)192 static inline void opa_vnic_set_pod_values(struct opa_vnic_adapter *adapter)
193 {
194 	adapter->info.vport.max_mac_tbl_ent = OPA_VNIC_MAC_TBL_MAX_ENTRIES;
195 	adapter->info.vport.max_smac_ent = OPA_VNIC_MAX_SMAC_LIMIT;
196 	adapter->info.vport.config_state = OPA_VNIC_STATE_DROP_ALL;
197 	adapter->info.vport.eth_link_status = OPA_VNIC_ETH_LINK_DOWN;
198 	adapter->info.vesw.eth_mtu = ETH_DATA_LEN;
199 }
200 
201 /* opa_vnic_set_mac_addr - change mac address */
opa_vnic_set_mac_addr(struct net_device * netdev,void * addr)202 static int opa_vnic_set_mac_addr(struct net_device *netdev, void *addr)
203 {
204 	struct opa_vnic_adapter *adapter = opa_vnic_priv(netdev);
205 	struct sockaddr *sa = addr;
206 	int rc;
207 
208 	if (!memcmp(netdev->dev_addr, sa->sa_data, ETH_ALEN))
209 		return 0;
210 
211 	mutex_lock(&adapter->lock);
212 	rc = eth_mac_addr(netdev, addr);
213 	mutex_unlock(&adapter->lock);
214 	if (rc)
215 		return rc;
216 
217 	adapter->info.vport.uc_macs_gen_count++;
218 	opa_vnic_vema_report_event(adapter,
219 				   OPA_VESWPORT_TRAP_IFACE_UCAST_MAC_CHANGE);
220 	return 0;
221 }
222 
223 /*
224  * opa_vnic_mac_send_event - post event on possible mac list exchange
225  *  Send trap when digest from uc/mc mac list differs from previous run.
226  *  Digest is evaluated similar to how cksum does.
227  */
opa_vnic_mac_send_event(struct net_device * netdev,u8 event)228 static void opa_vnic_mac_send_event(struct net_device *netdev, u8 event)
229 {
230 	struct opa_vnic_adapter *adapter = opa_vnic_priv(netdev);
231 	struct netdev_hw_addr *ha;
232 	struct netdev_hw_addr_list *hw_list;
233 	u32 *ref_crc;
234 	u32 l, crc = 0;
235 
236 	switch (event) {
237 	case OPA_VESWPORT_TRAP_IFACE_UCAST_MAC_CHANGE:
238 		hw_list = &netdev->uc;
239 		adapter->info.vport.uc_macs_gen_count++;
240 		ref_crc = &adapter->umac_hash;
241 		break;
242 	case OPA_VESWPORT_TRAP_IFACE_MCAST_MAC_CHANGE:
243 		hw_list = &netdev->mc;
244 		adapter->info.vport.mc_macs_gen_count++;
245 		ref_crc = &adapter->mmac_hash;
246 		break;
247 	default:
248 		return;
249 	}
250 	netdev_hw_addr_list_for_each(ha, hw_list) {
251 		crc = crc32_le(crc, ha->addr, ETH_ALEN);
252 	}
253 	l = netdev_hw_addr_list_count(hw_list) * ETH_ALEN;
254 	crc = ~crc32_le(crc, (void *)&l, sizeof(l));
255 
256 	if (crc != *ref_crc) {
257 		*ref_crc = crc;
258 		opa_vnic_vema_report_event(adapter, event);
259 	}
260 }
261 
262 /* opa_vnic_set_rx_mode - handle uc/mc mac list change */
opa_vnic_set_rx_mode(struct net_device * netdev)263 static void opa_vnic_set_rx_mode(struct net_device *netdev)
264 {
265 	opa_vnic_mac_send_event(netdev,
266 				OPA_VESWPORT_TRAP_IFACE_UCAST_MAC_CHANGE);
267 
268 	opa_vnic_mac_send_event(netdev,
269 				OPA_VESWPORT_TRAP_IFACE_MCAST_MAC_CHANGE);
270 }
271 
272 /* opa_netdev_open - activate network interface */
opa_netdev_open(struct net_device * netdev)273 static int opa_netdev_open(struct net_device *netdev)
274 {
275 	struct opa_vnic_adapter *adapter = opa_vnic_priv(netdev);
276 	int rc;
277 
278 	rc = adapter->rn_ops->ndo_open(adapter->netdev);
279 	if (rc) {
280 		v_dbg("open failed %d\n", rc);
281 		return rc;
282 	}
283 
284 	/* Update status and send trap */
285 	opa_vnic_update_state(adapter, true);
286 	opa_vnic_vema_report_event(adapter,
287 				   OPA_VESWPORT_TRAP_ETH_LINK_STATUS_CHANGE);
288 	return 0;
289 }
290 
291 /* opa_netdev_close - disable network interface */
opa_netdev_close(struct net_device * netdev)292 static int opa_netdev_close(struct net_device *netdev)
293 {
294 	struct opa_vnic_adapter *adapter = opa_vnic_priv(netdev);
295 	int rc;
296 
297 	rc = adapter->rn_ops->ndo_stop(adapter->netdev);
298 	if (rc) {
299 		v_dbg("close failed %d\n", rc);
300 		return rc;
301 	}
302 
303 	/* Update status and send trap */
304 	opa_vnic_update_state(adapter, false);
305 	opa_vnic_vema_report_event(adapter,
306 				   OPA_VESWPORT_TRAP_ETH_LINK_STATUS_CHANGE);
307 	return 0;
308 }
309 
310 /* netdev ops */
311 static const struct net_device_ops opa_netdev_ops = {
312 	.ndo_open = opa_netdev_open,
313 	.ndo_stop = opa_netdev_close,
314 	.ndo_start_xmit = opa_netdev_start_xmit,
315 	.ndo_get_stats64 = opa_vnic_get_stats64,
316 	.ndo_set_rx_mode = opa_vnic_set_rx_mode,
317 	.ndo_select_queue = opa_vnic_select_queue,
318 	.ndo_set_mac_address = opa_vnic_set_mac_addr,
319 };
320 
321 /* opa_vnic_add_netdev - create vnic netdev interface */
opa_vnic_add_netdev(struct ib_device * ibdev,u8 port_num,u8 vport_num)322 struct opa_vnic_adapter *opa_vnic_add_netdev(struct ib_device *ibdev,
323 					     u8 port_num, u8 vport_num)
324 {
325 	struct opa_vnic_adapter *adapter;
326 	struct net_device *netdev;
327 	struct rdma_netdev *rn;
328 	int rc;
329 
330 	netdev = ibdev->ops.alloc_rdma_netdev(ibdev, port_num,
331 					      RDMA_NETDEV_OPA_VNIC,
332 					      "veth%d", NET_NAME_UNKNOWN,
333 					      ether_setup);
334 	if (!netdev)
335 		return ERR_PTR(-ENOMEM);
336 	else if (IS_ERR(netdev))
337 		return ERR_CAST(netdev);
338 
339 	rn = netdev_priv(netdev);
340 	adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
341 	if (!adapter) {
342 		rc = -ENOMEM;
343 		goto adapter_err;
344 	}
345 
346 	rn->clnt_priv = adapter;
347 	rn->hca = ibdev;
348 	rn->port_num = port_num;
349 	adapter->netdev = netdev;
350 	adapter->ibdev = ibdev;
351 	adapter->port_num = port_num;
352 	adapter->vport_num = vport_num;
353 	adapter->rn_ops = netdev->netdev_ops;
354 
355 	netdev->netdev_ops = &opa_netdev_ops;
356 	netdev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
357 	netdev->hard_header_len += OPA_VNIC_SKB_HEADROOM;
358 	mutex_init(&adapter->lock);
359 	mutex_init(&adapter->mactbl_lock);
360 	spin_lock_init(&adapter->stats_lock);
361 
362 	SET_NETDEV_DEV(netdev, ibdev->dev.parent);
363 
364 	opa_vnic_set_ethtool_ops(netdev);
365 
366 	opa_vnic_set_pod_values(adapter);
367 
368 	rc = register_netdev(netdev);
369 	if (rc)
370 		goto netdev_err;
371 
372 	netif_carrier_off(netdev);
373 	netif_dormant_on(netdev);
374 	v_info("initialized\n");
375 
376 	return adapter;
377 netdev_err:
378 	mutex_destroy(&adapter->lock);
379 	mutex_destroy(&adapter->mactbl_lock);
380 	kfree(adapter);
381 adapter_err:
382 	rn->free_rdma_netdev(netdev);
383 
384 	return ERR_PTR(rc);
385 }
386 
387 /* opa_vnic_rem_netdev - remove vnic netdev interface */
opa_vnic_rem_netdev(struct opa_vnic_adapter * adapter)388 void opa_vnic_rem_netdev(struct opa_vnic_adapter *adapter)
389 {
390 	struct net_device *netdev = adapter->netdev;
391 	struct rdma_netdev *rn = netdev_priv(netdev);
392 
393 	v_info("removing\n");
394 	unregister_netdev(netdev);
395 	opa_vnic_release_mac_tbl(adapter);
396 	mutex_destroy(&adapter->lock);
397 	mutex_destroy(&adapter->mactbl_lock);
398 	kfree(adapter);
399 	rn->free_rdma_netdev(netdev);
400 }
401