1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright 2010 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
25 */
26
27 /*
28 * Xen inter-domain backend - GLDv3 driver edition.
29 *
30 * A traditional GLDv3 driver used to communicate with a guest
31 * domain. This driver is typically plumbed underneath the IP stack
32 * or a software ethernet bridge.
33 */
34
35 #include "xnb.h"
36
37 #include <sys/sunddi.h>
38 #include <sys/conf.h>
39 #include <sys/modctl.h>
40 #include <sys/strsubr.h>
41 #include <sys/dlpi.h>
42 #include <sys/pattr.h>
43 #include <sys/mac_provider.h>
44 #include <sys/mac_ether.h>
45 #include <xen/sys/xendev.h>
46 #include <sys/note.h>
47
48 /* Required driver entry points for GLDv3 */
49 static int xnbu_m_start(void *);
50 static void xnbu_m_stop(void *);
51 static int xnbu_m_set_mac_addr(void *, const uint8_t *);
52 static int xnbu_m_set_multicast(void *, boolean_t, const uint8_t *);
53 static int xnbu_m_set_promiscuous(void *, boolean_t);
54 static int xnbu_m_stat(void *, uint_t, uint64_t *);
55 static boolean_t xnbu_m_getcapab(void *, mac_capab_t, void *);
56 static mblk_t *xnbu_m_send(void *, mblk_t *);
57
58 typedef struct xnbu {
59 mac_handle_t u_mh;
60 boolean_t u_need_sched;
61 } xnbu_t;
62
63 static mac_callbacks_t xnbu_callbacks = {
64 MC_GETCAPAB,
65 xnbu_m_stat,
66 xnbu_m_start,
67 xnbu_m_stop,
68 xnbu_m_set_promiscuous,
69 xnbu_m_set_multicast,
70 xnbu_m_set_mac_addr,
71 xnbu_m_send,
72 NULL,
73 NULL,
74 xnbu_m_getcapab
75 };
76
77 static void
78 xnbu_to_host(xnb_t *xnbp, mblk_t *mp)
79 {
80 xnbu_t *xnbup = xnbp->xnb_flavour_data;
81 boolean_t sched = B_FALSE;
82
83 ASSERT(mp != NULL);
84
85 mac_rx(xnbup->u_mh, NULL, mp);
86
87 mutex_enter(&xnbp->xnb_rx_lock);
88
89 /*
90 * If a transmit attempt failed because we ran out of ring
91 * space and there is now some space, re-enable the transmit
92 * path.
93 */
94 if (xnbup->u_need_sched &&
95 RING_HAS_UNCONSUMED_REQUESTS(&xnbp->xnb_rx_ring)) {
96 sched = B_TRUE;
97 xnbup->u_need_sched = B_FALSE;
98 }
99
100 mutex_exit(&xnbp->xnb_rx_lock);
101
102 if (sched)
103 mac_tx_update(xnbup->u_mh);
104 }
105
106 static mblk_t *
107 xnbu_cksum_from_peer(xnb_t *xnbp, mblk_t *mp, uint16_t flags)
108 {
109 /*
110 * Take a conservative approach - if the checksum is blank
111 * then we fill it in.
112 *
113 * If the consumer of the packet is IP then we might actually
114 * only need fill it in if the data is not validated, but how
115 * do we know who might end up with the packet?
116 */
117
118 if ((flags & NETTXF_csum_blank) != 0) {
119 /*
120 * The checksum is blank. We must fill it in here.
121 */
122 mp = xnb_process_cksum_flags(xnbp, mp, 0);
123
124 /*
125 * Because we calculated the checksum ourselves we
126 * know that it must be good, so we assert this.
127 */
128 flags |= NETTXF_data_validated;
129 }
130
131 if ((flags & NETTXF_data_validated) != 0) {
132 /*
133 * The checksum is asserted valid.
134 */
135 mac_hcksum_set(mp, 0, 0, 0, 0, HCK_FULLCKSUM_OK);
136 }
137
138 return (mp);
139 }
140
141 static uint16_t
142 xnbu_cksum_to_peer(xnb_t *xnbp, mblk_t *mp)
143 {
144 _NOTE(ARGUNUSED(xnbp));
145 uint16_t r = 0;
146 uint32_t pflags;
147
148 mac_hcksum_get(mp, NULL, NULL, NULL, NULL, &pflags);
149
150 /*
151 * If the protocol stack has requested checksum
152 * offload, inform the peer that we have not
153 * calculated the checksum.
154 */
155 if ((pflags & HCK_FULLCKSUM) != 0)
156 r |= NETRXF_csum_blank;
157
158 return (r);
159 }
160
161 static boolean_t
162 xnbu_start_connect(xnb_t *xnbp)
163 {
164 xnbu_t *xnbup = xnbp->xnb_flavour_data;
165
166 mac_link_update(xnbup->u_mh, LINK_STATE_UP);
167 /*
168 * We are able to send packets now - bring them on.
169 */
170 mac_tx_update(xnbup->u_mh);
171
172 return (B_TRUE);
173 }
174
175 static boolean_t
176 xnbu_peer_connected(xnb_t *xnbp)
177 {
178 _NOTE(ARGUNUSED(xnbp));
179
180 return (B_TRUE);
181 }
182
183 static void
184 xnbu_peer_disconnected(xnb_t *xnbp)
185 {
186 xnbu_t *xnbup = xnbp->xnb_flavour_data;
187
188 mac_link_update(xnbup->u_mh, LINK_STATE_DOWN);
189 }
190
191 /*ARGSUSED*/
192 static boolean_t
193 xnbu_hotplug_connected(xnb_t *xnbp)
194 {
195 return (B_TRUE);
196 }
197
198 static mblk_t *
199 xnbu_m_send(void *arg, mblk_t *mp)
200 {
201 xnb_t *xnbp = arg;
202 xnbu_t *xnbup = xnbp->xnb_flavour_data;
203 boolean_t sched = B_FALSE;
204
205 mp = xnb_copy_to_peer(arg, mp);
206
207 mutex_enter(&xnbp->xnb_rx_lock);
208 /*
209 * If we consumed all of the mblk_t's offered, perhaps we need
210 * to indicate that we can accept more. Otherwise we are full
211 * and need to wait for space.
212 */
213 if (mp == NULL) {
214 sched = xnbup->u_need_sched;
215 xnbup->u_need_sched = B_FALSE;
216 } else {
217 xnbup->u_need_sched = B_TRUE;
218 }
219 mutex_exit(&xnbp->xnb_rx_lock);
220
221 /*
222 * If a previous transmit attempt failed because the ring
223 * was full, try again now.
224 */
225 if (sched)
226 mac_tx_update(xnbup->u_mh);
227
228 return (mp);
229 }
230
231 /*
232 * xnbu_m_set_mac_addr() -- set the physical network address on the board
233 */
234 /* ARGSUSED */
235 static int
236 xnbu_m_set_mac_addr(void *arg, const uint8_t *macaddr)
237 {
238 xnb_t *xnbp = arg;
239 xnbu_t *xnbup = xnbp->xnb_flavour_data;
240
241 bcopy(macaddr, xnbp->xnb_mac_addr, ETHERADDRL);
242 mac_unicst_update(xnbup->u_mh, xnbp->xnb_mac_addr);
243
244 return (0);
245 }
246
247 /*
248 * xnbu_m_set_multicast() -- set (enable) or disable a multicast address
249 */
250 /*ARGSUSED*/
251 static int
252 xnbu_m_set_multicast(void *arg, boolean_t add, const uint8_t *mca)
253 {
254 /*
255 * We always accept all packets from the peer, so nothing to
256 * do for enable or disable.
257 */
258 return (0);
259 }
260
261
262 /*
263 * xnbu_m_set_promiscuous() -- set or reset promiscuous mode on the board
264 */
265 /* ARGSUSED */
266 static int
267 xnbu_m_set_promiscuous(void *arg, boolean_t on)
268 {
269 /*
270 * We always accept all packets from the peer, so nothing to
271 * do for enable or disable.
272 */
273 return (0);
274 }
275
276 /*
277 * xnbu_m_start() -- start the board receiving and enable interrupts.
278 */
279 /*ARGSUSED*/
280 static int
281 xnbu_m_start(void *arg)
282 {
283 return (0);
284 }
285
286 /*
287 * xnbu_m_stop() - disable hardware
288 */
289 /*ARGSUSED*/
290 static void
291 xnbu_m_stop(void *arg)
292 {
293 }
294
295 static int
296 xnbu_m_stat(void *arg, uint_t stat, uint64_t *val)
297 {
298 xnb_t *xnbp = arg;
299
300 mutex_enter(&xnbp->xnb_tx_lock);
301 mutex_enter(&xnbp->xnb_rx_lock);
302
303 #define map_stat(q, r) \
304 case (MAC_STAT_##q): \
305 *val = xnbp->xnb_stat_##r; \
306 break
307
308 switch (stat) {
309
310 map_stat(IPACKETS, opackets);
311 map_stat(OPACKETS, ipackets);
312 map_stat(RBYTES, obytes);
313 map_stat(OBYTES, rbytes);
314
315 default:
316 mutex_exit(&xnbp->xnb_rx_lock);
317 mutex_exit(&xnbp->xnb_tx_lock);
318
319 return (ENOTSUP);
320 }
321
322 #undef map_stat
323
324 mutex_exit(&xnbp->xnb_rx_lock);
325 mutex_exit(&xnbp->xnb_tx_lock);
326
327 return (0);
328 }
329
330 static boolean_t
331 xnbu_m_getcapab(void *arg, mac_capab_t cap, void *cap_data)
332 {
333 _NOTE(ARGUNUSED(arg));
334
335 switch (cap) {
336 case MAC_CAPAB_HCKSUM: {
337 uint32_t *capab = cap_data;
338
339 *capab = HCKSUM_INET_PARTIAL;
340 break;
341 }
342 default:
343 return (B_FALSE);
344 }
345
346 return (B_TRUE);
347 }
348
349 /*
350 * All packets are passed to the peer, so adding and removing
351 * multicast addresses is meaningless.
352 */
353 static boolean_t
354 xnbu_mcast_add(xnb_t *xnbp, ether_addr_t *addr)
355 {
356 _NOTE(ARGUNUSED(xnbp, addr));
357
358 return (B_TRUE);
359 }
360
361 static boolean_t
362 xnbu_mcast_del(xnb_t *xnbp, ether_addr_t *addr)
363 {
364 _NOTE(ARGUNUSED(xnbp, addr));
365
366 return (B_TRUE);
367 }
368
369 static int
370 xnbu_attach(dev_info_t *dip, ddi_attach_cmd_t cmd)
371 {
372 static xnb_flavour_t flavour = {
373 xnbu_to_host, xnbu_peer_connected, xnbu_peer_disconnected,
374 xnbu_hotplug_connected, xnbu_start_connect,
375 xnbu_cksum_from_peer, xnbu_cksum_to_peer,
376 xnbu_mcast_add, xnbu_mcast_del,
377 };
378 xnbu_t *xnbup;
379 xnb_t *xnbp;
380 mac_register_t *mr;
381 int err;
382
383 switch (cmd) {
384 case DDI_ATTACH:
385 break;
386 case DDI_RESUME:
387 return (DDI_SUCCESS);
388 default:
389 return (DDI_FAILURE);
390 }
391
392 xnbup = kmem_zalloc(sizeof (*xnbup), KM_SLEEP);
393
394 if ((mr = mac_alloc(MAC_VERSION)) == NULL) {
395 kmem_free(xnbup, sizeof (*xnbup));
396 return (DDI_FAILURE);
397 }
398
399 if (xnb_attach(dip, &flavour, xnbup) != DDI_SUCCESS) {
400 mac_free(mr);
401 kmem_free(xnbup, sizeof (*xnbup));
402 return (DDI_FAILURE);
403 }
404
405 xnbp = ddi_get_driver_private(dip);
406 ASSERT(xnbp != NULL);
407
408 mr->m_dip = dip;
409 mr->m_driver = xnbp;
410
411 /*
412 * Initialize pointers to device specific functions which will be
413 * used by the generic layer.
414 */
415 mr->m_type_ident = MAC_PLUGIN_IDENT_ETHER;
416 mr->m_src_addr = xnbp->xnb_mac_addr;
417 mr->m_callbacks = &xnbu_callbacks;
418 mr->m_min_sdu = 0;
419 mr->m_max_sdu = XNBMAXPKT;
420 /*
421 * xnbu is a virtual device, and it is not associated with any
422 * physical device. Its margin size is determined by the maximum
423 * packet size it can handle, which is PAGESIZE.
424 */
425 mr->m_margin = PAGESIZE - XNBMAXPKT - sizeof (struct ether_header);
426
427 (void) memset(xnbp->xnb_mac_addr, 0xff, ETHERADDRL);
428 xnbp->xnb_mac_addr[0] &= 0xfe;
429 xnbup->u_need_sched = B_FALSE;
430
431 /*
432 * Register ourselves with the GLDv3 interface.
433 */
434 err = mac_register(mr, &xnbup->u_mh);
435 mac_free(mr);
436 if (err != 0) {
437 xnb_detach(dip);
438 kmem_free(xnbup, sizeof (*xnbup));
439 return (DDI_FAILURE);
440 }
441
442 mac_link_update(xnbup->u_mh, LINK_STATE_DOWN);
443
444 return (DDI_SUCCESS);
445 }
446
447 /*ARGSUSED*/
448 int
449 xnbu_detach(dev_info_t *dip, ddi_detach_cmd_t cmd)
450 {
451 xnb_t *xnbp = ddi_get_driver_private(dip);
452 xnbu_t *xnbup = xnbp->xnb_flavour_data;
453
454 switch (cmd) {
455 case DDI_DETACH:
456 break;
457 case DDI_SUSPEND:
458 return (DDI_SUCCESS);
459 default:
460 return (DDI_FAILURE);
461 }
462
463 ASSERT(xnbp != NULL);
464 ASSERT(xnbup != NULL);
465
466 mutex_enter(&xnbp->xnb_tx_lock);
467 mutex_enter(&xnbp->xnb_rx_lock);
468
469 if (!xnbp->xnb_detachable || xnbp->xnb_connected ||
470 (xnbp->xnb_tx_buf_count > 0)) {
471 mutex_exit(&xnbp->xnb_rx_lock);
472 mutex_exit(&xnbp->xnb_tx_lock);
473
474 return (DDI_FAILURE);
475 }
476
477 mutex_exit(&xnbp->xnb_rx_lock);
478 mutex_exit(&xnbp->xnb_tx_lock);
479
480 /*
481 * Attempt to unregister the mac.
482 */
483 if ((xnbup->u_mh != NULL) && (mac_unregister(xnbup->u_mh) != 0))
484 return (DDI_FAILURE);
485 kmem_free(xnbup, sizeof (*xnbup));
486
487 xnb_detach(dip);
488
489 return (DDI_SUCCESS);
490 }
491
492 DDI_DEFINE_STREAM_OPS(ops, nulldev, nulldev, xnbu_attach, xnbu_detach,
493 nodev, NULL, D_MP, NULL, ddi_quiesce_not_supported);
494
495 static struct modldrv modldrv = {
496 &mod_driverops, "xnbu driver", &ops
497 };
498
499 static struct modlinkage modlinkage = {
500 MODREV_1, &modldrv, NULL
501 };
502
503 int
504 _init(void)
505 {
506 int i;
507
508 mac_init_ops(&ops, "xnbu");
509
510 i = mod_install(&modlinkage);
511 if (i != DDI_SUCCESS)
512 mac_fini_ops(&ops);
513
514 return (i);
515 }
516
517 int
518 _fini(void)
519 {
520 int i;
521
522 i = mod_remove(&modlinkage);
523 if (i == DDI_SUCCESS)
524 mac_fini_ops(&ops);
525
526 return (i);
527 }
528
529 int
530 _info(struct modinfo *modinfop)
531 {
532 return (mod_info(&modlinkage, modinfop));
533 }