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  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /*
26  * Copyright (c) 2010, Intel Corporation.
27  * All rights reserved.
28  */
29 
30 #include <sys/types.h>
31 #include <sys/cmn_err.h>
32 #include <sys/conf.h>
33 #include <sys/debug.h>
34 #include <sys/errno.h>
35 #include <sys/note.h>
36 #include <sys/dditypes.h>
37 #include <sys/ddi.h>
38 #include <sys/sunddi.h>
39 #include <sys/sunndi.h>
40 #include <sys/ddi_impldefs.h>
41 #include <sys/ndi_impldefs.h>
42 #include <sys/varargs.h>
43 #include <sys/modctl.h>
44 #include <sys/kmem.h>
45 #include <sys/cpuvar.h>
46 #include <sys/cpupart.h>
47 #include <sys/mem_config.h>
48 #include <sys/mem_cage.h>
49 #include <sys/memnode.h>
50 #include <sys/callb.h>
51 #include <sys/ontrap.h>
52 #include <sys/obpdefs.h>
53 #include <sys/promif.h>
54 #include <sys/synch.h>
55 #include <sys/systm.h>
56 #include <sys/sysmacros.h>
57 #include <sys/archsystm.h>
58 #include <sys/machsystm.h>
59 #include <sys/x_call.h>
60 #include <sys/x86_archext.h>
61 #include <sys/fastboot_impl.h>
62 #include <sys/sysevent.h>
63 #include <sys/sysevent/dr.h>
64 #include <sys/sysevent/eventdefs.h>
65 #include <sys/acpi/acpi.h>
66 #include <sys/acpica.h>
67 #include <sys/acpidev.h>
68 #include <sys/acpidev_rsc.h>
69 #include <sys/acpidev_dr.h>
70 #include <sys/dr.h>
71 #include <sys/dr_util.h>
72 #include <sys/drmach.h>
73 #include "drmach_acpi.h"
74 
75 /* utility */
76 #define	MBYTE		(1048576ull)
77 #define	_ptob64(p)	((uint64_t)(p) << PAGESHIFT)
78 #define	_b64top(b)	((pgcnt_t)((b) >> PAGESHIFT))
79 
80 static int 		drmach_init(void);
81 static void 		drmach_fini(void);
82 static int		drmach_name2type_idx(char *);
83 static sbd_error_t	*drmach_mem_update_lgrp(drmachid_t);
84 
85 static void drmach_board_dispose(drmachid_t id);
86 static sbd_error_t *drmach_board_release(drmachid_t);
87 static sbd_error_t *drmach_board_status(drmachid_t, drmach_status_t *);
88 
89 static void drmach_io_dispose(drmachid_t);
90 static sbd_error_t *drmach_io_release(drmachid_t);
91 static sbd_error_t *drmach_io_status(drmachid_t, drmach_status_t *);
92 
93 static void drmach_cpu_dispose(drmachid_t);
94 static sbd_error_t *drmach_cpu_release(drmachid_t);
95 static sbd_error_t *drmach_cpu_status(drmachid_t, drmach_status_t *);
96 
97 static void drmach_mem_dispose(drmachid_t);
98 static sbd_error_t *drmach_mem_release(drmachid_t);
99 static sbd_error_t *drmach_mem_status(drmachid_t, drmach_status_t *);
100 
101 #ifdef DEBUG
102 int drmach_debug = 1;		 /* set to non-zero to enable debug messages */
103 #endif /* DEBUG */
104 
105 drmach_domain_info_t	 drmach_domain;
106 
107 static char		*drmach_ie_fmt = "drmach_acpi.c %d";
108 static drmach_array_t	*drmach_boards;
109 
110 /* rwlock to protect drmach_boards. */
111 static krwlock_t	 drmach_boards_rwlock;
112 
113 /* rwlock to block out CPR thread. */
114 static krwlock_t	 drmach_cpr_rwlock;
115 
116 /* CPR callb id. */
117 static callb_id_t	 drmach_cpr_cid;
118 
119 static struct {
120 	const char	*name;
121 	const char	*type;
122 	sbd_error_t	*(*new)(drmach_device_t *, drmachid_t *);
123 } drmach_name2type[] = {
124 	{ ACPIDEV_NODE_NAME_CPU,	DRMACH_DEVTYPE_CPU, drmach_cpu_new },
125 	{ ACPIDEV_NODE_NAME_MEMORY,	DRMACH_DEVTYPE_MEM, drmach_mem_new },
126 	{ ACPIDEV_NODE_NAME_PCI,	DRMACH_DEVTYPE_PCI, drmach_io_new  },
127 };
128 
129 /*
130  * drmach autoconfiguration data structures and interfaces
131  */
132 static struct modlmisc modlmisc = {
133 	&mod_miscops,
134 	"ACPI based DR v1.0"
135 };
136 
137 static struct modlinkage modlinkage = {
138 	MODREV_1,
139 	(void *)&modlmisc,
140 	NULL
141 };
142 
143 int
144 _init(void)
145 {
146 	int err;
147 
148 	if ((err = drmach_init()) != 0) {
149 		return (err);
150 	}
151 
152 	if ((err = mod_install(&modlinkage)) != 0) {
153 		drmach_fini();
154 	}
155 
156 	return (err);
157 }
158 
159 int
160 _fini(void)
161 {
162 	int	err;
163 
164 	if ((err = mod_remove(&modlinkage)) == 0) {
165 		drmach_fini();
166 	}
167 
168 	return (err);
169 }
170 
171 int
172 _info(struct modinfo *modinfop)
173 {
174 	return (mod_info(&modlinkage, modinfop));
175 }
176 
177 /*
178  * Internal support functions.
179  */
180 static DRMACH_HANDLE
181 drmach_node_acpi_get_dnode(drmach_node_t *np)
182 {
183 	return ((DRMACH_HANDLE)(uintptr_t)np->here);
184 }
185 
186 static dev_info_t *
187 drmach_node_acpi_get_dip(drmach_node_t *np)
188 {
189 	dev_info_t *dip = NULL;
190 
191 	if (ACPI_FAILURE(acpica_get_devinfo((DRMACH_HANDLE)(np->here), &dip))) {
192 		return (NULL);
193 	}
194 
195 	return (dip);
196 }
197 
198 static int
199 drmach_node_acpi_get_prop(drmach_node_t *np, char *name, void *buf, int len)
200 {
201 	int		rv = 0;
202 	DRMACH_HANDLE	hdl;
203 
204 	hdl = np->get_dnode(np);
205 	if (hdl == NULL) {
206 		DRMACH_PR("!drmach_node_acpi_get_prop: NULL handle");
207 		rv = -1;
208 	} else {
209 		rv = acpidev_dr_device_getprop(hdl, name, buf, len);
210 		if (rv >= 0) {
211 			ASSERT(rv <= len);
212 			rv = 0;
213 		}
214 	}
215 
216 	return (rv);
217 }
218 
219 static int
220 drmach_node_acpi_get_proplen(drmach_node_t *np, char *name, int *len)
221 {
222 	int		rv = 0;
223 	DRMACH_HANDLE	hdl;
224 
225 	hdl = np->get_dnode(np);
226 	if (hdl == NULL) {
227 		DRMACH_PR("!drmach_node_acpi_get_proplen: NULL handle");
228 		rv = -1;
229 	} else {
230 		rv = acpidev_dr_device_getprop(hdl, name, NULL, 0);
231 		if (rv >= 0) {
232 			*len = rv;
233 			return (0);
234 		}
235 	}
236 
237 	return (-1);
238 }
239 
240 static ACPI_STATUS
241 drmach_node_acpi_callback(ACPI_HANDLE hdl, uint_t lvl, void *ctx, void **retval)
242 {
243 	_NOTE(ARGUNUSED(lvl));
244 
245 	int rv;
246 	dev_info_t *dip;
247 	drmach_node_walk_args_t *argp = ctx;
248 	int (*cb)(drmach_node_walk_args_t *args);
249 	acpidev_class_id_t clsid;
250 
251 	ASSERT(hdl != NULL);
252 	ASSERT(ctx != NULL);
253 	ASSERT(retval != NULL);
254 
255 	/* Skip subtree if the device is not powered. */
256 	if (!acpidev_dr_device_is_powered(hdl)) {
257 		return (AE_CTRL_DEPTH);
258 	}
259 
260 	/*
261 	 * Keep scanning subtree if it fails to lookup device node.
262 	 * There may be some ACPI objects without device nodes created.
263 	 */
264 	if (ACPI_FAILURE(acpica_get_devinfo(hdl, &dip))) {
265 		return (AE_OK);
266 	}
267 
268 	argp->node->here = hdl;
269 	cb = (int (*)(drmach_node_walk_args_t *args))argp->func;
270 	rv = (*cb)(argp);
271 	argp->node->here = NULL;
272 	if (rv) {
273 		*(int *)retval = rv;
274 		return (AE_CTRL_TERMINATE);
275 	}
276 
277 	/*
278 	 * Skip descendants of PCI/PCIex host bridges.
279 	 * PCI/PCIex devices will be handled by pcihp.
280 	 */
281 	clsid = acpidev_dr_device_get_class(hdl);
282 	if (clsid == ACPIDEV_CLASS_ID_PCI || clsid == ACPIDEV_CLASS_ID_PCIEX) {
283 		return (AE_CTRL_DEPTH);
284 	}
285 
286 	return (AE_OK);
287 }
288 
289 static int
290 drmach_node_acpi_walk(drmach_node_t *np, void *data,
291     int (*cb)(drmach_node_walk_args_t *args))
292 {
293 	DRMACH_HANDLE		hdl;
294 	int			rv = 0;
295 	drmach_node_walk_args_t	args;
296 
297 	/* initialize the args structure for callback */
298 	args.node = np;
299 	args.data = data;
300 	args.func = (void *)cb;
301 
302 	/* save the handle, it will be modified when walking the tree. */
303 	hdl = np->get_dnode(np);
304 	if (hdl == NULL) {
305 		DRMACH_PR("!drmach_node_acpi_walk: failed to get device node.");
306 		return (EX86_INAPPROP);
307 	}
308 
309 	if (ACPI_FAILURE(acpidev_dr_device_walk_device(hdl,
310 	    ACPIDEV_MAX_ENUM_LEVELS, drmach_node_acpi_callback,
311 	    &args, (void *)&rv))) {
312 		/*
313 		 * If acpidev_dr_device_walk_device() itself fails, rv won't
314 		 * be set to suitable error code. Set it here.
315 		 */
316 		if (rv == 0) {
317 			cmn_err(CE_WARN, "!drmach_node_acpi_walk: failed to "
318 			    "walk ACPI namespace.");
319 			rv = EX86_ACPIWALK;
320 		}
321 	}
322 
323 	/* restore the handle to original value after walking the tree. */
324 	np->here = (void *)hdl;
325 
326 	return ((int)rv);
327 }
328 
329 static drmach_node_t *
330 drmach_node_new(void)
331 {
332 	drmach_node_t *np;
333 
334 	np = kmem_zalloc(sizeof (drmach_node_t), KM_SLEEP);
335 
336 	np->get_dnode = drmach_node_acpi_get_dnode;
337 	np->getdip = drmach_node_acpi_get_dip;
338 	np->getproplen = drmach_node_acpi_get_proplen;
339 	np->getprop = drmach_node_acpi_get_prop;
340 	np->walk = drmach_node_acpi_walk;
341 
342 	return (np);
343 }
344 
345 static drmachid_t
346 drmach_node_dup(drmach_node_t *np)
347 {
348 	drmach_node_t *dup;
349 
350 	dup = drmach_node_new();
351 	dup->here = np->here;
352 	dup->get_dnode = np->get_dnode;
353 	dup->getdip = np->getdip;
354 	dup->getproplen = np->getproplen;
355 	dup->getprop = np->getprop;
356 	dup->walk = np->walk;
357 
358 	return (dup);
359 }
360 
361 static void
362 drmach_node_dispose(drmach_node_t *np)
363 {
364 	kmem_free(np, sizeof (*np));
365 }
366 
367 static int
368 drmach_node_walk(drmach_node_t *np, void *param,
369 	int (*cb)(drmach_node_walk_args_t *args))
370 {
371 	return (np->walk(np, param, cb));
372 }
373 
374 static DRMACH_HANDLE
375 drmach_node_get_dnode(drmach_node_t *np)
376 {
377 	return (np->get_dnode(np));
378 }
379 
380 /*
381  * drmach_array provides convenient array construction, access,
382  * bounds checking and array destruction logic.
383  */
384 static drmach_array_t *
385 drmach_array_new(uint_t min_index, uint_t max_index)
386 {
387 	drmach_array_t *arr;
388 
389 	arr = kmem_zalloc(sizeof (drmach_array_t), KM_SLEEP);
390 
391 	arr->arr_sz = (max_index - min_index + 1) * sizeof (void *);
392 	if (arr->arr_sz > 0) {
393 		arr->min_index = min_index;
394 		arr->max_index = max_index;
395 
396 		arr->arr = kmem_zalloc(arr->arr_sz, KM_SLEEP);
397 		return (arr);
398 	} else {
399 		kmem_free(arr, sizeof (*arr));
400 		return (0);
401 	}
402 }
403 
404 static int
405 drmach_array_set(drmach_array_t *arr, uint_t idx, drmachid_t val)
406 {
407 	if (idx < arr->min_index || idx > arr->max_index)
408 		return (-1);
409 	arr->arr[idx - arr->min_index] = val;
410 	return (0);
411 }
412 
413 /*
414  * Get the item with index idx.
415  * Return 0 with the value stored in val if succeeds, otherwise return -1.
416  */
417 static int
418 drmach_array_get(drmach_array_t *arr, uint_t idx, drmachid_t *val)
419 {
420 	if (idx < arr->min_index || idx > arr->max_index)
421 		return (-1);
422 	*val = arr->arr[idx - arr->min_index];
423 	return (0);
424 }
425 
426 static int
427 drmach_array_first(drmach_array_t *arr, uint_t *idx, drmachid_t *val)
428 {
429 	int rv;
430 
431 	*idx = arr->min_index;
432 	while ((rv = drmach_array_get(arr, *idx, val)) == 0 && *val == NULL)
433 		*idx += 1;
434 
435 	return (rv);
436 }
437 
438 static int
439 drmach_array_next(drmach_array_t *arr, uint_t *idx, drmachid_t *val)
440 {
441 	int rv;
442 
443 	*idx += 1;
444 	while ((rv = drmach_array_get(arr, *idx, val)) == 0 && *val == NULL)
445 		*idx += 1;
446 
447 	return (rv);
448 }
449 
450 static void
451 drmach_array_dispose(drmach_array_t *arr, void (*disposer)(drmachid_t))
452 {
453 	drmachid_t	val;
454 	uint_t		idx;
455 	int		rv;
456 
457 	rv = drmach_array_first(arr, &idx, &val);
458 	while (rv == 0) {
459 		(*disposer)(val);
460 		rv = drmach_array_next(arr, &idx, &val);
461 	}
462 
463 	kmem_free(arr->arr, arr->arr_sz);
464 	kmem_free(arr, sizeof (*arr));
465 }
466 
467 static drmach_board_t *
468 drmach_get_board_by_bnum(uint_t bnum)
469 {
470 	drmachid_t id;
471 
472 	if (drmach_array_get(drmach_boards, bnum, &id) == 0)
473 		return ((drmach_board_t *)id);
474 	else
475 		return (NULL);
476 }
477 
478 sbd_error_t *
479 drmach_device_new(drmach_node_t *node,
480 	drmach_board_t *bp, int portid, drmachid_t *idp)
481 {
482 	int		 i;
483 	int		 rv;
484 	drmach_device_t	 proto;
485 	sbd_error_t	*err;
486 	char		 name[OBP_MAXDRVNAME];
487 
488 	rv = node->getprop(node, ACPIDEV_DR_PROP_DEVNAME, name, OBP_MAXDRVNAME);
489 	if (rv) {
490 		/* every node is expected to have a name */
491 		err = drerr_new(1, EX86_GETPROP, "device node %s: property %s",
492 		    ddi_node_name(node->getdip(node)),
493 		    ACPIDEV_DR_PROP_DEVNAME);
494 		return (err);
495 	}
496 
497 	/*
498 	 * The node currently being examined is not listed in the name2type[]
499 	 * array.  In this case, the node is no interest to drmach.  Both
500 	 * dp and err are initialized here to yield nothing (no device or
501 	 * error structure) for this case.
502 	 */
503 	i = drmach_name2type_idx(name);
504 	if (i < 0) {
505 		*idp = (drmachid_t)0;
506 		return (NULL);
507 	}
508 
509 	/* device specific new function will set unum */
510 	bzero(&proto, sizeof (proto));
511 	proto.type = drmach_name2type[i].type;
512 	proto.bp = bp;
513 	proto.node = node;
514 	proto.portid = portid;
515 
516 	return (drmach_name2type[i].new(&proto, idp));
517 }
518 
519 static void
520 drmach_device_dispose(drmachid_t id)
521 {
522 	drmach_device_t *self = id;
523 
524 	self->cm.dispose(id);
525 }
526 
527 static sbd_error_t *
528 drmach_device_status(drmachid_t id, drmach_status_t *stat)
529 {
530 	drmach_common_t *cp;
531 
532 	if (!DRMACH_IS_ID(id))
533 		return (drerr_new(0, EX86_NOTID, NULL));
534 	cp = id;
535 
536 	return (cp->status(id, stat));
537 }
538 
539 drmach_board_t *
540 drmach_board_new(uint_t bnum, int boot_board)
541 {
542 	sbd_error_t *err;
543 	drmach_board_t	*bp;
544 	dev_info_t *dip = NULL;
545 
546 	bp = kmem_zalloc(sizeof (drmach_board_t), KM_SLEEP);
547 	bp->cm.isa = (void *)drmach_board_new;
548 	bp->cm.release = drmach_board_release;
549 	bp->cm.status = drmach_board_status;
550 
551 	bp->bnum = bnum;
552 	bp->devices = NULL;
553 	bp->tree = drmach_node_new();
554 
555 	acpidev_dr_lock_all();
556 	if (ACPI_FAILURE(acpidev_dr_get_board_handle(bnum, &bp->tree->here))) {
557 		acpidev_dr_unlock_all();
558 		drmach_board_dispose(bp);
559 		return (NULL);
560 	}
561 	acpidev_dr_unlock_all();
562 	ASSERT(bp->tree->here != NULL);
563 
564 	err = drmach_board_name(bnum, bp->cm.name, sizeof (bp->cm.name));
565 	if (err != NULL) {
566 		sbd_err_clear(&err);
567 		drmach_board_dispose(bp);
568 		return (NULL);
569 	}
570 
571 	if (acpidev_dr_device_is_powered(bp->tree->here)) {
572 		bp->boot_board = boot_board;
573 		bp->powered = 1;
574 	} else {
575 		bp->boot_board = 0;
576 		bp->powered = 0;
577 	}
578 	bp->assigned = boot_board;
579 	if (ACPI_SUCCESS(acpica_get_devinfo(bp->tree->here, &dip))) {
580 		bp->connected = 1;
581 	} else {
582 		bp->connected = 0;
583 	}
584 
585 	(void) drmach_array_set(drmach_boards, bnum, bp);
586 
587 	return (bp);
588 }
589 
590 static void
591 drmach_board_dispose(drmachid_t id)
592 {
593 	drmach_board_t *bp;
594 
595 	ASSERT(DRMACH_IS_BOARD_ID(id));
596 	bp = id;
597 
598 	if (bp->tree)
599 		drmach_node_dispose(bp->tree);
600 
601 	if (bp->devices)
602 		drmach_array_dispose(bp->devices, drmach_device_dispose);
603 
604 	kmem_free(bp, sizeof (drmach_board_t));
605 }
606 
607 static sbd_error_t *
608 drmach_board_release(drmachid_t id)
609 {
610 	if (!DRMACH_IS_BOARD_ID(id))
611 		return (drerr_new(0, EX86_INAPPROP, NULL));
612 
613 	return (NULL);
614 }
615 
616 static int
617 drmach_board_check_power(drmach_board_t *bp)
618 {
619 	DRMACH_HANDLE	hdl;
620 
621 	hdl = drmach_node_get_dnode(bp->tree);
622 
623 	return (acpidev_dr_device_is_powered(hdl));
624 }
625 
626 struct drmach_board_list_dep_arg {
627 	int	count;
628 	size_t	len;
629 	ssize_t	off;
630 	char	*buf;
631 	char	temp[MAXPATHLEN];
632 };
633 
634 static ACPI_STATUS
635 drmach_board_generate_name(ACPI_HANDLE hdl, UINT32 lvl, void *ctx,
636     void **retval)
637 {
638 	_NOTE(ARGUNUSED(retval));
639 
640 	struct drmach_board_list_dep_arg *argp = ctx;
641 
642 	ASSERT(hdl != NULL);
643 	ASSERT(lvl == UINT32_MAX);
644 	ASSERT(ctx != NULL);
645 
646 	/* Skip non-board devices. */
647 	if (!acpidev_dr_device_is_board(hdl)) {
648 		return (AE_OK);
649 	}
650 
651 	if (ACPI_FAILURE(acpidev_dr_get_board_name(hdl, argp->temp,
652 	    sizeof (argp->temp)))) {
653 		DRMACH_PR("!drmach_board_generate_name: failed to "
654 		    "generate board name for handle %p.", hdl);
655 		/* Keep on walking. */
656 		return (AE_OK);
657 	}
658 	argp->count++;
659 	argp->off += snprintf(argp->buf + argp->off, argp->len - argp->off,
660 	    " %s", argp->temp);
661 	if (argp->off >= argp->len) {
662 		return (AE_CTRL_TERMINATE);
663 	}
664 
665 	return (AE_OK);
666 }
667 
668 static ssize_t
669 drmach_board_list_dependency(ACPI_HANDLE hdl, boolean_t edl, char *prefix,
670     char *buf, size_t len)
671 {
672 	ACPI_STATUS rc;
673 	ssize_t off;
674 	struct drmach_board_list_dep_arg *ap;
675 
676 	ASSERT(buf != NULL && len != 0);
677 	if (buf == NULL || len == 0) {
678 		return (-1);
679 	}
680 
681 	ap = kmem_zalloc(sizeof (*ap), KM_SLEEP);
682 	ap->buf = buf;
683 	ap->len = len;
684 	ap->off = snprintf(buf, len, "%s", prefix);
685 	if (ap->off >= len) {
686 		*buf = '\0';
687 		kmem_free(ap, sizeof (*ap));
688 		return (-1);
689 	}
690 
691 	/* Generate the device dependency list. */
692 	if (edl) {
693 		rc = acpidev_dr_device_walk_edl(hdl,
694 		    drmach_board_generate_name, ap, NULL);
695 	} else {
696 		rc = acpidev_dr_device_walk_ejd(hdl,
697 		    drmach_board_generate_name, ap, NULL);
698 	}
699 	if (ACPI_FAILURE(rc)) {
700 		*buf = '\0';
701 		ap->off = -1;
702 	/* No device has dependency on this board. */
703 	} else if (ap->count == 0) {
704 		*buf = '\0';
705 		ap->off = 0;
706 	}
707 
708 	off = ap->off;
709 	kmem_free(ap, sizeof (*ap));
710 
711 	return (off);
712 }
713 
714 static sbd_error_t *
715 drmach_board_status(drmachid_t id, drmach_status_t *stat)
716 {
717 	sbd_error_t	*err = NULL;
718 	drmach_board_t	*bp;
719 	DRMACH_HANDLE	hdl;
720 	size_t		off;
721 
722 	if (!DRMACH_IS_BOARD_ID(id))
723 		return (drerr_new(0, EX86_INAPPROP, NULL));
724 	bp = id;
725 
726 	if (bp->tree == NULL)
727 		return (drerr_new(0, EX86_INAPPROP, NULL));
728 	hdl = drmach_node_get_dnode(bp->tree);
729 	if (hdl == NULL)
730 		return (drerr_new(0, EX86_INAPPROP, NULL));
731 
732 	stat->busy = 0;			/* assume not busy */
733 	stat->configured = 0;		/* assume not configured */
734 	stat->assigned = bp->assigned;
735 	stat->powered = bp->powered = acpidev_dr_device_is_powered(hdl);
736 	stat->empty = !acpidev_dr_device_is_present(hdl);
737 	if (ACPI_SUCCESS(acpidev_dr_device_check_status(hdl))) {
738 		stat->cond = bp->cond = SBD_COND_OK;
739 	} else {
740 		stat->cond = bp->cond = SBD_COND_FAILED;
741 	}
742 	stat->info[0] = '\0';
743 
744 	/* Generate the eject device list. */
745 	if (drmach_board_list_dependency(hdl, B_TRUE, "EDL:",
746 	    stat->info, sizeof (stat->info)) < 0) {
747 		DRMACH_PR("!drmach_board_status: failed to generate "
748 		    "eject device list for board %d.", bp->bnum);
749 		stat->info[0] = '\0';
750 	}
751 	off = strlen(stat->info);
752 	if (off < sizeof (stat->info)) {
753 		if (drmach_board_list_dependency(hdl, B_FALSE,
754 		    off ? ", EJD:" : "EJD:",
755 		    stat->info + off, sizeof (stat->info) - off) < 0) {
756 			DRMACH_PR("!drmach_board_status: failed to generate "
757 			    "eject dependent device for board %d.", bp->bnum);
758 			stat->info[off] = '\0';
759 		}
760 	}
761 
762 	switch (acpidev_dr_get_board_type(bp->tree->get_dnode(bp->tree))) {
763 	case ACPIDEV_CPU_BOARD:
764 		(void) strlcpy(stat->type, "CPU Board", sizeof (stat->type));
765 		break;
766 	case ACPIDEV_MEMORY_BOARD:
767 		(void) strlcpy(stat->type, "MemoryBoard", sizeof (stat->type));
768 		break;
769 	case ACPIDEV_IO_BOARD:
770 		(void) strlcpy(stat->type, "IO Board", sizeof (stat->type));
771 		break;
772 	case ACPIDEV_SYSTEM_BOARD:
773 		/*FALLTHROUGH*/
774 	default:
775 		(void) strlcpy(stat->type, "SystemBoard", sizeof (stat->type));
776 		break;
777 	}
778 
779 	if (bp->devices) {
780 		int		 rv;
781 		uint_t		 d_idx;
782 		drmachid_t	 d_id;
783 
784 		rv = drmach_array_first(bp->devices, &d_idx, &d_id);
785 		while (rv == 0) {
786 			drmach_status_t	d_stat;
787 
788 			err = drmach_device_status(d_id, &d_stat);
789 			if (err)
790 				break;
791 
792 			stat->busy |= d_stat.busy;
793 			stat->configured |= d_stat.configured;
794 
795 			rv = drmach_array_next(bp->devices, &d_idx, &d_id);
796 		}
797 	}
798 
799 	return (err);
800 }
801 
802 /*
803  * When DR is initialized, we walk the device tree and acquire a hold on
804  * all the nodes that are interesting to DR. This is so that the corresponding
805  * branches cannot be deleted.
806  */
807 static int
808 drmach_hold_rele_devtree(dev_info_t *rdip, void *arg)
809 {
810 	int *holdp = (int *)arg;
811 	ACPI_HANDLE hdl = NULL;
812 	acpidev_data_handle_t dhdl;
813 
814 	/* Skip nodes and subtrees which are not created by acpidev. */
815 	if (ACPI_FAILURE(acpica_get_handle(rdip, &hdl))) {
816 		return (DDI_WALK_PRUNECHILD);
817 	}
818 	ASSERT(hdl != NULL);
819 	dhdl = acpidev_data_get_handle(hdl);
820 	if (dhdl == NULL) {
821 		return (DDI_WALK_PRUNECHILD);
822 	}
823 
824 	/* Hold/release devices which are interesting to DR operations. */
825 	if (acpidev_data_dr_ready(dhdl)) {
826 		if (*holdp) {
827 			ASSERT(!e_ddi_branch_held(rdip));
828 			e_ddi_branch_hold(rdip);
829 		} else {
830 			ASSERT(e_ddi_branch_held(rdip));
831 			e_ddi_branch_rele(rdip);
832 		}
833 	}
834 
835 	return (DDI_WALK_CONTINUE);
836 }
837 
838 static void
839 drmach_hold_devtree(void)
840 {
841 	dev_info_t *dip;
842 	int circ;
843 	int hold = 1;
844 
845 	dip = ddi_root_node();
846 	ndi_devi_enter(dip, &circ);
847 	ddi_walk_devs(ddi_get_child(dip), drmach_hold_rele_devtree, &hold);
848 	ndi_devi_exit(dip, circ);
849 }
850 
851 static void
852 drmach_release_devtree(void)
853 {
854 	dev_info_t *dip;
855 	int circ;
856 	int hold = 0;
857 
858 	dip = ddi_root_node();
859 	ndi_devi_enter(dip, &circ);
860 	ddi_walk_devs(ddi_get_child(dip), drmach_hold_rele_devtree, &hold);
861 	ndi_devi_exit(dip, circ);
862 }
863 
864 static boolean_t
865 drmach_cpr_callb(void *arg, int code)
866 {
867 	_NOTE(ARGUNUSED(arg));
868 
869 	if (code == CB_CODE_CPR_CHKPT) {
870 		/*
871 		 * Temporarily block CPR operations if there are DR operations
872 		 * ongoing.
873 		 */
874 		rw_enter(&drmach_cpr_rwlock, RW_WRITER);
875 	} else {
876 		rw_exit(&drmach_cpr_rwlock);
877 	}
878 
879 	return (B_TRUE);
880 }
881 
882 static int
883 drmach_init(void)
884 {
885 	DRMACH_HANDLE	hdl;
886 	drmachid_t	id;
887 	uint_t		bnum;
888 
889 	if (MAX_BOARDS > SHRT_MAX) {
890 		cmn_err(CE_WARN, "!drmach_init: system has too many (%d) "
891 		    "hotplug capable boards.", MAX_BOARDS);
892 		return (ENXIO);
893 	} else if (MAX_CMP_UNITS_PER_BOARD > 1) {
894 		cmn_err(CE_WARN, "!drmach_init: DR doesn't support multiple "
895 		    "(%d) physical processors on one board.",
896 		    MAX_CMP_UNITS_PER_BOARD);
897 		return (ENXIO);
898 	} else if (!ISP2(MAX_CORES_PER_CMP)) {
899 		cmn_err(CE_WARN, "!drmach_init: number of logical CPUs (%d) in "
900 		    "physical processor is not power of 2.",
901 		    MAX_CORES_PER_CMP);
902 		return (ENXIO);
903 	} else if (MAX_CPU_UNITS_PER_BOARD > DEVSET_CPU_NUMBER ||
904 	    MAX_MEM_UNITS_PER_BOARD > DEVSET_MEM_NUMBER ||
905 	    MAX_IO_UNITS_PER_BOARD > DEVSET_IO_NUMBER) {
906 		cmn_err(CE_WARN, "!drmach_init: system has more CPU/memory/IO "
907 		    "units than the DR driver can handle.");
908 		return (ENXIO);
909 	}
910 
911 	rw_init(&drmach_cpr_rwlock, NULL, RW_DEFAULT, NULL);
912 	drmach_cpr_cid = callb_add(drmach_cpr_callb, NULL,
913 	    CB_CL_CPR_PM, "drmach");
914 
915 	rw_init(&drmach_boards_rwlock, NULL, RW_DEFAULT, NULL);
916 	drmach_boards = drmach_array_new(0, MAX_BOARDS - 1);
917 	drmach_domain.allow_dr = acpidev_dr_capable();
918 
919 	for (bnum = 0; bnum < MAX_BOARDS; bnum++) {
920 		hdl = NULL;
921 		if (ACPI_FAILURE(acpidev_dr_get_board_handle(bnum, &hdl)) ||
922 		    hdl == NULL) {
923 			cmn_err(CE_WARN, "!drmach_init: failed to lookup ACPI "
924 			    "handle for board %d.", bnum);
925 			continue;
926 		}
927 		if (drmach_array_get(drmach_boards, bnum, &id) == -1) {
928 			DRMACH_PR("!drmach_init: failed to get handle "
929 			    "for board %d.", bnum);
930 			ASSERT(0);
931 			goto error;
932 		} else if (id == NULL) {
933 			(void) drmach_board_new(bnum, 1);
934 		}
935 	}
936 
937 	/*
938 	 * Walk descendants of the devinfo root node and hold
939 	 * all devinfo branches of interest.
940 	 */
941 	drmach_hold_devtree();
942 
943 	return (0);
944 
945 error:
946 	drmach_array_dispose(drmach_boards, drmach_board_dispose);
947 	rw_destroy(&drmach_boards_rwlock);
948 	rw_destroy(&drmach_cpr_rwlock);
949 	return (ENXIO);
950 }
951 
952 static void
953 drmach_fini(void)
954 {
955 	rw_enter(&drmach_boards_rwlock, RW_WRITER);
956 	if (drmach_boards != NULL) {
957 		drmach_array_dispose(drmach_boards, drmach_board_dispose);
958 		drmach_boards = NULL;
959 	}
960 	rw_exit(&drmach_boards_rwlock);
961 
962 	/*
963 	 * Walk descendants of the root devinfo node
964 	 * release holds acquired on branches in drmach_init()
965 	 */
966 	drmach_release_devtree();
967 
968 	(void) callb_delete(drmach_cpr_cid);
969 	rw_destroy(&drmach_cpr_rwlock);
970 	rw_destroy(&drmach_boards_rwlock);
971 }
972 
973 sbd_error_t *
974 drmach_io_new(drmach_device_t *proto, drmachid_t *idp)
975 {
976 	drmach_io_t	*ip;
977 	int		portid;
978 
979 	portid = proto->portid;
980 	ASSERT(portid != -1);
981 	proto->unum = portid;
982 
983 	ip = kmem_zalloc(sizeof (drmach_io_t), KM_SLEEP);
984 	bcopy(proto, &ip->dev, sizeof (ip->dev));
985 	ip->dev.node = drmach_node_dup(proto->node);
986 	ip->dev.cm.isa = (void *)drmach_io_new;
987 	ip->dev.cm.dispose = drmach_io_dispose;
988 	ip->dev.cm.release = drmach_io_release;
989 	ip->dev.cm.status = drmach_io_status;
990 	(void) snprintf(ip->dev.cm.name, sizeof (ip->dev.cm.name), "%s%d",
991 	    ip->dev.type, ip->dev.unum);
992 
993 	*idp = (drmachid_t)ip;
994 
995 	return (NULL);
996 }
997 
998 static void
999 drmach_io_dispose(drmachid_t id)
1000 {
1001 	drmach_io_t *self;
1002 
1003 	ASSERT(DRMACH_IS_IO_ID(id));
1004 
1005 	self = id;
1006 	if (self->dev.node)
1007 		drmach_node_dispose(self->dev.node);
1008 
1009 	kmem_free(self, sizeof (*self));
1010 }
1011 
1012 static sbd_error_t *
1013 drmach_io_release(drmachid_t id)
1014 {
1015 	if (!DRMACH_IS_IO_ID(id))
1016 		return (drerr_new(0, EX86_INAPPROP, NULL));
1017 
1018 	return (NULL);
1019 }
1020 
1021 static sbd_error_t *
1022 drmach_io_status(drmachid_t id, drmach_status_t *stat)
1023 {
1024 	drmach_device_t *dp;
1025 	sbd_error_t	*err;
1026 	int		 configured;
1027 
1028 	ASSERT(DRMACH_IS_IO_ID(id));
1029 	dp = id;
1030 
1031 	err = drmach_io_is_attached(id, &configured);
1032 	if (err)
1033 		return (err);
1034 
1035 	stat->assigned = dp->bp->assigned;
1036 	stat->powered = dp->bp->powered;
1037 	stat->configured = (configured != 0);
1038 	stat->busy = dp->busy;
1039 	(void) strlcpy(stat->type, dp->type, sizeof (stat->type));
1040 	stat->info[0] = '\0';
1041 
1042 	return (NULL);
1043 }
1044 
1045 sbd_error_t *
1046 drmach_cpu_new(drmach_device_t *proto, drmachid_t *idp)
1047 {
1048 	int		 portid;
1049 	processorid_t	 cpuid;
1050 	drmach_cpu_t	*cp = NULL;
1051 
1052 	/* the portid is APIC ID of the node */
1053 	portid = proto->portid;
1054 	ASSERT(portid != -1);
1055 
1056 	/*
1057 	 * Assume all CPUs are homogeneous and have the same number of
1058 	 * cores/threads.
1059 	 */
1060 	proto->unum = portid % MAX_CPU_UNITS_PER_BOARD;
1061 
1062 	cp = kmem_zalloc(sizeof (drmach_cpu_t), KM_SLEEP);
1063 	bcopy(proto, &cp->dev, sizeof (cp->dev));
1064 	cp->dev.node = drmach_node_dup(proto->node);
1065 	cp->dev.cm.isa = (void *)drmach_cpu_new;
1066 	cp->dev.cm.dispose = drmach_cpu_dispose;
1067 	cp->dev.cm.release = drmach_cpu_release;
1068 	cp->dev.cm.status = drmach_cpu_status;
1069 	(void) snprintf(cp->dev.cm.name, sizeof (cp->dev.cm.name), "%s%d",
1070 	    cp->dev.type, cp->dev.unum);
1071 
1072 	cp->apicid = portid;
1073 	if (ACPI_SUCCESS(acpica_get_cpu_id_by_object(
1074 	    drmach_node_get_dnode(proto->node), &cpuid))) {
1075 		cp->cpuid = cpuid;
1076 	} else {
1077 		cp->cpuid = -1;
1078 	}
1079 
1080 	/* Mark CPU0 as busy, many other components have dependency on it. */
1081 	if (cp->cpuid == 0) {
1082 		cp->dev.busy = 1;
1083 	}
1084 
1085 	*idp = (drmachid_t)cp;
1086 
1087 	return (NULL);
1088 }
1089 
1090 static void
1091 drmach_cpu_dispose(drmachid_t id)
1092 {
1093 	drmach_cpu_t	*self;
1094 
1095 	ASSERT(DRMACH_IS_CPU_ID(id));
1096 
1097 	self = id;
1098 	if (self->dev.node)
1099 		drmach_node_dispose(self->dev.node);
1100 
1101 	kmem_free(self, sizeof (*self));
1102 }
1103 
1104 static sbd_error_t *
1105 drmach_cpu_release(drmachid_t id)
1106 {
1107 	if (!DRMACH_IS_CPU_ID(id))
1108 		return (drerr_new(0, EX86_INAPPROP, NULL));
1109 
1110 	return (NULL);
1111 }
1112 
1113 static sbd_error_t *
1114 drmach_cpu_status(drmachid_t id, drmach_status_t *stat)
1115 {
1116 	drmach_cpu_t *cp;
1117 	drmach_device_t *dp;
1118 
1119 	ASSERT(DRMACH_IS_CPU_ID(id));
1120 	cp = (drmach_cpu_t *)id;
1121 	dp = &cp->dev;
1122 
1123 	stat->assigned = dp->bp->assigned;
1124 	stat->powered = dp->bp->powered;
1125 	mutex_enter(&cpu_lock);
1126 	stat->configured = (cpu_get(cp->cpuid) != NULL);
1127 	mutex_exit(&cpu_lock);
1128 	stat->busy = dp->busy;
1129 	(void) strlcpy(stat->type, dp->type, sizeof (stat->type));
1130 	stat->info[0] = '\0';
1131 
1132 	return (NULL);
1133 }
1134 
1135 static int
1136 drmach_setup_mc_info(DRMACH_HANDLE hdl, drmach_mem_t *mp)
1137 {
1138 	uint_t i, j, count;
1139 	struct memlist	*ml = NULL, *ml2 = NULL;
1140 	acpidev_regspec_t *regp;
1141 	uint64_t align, addr_min, addr_max, total_size, skipped_size;
1142 
1143 	if (hdl == NULL) {
1144 		return (-1);
1145 	} else if (ACPI_FAILURE(acpidev_dr_get_mem_alignment(hdl, &align))) {
1146 		return (-1);
1147 	} else {
1148 		ASSERT((align & (align - 1)) == 0);
1149 		mp->mem_alignment = align;
1150 	}
1151 
1152 	addr_min = UINT64_MAX;
1153 	addr_max = 0;
1154 	total_size = 0;
1155 	skipped_size = 0;
1156 	/*
1157 	 * There's a memory hole just below 4G on x86, which needs special
1158 	 * handling. All other addresses assigned to a specific memory device
1159 	 * should be contiguous.
1160 	 */
1161 	if (ACPI_FAILURE(acpidev_dr_device_get_regspec(hdl, TRUE, &regp,
1162 	    &count))) {
1163 		return (-1);
1164 	}
1165 	for (i = 0, j = 0; i < count; i++) {
1166 		uint64_t	addr, size;
1167 
1168 		addr  = (uint64_t)regp[i].phys_mid << 32;
1169 		addr |= (uint64_t)regp[i].phys_low;
1170 		size  = (uint64_t)regp[i].size_hi << 32;
1171 		size |= (uint64_t)regp[i].size_low;
1172 		if (size == 0)
1173 			continue;
1174 		else
1175 			j++;
1176 
1177 		total_size += size;
1178 		if (addr < addr_min)
1179 			addr_min = addr;
1180 		if (addr + size > addr_max)
1181 			addr_max = addr + size;
1182 		if (mp->dev.bp->boot_board ||
1183 		    j <= acpidev_dr_max_segments_per_mem_device()) {
1184 			ml = memlist_add_span(ml, addr, size);
1185 		} else {
1186 			skipped_size += size;
1187 		}
1188 	}
1189 	acpidev_dr_device_free_regspec(regp, count);
1190 
1191 	if (skipped_size != 0) {
1192 		cmn_err(CE_WARN, "!drmach: too many (%d) segments on memory "
1193 		    "device, max (%d) segments supported, 0x%" PRIx64 " bytes "
1194 		    "of memory skipped.",
1195 		    j, acpidev_dr_max_segments_per_mem_device(), skipped_size);
1196 	}
1197 
1198 	mp->slice_base = addr_min;
1199 	mp->slice_top = addr_max;
1200 	mp->slice_size = total_size;
1201 
1202 	if (mp->dev.bp->boot_board) {
1203 		uint64_t endpa = _ptob64(physmax + 1);
1204 
1205 		/*
1206 		 * we intersect phys_install to get base_pa.
1207 		 * This only works at boot-up time.
1208 		 */
1209 		memlist_read_lock();
1210 		ml2 = memlist_dup(phys_install);
1211 		memlist_read_unlock();
1212 
1213 		ml2 = memlist_del_span(ml2, 0ull, mp->slice_base);
1214 		if (ml2 && endpa > addr_max) {
1215 			ml2 = memlist_del_span(ml2, addr_max, endpa - addr_max);
1216 		}
1217 	}
1218 
1219 	/*
1220 	 * Create a memlist for the memory board.
1221 	 * The created memlist only contains configured memory if there's
1222 	 * configured memory on the board, otherwise it contains all memory
1223 	 * on the board.
1224 	 */
1225 	if (ml2) {
1226 		uint64_t nbytes = 0;
1227 		struct memlist *p;
1228 
1229 		for (p = ml2; p; p = p->ml_next) {
1230 			nbytes += p->ml_size;
1231 		}
1232 		if (nbytes == 0) {
1233 			memlist_delete(ml2);
1234 			ml2 = NULL;
1235 		} else {
1236 			/* Node has configured memory at boot time. */
1237 			mp->base_pa = ml2->ml_address;
1238 			mp->nbytes = nbytes;
1239 			mp->memlist = ml2;
1240 			if (ml)
1241 				memlist_delete(ml);
1242 		}
1243 	}
1244 	if (ml2 == NULL) {
1245 		/* Not configured at boot time. */
1246 		mp->base_pa = UINT64_MAX;
1247 		mp->nbytes = 0;
1248 		mp->memlist = ml;
1249 	}
1250 
1251 	return (0);
1252 }
1253 
1254 sbd_error_t *
1255 drmach_mem_new(drmach_device_t *proto, drmachid_t *idp)
1256 {
1257 	DRMACH_HANDLE	hdl;
1258 	drmach_mem_t	*mp;
1259 	int		portid;
1260 
1261 	mp = kmem_zalloc(sizeof (drmach_mem_t), KM_SLEEP);
1262 	portid = proto->portid;
1263 	ASSERT(portid != -1);
1264 	proto->unum = portid;
1265 
1266 	bcopy(proto, &mp->dev, sizeof (mp->dev));
1267 	mp->dev.node = drmach_node_dup(proto->node);
1268 	mp->dev.cm.isa = (void *)drmach_mem_new;
1269 	mp->dev.cm.dispose = drmach_mem_dispose;
1270 	mp->dev.cm.release = drmach_mem_release;
1271 	mp->dev.cm.status = drmach_mem_status;
1272 
1273 	(void) snprintf(mp->dev.cm.name, sizeof (mp->dev.cm.name), "%s%d",
1274 	    mp->dev.type, proto->unum);
1275 	hdl = mp->dev.node->get_dnode(mp->dev.node);
1276 	ASSERT(hdl != NULL);
1277 	if (drmach_setup_mc_info(hdl, mp) != 0) {
1278 		kmem_free(mp, sizeof (drmach_mem_t));
1279 		*idp = (drmachid_t)NULL;
1280 		return (drerr_new(1, EX86_MC_SETUP, NULL));
1281 	}
1282 
1283 	/* make sure we do not create memoryless nodes */
1284 	if (mp->nbytes == 0 && mp->slice_size == 0) {
1285 		kmem_free(mp, sizeof (drmach_mem_t));
1286 		*idp = (drmachid_t)NULL;
1287 	} else
1288 		*idp = (drmachid_t)mp;
1289 
1290 	return (NULL);
1291 }
1292 
1293 static void
1294 drmach_mem_dispose(drmachid_t id)
1295 {
1296 	drmach_mem_t *mp;
1297 
1298 	ASSERT(DRMACH_IS_MEM_ID(id));
1299 
1300 	mp = id;
1301 
1302 	if (mp->dev.node)
1303 		drmach_node_dispose(mp->dev.node);
1304 
1305 	if (mp->memlist) {
1306 		memlist_delete(mp->memlist);
1307 		mp->memlist = NULL;
1308 	}
1309 
1310 	kmem_free(mp, sizeof (*mp));
1311 }
1312 
1313 static sbd_error_t *
1314 drmach_mem_release(drmachid_t id)
1315 {
1316 	if (!DRMACH_IS_MEM_ID(id))
1317 		return (drerr_new(0, EX86_INAPPROP, NULL));
1318 
1319 	return (NULL);
1320 }
1321 
1322 static sbd_error_t *
1323 drmach_mem_status(drmachid_t id, drmach_status_t *stat)
1324 {
1325 	uint64_t	 pa;
1326 	drmach_mem_t	*dp;
1327 	struct memlist	*ml = NULL;
1328 
1329 	ASSERT(DRMACH_IS_MEM_ID(id));
1330 	dp = id;
1331 
1332 	/* get starting physical address of target memory */
1333 	pa = dp->base_pa;
1334 	/* round down to slice boundary */
1335 	pa &= ~(dp->mem_alignment - 1);
1336 
1337 	/* stop at first span that is in slice */
1338 	memlist_read_lock();
1339 	for (ml = phys_install; ml; ml = ml->ml_next)
1340 		if (ml->ml_address >= pa && ml->ml_address < dp->slice_top)
1341 			break;
1342 	memlist_read_unlock();
1343 
1344 	stat->assigned = dp->dev.bp->assigned;
1345 	stat->powered = dp->dev.bp->powered;
1346 	stat->configured = (ml != NULL);
1347 	stat->busy = dp->dev.busy;
1348 	(void) strlcpy(stat->type, dp->dev.type, sizeof (stat->type));
1349 	stat->info[0] = '\0';
1350 
1351 	return (NULL);
1352 }
1353 
1354 /*
1355  * Public interfaces exported to support platform independent dr driver.
1356  */
1357 uint_t
1358 drmach_max_boards(void)
1359 {
1360 	return (acpidev_dr_max_boards());
1361 }
1362 
1363 uint_t
1364 drmach_max_io_units_per_board(void)
1365 {
1366 	return (acpidev_dr_max_io_units_per_board());
1367 }
1368 
1369 uint_t
1370 drmach_max_cmp_units_per_board(void)
1371 {
1372 	return (acpidev_dr_max_cmp_units_per_board());
1373 }
1374 
1375 uint_t
1376 drmach_max_mem_units_per_board(void)
1377 {
1378 	return (acpidev_dr_max_mem_units_per_board());
1379 }
1380 
1381 uint_t
1382 drmach_max_core_per_cmp(void)
1383 {
1384 	return (acpidev_dr_max_cpu_units_per_cmp());
1385 }
1386 
1387 sbd_error_t *
1388 drmach_pre_op(int cmd, drmachid_t id, drmach_opts_t *opts, void *argp)
1389 {
1390 	drmach_board_t	*bp = (drmach_board_t *)id;
1391 	sbd_error_t	*err = NULL;
1392 
1393 	/* allow status and ncm operations to always succeed */
1394 	if ((cmd == SBD_CMD_STATUS) || (cmd == SBD_CMD_GETNCM)) {
1395 		return (NULL);
1396 	}
1397 
1398 	switch (cmd) {
1399 	case SBD_CMD_POWERON:
1400 	case SBD_CMD_POWEROFF:
1401 		/*
1402 		 * Disable fast reboot if CPU/MEM/IOH hotplug event happens.
1403 		 * Note: this is a temporary solution and will be revised when
1404 		 * fast reboot can support CPU/MEM/IOH DR operations in future.
1405 		 *
1406 		 * ACPI BIOS generates some static ACPI tables, such as MADT,
1407 		 * SRAT and SLIT, to describe system hardware configuration on
1408 		 * power-on. When CPU/MEM/IOH hotplug event happens, those
1409 		 * static tables won't be updated and will become stale.
1410 		 *
1411 		 * If we reset system by fast reboot, BIOS will have no chance
1412 		 * to regenerate those staled static tables. Fast reboot can't
1413 		 * tolerate such inconsistency between staled ACPI tables and
1414 		 * real hardware configuration yet.
1415 		 *
1416 		 * A temporary solution is introduced to disable fast reboot if
1417 		 * CPU/MEM/IOH hotplug event happens. This solution should be
1418 		 * revised when fast reboot is enhanced to support CPU/MEM/IOH
1419 		 * DR operations.
1420 		 */
1421 		fastreboot_disable(FBNS_HOTPLUG);
1422 		/*FALLTHROUGH*/
1423 
1424 	default:
1425 		/* Block out the CPR thread. */
1426 		rw_enter(&drmach_cpr_rwlock, RW_READER);
1427 		break;
1428 	}
1429 
1430 	/* check all other commands for the required option string */
1431 	if ((opts->size > 0) && (opts->copts != NULL)) {
1432 		if (strstr(opts->copts, ACPIDEV_CMD_OST_PREFIX) == NULL) {
1433 			err = drerr_new(1, EX86_SUPPORT, NULL);
1434 		}
1435 	} else {
1436 		err = drerr_new(1, EX86_SUPPORT, NULL);
1437 	}
1438 
1439 	if (!err && id && DRMACH_IS_BOARD_ID(id)) {
1440 		switch (cmd) {
1441 		case SBD_CMD_TEST:
1442 			break;
1443 		case SBD_CMD_CONNECT:
1444 			if (bp->connected)
1445 				err = drerr_new(0, ESBD_STATE, NULL);
1446 			else if (!drmach_domain.allow_dr)
1447 				err = drerr_new(1, EX86_SUPPORT, NULL);
1448 			break;
1449 		case SBD_CMD_DISCONNECT:
1450 			if (!bp->connected)
1451 				err = drerr_new(0, ESBD_STATE, NULL);
1452 			else if (!drmach_domain.allow_dr)
1453 				err = drerr_new(1, EX86_SUPPORT, NULL);
1454 			break;
1455 		default:
1456 			if (!drmach_domain.allow_dr)
1457 				err = drerr_new(1, EX86_SUPPORT, NULL);
1458 			break;
1459 
1460 		}
1461 	}
1462 
1463 	/*
1464 	 * CPU/memory/IO DR operations will be supported in stages on x86.
1465 	 * With early versions, some operations should be blocked here.
1466 	 * This temporary hook will be removed when all CPU/memory/IO DR
1467 	 * operations are supported on x86 systems.
1468 	 *
1469 	 * We only need to filter unsupported device types for
1470 	 * SBD_CMD_CONFIGURE/SBD_CMD_UNCONFIGURE commands, all other
1471 	 * commands are supported by all device types.
1472 	 */
1473 	if (!err && (cmd == SBD_CMD_CONFIGURE || cmd == SBD_CMD_UNCONFIGURE)) {
1474 		int		i;
1475 		dr_devset_t	*devsetp = (dr_devset_t *)argp;
1476 		dr_devset_t	devset = *devsetp;
1477 
1478 		switch (cmd) {
1479 		case SBD_CMD_CONFIGURE:
1480 			if (!plat_dr_support_cpu()) {
1481 				DEVSET_DEL(devset, SBD_COMP_CPU,
1482 				    DEVSET_ANYUNIT);
1483 			} else {
1484 				for (i = MAX_CPU_UNITS_PER_BOARD;
1485 				    i < DEVSET_CPU_NUMBER; i++) {
1486 					DEVSET_DEL(devset, SBD_COMP_CPU, i);
1487 				}
1488 			}
1489 
1490 			if (!plat_dr_support_memory()) {
1491 				DEVSET_DEL(devset, SBD_COMP_MEM,
1492 				    DEVSET_ANYUNIT);
1493 			} else {
1494 				for (i = MAX_MEM_UNITS_PER_BOARD;
1495 				    i < DEVSET_MEM_NUMBER; i++) {
1496 					DEVSET_DEL(devset, SBD_COMP_MEM, i);
1497 				}
1498 			}
1499 
1500 			/* No support of configuring IOH devices yet. */
1501 			DEVSET_DEL(devset, SBD_COMP_IO, DEVSET_ANYUNIT);
1502 			break;
1503 
1504 		case SBD_CMD_UNCONFIGURE:
1505 			if (!plat_dr_support_cpu()) {
1506 				DEVSET_DEL(devset, SBD_COMP_CPU,
1507 				    DEVSET_ANYUNIT);
1508 			} else {
1509 				for (i = MAX_CPU_UNITS_PER_BOARD;
1510 				    i < DEVSET_CPU_NUMBER; i++) {
1511 					DEVSET_DEL(devset, SBD_COMP_CPU, i);
1512 				}
1513 			}
1514 
1515 			/* No support of unconfiguring MEM/IOH devices yet. */
1516 			DEVSET_DEL(devset, SBD_COMP_MEM, DEVSET_ANYUNIT);
1517 			DEVSET_DEL(devset, SBD_COMP_IO, DEVSET_ANYUNIT);
1518 			break;
1519 		}
1520 
1521 		*devsetp = devset;
1522 		if (DEVSET_IS_NULL(devset)) {
1523 			err = drerr_new(1, EX86_SUPPORT, NULL);
1524 		}
1525 	}
1526 
1527 	return (err);
1528 }
1529 
1530 sbd_error_t *
1531 drmach_post_op(int cmd, drmachid_t id, drmach_opts_t *opts, int rv)
1532 {
1533 	_NOTE(ARGUNUSED(id, opts, rv));
1534 
1535 	switch (cmd) {
1536 	case SBD_CMD_STATUS:
1537 	case SBD_CMD_GETNCM:
1538 		break;
1539 
1540 	default:
1541 		rw_exit(&drmach_cpr_rwlock);
1542 		break;
1543 	}
1544 
1545 	return (NULL);
1546 }
1547 
1548 sbd_error_t *
1549 drmach_configure(drmachid_t id, int flags)
1550 {
1551 	_NOTE(ARGUNUSED(flags));
1552 
1553 	drmach_device_t		*dp;
1554 	sbd_error_t		*err = NULL;
1555 	dev_info_t		*rdip;
1556 	dev_info_t		*fdip = NULL;
1557 
1558 	if (!DRMACH_IS_DEVICE_ID(id))
1559 		return (drerr_new(0, EX86_INAPPROP, NULL));
1560 	dp = id;
1561 
1562 	rdip = dp->node->getdip(dp->node);
1563 	ASSERT(rdip);
1564 	ASSERT(e_ddi_branch_held(rdip));
1565 
1566 	/* allocate cpu id for the CPU device. */
1567 	if (DRMACH_IS_CPU_ID(id)) {
1568 		DRMACH_HANDLE hdl = drmach_node_get_dnode(dp->node);
1569 		ASSERT(hdl != NULL);
1570 		if (ACPI_FAILURE(acpidev_dr_allocate_cpuid(hdl, NULL))) {
1571 			err = drerr_new(1, EX86_ALLOC_CPUID, NULL);
1572 		}
1573 		return (err);
1574 	}
1575 
1576 	if (DRMACH_IS_MEM_ID(id)) {
1577 		err = drmach_mem_update_lgrp(id);
1578 		if (err)
1579 			return (err);
1580 	}
1581 
1582 	if (e_ddi_branch_configure(rdip, &fdip, 0) != 0) {
1583 		char *path = kmem_alloc(MAXPATHLEN, KM_SLEEP);
1584 		dev_info_t *dip = (fdip != NULL) ? fdip : rdip;
1585 
1586 		(void) ddi_pathname(dip, path);
1587 		err = drerr_new(1, EX86_DRVFAIL, path);
1588 		kmem_free(path, MAXPATHLEN);
1589 
1590 		/* If non-NULL, fdip is returned held and must be released */
1591 		if (fdip != NULL)
1592 			ddi_release_devi(fdip);
1593 	}
1594 
1595 	return (err);
1596 }
1597 
1598 sbd_error_t *
1599 drmach_unconfigure(drmachid_t id, int flags)
1600 {
1601 	_NOTE(ARGUNUSED(flags));
1602 
1603 	drmach_device_t *dp;
1604 	sbd_error_t	*err = NULL;
1605 	dev_info_t	*rdip, *fdip = NULL;
1606 
1607 	if (!DRMACH_IS_DEVICE_ID(id))
1608 		return (drerr_new(0, EX86_INAPPROP, NULL));
1609 	dp = id;
1610 
1611 	rdip = dp->node->getdip(dp->node);
1612 	ASSERT(rdip);
1613 	ASSERT(e_ddi_branch_held(rdip));
1614 
1615 	if (DRMACH_IS_CPU_ID(id)) {
1616 		DRMACH_HANDLE hdl = drmach_node_get_dnode(dp->node);
1617 		ASSERT(hdl != NULL);
1618 		if (ACPI_FAILURE(acpidev_dr_free_cpuid(hdl))) {
1619 			err = drerr_new(1, EX86_FREE_CPUID, NULL);
1620 		}
1621 		return (err);
1622 	}
1623 
1624 	/*
1625 	 * Note: FORCE flag is no longer necessary under devfs
1626 	 */
1627 	if (e_ddi_branch_unconfigure(rdip, &fdip, 0)) {
1628 		char		*path = kmem_alloc(MAXPATHLEN, KM_SLEEP);
1629 
1630 		/*
1631 		 * If non-NULL, fdip is returned held and must be released.
1632 		 */
1633 		if (fdip != NULL) {
1634 			(void) ddi_pathname(fdip, path);
1635 			ndi_rele_devi(fdip);
1636 		} else {
1637 			(void) ddi_pathname(rdip, path);
1638 		}
1639 
1640 		err = drerr_new(1, EX86_DRVFAIL, path);
1641 
1642 		kmem_free(path, MAXPATHLEN);
1643 	}
1644 
1645 	return (err);
1646 }
1647 
1648 sbd_error_t *
1649 drmach_get_dip(drmachid_t id, dev_info_t **dip)
1650 {
1651 	drmach_device_t	*dp;
1652 
1653 	if (!DRMACH_IS_DEVICE_ID(id))
1654 		return (drerr_new(0, EX86_INAPPROP, NULL));
1655 	dp = id;
1656 
1657 	*dip = dp->node->getdip(dp->node);
1658 
1659 	return (NULL);
1660 }
1661 
1662 sbd_error_t *
1663 drmach_release(drmachid_t id)
1664 {
1665 	drmach_common_t *cp;
1666 
1667 	if (!DRMACH_IS_DEVICE_ID(id))
1668 		return (drerr_new(0, EX86_INAPPROP, NULL));
1669 	cp = id;
1670 
1671 	return (cp->release(id));
1672 }
1673 
1674 sbd_error_t *
1675 drmach_status(drmachid_t id, drmach_status_t *stat)
1676 {
1677 	drmach_common_t *cp;
1678 	sbd_error_t	*err;
1679 
1680 	rw_enter(&drmach_boards_rwlock, RW_READER);
1681 	if (!DRMACH_IS_ID(id)) {
1682 		rw_exit(&drmach_boards_rwlock);
1683 		return (drerr_new(0, EX86_NOTID, NULL));
1684 	}
1685 	cp = (drmach_common_t *)id;
1686 	err = cp->status(id, stat);
1687 	rw_exit(&drmach_boards_rwlock);
1688 
1689 	return (err);
1690 }
1691 
1692 static sbd_error_t *
1693 drmach_update_acpi_status(drmachid_t id, drmach_opts_t *opts)
1694 {
1695 	char		*copts;
1696 	drmach_board_t	*bp;
1697 	DRMACH_HANDLE	hdl;
1698 	int		event, code;
1699 	boolean_t	inprogress = B_FALSE;
1700 
1701 	if (DRMACH_NULL_ID(id) || !DRMACH_IS_BOARD_ID(id))
1702 		return (drerr_new(0, EX86_INAPPROP, NULL));
1703 	bp = (drmach_board_t *)id;
1704 	hdl = drmach_node_get_dnode(bp->tree);
1705 	ASSERT(hdl != NULL);
1706 	if (hdl == NULL)
1707 		return (drerr_new(0, EX86_INAPPROP, NULL));
1708 
1709 	/* Get the status code. */
1710 	copts = opts->copts;
1711 	if (strncmp(copts, ACPIDEV_CMD_OST_INPROGRESS,
1712 	    strlen(ACPIDEV_CMD_OST_INPROGRESS)) == 0) {
1713 		inprogress = B_TRUE;
1714 		code = ACPI_OST_STA_INSERT_IN_PROGRESS;
1715 		copts += strlen(ACPIDEV_CMD_OST_INPROGRESS);
1716 	} else if (strncmp(copts, ACPIDEV_CMD_OST_SUCCESS,
1717 	    strlen(ACPIDEV_CMD_OST_SUCCESS)) == 0) {
1718 		code = ACPI_OST_STA_SUCCESS;
1719 		copts += strlen(ACPIDEV_CMD_OST_SUCCESS);
1720 	} else if (strncmp(copts, ACPIDEV_CMD_OST_FAILURE,
1721 	    strlen(ACPIDEV_CMD_OST_FAILURE)) == 0) {
1722 		code = ACPI_OST_STA_FAILURE;
1723 		copts += strlen(ACPIDEV_CMD_OST_FAILURE);
1724 	} else if (strncmp(copts, ACPIDEV_CMD_OST_NOOP,
1725 	    strlen(ACPIDEV_CMD_OST_NOOP)) == 0) {
1726 		return (NULL);
1727 	} else {
1728 		return (drerr_new(0, EX86_UNKPTCMD, opts->copts));
1729 	}
1730 
1731 	/* Get the event type. */
1732 	copts = strstr(copts, ACPIDEV_EVENT_TYPE_ATTR_NAME);
1733 	if (copts == NULL) {
1734 		return (drerr_new(0, EX86_UNKPTCMD, opts->copts));
1735 	}
1736 	copts += strlen(ACPIDEV_EVENT_TYPE_ATTR_NAME);
1737 	if (copts[0] != '=') {
1738 		return (drerr_new(0, EX86_UNKPTCMD, opts->copts));
1739 	}
1740 	copts += strlen("=");
1741 	if (strncmp(copts, ACPIDEV_EVENT_TYPE_BUS_CHECK,
1742 	    strlen(ACPIDEV_EVENT_TYPE_BUS_CHECK)) == 0) {
1743 		event = ACPI_NOTIFY_BUS_CHECK;
1744 	} else if (strncmp(copts, ACPIDEV_EVENT_TYPE_DEVICE_CHECK,
1745 	    strlen(ACPIDEV_EVENT_TYPE_DEVICE_CHECK)) == 0) {
1746 		event = ACPI_NOTIFY_DEVICE_CHECK;
1747 	} else if (strncmp(copts, ACPIDEV_EVENT_TYPE_DEVICE_CHECK_LIGHT,
1748 	    strlen(ACPIDEV_EVENT_TYPE_DEVICE_CHECK_LIGHT)) == 0) {
1749 		event = ACPI_NOTIFY_DEVICE_CHECK_LIGHT;
1750 	} else if (strncmp(copts, ACPIDEV_EVENT_TYPE_EJECT_REQUEST,
1751 	    strlen(ACPIDEV_EVENT_TYPE_EJECT_REQUEST)) == 0) {
1752 		event = ACPI_NOTIFY_EJECT_REQUEST;
1753 		if (inprogress) {
1754 			code = ACPI_OST_STA_EJECT_IN_PROGRESS;
1755 		}
1756 	} else {
1757 		return (drerr_new(0, EX86_UNKPTCMD, opts->copts));
1758 	}
1759 
1760 	(void) acpidev_eval_ost(hdl, event, code, NULL, 0);
1761 
1762 	return (NULL);
1763 }
1764 
1765 static struct {
1766 	const char	*name;
1767 	sbd_error_t	*(*handler)(drmachid_t id, drmach_opts_t *opts);
1768 } drmach_pt_arr[] = {
1769 	{ ACPIDEV_CMD_OST_PREFIX,	&drmach_update_acpi_status	},
1770 	/* the following line must always be last */
1771 	{ NULL,				NULL				}
1772 };
1773 
1774 sbd_error_t *
1775 drmach_passthru(drmachid_t id, drmach_opts_t *opts)
1776 {
1777 	int		i;
1778 	sbd_error_t	*err;
1779 
1780 	i = 0;
1781 	while (drmach_pt_arr[i].name != NULL) {
1782 		int len = strlen(drmach_pt_arr[i].name);
1783 
1784 		if (strncmp(drmach_pt_arr[i].name, opts->copts, len) == 0)
1785 			break;
1786 
1787 		i += 1;
1788 	}
1789 
1790 	if (drmach_pt_arr[i].name == NULL)
1791 		err = drerr_new(0, EX86_UNKPTCMD, opts->copts);
1792 	else
1793 		err = (*drmach_pt_arr[i].handler)(id, opts);
1794 
1795 	return (err);
1796 }
1797 
1798 /*
1799  * Board specific interfaces to support dr driver
1800  */
1801 static int
1802 drmach_get_portid(drmach_node_t *np)
1803 {
1804 	uint32_t	portid;
1805 
1806 	if (np->getprop(np, ACPIDEV_DR_PROP_PORTID,
1807 	    &portid, sizeof (portid)) == 0) {
1808 		/*
1809 		 * acpidev returns portid as uint32_t, validates it.
1810 		 */
1811 		if (portid > INT_MAX) {
1812 			return (-1);
1813 		} else {
1814 			return (portid);
1815 		}
1816 	}
1817 
1818 	return (-1);
1819 }
1820 
1821 /*
1822  * This is a helper function to determine if a given
1823  * node should be considered for a dr operation according
1824  * to predefined dr type nodes and the node's name.
1825  * Formal Parameter : The name of a device node.
1826  * Return Value: -1, name does not map to a valid dr type.
1827  *		 A value greater or equal to 0, name is a valid dr type.
1828  */
1829 static int
1830 drmach_name2type_idx(char *name)
1831 {
1832 	int 	index, ntypes;
1833 
1834 	if (name == NULL)
1835 		return (-1);
1836 
1837 	/*
1838 	 * Determine how many possible types are currently supported
1839 	 * for dr.
1840 	 */
1841 	ntypes = sizeof (drmach_name2type) / sizeof (drmach_name2type[0]);
1842 
1843 	/* Determine if the node's name correspond to a predefined type. */
1844 	for (index = 0; index < ntypes; index++) {
1845 		if (strcmp(drmach_name2type[index].name, name) == 0)
1846 			/* The node is an allowed type for dr. */
1847 			return (index);
1848 	}
1849 
1850 	/*
1851 	 * If the name of the node does not map to any of the
1852 	 * types in the array drmach_name2type then the node is not of
1853 	 * interest to dr.
1854 	 */
1855 	return (-1);
1856 }
1857 
1858 static int
1859 drmach_board_find_devices_cb(drmach_node_walk_args_t *args)
1860 {
1861 	drmach_node_t			*node = args->node;
1862 	drmach_board_cb_data_t		*data = args->data;
1863 	drmach_board_t			*obj = data->obj;
1864 
1865 	int		rv, portid;
1866 	uint32_t	bnum;
1867 	drmachid_t	id;
1868 	drmach_device_t	*device;
1869 	char		name[OBP_MAXDRVNAME];
1870 
1871 	portid = drmach_get_portid(node);
1872 	rv = node->getprop(node, ACPIDEV_DR_PROP_DEVNAME,
1873 	    name, OBP_MAXDRVNAME);
1874 	if (rv)
1875 		return (0);
1876 
1877 	rv = node->getprop(node, ACPIDEV_DR_PROP_BOARDNUM,
1878 	    &bnum, sizeof (bnum));
1879 	if (rv) {
1880 		return (0);
1881 	}
1882 	if (bnum > INT_MAX) {
1883 		return (0);
1884 	}
1885 
1886 	if (bnum != obj->bnum)
1887 		return (0);
1888 
1889 	if (drmach_name2type_idx(name) < 0) {
1890 		return (0);
1891 	}
1892 
1893 	/*
1894 	 * Create a device data structure from this node data.
1895 	 * The call may yield nothing if the node is not of interest
1896 	 * to drmach.
1897 	 */
1898 	data->err = drmach_device_new(node, obj, portid, &id);
1899 	if (data->err)
1900 		return (-1);
1901 	else if (!id) {
1902 		/*
1903 		 * drmach_device_new examined the node we passed in
1904 		 * and determined that it was one not of interest to
1905 		 * drmach.  So, it is skipped.
1906 		 */
1907 		return (0);
1908 	}
1909 
1910 	rv = drmach_array_set(obj->devices, data->ndevs++, id);
1911 	if (rv) {
1912 		data->err = DRMACH_INTERNAL_ERROR();
1913 		return (-1);
1914 	}
1915 	device = id;
1916 
1917 	data->err = (*data->found)(data->a, device->type, device->unum, id);
1918 
1919 	return (data->err == NULL ? 0 : -1);
1920 }
1921 
1922 sbd_error_t *
1923 drmach_board_find_devices(drmachid_t id, void *a,
1924 	sbd_error_t *(*found)(void *a, const char *, int, drmachid_t))
1925 {
1926 	drmach_board_t		*bp = (drmach_board_t *)id;
1927 	sbd_error_t		*err;
1928 	int			 max_devices;
1929 	int			 rv;
1930 	drmach_board_cb_data_t	data;
1931 
1932 	if (!DRMACH_IS_BOARD_ID(id))
1933 		return (drerr_new(0, EX86_INAPPROP, NULL));
1934 
1935 	max_devices  = MAX_CPU_UNITS_PER_BOARD;
1936 	max_devices += MAX_MEM_UNITS_PER_BOARD;
1937 	max_devices += MAX_IO_UNITS_PER_BOARD;
1938 
1939 	if (bp->devices == NULL)
1940 		bp->devices = drmach_array_new(0, max_devices);
1941 	ASSERT(bp->tree != NULL);
1942 
1943 	data.obj = bp;
1944 	data.ndevs = 0;
1945 	data.found = found;
1946 	data.a = a;
1947 	data.err = NULL;
1948 
1949 	acpidev_dr_lock_all();
1950 	rv = drmach_node_walk(bp->tree, &data, drmach_board_find_devices_cb);
1951 	acpidev_dr_unlock_all();
1952 	if (rv == 0) {
1953 		err = NULL;
1954 	} else {
1955 		drmach_array_dispose(bp->devices, drmach_device_dispose);
1956 		bp->devices = NULL;
1957 
1958 		if (data.err)
1959 			err = data.err;
1960 		else
1961 			err = DRMACH_INTERNAL_ERROR();
1962 	}
1963 
1964 	return (err);
1965 }
1966 
1967 int
1968 drmach_board_lookup(int bnum, drmachid_t *id)
1969 {
1970 	int	rv = 0;
1971 
1972 	if (bnum < 0) {
1973 		*id = 0;
1974 		return (-1);
1975 	}
1976 
1977 	rw_enter(&drmach_boards_rwlock, RW_READER);
1978 	if (drmach_array_get(drmach_boards, (uint_t)bnum, id)) {
1979 		*id = 0;
1980 		rv = -1;
1981 	}
1982 	rw_exit(&drmach_boards_rwlock);
1983 
1984 	return (rv);
1985 }
1986 
1987 sbd_error_t *
1988 drmach_board_name(int bnum, char *buf, int buflen)
1989 {
1990 	ACPI_HANDLE hdl;
1991 	sbd_error_t *err = NULL;
1992 
1993 	if (bnum < 0) {
1994 		return (drerr_new(1, EX86_BNUM, "%d", bnum));
1995 	}
1996 
1997 	acpidev_dr_lock_all();
1998 	if (ACPI_FAILURE(acpidev_dr_get_board_handle(bnum, &hdl))) {
1999 		DRMACH_PR("!drmach_board_name: failed to lookup ACPI handle "
2000 		    "for board %d.", bnum);
2001 		err = drerr_new(1, EX86_BNUM, "%d", bnum);
2002 	} else if (ACPI_FAILURE(acpidev_dr_get_board_name(hdl, buf, buflen))) {
2003 		DRMACH_PR("!drmach_board_name: failed to generate board name "
2004 		    "for board %d.", bnum);
2005 		err = drerr_new(0, EX86_INVALID_ARG,
2006 		    ": buffer is too small for board name.");
2007 	}
2008 	acpidev_dr_unlock_all();
2009 
2010 	return (err);
2011 }
2012 
2013 int
2014 drmach_board_is_floating(drmachid_t id)
2015 {
2016 	drmach_board_t *bp;
2017 
2018 	if (!DRMACH_IS_BOARD_ID(id))
2019 		return (0);
2020 
2021 	bp = (drmach_board_t *)id;
2022 
2023 	return ((drmach_domain.floating & (1ULL << bp->bnum)) ? 1 : 0);
2024 }
2025 
2026 static ACPI_STATUS
2027 drmach_board_check_dependent_cb(ACPI_HANDLE hdl, UINT32 lvl, void *ctx,
2028     void **retval)
2029 {
2030 	uint32_t bdnum;
2031 	drmach_board_t *bp;
2032 	ACPI_STATUS rc = AE_OK;
2033 	int cmd = (int)(intptr_t)ctx;
2034 
2035 	ASSERT(hdl != NULL);
2036 	ASSERT(lvl == UINT32_MAX);
2037 	ASSERT(retval != NULL);
2038 
2039 	/* Skip non-board devices. */
2040 	if (!acpidev_dr_device_is_board(hdl)) {
2041 		return (AE_OK);
2042 	} else if (ACPI_FAILURE(acpidev_dr_get_board_number(hdl, &bdnum))) {
2043 		DRMACH_PR("!drmach_board_check_dependent_cb: failed to get "
2044 		    "board number for object %p.\n", hdl);
2045 		return (AE_ERROR);
2046 	} else if (bdnum > MAX_BOARDS) {
2047 		DRMACH_PR("!drmach_board_check_dependent_cb: board number %u "
2048 		    "is too big, max %u.", bdnum, MAX_BOARDS);
2049 		return (AE_ERROR);
2050 	}
2051 
2052 	bp = drmach_get_board_by_bnum(bdnum);
2053 	switch (cmd) {
2054 	case SBD_CMD_CONNECT:
2055 		/*
2056 		 * Its parent board should be present, assigned, powered and
2057 		 * connected when connecting the child board.
2058 		 */
2059 		if (bp == NULL) {
2060 			*retval = hdl;
2061 			rc = AE_ERROR;
2062 		} else {
2063 			bp->powered = acpidev_dr_device_is_powered(hdl);
2064 			if (!bp->connected || !bp->powered || !bp->assigned) {
2065 				*retval = hdl;
2066 				rc = AE_ERROR;
2067 			}
2068 		}
2069 		break;
2070 
2071 	case SBD_CMD_POWERON:
2072 		/*
2073 		 * Its parent board should be present, assigned and powered when
2074 		 * powering on the child board.
2075 		 */
2076 		if (bp == NULL) {
2077 			*retval = hdl;
2078 			rc = AE_ERROR;
2079 		} else {
2080 			bp->powered = acpidev_dr_device_is_powered(hdl);
2081 			if (!bp->powered || !bp->assigned) {
2082 				*retval = hdl;
2083 				rc = AE_ERROR;
2084 			}
2085 		}
2086 		break;
2087 
2088 	case SBD_CMD_ASSIGN:
2089 		/*
2090 		 * Its parent board should be present and assigned when
2091 		 * assigning the child board.
2092 		 */
2093 		if (bp == NULL) {
2094 			*retval = hdl;
2095 			rc = AE_ERROR;
2096 		} else if (!bp->assigned) {
2097 			*retval = hdl;
2098 			rc = AE_ERROR;
2099 		}
2100 		break;
2101 
2102 	case SBD_CMD_DISCONNECT:
2103 		/*
2104 		 * The child board should be disconnected if present when
2105 		 * disconnecting its parent board.
2106 		 */
2107 		if (bp != NULL && bp->connected) {
2108 			*retval = hdl;
2109 			rc = AE_ERROR;
2110 		}
2111 		break;
2112 
2113 	case SBD_CMD_POWEROFF:
2114 		/*
2115 		 * The child board should be disconnected and powered off if
2116 		 * present when powering off its parent board.
2117 		 */
2118 		if (bp != NULL) {
2119 			bp->powered = acpidev_dr_device_is_powered(hdl);
2120 			if (bp->connected || bp->powered) {
2121 				*retval = hdl;
2122 				rc = AE_ERROR;
2123 			}
2124 		}
2125 		break;
2126 
2127 	case SBD_CMD_UNASSIGN:
2128 		/*
2129 		 * The child board should be disconnected, powered off and
2130 		 * unassigned if present when unassigning its parent board.
2131 		 */
2132 		if (bp != NULL) {
2133 			bp->powered = acpidev_dr_device_is_powered(hdl);
2134 			if (bp->connected || bp->powered || bp->assigned) {
2135 				*retval = hdl;
2136 				rc = AE_ERROR;
2137 			}
2138 		}
2139 		break;
2140 
2141 	default:
2142 		/* Return success for all other commands. */
2143 		break;
2144 	}
2145 
2146 	return (rc);
2147 }
2148 
2149 sbd_error_t *
2150 drmach_board_check_dependent(int cmd, drmach_board_t *bp)
2151 {
2152 	int reverse;
2153 	char *name;
2154 	sbd_error_t *err = NULL;
2155 	DRMACH_HANDLE hdl;
2156 	DRMACH_HANDLE dp = NULL;
2157 
2158 	ASSERT(bp != NULL);
2159 	ASSERT(DRMACH_IS_BOARD_ID(bp));
2160 	ASSERT(RW_LOCK_HELD(&drmach_boards_rwlock));
2161 
2162 	hdl = drmach_node_get_dnode(bp->tree);
2163 	if (hdl == NULL)
2164 		return (drerr_new(0, EX86_INAPPROP, NULL));
2165 
2166 	switch (cmd) {
2167 	case SBD_CMD_ASSIGN:
2168 	case SBD_CMD_POWERON:
2169 	case SBD_CMD_CONNECT:
2170 		if (ACPI_SUCCESS(acpidev_dr_device_walk_ejd(hdl,
2171 		    &drmach_board_check_dependent_cb,
2172 		    (void *)(intptr_t)cmd, &dp))) {
2173 			return (NULL);
2174 		}
2175 		reverse = 0;
2176 		break;
2177 
2178 	case SBD_CMD_UNASSIGN:
2179 	case SBD_CMD_POWEROFF:
2180 	case SBD_CMD_DISCONNECT:
2181 		if (ACPI_SUCCESS(acpidev_dr_device_walk_edl(hdl,
2182 		    &drmach_board_check_dependent_cb,
2183 		    (void *)(intptr_t)cmd, &dp))) {
2184 			return (NULL);
2185 		}
2186 		reverse = 1;
2187 		break;
2188 
2189 	default:
2190 		return (drerr_new(0, EX86_INAPPROP, NULL));
2191 	}
2192 
2193 	if (dp == NULL) {
2194 		return (drerr_new(1, EX86_WALK_DEPENDENCY, "%s", bp->cm.name));
2195 	}
2196 	name = kmem_zalloc(MAXPATHLEN, KM_SLEEP);
2197 	if (ACPI_FAILURE(acpidev_dr_get_board_name(dp, name, MAXPATHLEN))) {
2198 		err = drerr_new(1, EX86_WALK_DEPENDENCY, "%s", bp->cm.name);
2199 	} else if (reverse == 0) {
2200 		err = drerr_new(1, EX86_WALK_DEPENDENCY,
2201 		    "%s, depends on board %s", bp->cm.name, name);
2202 	} else {
2203 		err = drerr_new(1, EX86_WALK_DEPENDENCY,
2204 		    "board %s depends on %s", name, bp->cm.name);
2205 	}
2206 	kmem_free(name, MAXPATHLEN);
2207 
2208 	return (err);
2209 }
2210 
2211 sbd_error_t *
2212 drmach_board_assign(int bnum, drmachid_t *id)
2213 {
2214 	sbd_error_t	*err = NULL;
2215 
2216 	if (bnum < 0) {
2217 		return (drerr_new(1, EX86_BNUM, "%d", bnum));
2218 	}
2219 
2220 	rw_enter(&drmach_boards_rwlock, RW_WRITER);
2221 
2222 	if (drmach_array_get(drmach_boards, bnum, id) == -1) {
2223 		err = drerr_new(1, EX86_BNUM, "%d", bnum);
2224 	} else {
2225 		drmach_board_t	*bp;
2226 
2227 		/*
2228 		 * Board has already been created, downgrade to reader.
2229 		 */
2230 		if (*id)
2231 			rw_downgrade(&drmach_boards_rwlock);
2232 
2233 		bp = *id;
2234 		if (!(*id))
2235 			bp = *id  =
2236 			    (drmachid_t)drmach_board_new(bnum, 0);
2237 
2238 		if (bp == NULL) {
2239 			DRMACH_PR("!drmach_board_assign: failed to create "
2240 			    "object for board %d.", bnum);
2241 			err = drerr_new(1, EX86_BNUM, "%d", bnum);
2242 		} else {
2243 			err = drmach_board_check_dependent(SBD_CMD_ASSIGN, bp);
2244 			if (err == NULL)
2245 				bp->assigned = 1;
2246 		}
2247 	}
2248 
2249 	rw_exit(&drmach_boards_rwlock);
2250 
2251 	return (err);
2252 }
2253 
2254 sbd_error_t *
2255 drmach_board_unassign(drmachid_t id)
2256 {
2257 	drmach_board_t	*bp;
2258 	sbd_error_t	*err;
2259 	drmach_status_t	 stat;
2260 
2261 	if (DRMACH_NULL_ID(id))
2262 		return (NULL);
2263 
2264 	if (!DRMACH_IS_BOARD_ID(id)) {
2265 		return (drerr_new(0, EX86_INAPPROP, NULL));
2266 	}
2267 	bp = id;
2268 
2269 	rw_enter(&drmach_boards_rwlock, RW_WRITER);
2270 
2271 	err = drmach_board_status(id, &stat);
2272 	if (err) {
2273 		rw_exit(&drmach_boards_rwlock);
2274 		return (err);
2275 	}
2276 
2277 	if (stat.configured || stat.busy) {
2278 		err = drerr_new(0, EX86_CONFIGBUSY, bp->cm.name);
2279 	} else if (bp->connected) {
2280 		err = drerr_new(0, EX86_CONNECTBUSY, bp->cm.name);
2281 	} else if (stat.powered) {
2282 		err = drerr_new(0, EX86_POWERBUSY, bp->cm.name);
2283 	} else {
2284 		err = drmach_board_check_dependent(SBD_CMD_UNASSIGN, bp);
2285 		if (err == NULL) {
2286 			if (drmach_array_set(drmach_boards, bp->bnum, 0) != 0)
2287 				err = DRMACH_INTERNAL_ERROR();
2288 			else
2289 				drmach_board_dispose(bp);
2290 		}
2291 	}
2292 
2293 	rw_exit(&drmach_boards_rwlock);
2294 
2295 	return (err);
2296 }
2297 
2298 sbd_error_t *
2299 drmach_board_poweron(drmachid_t id)
2300 {
2301 	drmach_board_t	*bp;
2302 	sbd_error_t *err = NULL;
2303 	DRMACH_HANDLE hdl;
2304 
2305 	if (!DRMACH_IS_BOARD_ID(id))
2306 		return (drerr_new(0, EX86_INAPPROP, NULL));
2307 	bp = id;
2308 
2309 	hdl = drmach_node_get_dnode(bp->tree);
2310 	if (hdl == NULL)
2311 		return (drerr_new(0, EX86_INAPPROP, NULL));
2312 
2313 	bp->powered = drmach_board_check_power(bp);
2314 	if (bp->powered) {
2315 		return (NULL);
2316 	}
2317 
2318 	rw_enter(&drmach_boards_rwlock, RW_WRITER);
2319 	err = drmach_board_check_dependent(SBD_CMD_POWERON, bp);
2320 	if (err == NULL) {
2321 		acpidev_dr_lock_all();
2322 		if (ACPI_FAILURE(acpidev_dr_device_poweron(hdl)))
2323 			err = drerr_new(0, EX86_POWERON, NULL);
2324 		acpidev_dr_unlock_all();
2325 
2326 		/* Check whether the board is powered on. */
2327 		bp->powered = drmach_board_check_power(bp);
2328 		if (err == NULL && bp->powered == 0)
2329 			err = drerr_new(0, EX86_POWERON, NULL);
2330 	}
2331 	rw_exit(&drmach_boards_rwlock);
2332 
2333 	return (err);
2334 }
2335 
2336 sbd_error_t *
2337 drmach_board_poweroff(drmachid_t id)
2338 {
2339 	sbd_error_t	*err = NULL;
2340 	drmach_board_t	*bp;
2341 	drmach_status_t	 stat;
2342 	DRMACH_HANDLE	 hdl;
2343 
2344 	if (DRMACH_NULL_ID(id))
2345 		return (NULL);
2346 
2347 	if (!DRMACH_IS_BOARD_ID(id))
2348 		return (drerr_new(0, EX86_INAPPROP, NULL));
2349 	bp = id;
2350 
2351 	hdl = drmach_node_get_dnode(bp->tree);
2352 	if (hdl == NULL)
2353 		return (drerr_new(0, EX86_INAPPROP, NULL));
2354 
2355 	/* Check whether the board is busy, configured or connected. */
2356 	err = drmach_board_status(id, &stat);
2357 	if (err != NULL)
2358 		return (err);
2359 	if (stat.configured || stat.busy) {
2360 		return (drerr_new(0, EX86_CONFIGBUSY, bp->cm.name));
2361 	} else if (bp->connected) {
2362 		return (drerr_new(0, EX86_CONNECTBUSY, bp->cm.name));
2363 	}
2364 
2365 	bp->powered = drmach_board_check_power(bp);
2366 	if (bp->powered == 0) {
2367 		return (NULL);
2368 	}
2369 
2370 	rw_enter(&drmach_boards_rwlock, RW_WRITER);
2371 	err = drmach_board_check_dependent(SBD_CMD_POWEROFF, bp);
2372 	if (err == NULL) {
2373 		acpidev_dr_lock_all();
2374 		if (ACPI_FAILURE(acpidev_dr_device_poweroff(hdl)))
2375 			err = drerr_new(0, EX86_POWEROFF, NULL);
2376 		acpidev_dr_unlock_all();
2377 
2378 		bp->powered = drmach_board_check_power(bp);
2379 		if (err == NULL && bp->powered != 0)
2380 			err = drerr_new(0, EX86_POWEROFF, NULL);
2381 	}
2382 	rw_exit(&drmach_boards_rwlock);
2383 
2384 	return (err);
2385 }
2386 
2387 sbd_error_t *
2388 drmach_board_test(drmachid_t id, drmach_opts_t *opts, int force)
2389 {
2390 	_NOTE(ARGUNUSED(opts, force));
2391 
2392 	drmach_board_t	*bp;
2393 	DRMACH_HANDLE	 hdl;
2394 
2395 	if (DRMACH_NULL_ID(id))
2396 		return (NULL);
2397 
2398 	if (!DRMACH_IS_BOARD_ID(id))
2399 		return (drerr_new(0, EX86_INAPPROP, NULL));
2400 	bp = id;
2401 
2402 	hdl = drmach_node_get_dnode(bp->tree);
2403 	if (hdl == NULL)
2404 		return (drerr_new(0, EX86_INAPPROP, NULL));
2405 
2406 	if (ACPI_FAILURE(acpidev_dr_device_check_status(hdl)))
2407 		return (drerr_new(0, EX86_IN_FAILURE, NULL));
2408 
2409 	return (NULL);
2410 }
2411 
2412 sbd_error_t *
2413 drmach_board_connect(drmachid_t id, drmach_opts_t *opts)
2414 {
2415 	_NOTE(ARGUNUSED(opts));
2416 
2417 	sbd_error_t	*err = NULL;
2418 	drmach_board_t	*bp = (drmach_board_t *)id;
2419 	DRMACH_HANDLE	hdl;
2420 
2421 	if (!DRMACH_IS_BOARD_ID(id))
2422 		return (drerr_new(0, EX86_INAPPROP, NULL));
2423 	bp = (drmach_board_t *)id;
2424 
2425 	hdl = drmach_node_get_dnode(bp->tree);
2426 	if (hdl == NULL)
2427 		return (drerr_new(0, EX86_INAPPROP, NULL));
2428 
2429 	rw_enter(&drmach_boards_rwlock, RW_WRITER);
2430 	err = drmach_board_check_dependent(SBD_CMD_CONNECT, bp);
2431 	if (err == NULL) {
2432 		acpidev_dr_lock_all();
2433 		if (ACPI_FAILURE(acpidev_dr_device_insert(hdl))) {
2434 			(void) acpidev_dr_device_remove(hdl);
2435 			err = drerr_new(1, EX86_PROBE, NULL);
2436 		} else {
2437 			bp->connected = 1;
2438 		}
2439 		acpidev_dr_unlock_all();
2440 	}
2441 	rw_exit(&drmach_boards_rwlock);
2442 
2443 	return (err);
2444 }
2445 
2446 sbd_error_t *
2447 drmach_board_disconnect(drmachid_t id, drmach_opts_t *opts)
2448 {
2449 	_NOTE(ARGUNUSED(opts));
2450 
2451 	DRMACH_HANDLE hdl;
2452 	drmach_board_t *bp;
2453 	drmach_status_t	stat;
2454 	sbd_error_t *err = NULL;
2455 
2456 	if (DRMACH_NULL_ID(id))
2457 		return (NULL);
2458 	if (!DRMACH_IS_BOARD_ID(id))
2459 		return (drerr_new(0, EX86_INAPPROP, NULL));
2460 	bp = (drmach_board_t *)id;
2461 
2462 	hdl = drmach_node_get_dnode(bp->tree);
2463 	if (hdl == NULL)
2464 		return (drerr_new(0, EX86_INAPPROP, NULL));
2465 
2466 	/* Check whether the board is busy or configured. */
2467 	err = drmach_board_status(id, &stat);
2468 	if (err != NULL)
2469 		return (err);
2470 	if (stat.configured || stat.busy)
2471 		return (drerr_new(0, EX86_CONFIGBUSY, bp->cm.name));
2472 
2473 	rw_enter(&drmach_boards_rwlock, RW_WRITER);
2474 	err = drmach_board_check_dependent(SBD_CMD_DISCONNECT, bp);
2475 	if (err == NULL) {
2476 		acpidev_dr_lock_all();
2477 		if (ACPI_SUCCESS(acpidev_dr_device_remove(hdl))) {
2478 			bp->connected = 0;
2479 		} else {
2480 			err = drerr_new(1, EX86_DEPROBE, bp->cm.name);
2481 		}
2482 		acpidev_dr_unlock_all();
2483 	}
2484 	rw_exit(&drmach_boards_rwlock);
2485 
2486 	return (err);
2487 }
2488 
2489 sbd_error_t *
2490 drmach_board_deprobe(drmachid_t id)
2491 {
2492 	drmach_board_t	*bp;
2493 
2494 	if (!DRMACH_IS_BOARD_ID(id))
2495 		return (drerr_new(0, EX86_INAPPROP, NULL));
2496 	bp = id;
2497 
2498 	cmn_err(CE_CONT, "DR: detach board %d\n", bp->bnum);
2499 
2500 	if (bp->devices) {
2501 		drmach_array_dispose(bp->devices, drmach_device_dispose);
2502 		bp->devices = NULL;
2503 	}
2504 
2505 	bp->boot_board = 0;
2506 
2507 	return (NULL);
2508 }
2509 
2510 /*
2511  * CPU specific interfaces to support dr driver
2512  */
2513 sbd_error_t *
2514 drmach_cpu_disconnect(drmachid_t id)
2515 {
2516 	if (!DRMACH_IS_CPU_ID(id))
2517 		return (drerr_new(0, EX86_INAPPROP, NULL));
2518 
2519 	return (NULL);
2520 }
2521 
2522 sbd_error_t *
2523 drmach_cpu_get_id(drmachid_t id, processorid_t *cpuid)
2524 {
2525 	drmach_cpu_t *cpu;
2526 
2527 	if (!DRMACH_IS_CPU_ID(id))
2528 		return (drerr_new(0, EX86_INAPPROP, NULL));
2529 	cpu = (drmach_cpu_t *)id;
2530 
2531 	if (cpu->cpuid == -1) {
2532 		if (ACPI_SUCCESS(acpica_get_cpu_id_by_object(
2533 		    drmach_node_get_dnode(cpu->dev.node), cpuid))) {
2534 			cpu->cpuid = *cpuid;
2535 		} else {
2536 			*cpuid = -1;
2537 		}
2538 	} else {
2539 		*cpuid = cpu->cpuid;
2540 	}
2541 
2542 	return (NULL);
2543 }
2544 
2545 sbd_error_t *
2546 drmach_cpu_get_impl(drmachid_t id, int *ip)
2547 {
2548 	if (!DRMACH_IS_CPU_ID(id))
2549 		return (drerr_new(0, EX86_INAPPROP, NULL));
2550 
2551 	/* Assume all CPUs in system are homogeneous. */
2552 	*ip = X86_CPU_IMPL_UNKNOWN;
2553 
2554 	kpreempt_disable();
2555 	if (cpuid_getvendor(CPU) == X86_VENDOR_Intel) {
2556 		/* NHM-EX CPU */
2557 		if (cpuid_getfamily(CPU) == 0x6 &&
2558 		    cpuid_getmodel(CPU) == 0x2e) {
2559 			*ip = X86_CPU_IMPL_NEHALEM_EX;
2560 		}
2561 	}
2562 	kpreempt_enable();
2563 
2564 	return (NULL);
2565 }
2566 
2567 /*
2568  * Memory specific interfaces to support dr driver
2569  */
2570 
2571 /*
2572  * When drmach_mem_new() is called, the mp->base_pa field is set to the base
2573  * address of configured memory if there's configured memory on the board,
2574  * otherwise set to UINT64_MAX. For hot-added memory board, there's no
2575  * configured memory when drmach_mem_new() is called, so mp->base_pa is set
2576  * to UINT64_MAX and we need to set a correct value for it after memory
2577  * hot-add  operations.
2578  * A hot-added memory board may contain multiple memory segments,
2579  * drmach_mem_add_span() will be called once for each segment, so we can't
2580  * rely on the basepa argument. And it's possible that only part of a memory
2581  * segment is added into OS, so need to intersect with phys_installed list
2582  * to get the real base address of configured memory on the board.
2583  */
2584 sbd_error_t *
2585 drmach_mem_add_span(drmachid_t id, uint64_t basepa, uint64_t size)
2586 {
2587 	_NOTE(ARGUNUSED(basepa));
2588 
2589 	uint64_t	nbytes = 0;
2590 	uint64_t	endpa;
2591 	drmach_mem_t	*mp;
2592 	struct memlist	*ml2;
2593 	struct memlist	*p;
2594 
2595 	ASSERT(size != 0);
2596 
2597 	if (!DRMACH_IS_MEM_ID(id))
2598 		return (drerr_new(0, EX86_INAPPROP, NULL));
2599 	mp = (drmach_mem_t *)id;
2600 
2601 	/* Compute basepa and size of installed memory. */
2602 	endpa = _ptob64(physmax + 1);
2603 	memlist_read_lock();
2604 	ml2 = memlist_dup(phys_install);
2605 	memlist_read_unlock();
2606 	ml2 = memlist_del_span(ml2, 0ull, mp->slice_base);
2607 	if (ml2 && endpa > mp->slice_top) {
2608 		ml2 = memlist_del_span(ml2, mp->slice_top,
2609 		    endpa - mp->slice_top);
2610 	}
2611 
2612 	ASSERT(ml2);
2613 	if (ml2) {
2614 		for (p = ml2; p; p = p->ml_next) {
2615 			nbytes += p->ml_size;
2616 			if (mp->base_pa > p->ml_address)
2617 				mp->base_pa = p->ml_address;
2618 		}
2619 		ASSERT(nbytes > 0);
2620 		mp->nbytes += nbytes;
2621 		memlist_delete(ml2);
2622 	}
2623 
2624 	return (NULL);
2625 }
2626 
2627 static sbd_error_t *
2628 drmach_mem_update_lgrp(drmachid_t id)
2629 {
2630 	ACPI_STATUS	rc;
2631 	DRMACH_HANDLE	hdl;
2632 	void		*hdlp;
2633 	drmach_mem_t	*mp;
2634 	update_membounds_t umb;
2635 
2636 	if (!DRMACH_IS_MEM_ID(id))
2637 		return (drerr_new(0, EX86_INAPPROP, NULL));
2638 	mp = (drmach_mem_t *)id;
2639 	/* No need to update lgrp if memory is already installed. */
2640 	if (mp->nbytes != 0)
2641 		return (NULL);
2642 	/* No need to update lgrp if lgrp is disabled. */
2643 	if (max_mem_nodes == 1)
2644 		return (NULL);
2645 
2646 	/* Add memory to lgroup */
2647 	hdl = mp->dev.node->get_dnode(mp->dev.node);
2648 	rc = acpidev_dr_device_get_memory_index(hdl, &umb.u_device_id);
2649 	ASSERT(ACPI_SUCCESS(rc));
2650 	if (ACPI_FAILURE(rc)) {
2651 		cmn_err(CE_WARN, "drmach: failed to get device id of memory, "
2652 		    "can't update lgrp information.");
2653 		return (drerr_new(0, EX86_INTERNAL, NULL));
2654 	}
2655 	rc = acpidev_dr_get_mem_numa_info(hdl, mp->memlist, &hdlp,
2656 	    &umb.u_domain, &umb.u_sli_cnt, &umb.u_sli_ptr);
2657 	ASSERT(ACPI_SUCCESS(rc));
2658 	if (ACPI_FAILURE(rc)) {
2659 		cmn_err(CE_WARN, "drmach: failed to get lgrp info of memory, "
2660 		    "can't update lgrp information.");
2661 		return (drerr_new(0, EX86_INTERNAL, NULL));
2662 	}
2663 	umb.u_base = (uint64_t)mp->slice_base;
2664 	umb.u_length = (uint64_t)(mp->slice_top - mp->slice_base);
2665 	lgrp_plat_config(LGRP_CONFIG_MEM_ADD, (uintptr_t)&umb);
2666 	acpidev_dr_free_mem_numa_info(hdlp);
2667 
2668 	return (NULL);
2669 }
2670 
2671 sbd_error_t *
2672 drmach_mem_enable(drmachid_t id)
2673 {
2674 	if (!DRMACH_IS_MEM_ID(id))
2675 		return (drerr_new(0, EX86_INAPPROP, NULL));
2676 	else
2677 		return (NULL);
2678 }
2679 
2680 sbd_error_t *
2681 drmach_mem_get_info(drmachid_t id, drmach_mem_info_t *mem)
2682 {
2683 	drmach_mem_t *mp;
2684 
2685 	if (!DRMACH_IS_MEM_ID(id))
2686 		return (drerr_new(0, EX86_INAPPROP, NULL));
2687 	mp = (drmach_mem_t *)id;
2688 
2689 	/*
2690 	 * This is only used by dr to round up/down the memory
2691 	 * for copying.
2692 	 */
2693 	mem->mi_alignment_mask = mp->mem_alignment - 1;
2694 	mem->mi_basepa = mp->base_pa;
2695 	mem->mi_size = mp->nbytes;
2696 	mem->mi_slice_base = mp->slice_base;
2697 	mem->mi_slice_top = mp->slice_top;
2698 	mem->mi_slice_size = mp->slice_size;
2699 
2700 	return (NULL);
2701 }
2702 
2703 sbd_error_t *
2704 drmach_mem_get_slice_info(drmachid_t id,
2705     uint64_t *bp, uint64_t *ep, uint64_t *sp)
2706 {
2707 	drmach_mem_t *mp;
2708 
2709 	if (!DRMACH_IS_MEM_ID(id))
2710 		return (drerr_new(0, EX86_INAPPROP, NULL));
2711 	mp = (drmach_mem_t *)id;
2712 
2713 	if (bp)
2714 		*bp = mp->slice_base;
2715 	if (ep)
2716 		*ep = mp->slice_top;
2717 	if (sp)
2718 		*sp = mp->slice_size;
2719 
2720 	return (NULL);
2721 }
2722 
2723 sbd_error_t *
2724 drmach_mem_get_memlist(drmachid_t id, struct memlist **ml)
2725 {
2726 #ifdef	DEBUG
2727 	int		rv;
2728 #endif
2729 	drmach_mem_t	*mem;
2730 	struct memlist	*mlist;
2731 
2732 	if (!DRMACH_IS_MEM_ID(id))
2733 		return (drerr_new(0, EX86_INAPPROP, NULL));
2734 	mem = (drmach_mem_t *)id;
2735 
2736 	mlist = memlist_dup(mem->memlist);
2737 	*ml = mlist;
2738 
2739 #ifdef DEBUG
2740 	/*
2741 	 * Make sure the incoming memlist doesn't already
2742 	 * intersect with what's present in the system (phys_install).
2743 	 */
2744 	memlist_read_lock();
2745 	rv = memlist_intersect(phys_install, mlist);
2746 	memlist_read_unlock();
2747 	if (rv) {
2748 		DRMACH_PR("Derived memlist intersects with phys_install\n");
2749 		memlist_dump(mlist);
2750 
2751 		DRMACH_PR("phys_install memlist:\n");
2752 		memlist_dump(phys_install);
2753 
2754 		memlist_delete(mlist);
2755 		return (DRMACH_INTERNAL_ERROR());
2756 	}
2757 
2758 	DRMACH_PR("Derived memlist:");
2759 	memlist_dump(mlist);
2760 #endif
2761 
2762 	return (NULL);
2763 }
2764 
2765 processorid_t
2766 drmach_mem_cpu_affinity(drmachid_t id)
2767 {
2768 	_NOTE(ARGUNUSED(id));
2769 
2770 	return (CPU_CURRENT);
2771 }
2772 
2773 int
2774 drmach_copy_rename_need_suspend(drmachid_t id)
2775 {
2776 	_NOTE(ARGUNUSED(id));
2777 
2778 	return (0);
2779 }
2780 
2781 /*
2782  * IO specific interfaces to support dr driver
2783  */
2784 sbd_error_t *
2785 drmach_io_pre_release(drmachid_t id)
2786 {
2787 	if (!DRMACH_IS_IO_ID(id))
2788 		return (drerr_new(0, EX86_INAPPROP, NULL));
2789 
2790 	return (NULL);
2791 }
2792 
2793 sbd_error_t *
2794 drmach_io_unrelease(drmachid_t id)
2795 {
2796 	if (!DRMACH_IS_IO_ID(id))
2797 		return (drerr_new(0, EX86_INAPPROP, NULL));
2798 
2799 	return (NULL);
2800 }
2801 
2802 sbd_error_t *
2803 drmach_io_post_release(drmachid_t id)
2804 {
2805 	_NOTE(ARGUNUSED(id));
2806 
2807 	return (NULL);
2808 }
2809 
2810 sbd_error_t *
2811 drmach_io_post_attach(drmachid_t id)
2812 {
2813 	if (!DRMACH_IS_IO_ID(id))
2814 		return (drerr_new(0, EX86_INAPPROP, NULL));
2815 
2816 	return (NULL);
2817 }
2818 
2819 sbd_error_t *
2820 drmach_io_is_attached(drmachid_t id, int *yes)
2821 {
2822 	drmach_device_t *dp;
2823 	dev_info_t	*dip;
2824 	int		state;
2825 
2826 	if (!DRMACH_IS_IO_ID(id))
2827 		return (drerr_new(0, EX86_INAPPROP, NULL));
2828 	dp = id;
2829 
2830 	dip = dp->node->getdip(dp->node);
2831 	if (dip == NULL) {
2832 		*yes = 0;
2833 		return (NULL);
2834 	}
2835 
2836 	state = ddi_get_devstate(dip);
2837 	*yes = ((i_ddi_node_state(dip) >= DS_ATTACHED) ||
2838 	    (state == DDI_DEVSTATE_UP));
2839 
2840 	return (NULL);
2841 }
2842 
2843 /*
2844  * Miscellaneous interfaces to support dr driver
2845  */
2846 int
2847 drmach_verify_sr(dev_info_t *dip, int sflag)
2848 {
2849 	_NOTE(ARGUNUSED(dip, sflag));
2850 
2851 	return (0);
2852 }
2853 
2854 void
2855 drmach_suspend_last(void)
2856 {
2857 }
2858 
2859 void
2860 drmach_resume_first(void)
2861 {
2862 }
2863 
2864 /*
2865  * Log a DR sysevent.
2866  * Return value: 0 success, non-zero failure.
2867  */
2868 int
2869 drmach_log_sysevent(int board, char *hint, int flag, int verbose)
2870 {
2871 	sysevent_t			*ev = NULL;
2872 	sysevent_id_t			eid;
2873 	int				rv, km_flag;
2874 	sysevent_value_t		evnt_val;
2875 	sysevent_attr_list_t		*evnt_attr_list = NULL;
2876 	sbd_error_t			*err;
2877 	char				attach_pnt[MAXNAMELEN];
2878 
2879 	km_flag = (flag == SE_SLEEP) ? KM_SLEEP : KM_NOSLEEP;
2880 	attach_pnt[0] = '\0';
2881 	err = drmach_board_name(board, attach_pnt, MAXNAMELEN);
2882 	if (err != NULL) {
2883 		sbd_err_clear(&err);
2884 		rv = -1;
2885 		goto logexit;
2886 	}
2887 	if (verbose) {
2888 		DRMACH_PR("drmach_log_sysevent: %s %s, flag: %d, verbose: %d\n",
2889 		    attach_pnt, hint, flag, verbose);
2890 	}
2891 
2892 	if ((ev = sysevent_alloc(EC_DR, ESC_DR_AP_STATE_CHANGE,
2893 	    SUNW_KERN_PUB"dr", km_flag)) == NULL) {
2894 		rv = -2;
2895 		goto logexit;
2896 	}
2897 	evnt_val.value_type = SE_DATA_TYPE_STRING;
2898 	evnt_val.value.sv_string = attach_pnt;
2899 	if ((rv = sysevent_add_attr(&evnt_attr_list, DR_AP_ID, &evnt_val,
2900 	    km_flag)) != 0)
2901 		goto logexit;
2902 
2903 	evnt_val.value_type = SE_DATA_TYPE_STRING;
2904 	evnt_val.value.sv_string = hint;
2905 	if ((rv = sysevent_add_attr(&evnt_attr_list, DR_HINT, &evnt_val,
2906 	    km_flag)) != 0) {
2907 		sysevent_free_attr(evnt_attr_list);
2908 		goto logexit;
2909 	}
2910 
2911 	(void) sysevent_attach_attributes(ev, evnt_attr_list);
2912 
2913 	/*
2914 	 * Log the event but do not sleep waiting for its
2915 	 * delivery. This provides insulation from syseventd.
2916 	 */
2917 	rv = log_sysevent(ev, SE_NOSLEEP, &eid);
2918 
2919 logexit:
2920 	if (ev)
2921 		sysevent_free(ev);
2922 	if ((rv != 0) && verbose)
2923 		cmn_err(CE_WARN, "!drmach_log_sysevent failed (rv %d) for %s "
2924 		    " %s\n", rv, attach_pnt, hint);
2925 
2926 	return (rv);
2927 }
2928