xref: /illumos-gate/usr/src/uts/i86pc/io/mp_platform_common.c (revision bf5d9f18edeb77c14df996d367853599bdd43fd1)
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 (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
23  * Copyright 2016 Nexenta Systems, Inc.
24  * Copyright (c) 2017 by Delphix. All rights reserved.
25  * Copyright 2017 Joyent, Inc.
26  */
27 /*
28  * Copyright (c) 2010, Intel Corporation.
29  * All rights reserved.
30  */
31 
32 /*
33  * PSMI 1.1 extensions are supported only in 2.6 and later versions.
34  * PSMI 1.2 extensions are supported only in 2.7 and later versions.
35  * PSMI 1.3 and 1.4 extensions are supported in Solaris 10.
36  * PSMI 1.5 extensions are supported in Solaris Nevada.
37  * PSMI 1.6 extensions are supported in Solaris Nevada.
38  * PSMI 1.7 extensions are supported in Solaris Nevada.
39  */
40 #define	PSMI_1_7
41 
42 #include <sys/processor.h>
43 #include <sys/time.h>
44 #include <sys/psm.h>
45 #include <sys/smp_impldefs.h>
46 #include <sys/cram.h>
47 #include <sys/acpi/acpi.h>
48 #include <sys/acpica.h>
49 #include <sys/psm_common.h>
50 #include <sys/apic.h>
51 #include <sys/apic_timer.h>
52 #include <sys/pit.h>
53 #include <sys/ddi.h>
54 #include <sys/sunddi.h>
55 #include <sys/ddi_impldefs.h>
56 #include <sys/pci.h>
57 #include <sys/promif.h>
58 #include <sys/x86_archext.h>
59 #include <sys/cpc_impl.h>
60 #include <sys/uadmin.h>
61 #include <sys/panic.h>
62 #include <sys/debug.h>
63 #include <sys/archsystm.h>
64 #include <sys/trap.h>
65 #include <sys/machsystm.h>
66 #include <sys/cpuvar.h>
67 #include <sys/rm_platter.h>
68 #include <sys/privregs.h>
69 #include <sys/cyclic.h>
70 #include <sys/note.h>
71 #include <sys/pci_intr_lib.h>
72 #include <sys/sunndi.h>
73 #if !defined(__xpv)
74 #include <sys/hpet.h>
75 #include <sys/clock.h>
76 #endif
77 
78 /*
79  *	Local Function Prototypes
80  */
81 static int apic_handle_defconf();
82 static int apic_parse_mpct(caddr_t mpct, int bypass);
83 static struct apic_mpfps_hdr *apic_find_fps_sig(caddr_t fptr, int size);
84 static int apic_checksum(caddr_t bptr, int len);
85 static int apic_find_bus_type(char *bus);
86 static int apic_find_bus(int busid);
87 static struct apic_io_intr *apic_find_io_intr(int irqno);
88 static int apic_find_free_irq(int start, int end);
89 struct apic_io_intr *apic_find_io_intr_w_busid(int irqno, int busid);
90 static void apic_set_pwroff_method_from_mpcnfhdr(struct apic_mp_cnf_hdr *hdrp);
91 static void apic_free_apic_cpus(void);
92 static boolean_t apic_is_ioapic_AMD_813x(uint32_t physaddr);
93 static int apic_acpi_enter_apicmode(void);
94 
95 int apic_handle_pci_pci_bridge(dev_info_t *idip, int child_devno,
96     int child_ipin, struct apic_io_intr **intrp);
97 int apic_find_bus_id(int bustype);
98 int apic_find_intin(uchar_t ioapic, uchar_t intin);
99 void apic_record_rdt_entry(apic_irq_t *irqptr, int irq);
100 
101 int apic_debug_mps_id = 0;	/* 1 - print MPS ID strings */
102 
103 /* ACPI SCI interrupt configuration; -1 if SCI not used */
104 int apic_sci_vect = -1;
105 iflag_t apic_sci_flags;
106 
107 #if !defined(__xpv)
108 /* ACPI HPET interrupt configuration; -1 if HPET not used */
109 int apic_hpet_vect = -1;
110 iflag_t apic_hpet_flags;
111 #endif
112 
113 /*
114  * psm name pointer
115  */
116 char *psm_name;
117 
118 /* ACPI support routines */
119 static int acpi_probe(char *);
120 static int apic_acpi_irq_configure(acpi_psm_lnk_t *acpipsmlnkp, dev_info_t *dip,
121     int *pci_irqp, iflag_t *intr_flagp);
122 
123 int apic_acpi_translate_pci_irq(dev_info_t *dip, int busid, int devid,
124     int ipin, int *pci_irqp, iflag_t *intr_flagp);
125 uchar_t acpi_find_ioapic(int irq);
126 static int acpi_intr_compatible(iflag_t iflag1, iflag_t iflag2);
127 
128 /* Max wait time (in repetitions) for flags to clear in an RDT entry. */
129 int apic_max_reps_clear_pending = 1000;
130 
131 int	apic_intr_policy = INTR_ROUND_ROBIN;
132 
133 int	apic_next_bind_cpu = 1; /* For round robin assignment */
134 				/* start with cpu 1 */
135 
136 /*
137  * If enabled, the distribution works as follows:
138  * On every interrupt entry, the current ipl for the CPU is set in cpu_info
139  * and the irq corresponding to the ipl is also set in the aci_current array.
140  * interrupt exit and setspl (due to soft interrupts) will cause the current
141  * ipl to be be changed. This is cache friendly as these frequently used
142  * paths write into a per cpu structure.
143  *
144  * Sampling is done by checking the structures for all CPUs and incrementing
145  * the busy field of the irq (if any) executing on each CPU and the busy field
146  * of the corresponding CPU.
147  * In periodic mode this is done on every clock interrupt.
148  * In one-shot mode, this is done thru a cyclic with an interval of
149  * apic_redistribute_sample_interval (default 10 milli sec).
150  *
151  * Every apic_sample_factor_redistribution times we sample, we do computations
152  * to decide which interrupt needs to be migrated (see comments
153  * before apic_intr_redistribute().
154  */
155 
156 /*
157  * Following 3 variables start as % and can be patched or set using an
158  * API to be defined in future. They will be scaled to
159  * sample_factor_redistribution which is in turn set to hertz+1 (in periodic
160  * mode), or 101 in one-shot mode to stagger it away from one sec processing
161  */
162 
163 int	apic_int_busy_mark = 60;
164 int	apic_int_free_mark = 20;
165 int	apic_diff_for_redistribution = 10;
166 
167 /* sampling interval for interrupt redistribution for dynamic migration */
168 int	apic_redistribute_sample_interval = NANOSEC / 100; /* 10 millisec */
169 
170 /*
171  * number of times we sample before deciding to redistribute interrupts
172  * for dynamic migration
173  */
174 int	apic_sample_factor_redistribution = 101;
175 
176 int	apic_redist_cpu_skip = 0;
177 int	apic_num_imbalance = 0;
178 int	apic_num_rebind = 0;
179 
180 /*
181  * Maximum number of APIC CPUs in the system, -1 indicates that dynamic
182  * allocation of CPU ids is disabled.
183  */
184 int 	apic_max_nproc = -1;
185 int	apic_nproc = 0;
186 size_t	apic_cpus_size = 0;
187 int	apic_defconf = 0;
188 int	apic_irq_translate = 0;
189 int	apic_spec_rev = 0;
190 int	apic_imcrp = 0;
191 
192 int	apic_use_acpi = 1;	/* 1 = use ACPI, 0 = don't use ACPI */
193 int	apic_use_acpi_madt_only = 0;	/* 1=ONLY use MADT from ACPI */
194 
195 /*
196  * For interrupt link devices, if apic_unconditional_srs is set, an irq resource
197  * will be assigned (via _SRS). If it is not set, use the current
198  * irq setting (via _CRS), but only if that irq is in the set of possible
199  * irqs (returned by _PRS) for the device.
200  */
201 int	apic_unconditional_srs = 1;
202 
203 /*
204  * For interrupt link devices, if apic_prefer_crs is set when we are
205  * assigning an IRQ resource to a device, prefer the current IRQ setting
206  * over other possible irq settings under same conditions.
207  */
208 
209 int	apic_prefer_crs = 1;
210 
211 uchar_t apic_io_id[MAX_IO_APIC];
212 volatile uint32_t *apicioadr[MAX_IO_APIC];
213 uchar_t	apic_io_ver[MAX_IO_APIC];
214 uchar_t	apic_io_vectbase[MAX_IO_APIC];
215 uchar_t	apic_io_vectend[MAX_IO_APIC];
216 uchar_t apic_reserved_irqlist[MAX_ISA_IRQ + 1];
217 uint32_t apic_physaddr[MAX_IO_APIC];
218 
219 boolean_t ioapic_mask_workaround[MAX_IO_APIC];
220 
221 /*
222  * First available slot to be used as IRQ index into the apic_irq_table
223  * for those interrupts (like MSI/X) that don't have a physical IRQ.
224  */
225 int apic_first_avail_irq  = APIC_FIRST_FREE_IRQ;
226 
227 /*
228  * apic_ioapic_lock protects the ioapics (reg select), the status, temp_bound
229  * and bound elements of cpus_info and the temp_cpu element of irq_struct
230  */
231 lock_t	apic_ioapic_lock;
232 
233 int	apic_io_max = 0;	/* no. of i/o apics enabled */
234 
235 struct apic_io_intr *apic_io_intrp = NULL;
236 static	struct apic_bus	*apic_busp;
237 
238 uchar_t	apic_resv_vector[MAXIPL+1];
239 
240 char	apic_level_intr[APIC_MAX_VECTOR+1];
241 
242 uint32_t	eisa_level_intr_mask = 0;
243 	/* At least MSB will be set if EISA bus */
244 
245 int	apic_pci_bus_total = 0;
246 uchar_t	apic_single_pci_busid = 0;
247 
248 /*
249  * airq_mutex protects additions to the apic_irq_table - the first
250  * pointer and any airq_nexts off of that one. It also protects
251  * apic_max_device_irq & apic_min_device_irq. It also guarantees
252  * that share_id is unique as new ids are generated only when new
253  * irq_t structs are linked in. Once linked in the structs are never
254  * deleted. temp_cpu & mps_intr_index field indicate if it is programmed
255  * or allocated. Note that there is a slight gap between allocating in
256  * apic_introp_xlate and programming in addspl.
257  */
258 kmutex_t	airq_mutex;
259 apic_irq_t	*apic_irq_table[APIC_MAX_VECTOR+1];
260 int		apic_max_device_irq = 0;
261 int		apic_min_device_irq = APIC_MAX_VECTOR;
262 
263 typedef struct prs_irq_list_ent {
264 	int			list_prio;
265 	int32_t			irq;
266 	iflag_t			intrflags;
267 	acpi_prs_private_t	prsprv;
268 	struct prs_irq_list_ent	*next;
269 } prs_irq_list_t;
270 
271 
272 /*
273  * ACPI variables
274  */
275 /* 1 = acpi is enabled & working, 0 = acpi is not enabled or not there */
276 int apic_enable_acpi = 0;
277 
278 /* ACPI Multiple APIC Description Table ptr */
279 static	ACPI_TABLE_MADT *acpi_mapic_dtp = NULL;
280 
281 /* ACPI Interrupt Source Override Structure ptr */
282 ACPI_MADT_INTERRUPT_OVERRIDE *acpi_isop = NULL;
283 int acpi_iso_cnt = 0;
284 
285 /* ACPI Non-maskable Interrupt Sources ptr */
286 static	ACPI_MADT_NMI_SOURCE *acpi_nmi_sp = NULL;
287 static	int acpi_nmi_scnt = 0;
288 static	ACPI_MADT_LOCAL_APIC_NMI *acpi_nmi_cp = NULL;
289 static	int acpi_nmi_ccnt = 0;
290 
291 static	boolean_t acpi_found_smp_config = B_FALSE;
292 
293 /*
294  * The following added to identify a software poweroff method if available.
295  */
296 
297 static struct {
298 	int	poweroff_method;
299 	char	oem_id[APIC_MPS_OEM_ID_LEN + 1];	/* MAX + 1 for NULL */
300 	char	prod_id[APIC_MPS_PROD_ID_LEN + 1];	/* MAX + 1 for NULL */
301 } apic_mps_ids[] = {
302 	{ APIC_POWEROFF_VIA_RTC,	"INTEL",	"ALDER" },   /* 4300 */
303 	{ APIC_POWEROFF_VIA_RTC,	"NCR",		"AMC" },    /* 4300 */
304 	{ APIC_POWEROFF_VIA_ASPEN_BMC,	"INTEL",	"A450NX" },  /* 4400? */
305 	{ APIC_POWEROFF_VIA_ASPEN_BMC,	"INTEL",	"AD450NX" }, /* 4400 */
306 	{ APIC_POWEROFF_VIA_ASPEN_BMC,	"INTEL",	"AC450NX" }, /* 4400R */
307 	{ APIC_POWEROFF_VIA_SITKA_BMC,	"INTEL",	"S450NX" },  /* S50  */
308 	{ APIC_POWEROFF_VIA_SITKA_BMC,	"INTEL",	"SC450NX" }  /* S50? */
309 };
310 
311 int	apic_poweroff_method = APIC_POWEROFF_NONE;
312 
313 /*
314  * Auto-configuration routines
315  */
316 
317 /*
318  * Look at MPSpec 1.4 (Intel Order # 242016-005) for details of what we do here
319  * May work with 1.1 - but not guaranteed.
320  * According to the MP Spec, the MP floating pointer structure
321  * will be searched in the order described below:
322  * 1. In the first kilobyte of Extended BIOS Data Area (EBDA)
323  * 2. Within the last kilobyte of system base memory
324  * 3. In the BIOS ROM address space between 0F0000h and 0FFFFh
325  * Once we find the right signature with proper checksum, we call
326  * either handle_defconf or parse_mpct to get all info necessary for
327  * subsequent operations.
328  */
329 int
330 apic_probe_common(char *modname)
331 {
332 	uint32_t mpct_addr, ebda_start = 0, base_mem_end;
333 	caddr_t	biosdatap;
334 	caddr_t	mpct = NULL;
335 	caddr_t	fptr;
336 	int	i, mpct_size = 0, mapsize, retval = PSM_FAILURE;
337 	ushort_t	ebda_seg, base_mem_size;
338 	struct	apic_mpfps_hdr	*fpsp;
339 	struct	apic_mp_cnf_hdr	*hdrp;
340 	int bypass_cpu_and_ioapics_in_mptables;
341 	int acpi_user_options;
342 
343 	if (apic_forceload < 0)
344 		return (retval);
345 
346 	/*
347 	 * Remember who we are
348 	 */
349 	psm_name = modname;
350 
351 	/* Allow override for MADT-only mode */
352 	acpi_user_options = ddi_prop_get_int(DDI_DEV_T_ANY, ddi_root_node(), 0,
353 	    "acpi-user-options", 0);
354 	apic_use_acpi_madt_only = ((acpi_user_options & ACPI_OUSER_MADT) != 0);
355 
356 	/* Allow apic_use_acpi to override MADT-only mode */
357 	if (!apic_use_acpi)
358 		apic_use_acpi_madt_only = 0;
359 
360 	retval = acpi_probe(modname);
361 
362 	/* in UEFI system, there is no BIOS data */
363 	if (ddi_prop_exists(DDI_DEV_T_ANY, ddi_root_node(), 0, "efi-systab"))
364 		goto apic_ret;
365 
366 	/*
367 	 * mapin the bios data area 40:0
368 	 * 40:13h - two-byte location reports the base memory size
369 	 * 40:0Eh - two-byte location for the exact starting address of
370 	 *	    the EBDA segment for EISA
371 	 */
372 	biosdatap = psm_map_phys(0x400, 0x20, PROT_READ);
373 	if (!biosdatap)
374 		goto apic_ret;
375 	fpsp = (struct apic_mpfps_hdr *)NULL;
376 	mapsize = MPFPS_RAM_WIN_LEN;
377 	/*LINTED: pointer cast may result in improper alignment */
378 	ebda_seg = *((ushort_t *)(biosdatap+0xe));
379 	/* check the 1k of EBDA */
380 	if (ebda_seg) {
381 		ebda_start = ((uint32_t)ebda_seg) << 4;
382 		fptr = psm_map_phys(ebda_start, MPFPS_RAM_WIN_LEN, PROT_READ);
383 		if (fptr) {
384 			if (!(fpsp =
385 			    apic_find_fps_sig(fptr, MPFPS_RAM_WIN_LEN)))
386 				psm_unmap_phys(fptr, MPFPS_RAM_WIN_LEN);
387 		}
388 	}
389 	/* If not in EBDA, check the last k of system base memory */
390 	if (!fpsp) {
391 		/*LINTED: pointer cast may result in improper alignment */
392 		base_mem_size = *((ushort_t *)(biosdatap + 0x13));
393 
394 		if (base_mem_size > 512)
395 			base_mem_end = 639 * 1024;
396 		else
397 			base_mem_end = 511 * 1024;
398 		/* if ebda == last k of base mem, skip to check BIOS ROM */
399 		if (base_mem_end != ebda_start) {
400 
401 			fptr = psm_map_phys(base_mem_end, MPFPS_RAM_WIN_LEN,
402 			    PROT_READ);
403 
404 			if (fptr) {
405 				if (!(fpsp = apic_find_fps_sig(fptr,
406 				    MPFPS_RAM_WIN_LEN)))
407 					psm_unmap_phys(fptr, MPFPS_RAM_WIN_LEN);
408 			}
409 		}
410 	}
411 	psm_unmap_phys(biosdatap, 0x20);
412 
413 	/* If still cannot find it, check the BIOS ROM space */
414 	if (!fpsp) {
415 		mapsize = MPFPS_ROM_WIN_LEN;
416 		fptr = psm_map_phys(MPFPS_ROM_WIN_START,
417 		    MPFPS_ROM_WIN_LEN, PROT_READ);
418 		if (fptr) {
419 			if (!(fpsp =
420 			    apic_find_fps_sig(fptr, MPFPS_ROM_WIN_LEN))) {
421 				psm_unmap_phys(fptr, MPFPS_ROM_WIN_LEN);
422 				goto apic_ret;
423 			}
424 		}
425 	}
426 
427 	if (apic_checksum((caddr_t)fpsp, fpsp->mpfps_length * 16) != 0) {
428 		psm_unmap_phys(fptr, MPFPS_ROM_WIN_LEN);
429 		goto apic_ret;
430 	}
431 
432 	apic_spec_rev = fpsp->mpfps_spec_rev;
433 	if ((apic_spec_rev != 04) && (apic_spec_rev != 01)) {
434 		psm_unmap_phys(fptr, MPFPS_ROM_WIN_LEN);
435 		goto apic_ret;
436 	}
437 
438 	/* check IMCR is present or not */
439 	apic_imcrp = fpsp->mpfps_featinfo2 & MPFPS_FEATINFO2_IMCRP;
440 
441 	/* check default configuration (dual CPUs) */
442 	if ((apic_defconf = fpsp->mpfps_featinfo1) != 0) {
443 		psm_unmap_phys(fptr, mapsize);
444 		if ((retval = apic_handle_defconf()) != PSM_SUCCESS)
445 			return (retval);
446 
447 		goto apic_ret;
448 	}
449 
450 	/* MP Configuration Table */
451 	mpct_addr = (uint32_t)(fpsp->mpfps_mpct_paddr);
452 
453 	psm_unmap_phys(fptr, mapsize); /* unmap floating ptr struct */
454 
455 	/*
456 	 * Map in enough memory for the MP Configuration Table Header.
457 	 * Use this table to read the total length of the BIOS data and
458 	 * map in all the info
459 	 */
460 	/*LINTED: pointer cast may result in improper alignment */
461 	hdrp = (struct apic_mp_cnf_hdr *)psm_map_phys(mpct_addr,
462 	    sizeof (struct apic_mp_cnf_hdr), PROT_READ);
463 	if (!hdrp)
464 		goto apic_ret;
465 
466 	/* check mp configuration table signature PCMP */
467 	if (hdrp->mpcnf_sig != 0x504d4350) {
468 		psm_unmap_phys((caddr_t)hdrp, sizeof (struct apic_mp_cnf_hdr));
469 		goto apic_ret;
470 	}
471 	mpct_size = (int)hdrp->mpcnf_tbl_length;
472 
473 	apic_set_pwroff_method_from_mpcnfhdr(hdrp);
474 
475 	psm_unmap_phys((caddr_t)hdrp, sizeof (struct apic_mp_cnf_hdr));
476 
477 	if ((retval == PSM_SUCCESS) && !apic_use_acpi_madt_only) {
478 		/* This is an ACPI machine No need for further checks */
479 		goto apic_ret;
480 	}
481 
482 	/*
483 	 * Map in the entries for this machine, ie. Processor
484 	 * Entry Tables, Bus Entry Tables, etc.
485 	 * They are in fixed order following one another
486 	 */
487 	mpct = psm_map_phys(mpct_addr, mpct_size, PROT_READ);
488 	if (!mpct)
489 		goto apic_ret;
490 
491 	if (apic_checksum(mpct, mpct_size) != 0)
492 		goto apic_fail1;
493 
494 	/*LINTED: pointer cast may result in improper alignment */
495 	hdrp = (struct apic_mp_cnf_hdr *)mpct;
496 	apicadr = (uint32_t *)mapin_apic((uint32_t)hdrp->mpcnf_local_apic,
497 	    APIC_LOCAL_MEMLEN, PROT_READ | PROT_WRITE);
498 	if (!apicadr)
499 		goto apic_fail1;
500 
501 	/* Parse all information in the tables */
502 	bypass_cpu_and_ioapics_in_mptables = (retval == PSM_SUCCESS);
503 	if (apic_parse_mpct(mpct, bypass_cpu_and_ioapics_in_mptables) ==
504 	    PSM_SUCCESS) {
505 		retval = PSM_SUCCESS;
506 		goto apic_ret;
507 	}
508 
509 apic_fail1:
510 	psm_unmap_phys(mpct, mpct_size);
511 	mpct = NULL;
512 
513 apic_ret:
514 	if (retval == PSM_SUCCESS) {
515 		extern int apic_ioapic_method_probe();
516 
517 		if ((retval = apic_ioapic_method_probe()) == PSM_SUCCESS)
518 			return (PSM_SUCCESS);
519 	}
520 
521 	for (i = 0; i < apic_io_max; i++)
522 		mapout_ioapic((caddr_t)apicioadr[i], APIC_IO_MEMLEN);
523 	if (apic_cpus) {
524 		kmem_free(apic_cpus, apic_cpus_size);
525 		apic_cpus = NULL;
526 	}
527 	if (apicadr) {
528 		mapout_apic((caddr_t)apicadr, APIC_LOCAL_MEMLEN);
529 		apicadr = NULL;
530 	}
531 	if (mpct)
532 		psm_unmap_phys(mpct, mpct_size);
533 
534 	return (retval);
535 }
536 
537 static void
538 apic_set_pwroff_method_from_mpcnfhdr(struct apic_mp_cnf_hdr *hdrp)
539 {
540 	int	i;
541 
542 	for (i = 0; i < (sizeof (apic_mps_ids) / sizeof (apic_mps_ids[0]));
543 	    i++) {
544 		if ((strncmp(hdrp->mpcnf_oem_str, apic_mps_ids[i].oem_id,
545 		    strlen(apic_mps_ids[i].oem_id)) == 0) &&
546 		    (strncmp(hdrp->mpcnf_prod_str, apic_mps_ids[i].prod_id,
547 		    strlen(apic_mps_ids[i].prod_id)) == 0)) {
548 
549 			apic_poweroff_method = apic_mps_ids[i].poweroff_method;
550 			break;
551 		}
552 	}
553 
554 	if (apic_debug_mps_id != 0) {
555 		cmn_err(CE_CONT, "%s: MPS OEM ID = '%c%c%c%c%c%c%c%c'"
556 		    "Product ID = '%c%c%c%c%c%c%c%c%c%c%c%c'\n",
557 		    psm_name,
558 		    hdrp->mpcnf_oem_str[0],
559 		    hdrp->mpcnf_oem_str[1],
560 		    hdrp->mpcnf_oem_str[2],
561 		    hdrp->mpcnf_oem_str[3],
562 		    hdrp->mpcnf_oem_str[4],
563 		    hdrp->mpcnf_oem_str[5],
564 		    hdrp->mpcnf_oem_str[6],
565 		    hdrp->mpcnf_oem_str[7],
566 		    hdrp->mpcnf_prod_str[0],
567 		    hdrp->mpcnf_prod_str[1],
568 		    hdrp->mpcnf_prod_str[2],
569 		    hdrp->mpcnf_prod_str[3],
570 		    hdrp->mpcnf_prod_str[4],
571 		    hdrp->mpcnf_prod_str[5],
572 		    hdrp->mpcnf_prod_str[6],
573 		    hdrp->mpcnf_prod_str[7],
574 		    hdrp->mpcnf_prod_str[8],
575 		    hdrp->mpcnf_prod_str[9],
576 		    hdrp->mpcnf_prod_str[10],
577 		    hdrp->mpcnf_prod_str[11]);
578 	}
579 }
580 
581 static void
582 apic_free_apic_cpus(void)
583 {
584 	if (apic_cpus != NULL) {
585 		kmem_free(apic_cpus, apic_cpus_size);
586 		apic_cpus = NULL;
587 		apic_cpus_size = 0;
588 	}
589 }
590 
591 static int
592 acpi_probe(char *modname)
593 {
594 	int			i, intmax, index;
595 	uint32_t		id, ver;
596 	int			acpi_verboseflags = 0;
597 	int			madt_seen, madt_size;
598 	ACPI_SUBTABLE_HEADER		*ap;
599 	ACPI_MADT_LOCAL_APIC	*mpa;
600 	ACPI_MADT_LOCAL_X2APIC	*mpx2a;
601 	ACPI_MADT_IO_APIC		*mia;
602 	ACPI_MADT_IO_SAPIC		*misa;
603 	ACPI_MADT_INTERRUPT_OVERRIDE	*mio;
604 	ACPI_MADT_NMI_SOURCE		*mns;
605 	ACPI_MADT_INTERRUPT_SOURCE	*mis;
606 	ACPI_MADT_LOCAL_APIC_NMI	*mlan;
607 	ACPI_MADT_LOCAL_X2APIC_NMI	*mx2alan;
608 	ACPI_MADT_LOCAL_APIC_OVERRIDE	*mao;
609 	int			sci;
610 	iflag_t			sci_flags;
611 	volatile uint32_t	*ioapic;
612 	int			ioapic_ix;
613 	uint32_t		*local_ids;
614 	uint32_t		*proc_ids;
615 	uchar_t			hid;
616 	int			warned = 0;
617 
618 	if (!apic_use_acpi)
619 		return (PSM_FAILURE);
620 
621 	if (AcpiGetTable(ACPI_SIG_MADT, 1,
622 	    (ACPI_TABLE_HEADER **) &acpi_mapic_dtp) != AE_OK) {
623 		cmn_err(CE_WARN, "!acpi_probe: No MADT found!");
624 		return (PSM_FAILURE);
625 	}
626 
627 	apicadr = mapin_apic((uint32_t)acpi_mapic_dtp->Address,
628 	    APIC_LOCAL_MEMLEN, PROT_READ | PROT_WRITE);
629 	if (!apicadr)
630 		return (PSM_FAILURE);
631 
632 	if ((local_ids = (uint32_t *)kmem_zalloc(NCPU * sizeof (uint32_t),
633 	    KM_NOSLEEP)) == NULL)
634 		return (PSM_FAILURE);
635 
636 	if ((proc_ids = (uint32_t *)kmem_zalloc(NCPU * sizeof (uint32_t),
637 	    KM_NOSLEEP)) == NULL) {
638 		kmem_free(local_ids, NCPU * sizeof (uint32_t));
639 		return (PSM_FAILURE);
640 	}
641 
642 	id = apic_reg_ops->apic_read(APIC_LID_REG);
643 	local_ids[0] = (uchar_t)(id >> 24);
644 	apic_nproc = index = 1;
645 	apic_io_max = 0;
646 
647 	ap = (ACPI_SUBTABLE_HEADER *) (acpi_mapic_dtp + 1);
648 	madt_size = acpi_mapic_dtp->Header.Length;
649 	madt_seen = sizeof (*acpi_mapic_dtp);
650 
651 	while (madt_seen < madt_size) {
652 		switch (ap->Type) {
653 		case ACPI_MADT_TYPE_LOCAL_APIC:
654 			mpa = (ACPI_MADT_LOCAL_APIC *) ap;
655 			if (mpa->LapicFlags & ACPI_MADT_ENABLED) {
656 				if (mpa->Id == 255) {
657 					cmn_err(CE_WARN, "!%s: encountered "
658 					    "invalid entry in MADT: CPU %d "
659 					    "has Local APIC Id equal to 255 ",
660 					    psm_name, mpa->ProcessorId);
661 				}
662 				if (mpa->Id == local_ids[0]) {
663 					ASSERT(index == 1);
664 					proc_ids[0] = mpa->ProcessorId;
665 				} else if (apic_nproc < NCPU && use_mp &&
666 				    apic_nproc < boot_ncpus) {
667 					local_ids[index] = mpa->Id;
668 					proc_ids[index] = mpa->ProcessorId;
669 					index++;
670 					apic_nproc++;
671 				} else if (apic_nproc == NCPU && !warned) {
672 					cmn_err(CE_WARN, "%s: CPU limit "
673 					    "exceeded"
674 #if !defined(__amd64)
675 					    " for 32-bit mode"
676 #endif
677 					    "; Solaris will use %d CPUs.",
678 					    psm_name,  NCPU);
679 					warned = 1;
680 				}
681 			}
682 			break;
683 
684 		case ACPI_MADT_TYPE_IO_APIC:
685 			mia = (ACPI_MADT_IO_APIC *) ap;
686 			if (apic_io_max < MAX_IO_APIC) {
687 				ioapic_ix = apic_io_max;
688 				apic_io_id[apic_io_max] = mia->Id;
689 				apic_io_vectbase[apic_io_max] =
690 				    mia->GlobalIrqBase;
691 				apic_physaddr[apic_io_max] =
692 				    (uint32_t)mia->Address;
693 				ioapic = apicioadr[apic_io_max] =
694 				    mapin_ioapic((uint32_t)mia->Address,
695 				    APIC_IO_MEMLEN, PROT_READ | PROT_WRITE);
696 				if (!ioapic)
697 					goto cleanup;
698 				ioapic_mask_workaround[apic_io_max] =
699 				    apic_is_ioapic_AMD_813x(mia->Address);
700 				apic_io_max++;
701 			}
702 			break;
703 
704 		case ACPI_MADT_TYPE_INTERRUPT_OVERRIDE:
705 			mio = (ACPI_MADT_INTERRUPT_OVERRIDE *) ap;
706 			if (acpi_isop == NULL)
707 				acpi_isop = mio;
708 			acpi_iso_cnt++;
709 			break;
710 
711 		case ACPI_MADT_TYPE_NMI_SOURCE:
712 			/* UNIMPLEMENTED */
713 			mns = (ACPI_MADT_NMI_SOURCE *) ap;
714 			if (acpi_nmi_sp == NULL)
715 				acpi_nmi_sp = mns;
716 			acpi_nmi_scnt++;
717 
718 			cmn_err(CE_NOTE, "!apic: nmi source: %d 0x%x\n",
719 			    mns->GlobalIrq, mns->IntiFlags);
720 			break;
721 
722 		case ACPI_MADT_TYPE_LOCAL_APIC_NMI:
723 			/* UNIMPLEMENTED */
724 			mlan = (ACPI_MADT_LOCAL_APIC_NMI *) ap;
725 			if (acpi_nmi_cp == NULL)
726 				acpi_nmi_cp = mlan;
727 			acpi_nmi_ccnt++;
728 
729 			cmn_err(CE_NOTE, "!apic: local nmi: %d 0x%x %d\n",
730 			    mlan->ProcessorId, mlan->IntiFlags,
731 			    mlan->Lint);
732 			break;
733 
734 		case ACPI_MADT_TYPE_LOCAL_APIC_OVERRIDE:
735 			/* UNIMPLEMENTED */
736 			mao = (ACPI_MADT_LOCAL_APIC_OVERRIDE *) ap;
737 			cmn_err(CE_NOTE, "!apic: address override: %lx\n",
738 			    (long)mao->Address);
739 			break;
740 
741 		case ACPI_MADT_TYPE_IO_SAPIC:
742 			/* UNIMPLEMENTED */
743 			misa = (ACPI_MADT_IO_SAPIC *) ap;
744 
745 			cmn_err(CE_NOTE, "!apic: io sapic: %d %d %lx\n",
746 			    misa->Id, misa->GlobalIrqBase,
747 			    (long)misa->Address);
748 			break;
749 
750 		case ACPI_MADT_TYPE_INTERRUPT_SOURCE:
751 			/* UNIMPLEMENTED */
752 			mis = (ACPI_MADT_INTERRUPT_SOURCE *) ap;
753 
754 			cmn_err(CE_NOTE,
755 			    "!apic: irq source: %d %d %d 0x%x %d %d\n",
756 			    mis->Id, mis->Eid, mis->GlobalIrq,
757 			    mis->IntiFlags, mis->Type,
758 			    mis->IoSapicVector);
759 			break;
760 
761 		case ACPI_MADT_TYPE_LOCAL_X2APIC:
762 			mpx2a = (ACPI_MADT_LOCAL_X2APIC *) ap;
763 
764 			if (mpx2a->LapicFlags & ACPI_MADT_ENABLED) {
765 				if (mpx2a->LocalApicId == local_ids[0]) {
766 					ASSERT(index == 1);
767 					proc_ids[0] = mpx2a->Uid;
768 				} else if (apic_nproc < NCPU && use_mp &&
769 				    apic_nproc < boot_ncpus) {
770 					local_ids[index] = mpx2a->LocalApicId;
771 					proc_ids[index] = mpx2a->Uid;
772 					index++;
773 					apic_nproc++;
774 				} else if (apic_nproc == NCPU && !warned) {
775 					cmn_err(CE_WARN, "%s: CPU limit "
776 					    "exceeded"
777 #if !defined(__amd64)
778 					    " for 32-bit mode"
779 #endif
780 					    "; Solaris will use %d CPUs.",
781 					    psm_name,  NCPU);
782 					warned = 1;
783 				}
784 			}
785 
786 			break;
787 
788 		case ACPI_MADT_TYPE_LOCAL_X2APIC_NMI:
789 			/* UNIMPLEMENTED */
790 			mx2alan = (ACPI_MADT_LOCAL_X2APIC_NMI *) ap;
791 			if (mx2alan->Uid >> 8)
792 				acpi_nmi_ccnt++;
793 
794 #ifdef	DEBUG
795 			cmn_err(CE_NOTE,
796 			    "!apic: local x2apic nmi: %d 0x%x %d\n",
797 			    mx2alan->Uid, mx2alan->IntiFlags, mx2alan->Lint);
798 #endif
799 
800 			break;
801 
802 		case ACPI_MADT_TYPE_RESERVED:
803 		default:
804 			break;
805 		}
806 
807 		/* advance to next entry */
808 		madt_seen += ap->Length;
809 		ap = (ACPI_SUBTABLE_HEADER *)(((char *)ap) + ap->Length);
810 	}
811 
812 	/* We found multiple enabled cpus via MADT */
813 	if ((apic_nproc > 1) && (apic_io_max > 0)) {
814 		acpi_found_smp_config = B_TRUE;
815 		cmn_err(CE_NOTE,
816 		    "!apic: Using ACPI (MADT) for SMP configuration");
817 	}
818 
819 	/*
820 	 * allocate enough space for possible hot-adding of CPUs.
821 	 * max_ncpus may be less than apic_nproc if it's set by user.
822 	 */
823 	if (plat_dr_support_cpu()) {
824 		apic_max_nproc = max_ncpus;
825 	}
826 	apic_cpus_size = max(apic_nproc, max_ncpus) * sizeof (*apic_cpus);
827 	if ((apic_cpus = kmem_zalloc(apic_cpus_size, KM_NOSLEEP)) == NULL)
828 		goto cleanup;
829 
830 	/*
831 	 * ACPI doesn't provide the local apic ver, get it directly from the
832 	 * local apic
833 	 */
834 	ver = apic_reg_ops->apic_read(APIC_VERS_REG);
835 	for (i = 0; i < apic_nproc; i++) {
836 		apic_cpus[i].aci_local_id = local_ids[i];
837 		apic_cpus[i].aci_local_ver = (uchar_t)(ver & 0xFF);
838 		apic_cpus[i].aci_processor_id = proc_ids[i];
839 		/* Only build mapping info for CPUs present at boot. */
840 		if (i < boot_ncpus)
841 			(void) acpica_map_cpu(i, proc_ids[i]);
842 	}
843 
844 	/*
845 	 * To support CPU dynamic reconfiguration, the apic CPU info structure
846 	 * for each possible CPU will be pre-allocated at boot time.
847 	 * The state for each apic CPU info structure will be assigned according
848 	 * to the following rules:
849 	 * Rule 1:
850 	 * 	Slot index range: [0, min(apic_nproc, boot_ncpus))
851 	 *	State flags: 0
852 	 *	Note: cpu exists and will be configured/enabled at boot time
853 	 * Rule 2:
854 	 * 	Slot index range: [boot_ncpus, apic_nproc)
855 	 *	State flags: APIC_CPU_FREE | APIC_CPU_DIRTY
856 	 *	Note: cpu exists but won't be configured/enabled at boot time
857 	 * Rule 3:
858 	 * 	Slot index range: [apic_nproc, boot_ncpus)
859 	 *	State flags: APIC_CPU_FREE
860 	 *	Note: cpu doesn't exist at boot time
861 	 * Rule 4:
862 	 * 	Slot index range: [max(apic_nproc, boot_ncpus), max_ncpus)
863 	 *	State flags: APIC_CPU_FREE
864 	 *	Note: cpu doesn't exist at boot time
865 	 */
866 	CPUSET_ZERO(apic_cpumask);
867 	for (i = 0; i < min(boot_ncpus, apic_nproc); i++) {
868 		CPUSET_ADD(apic_cpumask, i);
869 		apic_cpus[i].aci_status = 0;
870 	}
871 	for (i = boot_ncpus; i < apic_nproc; i++) {
872 		apic_cpus[i].aci_status = APIC_CPU_FREE | APIC_CPU_DIRTY;
873 	}
874 	for (i = apic_nproc; i < boot_ncpus; i++) {
875 		apic_cpus[i].aci_status = APIC_CPU_FREE;
876 	}
877 	for (i = max(boot_ncpus, apic_nproc); i < max_ncpus; i++) {
878 		apic_cpus[i].aci_status = APIC_CPU_FREE;
879 	}
880 
881 	for (i = 0; i < apic_io_max; i++) {
882 		ioapic_ix = i;
883 
884 		/*
885 		 * need to check Sitka on the following acpi problem
886 		 * On the Sitka, the ioapic's apic_id field isn't reporting
887 		 * the actual io apic id. We have reported this problem
888 		 * to Intel. Until they fix the problem, we will get the
889 		 * actual id directly from the ioapic.
890 		 */
891 		id = ioapic_read(ioapic_ix, APIC_ID_CMD);
892 		hid = (uchar_t)(id >> 24);
893 
894 		if (hid != apic_io_id[i]) {
895 			if (apic_io_id[i] == 0)
896 				apic_io_id[i] = hid;
897 			else { /* set ioapic id to whatever reported by ACPI */
898 				id = ((uint32_t)apic_io_id[i]) << 24;
899 				ioapic_write(ioapic_ix, APIC_ID_CMD, id);
900 			}
901 		}
902 		ver = ioapic_read(ioapic_ix, APIC_VERS_CMD);
903 		apic_io_ver[i] = (uchar_t)(ver & 0xff);
904 		intmax = (ver >> 16) & 0xff;
905 		apic_io_vectend[i] = apic_io_vectbase[i] + intmax;
906 		if (apic_first_avail_irq <= apic_io_vectend[i])
907 			apic_first_avail_irq = apic_io_vectend[i] + 1;
908 	}
909 
910 
911 	/*
912 	 * Process SCI configuration here
913 	 * An error may be returned here if
914 	 * acpi-user-options specifies legacy mode
915 	 * (no SCI, no ACPI mode)
916 	 */
917 	if (acpica_get_sci(&sci, &sci_flags) != AE_OK)
918 		sci = -1;
919 
920 	/*
921 	 * Now call acpi_init() to generate namespaces
922 	 * If this fails, we don't attempt to use ACPI
923 	 * even if we were able to get a MADT above
924 	 */
925 	if (acpica_init() != AE_OK) {
926 		cmn_err(CE_WARN, "!apic: Failed to initialize acpica!");
927 		goto cleanup;
928 	}
929 
930 	/*
931 	 * Call acpica_build_processor_map() now that we have
932 	 * ACPI namesspace access
933 	 */
934 	(void) acpica_build_processor_map();
935 
936 	/*
937 	 * Squirrel away the SCI and flags for later on
938 	 * in apic_picinit() when we're ready
939 	 */
940 	apic_sci_vect = sci;
941 	apic_sci_flags = sci_flags;
942 
943 	if (apic_verbose & APIC_VERBOSE_IRQ_FLAG)
944 		acpi_verboseflags |= PSM_VERBOSE_IRQ_FLAG;
945 
946 	if (apic_verbose & APIC_VERBOSE_POWEROFF_FLAG)
947 		acpi_verboseflags |= PSM_VERBOSE_POWEROFF_FLAG;
948 
949 	if (apic_verbose & APIC_VERBOSE_POWEROFF_PAUSE_FLAG)
950 		acpi_verboseflags |= PSM_VERBOSE_POWEROFF_PAUSE_FLAG;
951 
952 	if (acpi_psm_init(modname, acpi_verboseflags) == ACPI_PSM_FAILURE)
953 		goto cleanup;
954 
955 	/* Enable ACPI APIC interrupt routing */
956 	if (apic_acpi_enter_apicmode() != PSM_FAILURE) {
957 		cmn_err(CE_NOTE, "!apic: Using APIC interrupt routing mode");
958 		build_reserved_irqlist((uchar_t *)apic_reserved_irqlist);
959 		apic_enable_acpi = 1;
960 		if (apic_sci_vect > 0) {
961 			acpica_set_core_feature(ACPI_FEATURE_SCI_EVENT);
962 		}
963 		if (apic_use_acpi_madt_only) {
964 			cmn_err(CE_CONT,
965 			    "?Using ACPI for CPU/IOAPIC information ONLY\n");
966 		}
967 
968 #if !defined(__xpv)
969 		/*
970 		 * probe ACPI for hpet information here which is used later
971 		 * in apic_picinit().
972 		 */
973 		if (hpet_acpi_init(&apic_hpet_vect, &apic_hpet_flags) < 0) {
974 			cmn_err(CE_NOTE, "!ACPI HPET table query failed\n");
975 		}
976 #endif
977 
978 		kmem_free(local_ids, NCPU * sizeof (uint32_t));
979 		kmem_free(proc_ids, NCPU * sizeof (uint32_t));
980 		return (PSM_SUCCESS);
981 	}
982 	/* if setting APIC mode failed above, we fall through to cleanup */
983 
984 cleanup:
985 	cmn_err(CE_WARN, "!apic: Failed acpi_probe, SMP config was %s",
986 	    acpi_found_smp_config ? "found" : "not found");
987 	apic_free_apic_cpus();
988 	if (apicadr != NULL) {
989 		mapout_apic((caddr_t)apicadr, APIC_LOCAL_MEMLEN);
990 		apicadr = NULL;
991 	}
992 	apic_max_nproc = -1;
993 	apic_nproc = 0;
994 	for (i = 0; i < apic_io_max; i++) {
995 		mapout_ioapic((caddr_t)apicioadr[i], APIC_IO_MEMLEN);
996 		apicioadr[i] = NULL;
997 	}
998 	apic_io_max = 0;
999 	acpi_isop = NULL;
1000 	acpi_iso_cnt = 0;
1001 	acpi_nmi_sp = NULL;
1002 	acpi_nmi_scnt = 0;
1003 	acpi_nmi_cp = NULL;
1004 	acpi_nmi_ccnt = 0;
1005 	acpi_found_smp_config = B_FALSE;
1006 	kmem_free(local_ids, NCPU * sizeof (uint32_t));
1007 	kmem_free(proc_ids, NCPU * sizeof (uint32_t));
1008 	return (PSM_FAILURE);
1009 }
1010 
1011 /*
1012  * Handle default configuration. Fill in reqd global variables & tables
1013  * Fill all details as MP table does not give any more info
1014  */
1015 static int
1016 apic_handle_defconf()
1017 {
1018 	uint_t	lid;
1019 
1020 	/* Failed to probe ACPI MADT tables, disable CPU DR. */
1021 	apic_max_nproc = -1;
1022 	apic_free_apic_cpus();
1023 	plat_dr_disable_cpu();
1024 
1025 	apicioadr[0] = (void *)mapin_ioapic(APIC_IO_ADDR,
1026 	    APIC_IO_MEMLEN, PROT_READ | PROT_WRITE);
1027 	apicadr = (void *)psm_map_phys(APIC_LOCAL_ADDR,
1028 	    APIC_LOCAL_MEMLEN, PROT_READ);
1029 	apic_cpus_size = 2 * sizeof (*apic_cpus);
1030 	apic_cpus = (apic_cpus_info_t *)
1031 	    kmem_zalloc(apic_cpus_size, KM_NOSLEEP);
1032 	if ((!apicadr) || (!apicioadr[0]) || (!apic_cpus))
1033 		goto apic_handle_defconf_fail;
1034 	CPUSET_ONLY(apic_cpumask, 0);
1035 	CPUSET_ADD(apic_cpumask, 1);
1036 	apic_nproc = 2;
1037 	lid = apic_reg_ops->apic_read(APIC_LID_REG);
1038 	apic_cpus[0].aci_local_id = (uchar_t)(lid >> APIC_ID_BIT_OFFSET);
1039 	/*
1040 	 * According to the PC+MP spec 1.1, the local ids
1041 	 * for the default configuration has to be 0 or 1
1042 	 */
1043 	if (apic_cpus[0].aci_local_id == 1)
1044 		apic_cpus[1].aci_local_id = 0;
1045 	else if (apic_cpus[0].aci_local_id == 0)
1046 		apic_cpus[1].aci_local_id = 1;
1047 	else
1048 		goto apic_handle_defconf_fail;
1049 
1050 	apic_io_id[0] = 2;
1051 	apic_io_max = 1;
1052 	if (apic_defconf >= 5) {
1053 		apic_cpus[0].aci_local_ver = APIC_INTEGRATED_VERS;
1054 		apic_cpus[1].aci_local_ver = APIC_INTEGRATED_VERS;
1055 		apic_io_ver[0] = APIC_INTEGRATED_VERS;
1056 	} else {
1057 		apic_cpus[0].aci_local_ver = 0;		/* 82489 DX */
1058 		apic_cpus[1].aci_local_ver = 0;
1059 		apic_io_ver[0] = 0;
1060 	}
1061 	if (apic_defconf == 2 || apic_defconf == 3 || apic_defconf == 6)
1062 		eisa_level_intr_mask = (inb(EISA_LEVEL_CNTL + 1) << 8) |
1063 		    inb(EISA_LEVEL_CNTL) | ((uint_t)INT32_MAX + 1);
1064 	return (PSM_SUCCESS);
1065 
1066 apic_handle_defconf_fail:
1067 	if (apicadr)
1068 		mapout_apic((caddr_t)apicadr, APIC_LOCAL_MEMLEN);
1069 	if (apicioadr[0])
1070 		mapout_ioapic((caddr_t)apicioadr[0], APIC_IO_MEMLEN);
1071 	return (PSM_FAILURE);
1072 }
1073 
1074 /* Parse the entries in MP configuration table and collect info that we need */
1075 static int
1076 apic_parse_mpct(caddr_t mpct, int bypass_cpus_and_ioapics)
1077 {
1078 	struct	apic_procent	*procp;
1079 	struct	apic_bus	*busp;
1080 	struct	apic_io_entry	*ioapicp;
1081 	struct	apic_io_intr	*intrp;
1082 	int			ioapic_ix;
1083 	uint_t	lid;
1084 	uint32_t	id;
1085 	uchar_t hid;
1086 	int	warned = 0;
1087 
1088 	/*LINTED: pointer cast may result in improper alignment */
1089 	procp = (struct apic_procent *)(mpct + sizeof (struct apic_mp_cnf_hdr));
1090 
1091 	/* No need to count cpu entries if we won't use them */
1092 	if (!bypass_cpus_and_ioapics) {
1093 
1094 		/* Find max # of CPUS and allocate structure accordingly */
1095 		apic_nproc = 0;
1096 		CPUSET_ZERO(apic_cpumask);
1097 		while (procp->proc_entry == APIC_CPU_ENTRY) {
1098 			if (procp->proc_cpuflags & CPUFLAGS_EN) {
1099 				if (apic_nproc < NCPU && use_mp &&
1100 				    apic_nproc < boot_ncpus) {
1101 					CPUSET_ADD(apic_cpumask, apic_nproc);
1102 					apic_nproc++;
1103 				} else if (apic_nproc == NCPU && !warned) {
1104 					cmn_err(CE_WARN, "%s: CPU limit "
1105 					    "exceeded"
1106 #if !defined(__amd64)
1107 					    " for 32-bit mode"
1108 #endif
1109 					    "; Solaris will use %d CPUs.",
1110 					    psm_name,  NCPU);
1111 					warned = 1;
1112 				}
1113 
1114 			}
1115 			procp++;
1116 		}
1117 		apic_cpus_size = apic_nproc * sizeof (*apic_cpus);
1118 		if (!apic_nproc || !(apic_cpus = (apic_cpus_info_t *)
1119 		    kmem_zalloc(apic_cpus_size, KM_NOSLEEP)))
1120 			return (PSM_FAILURE);
1121 	}
1122 
1123 	/*LINTED: pointer cast may result in improper alignment */
1124 	procp = (struct apic_procent *)(mpct + sizeof (struct apic_mp_cnf_hdr));
1125 
1126 	/*
1127 	 * start with index 1 as 0 needs to be filled in with Boot CPU, but
1128 	 * if we're bypassing this information, it has already been filled
1129 	 * in by acpi_probe(), so don't overwrite it.
1130 	 */
1131 	if (!bypass_cpus_and_ioapics)
1132 		apic_nproc = 1;
1133 
1134 	while (procp->proc_entry == APIC_CPU_ENTRY) {
1135 		/* check whether the cpu exists or not */
1136 		if (!bypass_cpus_and_ioapics &&
1137 		    procp->proc_cpuflags & CPUFLAGS_EN) {
1138 			if (procp->proc_cpuflags & CPUFLAGS_BP) { /* Boot CPU */
1139 				lid = apic_reg_ops->apic_read(APIC_LID_REG);
1140 				apic_cpus[0].aci_local_id = procp->proc_apicid;
1141 				if (apic_cpus[0].aci_local_id !=
1142 				    (uchar_t)(lid >> APIC_ID_BIT_OFFSET)) {
1143 					return (PSM_FAILURE);
1144 				}
1145 				apic_cpus[0].aci_local_ver =
1146 				    procp->proc_version;
1147 			} else if (apic_nproc < NCPU && use_mp &&
1148 			    apic_nproc < boot_ncpus) {
1149 				apic_cpus[apic_nproc].aci_local_id =
1150 				    procp->proc_apicid;
1151 
1152 				apic_cpus[apic_nproc].aci_local_ver =
1153 				    procp->proc_version;
1154 				apic_nproc++;
1155 
1156 			}
1157 		}
1158 		procp++;
1159 	}
1160 
1161 	/*
1162 	 * Save start of bus entries for later use.
1163 	 * Get EISA level cntrl if EISA bus is present.
1164 	 * Also get the CPI bus id for single CPI bus case
1165 	 */
1166 	apic_busp = busp = (struct apic_bus *)procp;
1167 	while (busp->bus_entry == APIC_BUS_ENTRY) {
1168 		lid = apic_find_bus_type((char *)&busp->bus_str1);
1169 		if (lid	== BUS_EISA) {
1170 			eisa_level_intr_mask = (inb(EISA_LEVEL_CNTL + 1) << 8) |
1171 			    inb(EISA_LEVEL_CNTL) | ((uint_t)INT32_MAX + 1);
1172 		} else if (lid == BUS_PCI) {
1173 			/*
1174 			 * apic_single_pci_busid will be used only if
1175 			 * apic_pic_bus_total is equal to 1
1176 			 */
1177 			apic_pci_bus_total++;
1178 			apic_single_pci_busid = busp->bus_id;
1179 		}
1180 		busp++;
1181 	}
1182 
1183 	ioapicp = (struct apic_io_entry *)busp;
1184 
1185 	if (!bypass_cpus_and_ioapics)
1186 		apic_io_max = 0;
1187 	do {
1188 		if (!bypass_cpus_and_ioapics && apic_io_max < MAX_IO_APIC) {
1189 			if (ioapicp->io_flags & IOAPIC_FLAGS_EN) {
1190 				apic_io_id[apic_io_max] = ioapicp->io_apicid;
1191 				apic_io_ver[apic_io_max] = ioapicp->io_version;
1192 				apicioadr[apic_io_max] =
1193 				    (void *)mapin_ioapic(
1194 				    (uint32_t)ioapicp->io_apic_addr,
1195 				    APIC_IO_MEMLEN, PROT_READ | PROT_WRITE);
1196 
1197 				if (!apicioadr[apic_io_max])
1198 					return (PSM_FAILURE);
1199 
1200 				ioapic_mask_workaround[apic_io_max] =
1201 				    apic_is_ioapic_AMD_813x(
1202 				    ioapicp->io_apic_addr);
1203 
1204 				ioapic_ix = apic_io_max;
1205 				id = ioapic_read(ioapic_ix, APIC_ID_CMD);
1206 				hid = (uchar_t)(id >> 24);
1207 
1208 				if (hid != apic_io_id[apic_io_max]) {
1209 					if (apic_io_id[apic_io_max] == 0)
1210 						apic_io_id[apic_io_max] = hid;
1211 					else {
1212 						/*
1213 						 * set ioapic id to whatever
1214 						 * reported by MPS
1215 						 *
1216 						 * may not need to set index
1217 						 * again ???
1218 						 * take it out and try
1219 						 */
1220 
1221 						id = ((uint32_t)
1222 						    apic_io_id[apic_io_max]) <<
1223 						    24;
1224 
1225 						ioapic_write(ioapic_ix,
1226 						    APIC_ID_CMD, id);
1227 					}
1228 				}
1229 				apic_io_max++;
1230 			}
1231 		}
1232 		ioapicp++;
1233 	} while (ioapicp->io_entry == APIC_IO_ENTRY);
1234 
1235 	apic_io_intrp = (struct apic_io_intr *)ioapicp;
1236 
1237 	intrp = apic_io_intrp;
1238 	while (intrp->intr_entry == APIC_IO_INTR_ENTRY) {
1239 		if ((intrp->intr_irq > APIC_MAX_ISA_IRQ) ||
1240 		    (apic_find_bus(intrp->intr_busid) == BUS_PCI)) {
1241 			apic_irq_translate = 1;
1242 			break;
1243 		}
1244 		intrp++;
1245 	}
1246 
1247 	return (PSM_SUCCESS);
1248 }
1249 
1250 boolean_t
1251 apic_cpu_in_range(int cpu)
1252 {
1253 	cpu &= ~IRQ_USER_BOUND;
1254 	/* Check whether cpu id is in valid range. */
1255 	if (cpu < 0 || cpu >= apic_nproc) {
1256 		return (B_FALSE);
1257 	} else if (apic_max_nproc != -1 && cpu >= apic_max_nproc) {
1258 		/*
1259 		 * Check whether cpuid is in valid range if CPU DR is enabled.
1260 		 */
1261 		return (B_FALSE);
1262 	} else if (!CPU_IN_SET(apic_cpumask, cpu)) {
1263 		return (B_FALSE);
1264 	}
1265 
1266 	return (B_TRUE);
1267 }
1268 
1269 processorid_t
1270 apic_get_next_bind_cpu(void)
1271 {
1272 	int i, count;
1273 	processorid_t cpuid = 0;
1274 
1275 	for (count = 0; count < apic_nproc; count++) {
1276 		if (apic_next_bind_cpu >= apic_nproc) {
1277 			apic_next_bind_cpu = 0;
1278 		}
1279 		i = apic_next_bind_cpu++;
1280 		if (apic_cpu_in_range(i)) {
1281 			cpuid = i;
1282 			break;
1283 		}
1284 	}
1285 
1286 	return (cpuid);
1287 }
1288 
1289 uint16_t
1290 apic_get_apic_version()
1291 {
1292 	int i;
1293 	uchar_t min_io_apic_ver = 0;
1294 	static uint16_t version;		/* Cache as value is constant */
1295 	static boolean_t found = B_FALSE;	/* Accomodate zero version */
1296 
1297 	if (found == B_FALSE) {
1298 		found = B_TRUE;
1299 
1300 		/*
1301 		 * Don't assume all IO APICs in the system are the same.
1302 		 *
1303 		 * Set to the minimum version.
1304 		 */
1305 		for (i = 0; i < apic_io_max; i++) {
1306 			if ((apic_io_ver[i] != 0) &&
1307 			    ((min_io_apic_ver == 0) ||
1308 			    (min_io_apic_ver >= apic_io_ver[i])))
1309 				min_io_apic_ver = apic_io_ver[i];
1310 		}
1311 
1312 		/* Assume all local APICs are of the same version. */
1313 		version = (min_io_apic_ver << 8) | apic_cpus[0].aci_local_ver;
1314 	}
1315 	return (version);
1316 }
1317 
1318 static struct apic_mpfps_hdr *
1319 apic_find_fps_sig(caddr_t cptr, int len)
1320 {
1321 	int	i;
1322 
1323 	/* Look for the pattern "_MP_" */
1324 	for (i = 0; i < len; i += 16) {
1325 		if ((*(cptr+i) == '_') &&
1326 		    (*(cptr+i+1) == 'M') &&
1327 		    (*(cptr+i+2) == 'P') &&
1328 		    (*(cptr+i+3) == '_'))
1329 		    /*LINTED: pointer cast may result in improper alignment */
1330 			return ((struct apic_mpfps_hdr *)(cptr + i));
1331 	}
1332 	return (NULL);
1333 }
1334 
1335 static int
1336 apic_checksum(caddr_t bptr, int len)
1337 {
1338 	int	i;
1339 	uchar_t	cksum;
1340 
1341 	cksum = 0;
1342 	for (i = 0; i < len; i++)
1343 		cksum += *bptr++;
1344 	return ((int)cksum);
1345 }
1346 
1347 /*
1348  * On machines with PCI-PCI bridges, a device behind a PCI-PCI bridge
1349  * needs special handling.  We may need to chase up the device tree,
1350  * using the PCI-PCI Bridge specification's "rotating IPIN assumptions",
1351  * to find the IPIN at the root bus that relates to the IPIN on the
1352  * subsidiary bus (for ACPI or MP).  We may, however, have an entry
1353  * in the MP table or the ACPI namespace for this device itself.
1354  * We handle both cases in the search below.
1355  */
1356 /* this is the non-acpi version */
1357 int
1358 apic_handle_pci_pci_bridge(dev_info_t *idip, int child_devno, int child_ipin,
1359     struct apic_io_intr **intrp)
1360 {
1361 	dev_info_t *dipp, *dip;
1362 	int pci_irq;
1363 	ddi_acc_handle_t cfg_handle;
1364 	int bridge_devno, bridge_bus;
1365 	int ipin;
1366 
1367 	dip = idip;
1368 
1369 	/*CONSTCOND*/
1370 	while (1) {
1371 		if (((dipp = ddi_get_parent(dip)) == (dev_info_t *)NULL) ||
1372 		    (pci_config_setup(dipp, &cfg_handle) != DDI_SUCCESS))
1373 			return (-1);
1374 		if ((pci_config_get8(cfg_handle, PCI_CONF_BASCLASS) ==
1375 		    PCI_CLASS_BRIDGE) && (pci_config_get8(cfg_handle,
1376 		    PCI_CONF_SUBCLASS) == PCI_BRIDGE_PCI)) {
1377 			pci_config_teardown(&cfg_handle);
1378 			if (acpica_get_bdf(dipp, &bridge_bus, &bridge_devno,
1379 			    NULL) != 0)
1380 				return (-1);
1381 			/*
1382 			 * This is the rotating scheme documented in the
1383 			 * PCI-to-PCI spec.  If the PCI-to-PCI bridge is
1384 			 * behind another PCI-to-PCI bridge, then it needs
1385 			 * to keep ascending until an interrupt entry is
1386 			 * found or the root is reached.
1387 			 */
1388 			ipin = (child_devno + child_ipin) % PCI_INTD;
1389 				if (bridge_bus == 0 && apic_pci_bus_total == 1)
1390 					bridge_bus = (int)apic_single_pci_busid;
1391 				pci_irq = ((bridge_devno & 0x1f) << 2) |
1392 				    (ipin & 0x3);
1393 				if ((*intrp = apic_find_io_intr_w_busid(pci_irq,
1394 				    bridge_bus)) != NULL) {
1395 					return (pci_irq);
1396 				}
1397 			dip = dipp;
1398 			child_devno = bridge_devno;
1399 			child_ipin = ipin;
1400 		} else {
1401 			pci_config_teardown(&cfg_handle);
1402 			return (-1);
1403 		}
1404 	}
1405 	/*LINTED: function will not fall off the bottom */
1406 }
1407 
1408 uchar_t
1409 acpi_find_ioapic(int irq)
1410 {
1411 	int i;
1412 
1413 	for (i = 0; i < apic_io_max; i++) {
1414 		if (irq >= apic_io_vectbase[i] && irq <= apic_io_vectend[i])
1415 			return ((uchar_t)i);
1416 	}
1417 	return (0xFF);	/* shouldn't happen */
1418 }
1419 
1420 /*
1421  * See if two irqs are compatible for sharing a vector.
1422  * Currently we only support sharing of PCI devices.
1423  */
1424 static int
1425 acpi_intr_compatible(iflag_t iflag1, iflag_t iflag2)
1426 {
1427 	uint_t	level1, po1;
1428 	uint_t	level2, po2;
1429 
1430 	/* Assume active high by default */
1431 	po1 = 0;
1432 	po2 = 0;
1433 
1434 	if (iflag1.bustype != iflag2.bustype || iflag1.bustype != BUS_PCI)
1435 		return (0);
1436 
1437 	if (iflag1.intr_el == INTR_EL_CONFORM)
1438 		level1 = AV_LEVEL;
1439 	else
1440 		level1 = (iflag1.intr_el == INTR_EL_LEVEL) ? AV_LEVEL : 0;
1441 
1442 	if (level1 && ((iflag1.intr_po == INTR_PO_ACTIVE_LOW) ||
1443 	    (iflag1.intr_po == INTR_PO_CONFORM)))
1444 		po1 = AV_ACTIVE_LOW;
1445 
1446 	if (iflag2.intr_el == INTR_EL_CONFORM)
1447 		level2 = AV_LEVEL;
1448 	else
1449 		level2 = (iflag2.intr_el == INTR_EL_LEVEL) ? AV_LEVEL : 0;
1450 
1451 	if (level2 && ((iflag2.intr_po == INTR_PO_ACTIVE_LOW) ||
1452 	    (iflag2.intr_po == INTR_PO_CONFORM)))
1453 		po2 = AV_ACTIVE_LOW;
1454 
1455 	if ((level1 == level2) && (po1 == po2))
1456 		return (1);
1457 
1458 	return (0);
1459 }
1460 
1461 struct apic_io_intr *
1462 apic_find_io_intr_w_busid(int irqno, int busid)
1463 {
1464 	struct	apic_io_intr	*intrp;
1465 
1466 	/*
1467 	 * It can have more than 1 entry with same source bus IRQ,
1468 	 * but unique with the source bus id
1469 	 */
1470 	intrp = apic_io_intrp;
1471 	if (intrp != NULL) {
1472 		while (intrp->intr_entry == APIC_IO_INTR_ENTRY) {
1473 			if (intrp->intr_irq == irqno &&
1474 			    intrp->intr_busid == busid &&
1475 			    intrp->intr_type == IO_INTR_INT)
1476 				return (intrp);
1477 			intrp++;
1478 		}
1479 	}
1480 	APIC_VERBOSE_IOAPIC((CE_NOTE, "Did not find io intr for irqno:"
1481 	    "busid %x:%x\n", irqno, busid));
1482 	return ((struct apic_io_intr *)NULL);
1483 }
1484 
1485 
1486 struct mps_bus_info {
1487 	char	*bus_name;
1488 	int	bus_id;
1489 } bus_info_array[] = {
1490 	"ISA ", BUS_ISA,
1491 	"PCI ", BUS_PCI,
1492 	"EISA ", BUS_EISA,
1493 	"XPRESS", BUS_XPRESS,
1494 	"PCMCIA", BUS_PCMCIA,
1495 	"VL ", BUS_VL,
1496 	"CBUS ", BUS_CBUS,
1497 	"CBUSII", BUS_CBUSII,
1498 	"FUTURE", BUS_FUTURE,
1499 	"INTERN", BUS_INTERN,
1500 	"MBI ", BUS_MBI,
1501 	"MBII ", BUS_MBII,
1502 	"MPI ", BUS_MPI,
1503 	"MPSA ", BUS_MPSA,
1504 	"NUBUS ", BUS_NUBUS,
1505 	"TC ", BUS_TC,
1506 	"VME ", BUS_VME,
1507 	"PCI-E ", BUS_PCIE
1508 };
1509 
1510 static int
1511 apic_find_bus_type(char *bus)
1512 {
1513 	int	i = 0;
1514 
1515 	for (; i < sizeof (bus_info_array)/sizeof (struct mps_bus_info); i++)
1516 		if (strncmp(bus, bus_info_array[i].bus_name,
1517 		    strlen(bus_info_array[i].bus_name)) == 0)
1518 			return (bus_info_array[i].bus_id);
1519 	APIC_VERBOSE_IOAPIC((CE_WARN, "Did not find bus type for bus %s", bus));
1520 	return (0);
1521 }
1522 
1523 static int
1524 apic_find_bus(int busid)
1525 {
1526 	struct	apic_bus	*busp;
1527 
1528 	busp = apic_busp;
1529 	while (busp->bus_entry == APIC_BUS_ENTRY) {
1530 		if (busp->bus_id == busid)
1531 			return (apic_find_bus_type((char *)&busp->bus_str1));
1532 		busp++;
1533 	}
1534 	APIC_VERBOSE_IOAPIC((CE_WARN, "Did not find bus for bus id %x", busid));
1535 	return (0);
1536 }
1537 
1538 int
1539 apic_find_bus_id(int bustype)
1540 {
1541 	struct	apic_bus	*busp;
1542 
1543 	busp = apic_busp;
1544 	while (busp->bus_entry == APIC_BUS_ENTRY) {
1545 		if (apic_find_bus_type((char *)&busp->bus_str1) == bustype)
1546 			return (busp->bus_id);
1547 		busp++;
1548 	}
1549 	APIC_VERBOSE_IOAPIC((CE_WARN, "Did not find bus id for bustype %x",
1550 	    bustype));
1551 	return (-1);
1552 }
1553 
1554 /*
1555  * Check if a particular irq need to be reserved for any io_intr
1556  */
1557 static struct apic_io_intr *
1558 apic_find_io_intr(int irqno)
1559 {
1560 	struct	apic_io_intr	*intrp;
1561 
1562 	intrp = apic_io_intrp;
1563 	if (intrp != NULL) {
1564 		while (intrp->intr_entry == APIC_IO_INTR_ENTRY) {
1565 			if (intrp->intr_irq == irqno &&
1566 			    intrp->intr_type == IO_INTR_INT)
1567 				return (intrp);
1568 			intrp++;
1569 		}
1570 	}
1571 	return ((struct apic_io_intr *)NULL);
1572 }
1573 
1574 /*
1575  * Check if the given ioapicindex intin combination has already been assigned
1576  * an irq. If so return irqno. Else -1
1577  */
1578 int
1579 apic_find_intin(uchar_t ioapic, uchar_t intin)
1580 {
1581 	apic_irq_t *irqptr;
1582 	int	i;
1583 
1584 	/* find ioapic and intin in the apic_irq_table[] and return the index */
1585 	for (i = apic_min_device_irq; i <= apic_max_device_irq; i++) {
1586 		irqptr = apic_irq_table[i];
1587 		while (irqptr) {
1588 			if ((irqptr->airq_mps_intr_index >= 0) &&
1589 			    (irqptr->airq_intin_no == intin) &&
1590 			    (irqptr->airq_ioapicindex == ioapic)) {
1591 				APIC_VERBOSE_IOAPIC((CE_NOTE, "!Found irq "
1592 				    "entry for ioapic:intin %x:%x "
1593 				    "shared interrupts ?", ioapic, intin));
1594 				return (i);
1595 			}
1596 			irqptr = irqptr->airq_next;
1597 		}
1598 	}
1599 	return (-1);
1600 }
1601 
1602 int
1603 apic_allocate_irq(int irq)
1604 {
1605 	int	freeirq, i;
1606 
1607 	if ((freeirq = apic_find_free_irq(irq, (APIC_RESV_IRQ - 1))) == -1)
1608 		if ((freeirq = apic_find_free_irq(APIC_FIRST_FREE_IRQ,
1609 		    (irq - 1))) == -1) {
1610 			/*
1611 			 * if BIOS really defines every single irq in the mps
1612 			 * table, then don't worry about conflicting with
1613 			 * them, just use any free slot in apic_irq_table
1614 			 */
1615 			for (i = APIC_FIRST_FREE_IRQ; i < APIC_RESV_IRQ; i++) {
1616 				if ((apic_irq_table[i] == NULL) ||
1617 				    apic_irq_table[i]->airq_mps_intr_index ==
1618 				    FREE_INDEX) {
1619 				freeirq = i;
1620 				break;
1621 			}
1622 		}
1623 		if (freeirq == -1) {
1624 			/* This shouldn't happen, but just in case */
1625 			cmn_err(CE_WARN, "%s: NO available IRQ", psm_name);
1626 			return (-1);
1627 		}
1628 	}
1629 	if (apic_irq_table[freeirq] == NULL) {
1630 		apic_irq_table[freeirq] =
1631 		    kmem_zalloc(sizeof (apic_irq_t), KM_NOSLEEP);
1632 		if (apic_irq_table[freeirq] == NULL) {
1633 			cmn_err(CE_WARN, "%s: NO memory to allocate IRQ",
1634 			    psm_name);
1635 			return (-1);
1636 		}
1637 		apic_irq_table[freeirq]->airq_temp_cpu = IRQ_UNINIT;
1638 		apic_irq_table[freeirq]->airq_mps_intr_index = FREE_INDEX;
1639 	}
1640 	return (freeirq);
1641 }
1642 
1643 static int
1644 apic_find_free_irq(int start, int end)
1645 {
1646 	int	i;
1647 
1648 	for (i = start; i <= end; i++)
1649 		/* Check if any I/O entry needs this IRQ */
1650 		if (apic_find_io_intr(i) == NULL) {
1651 			/* Then see if it is free */
1652 			if ((apic_irq_table[i] == NULL) ||
1653 			    (apic_irq_table[i]->airq_mps_intr_index ==
1654 			    FREE_INDEX)) {
1655 				return (i);
1656 			}
1657 		}
1658 	return (-1);
1659 }
1660 
1661 /*
1662  * compute the polarity, trigger mode and vector for programming into
1663  * the I/O apic and record in airq_rdt_entry.
1664  */
1665 void
1666 apic_record_rdt_entry(apic_irq_t *irqptr, int irq)
1667 {
1668 	int	ioapicindex, bus_type, vector;
1669 	short	intr_index;
1670 	uint_t	level, po, io_po;
1671 	struct apic_io_intr *iointrp;
1672 
1673 	intr_index = irqptr->airq_mps_intr_index;
1674 	DDI_INTR_IMPLDBG((CE_CONT, "apic_record_rdt_entry: intr_index=%d "
1675 	    "irq = 0x%x dip = 0x%p vector = 0x%x\n", intr_index, irq,
1676 	    (void *)irqptr->airq_dip, irqptr->airq_vector));
1677 
1678 	if (intr_index == RESERVE_INDEX) {
1679 		apic_error |= APIC_ERR_INVALID_INDEX;
1680 		return;
1681 	} else if (APIC_IS_MSI_OR_MSIX_INDEX(intr_index)) {
1682 		return;
1683 	}
1684 
1685 	vector = irqptr->airq_vector;
1686 	ioapicindex = irqptr->airq_ioapicindex;
1687 	/* Assume edge triggered by default */
1688 	level = 0;
1689 	/* Assume active high by default */
1690 	po = 0;
1691 
1692 	if (intr_index == DEFAULT_INDEX || intr_index == FREE_INDEX) {
1693 		ASSERT(irq < 16);
1694 		if (eisa_level_intr_mask & (1 << irq))
1695 			level = AV_LEVEL;
1696 		if (intr_index == FREE_INDEX && apic_defconf == 0)
1697 			apic_error |= APIC_ERR_INVALID_INDEX;
1698 	} else if (intr_index == ACPI_INDEX) {
1699 		bus_type = irqptr->airq_iflag.bustype;
1700 		if (irqptr->airq_iflag.intr_el == INTR_EL_CONFORM) {
1701 			if (bus_type == BUS_PCI)
1702 				level = AV_LEVEL;
1703 		} else
1704 			level = (irqptr->airq_iflag.intr_el == INTR_EL_LEVEL) ?
1705 			    AV_LEVEL : 0;
1706 		if (level &&
1707 		    ((irqptr->airq_iflag.intr_po == INTR_PO_ACTIVE_LOW) ||
1708 		    (irqptr->airq_iflag.intr_po == INTR_PO_CONFORM &&
1709 		    bus_type == BUS_PCI)))
1710 			po = AV_ACTIVE_LOW;
1711 	} else {
1712 		iointrp = apic_io_intrp + intr_index;
1713 		bus_type = apic_find_bus(iointrp->intr_busid);
1714 		if (iointrp->intr_el == INTR_EL_CONFORM) {
1715 			if ((irq < 16) && (eisa_level_intr_mask & (1 << irq)))
1716 				level = AV_LEVEL;
1717 			else if (bus_type == BUS_PCI)
1718 				level = AV_LEVEL;
1719 		} else
1720 			level = (iointrp->intr_el == INTR_EL_LEVEL) ?
1721 			    AV_LEVEL : 0;
1722 		if (level && ((iointrp->intr_po == INTR_PO_ACTIVE_LOW) ||
1723 		    (iointrp->intr_po == INTR_PO_CONFORM &&
1724 		    bus_type == BUS_PCI)))
1725 			po = AV_ACTIVE_LOW;
1726 	}
1727 	if (level)
1728 		apic_level_intr[irq] = 1;
1729 	/*
1730 	 * The 82489DX External APIC cannot do active low polarity interrupts.
1731 	 */
1732 	if (po && (apic_io_ver[ioapicindex] != IOAPIC_VER_82489DX))
1733 		io_po = po;
1734 	else
1735 		io_po = 0;
1736 
1737 	if (apic_verbose & APIC_VERBOSE_IOAPIC_FLAG)
1738 		prom_printf("setio: ioapic=0x%x intin=0x%x level=0x%x po=0x%x "
1739 		    "vector=0x%x cpu=0x%x\n\n", ioapicindex,
1740 		    irqptr->airq_intin_no, level, io_po, vector,
1741 		    irqptr->airq_cpu);
1742 
1743 	irqptr->airq_rdt_entry = level|io_po|vector;
1744 }
1745 
1746 int
1747 apic_acpi_translate_pci_irq(dev_info_t *dip, int busid, int devid,
1748     int ipin, int *pci_irqp, iflag_t *intr_flagp)
1749 {
1750 
1751 	int status;
1752 	acpi_psm_lnk_t acpipsmlnk;
1753 
1754 	if ((status = acpi_get_irq_cache_ent(busid, devid, ipin, pci_irqp,
1755 	    intr_flagp)) == ACPI_PSM_SUCCESS) {
1756 		APIC_VERBOSE_IRQ((CE_CONT, "!%s: Found irqno %d "
1757 		    "from cache for device %s, instance #%d\n", psm_name,
1758 		    *pci_irqp, ddi_get_name(dip), ddi_get_instance(dip)));
1759 		return (status);
1760 	}
1761 
1762 	bzero(&acpipsmlnk, sizeof (acpi_psm_lnk_t));
1763 
1764 	if ((status = acpi_translate_pci_irq(dip, ipin, pci_irqp, intr_flagp,
1765 	    &acpipsmlnk)) == ACPI_PSM_FAILURE) {
1766 		APIC_VERBOSE_IRQ((CE_WARN, "%s: "
1767 		    " acpi_translate_pci_irq failed for device %s, instance"
1768 		    " #%d", psm_name, ddi_get_name(dip),
1769 		    ddi_get_instance(dip)));
1770 		return (status);
1771 	}
1772 
1773 	if (status == ACPI_PSM_PARTIAL && acpipsmlnk.lnkobj != NULL) {
1774 		status = apic_acpi_irq_configure(&acpipsmlnk, dip, pci_irqp,
1775 		    intr_flagp);
1776 		if (status != ACPI_PSM_SUCCESS) {
1777 			status = acpi_get_current_irq_resource(&acpipsmlnk,
1778 			    pci_irqp, intr_flagp);
1779 		}
1780 	}
1781 
1782 	if (status == ACPI_PSM_SUCCESS) {
1783 		acpi_new_irq_cache_ent(busid, devid, ipin, *pci_irqp,
1784 		    intr_flagp, &acpipsmlnk);
1785 
1786 		APIC_VERBOSE_IRQ((CE_CONT, "%s: [ACPI] "
1787 		    "new irq %d for device %s, instance #%d\n", psm_name,
1788 		    *pci_irqp, ddi_get_name(dip), ddi_get_instance(dip)));
1789 	}
1790 
1791 	return (status);
1792 }
1793 
1794 /*
1795  * Adds an entry to the irq list passed in, and returns the new list.
1796  * Entries are added in priority order (lower numerical priorities are
1797  * placed closer to the head of the list)
1798  */
1799 static prs_irq_list_t *
1800 acpi_insert_prs_irq_ent(prs_irq_list_t *listp, int priority, int irq,
1801     iflag_t *iflagp, acpi_prs_private_t *prsprvp)
1802 {
1803 	struct prs_irq_list_ent *newent, *prevp = NULL, *origlistp;
1804 
1805 	newent = kmem_zalloc(sizeof (struct prs_irq_list_ent), KM_SLEEP);
1806 
1807 	newent->list_prio = priority;
1808 	newent->irq = irq;
1809 	newent->intrflags = *iflagp;
1810 	newent->prsprv = *prsprvp;
1811 	/* ->next is NULL from kmem_zalloc */
1812 
1813 	/*
1814 	 * New list -- return the new entry as the list.
1815 	 */
1816 	if (listp == NULL)
1817 		return (newent);
1818 
1819 	/*
1820 	 * Save original list pointer for return (since we're not modifying
1821 	 * the head)
1822 	 */
1823 	origlistp = listp;
1824 
1825 	/*
1826 	 * Insertion sort, with entries with identical keys stored AFTER
1827 	 * existing entries (the less-than-or-equal test of priority does
1828 	 * this for us).
1829 	 */
1830 	while (listp != NULL && listp->list_prio <= priority) {
1831 		prevp = listp;
1832 		listp = listp->next;
1833 	}
1834 
1835 	newent->next = listp;
1836 
1837 	if (prevp == NULL) { /* Add at head of list (newent is the new head) */
1838 		return (newent);
1839 	} else {
1840 		prevp->next = newent;
1841 		return (origlistp);
1842 	}
1843 }
1844 
1845 /*
1846  * Frees the list passed in, deallocating all memory and leaving *listpp
1847  * set to NULL.
1848  */
1849 static void
1850 acpi_destroy_prs_irq_list(prs_irq_list_t **listpp)
1851 {
1852 	struct prs_irq_list_ent *nextp;
1853 
1854 	ASSERT(listpp != NULL);
1855 
1856 	while (*listpp != NULL) {
1857 		nextp = (*listpp)->next;
1858 		kmem_free(*listpp, sizeof (struct prs_irq_list_ent));
1859 		*listpp = nextp;
1860 	}
1861 }
1862 
1863 /*
1864  * apic_choose_irqs_from_prs returns a list of irqs selected from the list of
1865  * irqs returned by the link device's _PRS method.  The irqs are chosen
1866  * to minimize contention in situations where the interrupt link device
1867  * can be programmed to steer interrupts to different interrupt controller
1868  * inputs (some of which may already be in use).  The list is sorted in order
1869  * of irqs to use, with the highest priority given to interrupt controller
1870  * inputs that are not shared.   When an interrupt controller input
1871  * must be shared, apic_choose_irqs_from_prs adds the possible irqs to the
1872  * returned list in the order that minimizes sharing (thereby ensuring lowest
1873  * possible latency from interrupt trigger time to ISR execution time).
1874  */
1875 static prs_irq_list_t *
1876 apic_choose_irqs_from_prs(acpi_irqlist_t *irqlistent, dev_info_t *dip,
1877     int crs_irq)
1878 {
1879 	int32_t irq;
1880 	int i;
1881 	prs_irq_list_t *prsirqlistp = NULL;
1882 	iflag_t iflags;
1883 
1884 	while (irqlistent != NULL) {
1885 		irqlistent->intr_flags.bustype = BUS_PCI;
1886 
1887 		for (i = 0; i < irqlistent->num_irqs; i++) {
1888 
1889 			irq = irqlistent->irqs[i];
1890 
1891 			if (irq <= 0) {
1892 				/* invalid irq number */
1893 				continue;
1894 			}
1895 
1896 			if ((irq < 16) && (apic_reserved_irqlist[irq]))
1897 				continue;
1898 
1899 			if ((apic_irq_table[irq] == NULL) ||
1900 			    (apic_irq_table[irq]->airq_dip == dip)) {
1901 
1902 				prsirqlistp = acpi_insert_prs_irq_ent(
1903 				    prsirqlistp, 0 /* Highest priority */, irq,
1904 				    &irqlistent->intr_flags,
1905 				    &irqlistent->acpi_prs_prv);
1906 
1907 				/*
1908 				 * If we do not prefer the current irq from _CRS
1909 				 * or if we do and this irq is the same as the
1910 				 * current irq from _CRS, this is the one
1911 				 * to pick.
1912 				 */
1913 				if (!(apic_prefer_crs) || (irq == crs_irq)) {
1914 					return (prsirqlistp);
1915 				}
1916 				continue;
1917 			}
1918 
1919 			/*
1920 			 * Edge-triggered interrupts cannot be shared
1921 			 */
1922 			if (irqlistent->intr_flags.intr_el == INTR_EL_EDGE)
1923 				continue;
1924 
1925 			/*
1926 			 * To work around BIOSes that contain incorrect
1927 			 * interrupt polarity information in interrupt
1928 			 * descriptors returned by _PRS, we assume that
1929 			 * the polarity of the other device sharing this
1930 			 * interrupt controller input is compatible.
1931 			 * If it's not, the caller will catch it when
1932 			 * the caller invokes the link device's _CRS method
1933 			 * (after invoking its _SRS method).
1934 			 */
1935 			iflags = irqlistent->intr_flags;
1936 			iflags.intr_po =
1937 			    apic_irq_table[irq]->airq_iflag.intr_po;
1938 
1939 			if (!acpi_intr_compatible(iflags,
1940 			    apic_irq_table[irq]->airq_iflag)) {
1941 				APIC_VERBOSE_IRQ((CE_CONT, "!%s: irq %d "
1942 				    "not compatible [%x:%x:%x !~ %x:%x:%x]",
1943 				    psm_name, irq,
1944 				    iflags.intr_po,
1945 				    iflags.intr_el,
1946 				    iflags.bustype,
1947 				    apic_irq_table[irq]->airq_iflag.intr_po,
1948 				    apic_irq_table[irq]->airq_iflag.intr_el,
1949 				    apic_irq_table[irq]->airq_iflag.bustype));
1950 				continue;
1951 			}
1952 
1953 			/*
1954 			 * If we prefer the irq from _CRS, no need
1955 			 * to search any further (and make sure
1956 			 * to add this irq with the highest priority
1957 			 * so it's tried first).
1958 			 */
1959 			if (crs_irq == irq && apic_prefer_crs) {
1960 
1961 				return (acpi_insert_prs_irq_ent(
1962 				    prsirqlistp,
1963 				    0 /* Highest priority */,
1964 				    irq, &iflags,
1965 				    &irqlistent->acpi_prs_prv));
1966 			}
1967 
1968 			/*
1969 			 * Priority is equal to the share count (lower
1970 			 * share count is higher priority). Note that
1971 			 * the intr flags passed in here are the ones we
1972 			 * changed above -- if incorrect, it will be
1973 			 * caught by the caller's _CRS flags comparison.
1974 			 */
1975 			prsirqlistp = acpi_insert_prs_irq_ent(
1976 			    prsirqlistp,
1977 			    apic_irq_table[irq]->airq_share, irq,
1978 			    &iflags, &irqlistent->acpi_prs_prv);
1979 		}
1980 
1981 		/* Go to the next irqlist entry */
1982 		irqlistent = irqlistent->next;
1983 	}
1984 
1985 	return (prsirqlistp);
1986 }
1987 
1988 /*
1989  * Configures the irq for the interrupt link device identified by
1990  * acpipsmlnkp.
1991  *
1992  * Gets the current and the list of possible irq settings for the
1993  * device. If apic_unconditional_srs is not set, and the current
1994  * resource setting is in the list of possible irq settings,
1995  * current irq resource setting is passed to the caller.
1996  *
1997  * Otherwise, picks an irq number from the list of possible irq
1998  * settings, and sets the irq of the device to this value.
1999  * If prefer_crs is set, among a set of irq numbers in the list that have
2000  * the least number of devices sharing the interrupt, we pick current irq
2001  * resource setting if it is a member of this set.
2002  *
2003  * Passes the irq number in the value pointed to by pci_irqp, and
2004  * polarity and sensitivity in the structure pointed to by dipintrflagp
2005  * to the caller.
2006  *
2007  * Note that if setting the irq resource failed, but successfuly obtained
2008  * the current irq resource settings, passes the current irq resources
2009  * and considers it a success.
2010  *
2011  * Returns:
2012  * ACPI_PSM_SUCCESS on success.
2013  *
2014  * ACPI_PSM_FAILURE if an error occured during the configuration or
2015  * if a suitable irq was not found for this device, or if setting the
2016  * irq resource and obtaining the current resource fails.
2017  *
2018  */
2019 static int
2020 apic_acpi_irq_configure(acpi_psm_lnk_t *acpipsmlnkp, dev_info_t *dip,
2021     int *pci_irqp, iflag_t *dipintr_flagp)
2022 {
2023 	int32_t irq;
2024 	int cur_irq = -1;
2025 	acpi_irqlist_t *irqlistp;
2026 	prs_irq_list_t *prs_irq_listp, *prs_irq_entp;
2027 	boolean_t found_irq = B_FALSE;
2028 
2029 	dipintr_flagp->bustype = BUS_PCI;
2030 
2031 	if ((acpi_get_possible_irq_resources(acpipsmlnkp, &irqlistp))
2032 	    == ACPI_PSM_FAILURE) {
2033 		APIC_VERBOSE_IRQ((CE_WARN, "!%s: Unable to determine "
2034 		    "or assign IRQ for device %s, instance #%d: The system was "
2035 		    "unable to get the list of potential IRQs from ACPI.",
2036 		    psm_name, ddi_get_name(dip), ddi_get_instance(dip)));
2037 
2038 		return (ACPI_PSM_FAILURE);
2039 	}
2040 
2041 	if ((acpi_get_current_irq_resource(acpipsmlnkp, &cur_irq,
2042 	    dipintr_flagp) == ACPI_PSM_SUCCESS) && (!apic_unconditional_srs) &&
2043 	    (cur_irq > 0)) {
2044 		/*
2045 		 * If an IRQ is set in CRS and that IRQ exists in the set
2046 		 * returned from _PRS, return that IRQ, otherwise print
2047 		 * a warning
2048 		 */
2049 
2050 		if (acpi_irqlist_find_irq(irqlistp, cur_irq, NULL)
2051 		    == ACPI_PSM_SUCCESS) {
2052 
2053 			ASSERT(pci_irqp != NULL);
2054 			*pci_irqp = cur_irq;
2055 			acpi_free_irqlist(irqlistp);
2056 			return (ACPI_PSM_SUCCESS);
2057 		}
2058 
2059 		APIC_VERBOSE_IRQ((CE_WARN, "!%s: Could not find the "
2060 		    "current irq %d for device %s, instance #%d in ACPI's "
2061 		    "list of possible irqs for this device. Picking one from "
2062 		    " the latter list.", psm_name, cur_irq, ddi_get_name(dip),
2063 		    ddi_get_instance(dip)));
2064 	}
2065 
2066 	if ((prs_irq_listp = apic_choose_irqs_from_prs(irqlistp, dip,
2067 	    cur_irq)) == NULL) {
2068 
2069 		APIC_VERBOSE_IRQ((CE_WARN, "!%s: Could not find a "
2070 		    "suitable irq from the list of possible irqs for device "
2071 		    "%s, instance #%d in ACPI's list of possible irqs",
2072 		    psm_name, ddi_get_name(dip), ddi_get_instance(dip)));
2073 
2074 		acpi_free_irqlist(irqlistp);
2075 		return (ACPI_PSM_FAILURE);
2076 	}
2077 
2078 	acpi_free_irqlist(irqlistp);
2079 
2080 	for (prs_irq_entp = prs_irq_listp;
2081 	    prs_irq_entp != NULL && found_irq == B_FALSE;
2082 	    prs_irq_entp = prs_irq_entp->next) {
2083 
2084 		acpipsmlnkp->acpi_prs_prv = prs_irq_entp->prsprv;
2085 		irq = prs_irq_entp->irq;
2086 
2087 		APIC_VERBOSE_IRQ((CE_CONT, "!%s: Setting irq %d for "
2088 		    "device %s instance #%d\n", psm_name, irq,
2089 		    ddi_get_name(dip), ddi_get_instance(dip)));
2090 
2091 		if ((acpi_set_irq_resource(acpipsmlnkp, irq))
2092 		    == ACPI_PSM_SUCCESS) {
2093 			/*
2094 			 * setting irq was successful, check to make sure CRS
2095 			 * reflects that. If CRS does not agree with what we
2096 			 * set, return the irq that was set.
2097 			 */
2098 
2099 			if (acpi_get_current_irq_resource(acpipsmlnkp, &cur_irq,
2100 			    dipintr_flagp) == ACPI_PSM_SUCCESS) {
2101 
2102 				if (cur_irq != irq)
2103 					APIC_VERBOSE_IRQ((CE_WARN,
2104 					    "!%s: IRQ resource set "
2105 					    "(irqno %d) for device %s "
2106 					    "instance #%d, differs from "
2107 					    "current setting irqno %d",
2108 					    psm_name, irq, ddi_get_name(dip),
2109 					    ddi_get_instance(dip), cur_irq));
2110 			} else {
2111 				/*
2112 				 * On at least one system, there was a bug in
2113 				 * a DSDT method called by _STA, causing _STA to
2114 				 * indicate that the link device was disabled
2115 				 * (when, in fact, it was enabled).  Since _SRS
2116 				 * succeeded, assume that _CRS is lying and use
2117 				 * the iflags from this _PRS interrupt choice.
2118 				 * If we're wrong about the flags, the polarity
2119 				 * will be incorrect and we may get an interrupt
2120 				 * storm, but there's not much else we can do
2121 				 * at this point.
2122 				 */
2123 				*dipintr_flagp = prs_irq_entp->intrflags;
2124 			}
2125 
2126 			/*
2127 			 * Return the irq that was set, and not what _CRS
2128 			 * reports, since _CRS has been seen to return
2129 			 * different IRQs than what was passed to _SRS on some
2130 			 * systems (and just not return successfully on others).
2131 			 */
2132 			cur_irq = irq;
2133 			found_irq = B_TRUE;
2134 		} else {
2135 			APIC_VERBOSE_IRQ((CE_WARN, "!%s: set resource "
2136 			    "irq %d failed for device %s instance #%d",
2137 			    psm_name, irq, ddi_get_name(dip),
2138 			    ddi_get_instance(dip)));
2139 
2140 			if (cur_irq == -1) {
2141 				acpi_destroy_prs_irq_list(&prs_irq_listp);
2142 				return (ACPI_PSM_FAILURE);
2143 			}
2144 		}
2145 	}
2146 
2147 	acpi_destroy_prs_irq_list(&prs_irq_listp);
2148 
2149 	if (!found_irq)
2150 		return (ACPI_PSM_FAILURE);
2151 
2152 	ASSERT(pci_irqp != NULL);
2153 	*pci_irqp = cur_irq;
2154 	return (ACPI_PSM_SUCCESS);
2155 }
2156 
2157 void
2158 ioapic_disable_redirection()
2159 {
2160 	int ioapic_ix;
2161 	int intin_max;
2162 	int intin_ix;
2163 
2164 	/* Disable the I/O APIC redirection entries */
2165 	for (ioapic_ix = 0; ioapic_ix < apic_io_max; ioapic_ix++) {
2166 
2167 		/* Bits 23-16 define the maximum redirection entries */
2168 		intin_max = (ioapic_read(ioapic_ix, APIC_VERS_CMD) >> 16)
2169 		    & 0xff;
2170 
2171 		for (intin_ix = 0; intin_ix <= intin_max; intin_ix++) {
2172 			/*
2173 			 * The assumption here is that this is safe, even for
2174 			 * systems with IOAPICs that suffer from the hardware
2175 			 * erratum because all devices have been quiesced before
2176 			 * this function is called from apic_shutdown()
2177 			 * (or equivalent). If that assumption turns out to be
2178 			 * false, this mask operation can induce the same
2179 			 * erratum result we're trying to avoid.
2180 			 */
2181 			ioapic_write(ioapic_ix, APIC_RDT_CMD + 2 * intin_ix,
2182 			    AV_MASK);
2183 		}
2184 	}
2185 }
2186 
2187 /*
2188  * Looks for an IOAPIC with the specified physical address in the /ioapics
2189  * node in the device tree (created by the PCI enumerator).
2190  */
2191 static boolean_t
2192 apic_is_ioapic_AMD_813x(uint32_t physaddr)
2193 {
2194 	/*
2195 	 * Look in /ioapics, for the ioapic with
2196 	 * the physical address given
2197 	 */
2198 	dev_info_t *ioapicsnode = ddi_find_devinfo(IOAPICS_NODE_NAME, -1, 0);
2199 	dev_info_t *ioapic_child;
2200 	boolean_t rv = B_FALSE;
2201 	int vid, did;
2202 	uint64_t ioapic_paddr;
2203 	boolean_t done = B_FALSE;
2204 
2205 	if (ioapicsnode == NULL)
2206 		return (B_FALSE);
2207 
2208 	/* Load first child: */
2209 	ioapic_child = ddi_get_child(ioapicsnode);
2210 	while (!done && ioapic_child != 0) { /* Iterate over children */
2211 
2212 		if ((ioapic_paddr = (uint64_t)ddi_prop_get_int64(DDI_DEV_T_ANY,
2213 		    ioapic_child, DDI_PROP_DONTPASS, "reg", 0))
2214 		    != 0 && physaddr == ioapic_paddr) {
2215 
2216 			vid = ddi_prop_get_int(DDI_DEV_T_ANY, ioapic_child,
2217 			    DDI_PROP_DONTPASS, IOAPICS_PROP_VENID, 0);
2218 
2219 			if (vid == VENID_AMD) {
2220 
2221 				did = ddi_prop_get_int(DDI_DEV_T_ANY,
2222 				    ioapic_child, DDI_PROP_DONTPASS,
2223 				    IOAPICS_PROP_DEVID, 0);
2224 
2225 				if (did == DEVID_8131_IOAPIC ||
2226 				    did == DEVID_8132_IOAPIC) {
2227 					rv = B_TRUE;
2228 					done = B_TRUE;
2229 				}
2230 			}
2231 		}
2232 
2233 		if (!done)
2234 			ioapic_child = ddi_get_next_sibling(ioapic_child);
2235 	}
2236 
2237 	/* The ioapics node was held by ddi_find_devinfo, so release it */
2238 	ndi_rele_devi(ioapicsnode);
2239 	return (rv);
2240 }
2241 
2242 struct apic_state {
2243 	int32_t as_task_reg;
2244 	int32_t as_dest_reg;
2245 	int32_t as_format_reg;
2246 	int32_t as_local_timer;
2247 	int32_t as_pcint_vect;
2248 	int32_t as_int_vect0;
2249 	int32_t as_int_vect1;
2250 	int32_t as_err_vect;
2251 	int32_t as_init_count;
2252 	int32_t as_divide_reg;
2253 	int32_t as_spur_int_reg;
2254 	uint32_t as_ioapic_ids[MAX_IO_APIC];
2255 };
2256 
2257 
2258 static int
2259 apic_acpi_enter_apicmode(void)
2260 {
2261 	ACPI_OBJECT_LIST	arglist;
2262 	ACPI_OBJECT		arg;
2263 	ACPI_STATUS		status;
2264 
2265 	/* Setup parameter object */
2266 	arglist.Count = 1;
2267 	arglist.Pointer = &arg;
2268 	arg.Type = ACPI_TYPE_INTEGER;
2269 	arg.Integer.Value = ACPI_APIC_MODE;
2270 
2271 	status = AcpiEvaluateObject(NULL, "\\_PIC", &arglist, NULL);
2272 	/*
2273 	 * Per ACPI spec - section 5.8.1 _PIC Method
2274 	 * calling the \_PIC control method is optional for the OS
2275 	 * and might not be found. It's ok to not fail in such cases.
2276 	 * This is the case on linux KVM and qemu (status AE_NOT_FOUND)
2277 	 */
2278 	if (ACPI_FAILURE(status) && (status != AE_NOT_FOUND)) {
2279 		cmn_err(CE_NOTE,
2280 		    "!apic: Reporting APIC mode failed (via _PIC), err: 0x%x",
2281 		    ACPI_FAILURE(status));
2282 		return (PSM_FAILURE);
2283 	} else {
2284 		return (PSM_SUCCESS);
2285 	}
2286 }
2287 
2288 
2289 static void
2290 apic_save_state(struct apic_state *sp)
2291 {
2292 	int	i, cpuid;
2293 	ulong_t	iflag;
2294 
2295 	PMD(PMD_SX, ("apic_save_state %p\n", (void *)sp))
2296 	/*
2297 	 * First the local APIC.
2298 	 */
2299 	sp->as_task_reg = apic_reg_ops->apic_get_pri();
2300 	sp->as_dest_reg =  apic_reg_ops->apic_read(APIC_DEST_REG);
2301 	if (apic_mode == LOCAL_APIC)
2302 		sp->as_format_reg = apic_reg_ops->apic_read(APIC_FORMAT_REG);
2303 	sp->as_local_timer = apic_reg_ops->apic_read(APIC_LOCAL_TIMER);
2304 	sp->as_pcint_vect = apic_reg_ops->apic_read(APIC_PCINT_VECT);
2305 	sp->as_int_vect0 = apic_reg_ops->apic_read(APIC_INT_VECT0);
2306 	sp->as_int_vect1 = apic_reg_ops->apic_read(APIC_INT_VECT1);
2307 	sp->as_err_vect = apic_reg_ops->apic_read(APIC_ERR_VECT);
2308 	sp->as_init_count = apic_reg_ops->apic_read(APIC_INIT_COUNT);
2309 	sp->as_divide_reg = apic_reg_ops->apic_read(APIC_DIVIDE_REG);
2310 	sp->as_spur_int_reg = apic_reg_ops->apic_read(APIC_SPUR_INT_REG);
2311 
2312 	/*
2313 	 * If on the boot processor then save the IOAPICs' IDs
2314 	 */
2315 	if ((cpuid = psm_get_cpu_id()) == 0) {
2316 
2317 		iflag = intr_clear();
2318 		lock_set(&apic_ioapic_lock);
2319 
2320 		for (i = 0; i < apic_io_max; i++)
2321 			sp->as_ioapic_ids[i] = ioapic_read(i, APIC_ID_CMD);
2322 
2323 		lock_clear(&apic_ioapic_lock);
2324 		intr_restore(iflag);
2325 	}
2326 
2327 	/* apic_state() is currently invoked only in Suspend/Resume */
2328 	apic_cpus[cpuid].aci_status |= APIC_CPU_SUSPEND;
2329 }
2330 
2331 static void
2332 apic_restore_state(struct apic_state *sp)
2333 {
2334 	int	i;
2335 	ulong_t	iflag;
2336 
2337 	/*
2338 	 * First the local APIC.
2339 	 */
2340 	apic_reg_ops->apic_write_task_reg(sp->as_task_reg);
2341 	if (apic_mode == LOCAL_APIC) {
2342 		apic_reg_ops->apic_write(APIC_DEST_REG, sp->as_dest_reg);
2343 		apic_reg_ops->apic_write(APIC_FORMAT_REG, sp->as_format_reg);
2344 	}
2345 	apic_reg_ops->apic_write(APIC_LOCAL_TIMER, sp->as_local_timer);
2346 	apic_reg_ops->apic_write(APIC_PCINT_VECT, sp->as_pcint_vect);
2347 	apic_reg_ops->apic_write(APIC_INT_VECT0, sp->as_int_vect0);
2348 	apic_reg_ops->apic_write(APIC_INT_VECT1, sp->as_int_vect1);
2349 	apic_reg_ops->apic_write(APIC_ERR_VECT, sp->as_err_vect);
2350 	apic_reg_ops->apic_write(APIC_INIT_COUNT, sp->as_init_count);
2351 	apic_reg_ops->apic_write(APIC_DIVIDE_REG, sp->as_divide_reg);
2352 	apic_reg_ops->apic_write(APIC_SPUR_INT_REG, sp->as_spur_int_reg);
2353 
2354 	/*
2355 	 * the following only needs to be done once, so we do it on the
2356 	 * boot processor, since we know that we only have one of those
2357 	 */
2358 	if (psm_get_cpu_id() == 0) {
2359 
2360 		iflag = intr_clear();
2361 		lock_set(&apic_ioapic_lock);
2362 
2363 		/* Restore IOAPICs' APIC IDs */
2364 		for (i = 0; i < apic_io_max; i++) {
2365 			ioapic_write(i, APIC_ID_CMD, sp->as_ioapic_ids[i]);
2366 		}
2367 
2368 		lock_clear(&apic_ioapic_lock);
2369 		intr_restore(iflag);
2370 
2371 		/*
2372 		 * Reenter APIC mode before restoring LNK devices
2373 		 */
2374 		(void) apic_acpi_enter_apicmode();
2375 
2376 		/*
2377 		 * restore acpi link device mappings
2378 		 */
2379 		acpi_restore_link_devices();
2380 	}
2381 }
2382 
2383 /*
2384  * Returns 0 on success
2385  */
2386 int
2387 apic_state(psm_state_request_t *rp)
2388 {
2389 	PMD(PMD_SX, ("apic_state "))
2390 	switch (rp->psr_cmd) {
2391 	case PSM_STATE_ALLOC:
2392 		rp->req.psm_state_req.psr_state =
2393 		    kmem_zalloc(sizeof (struct apic_state), KM_NOSLEEP);
2394 		if (rp->req.psm_state_req.psr_state == NULL)
2395 			return (ENOMEM);
2396 		rp->req.psm_state_req.psr_state_size =
2397 		    sizeof (struct apic_state);
2398 		PMD(PMD_SX, (":STATE_ALLOC: state %p, size %lx\n",
2399 		    rp->req.psm_state_req.psr_state,
2400 		    rp->req.psm_state_req.psr_state_size))
2401 		return (0);
2402 
2403 	case PSM_STATE_FREE:
2404 		kmem_free(rp->req.psm_state_req.psr_state,
2405 		    rp->req.psm_state_req.psr_state_size);
2406 		PMD(PMD_SX, (" STATE_FREE: state %p, size %lx\n",
2407 		    rp->req.psm_state_req.psr_state,
2408 		    rp->req.psm_state_req.psr_state_size))
2409 		return (0);
2410 
2411 	case PSM_STATE_SAVE:
2412 		PMD(PMD_SX, (" STATE_SAVE: state %p, size %lx\n",
2413 		    rp->req.psm_state_req.psr_state,
2414 		    rp->req.psm_state_req.psr_state_size))
2415 		apic_save_state(rp->req.psm_state_req.psr_state);
2416 		return (0);
2417 
2418 	case PSM_STATE_RESTORE:
2419 		apic_restore_state(rp->req.psm_state_req.psr_state);
2420 		PMD(PMD_SX, (" STATE_RESTORE: state %p, size %lx\n",
2421 		    rp->req.psm_state_req.psr_state,
2422 		    rp->req.psm_state_req.psr_state_size))
2423 		return (0);
2424 
2425 	default:
2426 		return (EINVAL);
2427 	}
2428 }
2429