1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright 2010 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
25 *
26 * Copyright (c) 2010, Intel Corporation.
27 * All rights reserved.
28 *
29 * Copyright 2020 Joyent, Inc.
30 */
31
32 /*
33 * This file contains the functionality that mimics the boot operations
34 * on SPARC systems or the old boot.bin/multiboot programs on x86 systems.
35 * The x86 kernel now does everything on its own.
36 */
37
38 #include <sys/types.h>
39 #include <sys/bootconf.h>
40 #include <sys/bootsvcs.h>
41 #include <sys/bootinfo.h>
42 #include <sys/multiboot.h>
43 #include <sys/multiboot2.h>
44 #include <sys/multiboot2_impl.h>
45 #include <sys/bootvfs.h>
46 #include <sys/bootprops.h>
47 #include <sys/varargs.h>
48 #include <sys/param.h>
49 #include <sys/machparam.h>
50 #include <sys/machsystm.h>
51 #include <sys/archsystm.h>
52 #include <sys/boot_console.h>
53 #include <sys/framebuffer.h>
54 #include <sys/cmn_err.h>
55 #include <sys/systm.h>
56 #include <sys/promif.h>
57 #include <sys/archsystm.h>
58 #include <sys/x86_archext.h>
59 #include <sys/kobj.h>
60 #include <sys/privregs.h>
61 #include <sys/sysmacros.h>
62 #include <sys/ctype.h>
63 #include <sys/fastboot.h>
64 #ifdef __xpv
65 #include <sys/hypervisor.h>
66 #include <net/if.h>
67 #endif
68 #include <vm/kboot_mmu.h>
69 #include <vm/hat_pte.h>
70 #include <sys/kobj.h>
71 #include <sys/kobj_lex.h>
72 #include <sys/pci_cfgspace_impl.h>
73 #include <sys/fastboot_impl.h>
74 #include <sys/acpi/acconfig.h>
75 #include <sys/acpi/acpi.h>
76 #include <sys/ddipropdefs.h> /* For DDI prop types */
77
78 static int have_console = 0; /* set once primitive console is initialized */
79 static char *boot_args = "";
80
81 /*
82 * Debugging macros
83 */
84 static uint_t kbm_debug = 0;
85 #define DBG_MSG(s) { if (kbm_debug) bop_printf(NULL, "%s", s); }
86 #define DBG(x) { if (kbm_debug) \
87 bop_printf(NULL, "%s is %" PRIx64 "\n", #x, (uint64_t)(x)); \
88 }
89
90 #define PUT_STRING(s) { \
91 char *cp; \
92 for (cp = (s); *cp; ++cp) \
93 bcons_putchar(*cp); \
94 }
95
96 /* callback to boot_fb to set shadow frame buffer */
97 extern void boot_fb_shadow_init(bootops_t *);
98
99 bootops_t bootop; /* simple bootops we'll pass on to kernel */
100 struct bsys_mem bm;
101
102 /*
103 * Boot info from "glue" code in low memory. xbootp is used by:
104 * do_bop_phys_alloc(), do_bsys_alloc() and read_bootenvrc().
105 */
106 static struct xboot_info *xbootp;
107 static uintptr_t next_virt; /* next available virtual address */
108 static paddr_t next_phys; /* next available physical address from dboot */
109 static paddr_t high_phys = -(paddr_t)1; /* last used physical address */
110
111 /*
112 * buffer for vsnprintf for console I/O
113 */
114 #define BUFFERSIZE 512
115 static char buffer[BUFFERSIZE];
116
117 /*
118 * stuff to store/report/manipulate boot property settings.
119 */
120 typedef struct bootprop {
121 struct bootprop *bp_next;
122 char *bp_name;
123 int bp_flags; /* DDI prop type */
124 uint_t bp_vlen; /* 0 for boolean */
125 char *bp_value;
126 } bootprop_t;
127
128 static bootprop_t *bprops = NULL;
129 static char *curr_page = NULL; /* ptr to avail bprop memory */
130 static int curr_space = 0; /* amount of memory at curr_page */
131
132 #ifdef __xpv
133 start_info_t *xen_info;
134 shared_info_t *HYPERVISOR_shared_info;
135 #endif
136
137 /*
138 * some allocator statistics
139 */
140 static ulong_t total_bop_alloc_scratch = 0;
141 static ulong_t total_bop_alloc_kernel = 0;
142
143 static void build_firmware_properties(struct xboot_info *);
144
145 static int early_allocation = 1;
146
147 int force_fastreboot = 0;
148 volatile int fastreboot_onpanic = 0;
149 int post_fastreboot = 0;
150 #ifdef __xpv
151 volatile int fastreboot_capable = 0;
152 #else
153 volatile int fastreboot_capable = 1;
154 #endif
155
156 /*
157 * Information saved from current boot for fast reboot.
158 * If the information size exceeds what we have allocated, fast reboot
159 * will not be supported.
160 */
161 multiboot_info_t saved_mbi;
162 mb_memory_map_t saved_mmap[FASTBOOT_SAVED_MMAP_COUNT];
163 uint8_t saved_drives[FASTBOOT_SAVED_DRIVES_SIZE];
164 char saved_cmdline[FASTBOOT_SAVED_CMDLINE_LEN];
165 int saved_cmdline_len = 0;
166 size_t saved_file_size[FASTBOOT_MAX_FILES_MAP];
167
168 /*
169 * Turn off fastreboot_onpanic to avoid panic loop.
170 */
171 char fastreboot_onpanic_cmdline[FASTBOOT_SAVED_CMDLINE_LEN];
172 static const char fastreboot_onpanic_args[] = " -B fastreboot_onpanic=0";
173
174 /*
175 * Pointers to where System Resource Affinity Table (SRAT), System Locality
176 * Information Table (SLIT) and Maximum System Capability Table (MSCT)
177 * are mapped into virtual memory
178 */
179 ACPI_TABLE_SRAT *srat_ptr = NULL;
180 ACPI_TABLE_SLIT *slit_ptr = NULL;
181 ACPI_TABLE_MSCT *msct_ptr = NULL;
182
183 /*
184 * Arbitrary limit on number of localities we handle; if
185 * this limit is raised to more than UINT16_MAX, make sure
186 * process_slit() knows how to handle it.
187 */
188 #define SLIT_LOCALITIES_MAX (4096)
189
190 #define SLIT_NUM_PROPNAME "acpi-slit-localities"
191 #define SLIT_PROPNAME "acpi-slit"
192
193 /*
194 * Allocate aligned physical memory at boot time. This allocator allocates
195 * from the highest possible addresses. This avoids exhausting memory that
196 * would be useful for DMA buffers.
197 */
198 paddr_t
199 do_bop_phys_alloc(uint64_t size, uint64_t align)
200 {
201 paddr_t pa = 0;
202 paddr_t start;
203 paddr_t end;
204 struct memlist *ml = (struct memlist *)xbootp->bi_phys_install;
205
206 /*
207 * Be careful if high memory usage is limited in startup.c
208 * Since there are holes in the low part of the physical address
209 * space we can treat physmem as a pfn (not just a pgcnt) and
210 * get a conservative upper limit.
211 */
212 if (physmem != 0 && high_phys > pfn_to_pa(physmem))
213 high_phys = pfn_to_pa(physmem);
214
215 /*
216 * find the highest available memory in physinstalled
217 */
218 size = P2ROUNDUP(size, align);
219 for (; ml; ml = ml->ml_next) {
220 start = P2ROUNDUP(ml->ml_address, align);
221 end = P2ALIGN(ml->ml_address + ml->ml_size, align);
222 if (start < next_phys)
223 start = P2ROUNDUP(next_phys, align);
224 if (end > high_phys)
225 end = P2ALIGN(high_phys, align);
226
227 if (end <= start)
228 continue;
229 if (end - start < size)
230 continue;
231
232 /*
233 * Early allocations need to use low memory, since
234 * physmem might be further limited by bootenv.rc
235 */
236 if (early_allocation) {
237 if (pa == 0 || start < pa)
238 pa = start;
239 } else {
240 if (end - size > pa)
241 pa = end - size;
242 }
243 }
244 if (pa != 0) {
245 if (early_allocation)
246 next_phys = pa + size;
247 else
248 high_phys = pa;
249 return (pa);
250 }
251 bop_panic("do_bop_phys_alloc(0x%" PRIx64 ", 0x%" PRIx64
252 ") Out of memory\n", size, align);
253 /*NOTREACHED*/
254 }
255
256 uintptr_t
257 alloc_vaddr(size_t size, paddr_t align)
258 {
259 uintptr_t rv;
260
261 next_virt = P2ROUNDUP(next_virt, (uintptr_t)align);
262 rv = (uintptr_t)next_virt;
263 next_virt += size;
264 return (rv);
265 }
266
267 /*
268 * Allocate virtual memory. The size is always rounded up to a multiple
269 * of base pagesize.
270 */
271
272 /*ARGSUSED*/
273 static caddr_t
274 do_bsys_alloc(bootops_t *bop, caddr_t virthint, size_t size, int align)
275 {
276 paddr_t a = align; /* same type as pa for masking */
277 uint_t pgsize;
278 paddr_t pa;
279 uintptr_t va;
280 ssize_t s; /* the aligned size */
281 uint_t level;
282 uint_t is_kernel = (virthint != 0);
283
284 if (a < MMU_PAGESIZE)
285 a = MMU_PAGESIZE;
286 else if (!ISP2(a))
287 prom_panic("do_bsys_alloc() incorrect alignment");
288 size = P2ROUNDUP(size, MMU_PAGESIZE);
289
290 /*
291 * Use the next aligned virtual address if we weren't given one.
292 */
293 if (virthint == NULL) {
294 virthint = (caddr_t)alloc_vaddr(size, a);
295 total_bop_alloc_scratch += size;
296 } else {
297 total_bop_alloc_kernel += size;
298 }
299
300 /*
301 * allocate the physical memory
302 */
303 pa = do_bop_phys_alloc(size, a);
304
305 /*
306 * Add the mappings to the page tables, try large pages first.
307 */
308 va = (uintptr_t)virthint;
309 s = size;
310 level = 1;
311 pgsize = xbootp->bi_use_pae ? TWO_MEG : FOUR_MEG;
312 if (xbootp->bi_use_largepage && a == pgsize) {
313 while (IS_P2ALIGNED(pa, pgsize) && IS_P2ALIGNED(va, pgsize) &&
314 s >= pgsize) {
315 kbm_map(va, pa, level, is_kernel);
316 va += pgsize;
317 pa += pgsize;
318 s -= pgsize;
319 }
320 }
321
322 /*
323 * Map remaining pages use small mappings
324 */
325 level = 0;
326 pgsize = MMU_PAGESIZE;
327 while (s > 0) {
328 kbm_map(va, pa, level, is_kernel);
329 va += pgsize;
330 pa += pgsize;
331 s -= pgsize;
332 }
333 return (virthint);
334 }
335
336 /*
337 * Free virtual memory - we'll just ignore these.
338 */
339 /*ARGSUSED*/
340 static void
341 do_bsys_free(bootops_t *bop, caddr_t virt, size_t size)
342 {
343 bop_printf(NULL, "do_bsys_free(virt=0x%p, size=0x%lx) ignored\n",
344 (void *)virt, size);
345 }
346
347 /*
348 * Old interface
349 */
350 /*ARGSUSED*/
351 static caddr_t
352 do_bsys_ealloc(bootops_t *bop, caddr_t virthint, size_t size,
353 int align, int flags)
354 {
355 prom_panic("unsupported call to BOP_EALLOC()\n");
356 return (0);
357 }
358
359
360 static void
361 bsetprop(int flags, char *name, int nlen, void *value, int vlen)
362 {
363 uint_t size;
364 uint_t need_size;
365 bootprop_t *b;
366
367 /*
368 * align the size to 16 byte boundary
369 */
370 size = sizeof (bootprop_t) + nlen + 1 + vlen;
371 size = (size + 0xf) & ~0xf;
372 if (size > curr_space) {
373 need_size = (size + (MMU_PAGEOFFSET)) & MMU_PAGEMASK;
374 curr_page = do_bsys_alloc(NULL, 0, need_size, MMU_PAGESIZE);
375 curr_space = need_size;
376 }
377
378 /*
379 * use a bootprop_t at curr_page and link into list
380 */
381 b = (bootprop_t *)curr_page;
382 curr_page += sizeof (bootprop_t);
383 curr_space -= sizeof (bootprop_t);
384 b->bp_next = bprops;
385 bprops = b;
386
387 /*
388 * follow by name and ending zero byte
389 */
390 b->bp_name = curr_page;
391 bcopy(name, curr_page, nlen);
392 curr_page += nlen;
393 *curr_page++ = 0;
394 curr_space -= nlen + 1;
395
396 /*
397 * set the property type
398 */
399 b->bp_flags = flags & DDI_PROP_TYPE_MASK;
400
401 /*
402 * copy in value, but no ending zero byte
403 */
404 b->bp_value = curr_page;
405 b->bp_vlen = vlen;
406 if (vlen > 0) {
407 bcopy(value, curr_page, vlen);
408 curr_page += vlen;
409 curr_space -= vlen;
410 }
411
412 /*
413 * align new values of curr_page, curr_space
414 */
415 while (curr_space & 0xf) {
416 ++curr_page;
417 --curr_space;
418 }
419 }
420
421 static void
422 bsetprops(char *name, char *value)
423 {
424 bsetprop(DDI_PROP_TYPE_STRING, name, strlen(name),
425 value, strlen(value) + 1);
426 }
427
428 static void
429 bsetprop32(char *name, uint32_t value)
430 {
431 bsetprop(DDI_PROP_TYPE_INT, name, strlen(name),
432 (void *)&value, sizeof (value));
433 }
434
435 static void
436 bsetprop64(char *name, uint64_t value)
437 {
438 bsetprop(DDI_PROP_TYPE_INT64, name, strlen(name),
439 (void *)&value, sizeof (value));
440 }
441
442 static void
443 bsetpropsi(char *name, int value)
444 {
445 char prop_val[32];
446
447 (void) snprintf(prop_val, sizeof (prop_val), "%d", value);
448 bsetprops(name, prop_val);
449 }
450
451 /*
452 * to find the type of the value associated with this name
453 */
454 /*ARGSUSED*/
455 int
456 do_bsys_getproptype(bootops_t *bop, const char *name)
457 {
458 bootprop_t *b;
459
460 for (b = bprops; b != NULL; b = b->bp_next) {
461 if (strcmp(name, b->bp_name) != 0)
462 continue;
463 return (b->bp_flags);
464 }
465 return (-1);
466 }
467
468 /*
469 * to find the size of the buffer to allocate
470 */
471 /*ARGSUSED*/
472 int
473 do_bsys_getproplen(bootops_t *bop, const char *name)
474 {
475 bootprop_t *b;
476
477 for (b = bprops; b; b = b->bp_next) {
478 if (strcmp(name, b->bp_name) != 0)
479 continue;
480 return (b->bp_vlen);
481 }
482 return (-1);
483 }
484
485 /*
486 * get the value associated with this name
487 */
488 /*ARGSUSED*/
489 int
490 do_bsys_getprop(bootops_t *bop, const char *name, void *value)
491 {
492 bootprop_t *b;
493
494 for (b = bprops; b; b = b->bp_next) {
495 if (strcmp(name, b->bp_name) != 0)
496 continue;
497 bcopy(b->bp_value, value, b->bp_vlen);
498 return (0);
499 }
500 return (-1);
501 }
502
503 /*
504 * get the name of the next property in succession from the standalone
505 */
506 /*ARGSUSED*/
507 static char *
508 do_bsys_nextprop(bootops_t *bop, char *name)
509 {
510 bootprop_t *b;
511
512 /*
513 * A null name is a special signal for the 1st boot property
514 */
515 if (name == NULL || strlen(name) == 0) {
516 if (bprops == NULL)
517 return (NULL);
518 return (bprops->bp_name);
519 }
520
521 for (b = bprops; b; b = b->bp_next) {
522 if (name != b->bp_name)
523 continue;
524 b = b->bp_next;
525 if (b == NULL)
526 return (NULL);
527 return (b->bp_name);
528 }
529 return (NULL);
530 }
531
532 /*
533 * Parse numeric value from a string. Understands decimal, hex, octal, - and ~
534 */
535 static int
536 parse_value(char *p, uint64_t *retval)
537 {
538 int adjust = 0;
539 uint64_t tmp = 0;
540 int digit;
541 int radix = 10;
542
543 *retval = 0;
544 if (*p == '-' || *p == '~')
545 adjust = *p++;
546
547 if (*p == '0') {
548 ++p;
549 if (*p == 0)
550 return (0);
551 if (*p == 'x' || *p == 'X') {
552 radix = 16;
553 ++p;
554 } else {
555 radix = 8;
556 ++p;
557 }
558 }
559 while (*p) {
560 if ('0' <= *p && *p <= '9')
561 digit = *p - '0';
562 else if ('a' <= *p && *p <= 'f')
563 digit = 10 + *p - 'a';
564 else if ('A' <= *p && *p <= 'F')
565 digit = 10 + *p - 'A';
566 else
567 return (-1);
568 if (digit >= radix)
569 return (-1);
570 tmp = tmp * radix + digit;
571 ++p;
572 }
573 if (adjust == '-')
574 tmp = -tmp;
575 else if (adjust == '~')
576 tmp = ~tmp;
577 *retval = tmp;
578 return (0);
579 }
580
581 static boolean_t
582 unprintable(char *value, int size)
583 {
584 int i;
585
586 if (size <= 0 || value[0] == '\0')
587 return (B_TRUE);
588
589 for (i = 0; i < size; i++) {
590 if (value[i] == '\0')
591 return (i != (size - 1));
592
593 if (!isprint(value[i]))
594 return (B_TRUE);
595 }
596 return (B_FALSE);
597 }
598
599 /*
600 * Print out information about all boot properties.
601 * buffer is pointer to pre-allocated space to be used as temporary
602 * space for property values.
603 */
604 static void
605 boot_prop_display(char *buffer)
606 {
607 char *name = "";
608 int i, len, flags, *buf32;
609 int64_t *buf64;
610
611 bop_printf(NULL, "\nBoot properties:\n");
612
613 while ((name = do_bsys_nextprop(NULL, name)) != NULL) {
614 bop_printf(NULL, "\t0x%p %s = ", (void *)name, name);
615 (void) do_bsys_getprop(NULL, name, buffer);
616 len = do_bsys_getproplen(NULL, name);
617 flags = do_bsys_getproptype(NULL, name);
618 bop_printf(NULL, "len=%d ", len);
619
620 switch (flags) {
621 case DDI_PROP_TYPE_INT:
622 len = len / sizeof (int);
623 buf32 = (int *)buffer;
624 for (i = 0; i < len; i++) {
625 bop_printf(NULL, "%08x", buf32[i]);
626 if (i < len - 1)
627 bop_printf(NULL, ".");
628 }
629 break;
630 case DDI_PROP_TYPE_STRING:
631 bop_printf(NULL, "%s", buffer);
632 break;
633 case DDI_PROP_TYPE_INT64:
634 len = len / sizeof (int64_t);
635 buf64 = (int64_t *)buffer;
636 for (i = 0; i < len; i++) {
637 bop_printf(NULL, "%016" PRIx64, buf64[i]);
638 if (i < len - 1)
639 bop_printf(NULL, ".");
640 }
641 break;
642 default:
643 if (!unprintable(buffer, len)) {
644 buffer[len] = 0;
645 bop_printf(NULL, "%s", buffer);
646 break;
647 }
648 for (i = 0; i < len; i++) {
649 bop_printf(NULL, "%02x", buffer[i] & 0xff);
650 if (i < len - 1)
651 bop_printf(NULL, ".");
652 }
653 break;
654 }
655 bop_printf(NULL, "\n");
656 }
657 }
658
659 /*
660 * 2nd part of building the table of boot properties. This includes:
661 * - values from /boot/solaris/bootenv.rc (ie. eeprom(1m) values)
662 *
663 * lines look like one of:
664 * ^$
665 * ^# comment till end of line
666 * setprop name 'value'
667 * setprop name value
668 * setprop name "value"
669 *
670 * we do single character I/O since this is really just looking at memory
671 */
672 void
673 read_bootenvrc(void)
674 {
675 int fd;
676 char *line;
677 int c;
678 int bytes_read;
679 char *name;
680 int n_len;
681 char *value;
682 int v_len;
683 char *inputdev; /* these override the command line if serial ports */
684 char *outputdev;
685 char *consoledev;
686 uint64_t lvalue;
687 int use_xencons = 0;
688 extern int bootrd_debug;
689
690 #ifdef __xpv
691 if (!DOMAIN_IS_INITDOMAIN(xen_info))
692 use_xencons = 1;
693 #endif /* __xpv */
694
695 DBG_MSG("Opening /boot/solaris/bootenv.rc\n");
696 fd = BRD_OPEN(bfs_ops, "/boot/solaris/bootenv.rc", 0);
697 DBG(fd);
698
699 line = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
700 while (fd >= 0) {
701
702 /*
703 * get a line
704 */
705 for (c = 0; ; ++c) {
706 bytes_read = BRD_READ(bfs_ops, fd, line + c, 1);
707 if (bytes_read == 0) {
708 if (c == 0)
709 goto done;
710 break;
711 }
712 if (line[c] == '\n')
713 break;
714 }
715 line[c] = 0;
716
717 /*
718 * ignore comment lines
719 */
720 c = 0;
721 while (ISSPACE(line[c]))
722 ++c;
723 if (line[c] == '#' || line[c] == 0)
724 continue;
725
726 /*
727 * must have "setprop " or "setprop\t"
728 */
729 if (strncmp(line + c, "setprop ", 8) != 0 &&
730 strncmp(line + c, "setprop\t", 8) != 0)
731 continue;
732 c += 8;
733 while (ISSPACE(line[c]))
734 ++c;
735 if (line[c] == 0)
736 continue;
737
738 /*
739 * gather up the property name
740 */
741 name = line + c;
742 n_len = 0;
743 while (line[c] && !ISSPACE(line[c]))
744 ++n_len, ++c;
745
746 /*
747 * gather up the value, if any
748 */
749 value = "";
750 v_len = 0;
751 while (ISSPACE(line[c]))
752 ++c;
753 if (line[c] != 0) {
754 value = line + c;
755 while (line[c] && !ISSPACE(line[c]))
756 ++v_len, ++c;
757 }
758
759 if (v_len >= 2 && value[0] == value[v_len - 1] &&
760 (value[0] == '\'' || value[0] == '"')) {
761 ++value;
762 v_len -= 2;
763 }
764 name[n_len] = 0;
765 if (v_len > 0)
766 value[v_len] = 0;
767 else
768 continue;
769
770 /*
771 * ignore "boot-file" property, it's now meaningless
772 */
773 if (strcmp(name, "boot-file") == 0)
774 continue;
775 if (strcmp(name, "boot-args") == 0 &&
776 strlen(boot_args) > 0)
777 continue;
778
779 /*
780 * If a property was explicitly set on the command line
781 * it will override a setting in bootenv.rc. We make an
782 * exception for a property from the bootloader such as:
783 *
784 * console="text,ttya,ttyb,ttyc,ttyd"
785 *
786 * In such a case, picking the first value here (as
787 * lookup_console_devices() does) is at best a guess; if
788 * bootenv.rc has a value, it's probably better.
789 */
790 if (strcmp(name, "console") == 0) {
791 char propval[BP_MAX_STRLEN] = "";
792
793 if (do_bsys_getprop(NULL, name, propval) == -1 ||
794 strchr(propval, ',') != NULL)
795 bsetprops(name, value);
796 continue;
797 }
798
799 if (do_bsys_getproplen(NULL, name) == -1)
800 bsetprops(name, value);
801 }
802 done:
803 if (fd >= 0)
804 (void) BRD_CLOSE(bfs_ops, fd);
805
806
807 /*
808 * Check if we have to limit the boot time allocator
809 */
810 if (do_bsys_getproplen(NULL, "physmem") != -1 &&
811 do_bsys_getprop(NULL, "physmem", line) >= 0 &&
812 parse_value(line, &lvalue) != -1) {
813 if (0 < lvalue && (lvalue < physmem || physmem == 0)) {
814 physmem = (pgcnt_t)lvalue;
815 DBG(physmem);
816 }
817 }
818 early_allocation = 0;
819
820 /*
821 * Check for bootrd_debug.
822 */
823 if (find_boot_prop("bootrd_debug"))
824 bootrd_debug = 1;
825
826 /*
827 * check to see if we have to override the default value of the console
828 */
829 if (!use_xencons) {
830 inputdev = line;
831 v_len = do_bsys_getproplen(NULL, "input-device");
832 if (v_len > 0)
833 (void) do_bsys_getprop(NULL, "input-device", inputdev);
834 else
835 v_len = 0;
836 inputdev[v_len] = 0;
837
838 outputdev = inputdev + v_len + 1;
839 v_len = do_bsys_getproplen(NULL, "output-device");
840 if (v_len > 0)
841 (void) do_bsys_getprop(NULL, "output-device",
842 outputdev);
843 else
844 v_len = 0;
845 outputdev[v_len] = 0;
846
847 consoledev = outputdev + v_len + 1;
848 v_len = do_bsys_getproplen(NULL, "console");
849 if (v_len > 0) {
850 (void) do_bsys_getprop(NULL, "console", consoledev);
851 if (post_fastreboot &&
852 strcmp(consoledev, "graphics") == 0) {
853 bsetprops("console", "text");
854 v_len = strlen("text");
855 bcopy("text", consoledev, v_len);
856 }
857 } else {
858 v_len = 0;
859 }
860 consoledev[v_len] = 0;
861 bcons_post_bootenvrc(inputdev, outputdev, consoledev);
862 } else {
863 /*
864 * Ensure console property exists
865 * If not create it as "hypervisor"
866 */
867 v_len = do_bsys_getproplen(NULL, "console");
868 if (v_len < 0)
869 bsetprops("console", "hypervisor");
870 inputdev = outputdev = consoledev = "hypervisor";
871 bcons_post_bootenvrc(inputdev, outputdev, consoledev);
872 }
873
874 if (find_boot_prop("prom_debug") || kbm_debug)
875 boot_prop_display(line);
876 }
877
878 /*
879 * print formatted output
880 */
881 /*ARGSUSED*/
882 void
883 vbop_printf(void *ptr, const char *fmt, va_list ap)
884 {
885 if (have_console == 0)
886 return;
887
888 (void) vsnprintf(buffer, BUFFERSIZE, fmt, ap);
889 PUT_STRING(buffer);
890 }
891
892 /*PRINTFLIKE2*/
893 void
894 bop_printf(void *bop, const char *fmt, ...)
895 {
896 va_list ap;
897
898 va_start(ap, fmt);
899 vbop_printf(bop, fmt, ap);
900 va_end(ap);
901 }
902
903 /*
904 * Another panic() variant; this one can be used even earlier during boot than
905 * prom_panic().
906 */
907 /*PRINTFLIKE1*/
908 void
909 bop_panic(const char *fmt, ...)
910 {
911 va_list ap;
912
913 va_start(ap, fmt);
914 bop_printf(NULL, fmt, ap);
915 va_end(ap);
916
917 bop_printf(NULL, "\nPress any key to reboot.\n");
918 (void) bcons_getchar();
919 bop_printf(NULL, "Resetting...\n");
920 pc_reset();
921 }
922
923 /*
924 * Do a real mode interrupt BIOS call
925 */
926 typedef struct bios_regs {
927 unsigned short ax, bx, cx, dx, si, di, bp, es, ds;
928 } bios_regs_t;
929 typedef int (*bios_func_t)(int, bios_regs_t *);
930
931 /*ARGSUSED*/
932 static void
933 do_bsys_doint(bootops_t *bop, int intnum, struct bop_regs *rp)
934 {
935 #if defined(__xpv)
936 prom_panic("unsupported call to BOP_DOINT()\n");
937 #else /* __xpv */
938 static int firsttime = 1;
939 bios_func_t bios_func = (bios_func_t)(void *)(uintptr_t)0x5000;
940 bios_regs_t br;
941
942 /*
943 * We're about to disable paging; we shouldn't be PCID enabled.
944 */
945 if (getcr4() & CR4_PCIDE)
946 prom_panic("do_bsys_doint() with PCID enabled\n");
947
948 /*
949 * The first time we do this, we have to copy the pre-packaged
950 * low memory bios call code image into place.
951 */
952 if (firsttime) {
953 extern char bios_image[];
954 extern uint32_t bios_size;
955
956 bcopy(bios_image, (void *)bios_func, bios_size);
957 firsttime = 0;
958 }
959
960 br.ax = rp->eax.word.ax;
961 br.bx = rp->ebx.word.bx;
962 br.cx = rp->ecx.word.cx;
963 br.dx = rp->edx.word.dx;
964 br.bp = rp->ebp.word.bp;
965 br.si = rp->esi.word.si;
966 br.di = rp->edi.word.di;
967 br.ds = rp->ds;
968 br.es = rp->es;
969
970 DBG_MSG("Doing BIOS call...");
971 DBG(br.ax);
972 DBG(br.bx);
973 DBG(br.dx);
974 rp->eflags = bios_func(intnum, &br);
975 DBG_MSG("done\n");
976
977 rp->eax.word.ax = br.ax;
978 rp->ebx.word.bx = br.bx;
979 rp->ecx.word.cx = br.cx;
980 rp->edx.word.dx = br.dx;
981 rp->ebp.word.bp = br.bp;
982 rp->esi.word.si = br.si;
983 rp->edi.word.di = br.di;
984 rp->ds = br.ds;
985 rp->es = br.es;
986 #endif /* __xpv */
987 }
988
989 static struct boot_syscalls bop_sysp = {
990 bcons_getchar,
991 bcons_putchar,
992 bcons_ischar,
993 };
994
995 static char *whoami;
996
997 #define BUFLEN 64
998
999 #if defined(__xpv)
1000
1001 static char namebuf[32];
1002
1003 static void
1004 xen_parse_props(char *s, char *prop_map[], int n_prop)
1005 {
1006 char **prop_name = prop_map;
1007 char *cp = s, *scp;
1008
1009 do {
1010 scp = cp;
1011 while ((*cp != '\0') && (*cp != ':'))
1012 cp++;
1013
1014 if ((scp != cp) && (*prop_name != NULL)) {
1015 *cp = '\0';
1016 bsetprops(*prop_name, scp);
1017 }
1018
1019 cp++;
1020 prop_name++;
1021 n_prop--;
1022 } while (n_prop > 0);
1023 }
1024
1025 #define VBDPATHLEN 64
1026
1027 /*
1028 * parse the 'xpv-root' property to create properties used by
1029 * ufs_mountroot.
1030 */
1031 static void
1032 xen_vbdroot_props(char *s)
1033 {
1034 char vbdpath[VBDPATHLEN] = "/xpvd/xdf@";
1035 const char lnamefix[] = "/dev/dsk/c0d";
1036 char *pnp;
1037 char *prop_p;
1038 char mi;
1039 short minor;
1040 long addr = 0;
1041
1042 pnp = vbdpath + strlen(vbdpath);
1043 prop_p = s + strlen(lnamefix);
1044 while ((*prop_p != '\0') && (*prop_p != 's') && (*prop_p != 'p'))
1045 addr = addr * 10 + *prop_p++ - '0';
1046 (void) snprintf(pnp, VBDPATHLEN, "%lx", addr);
1047 pnp = vbdpath + strlen(vbdpath);
1048 if (*prop_p == 's')
1049 mi = 'a';
1050 else if (*prop_p == 'p')
1051 mi = 'q';
1052 else
1053 ASSERT(0); /* shouldn't be here */
1054 prop_p++;
1055 ASSERT(*prop_p != '\0');
1056 if (ISDIGIT(*prop_p)) {
1057 minor = *prop_p - '0';
1058 prop_p++;
1059 if (ISDIGIT(*prop_p)) {
1060 minor = minor * 10 + *prop_p - '0';
1061 }
1062 } else {
1063 /* malformed root path, use 0 as default */
1064 minor = 0;
1065 }
1066 ASSERT(minor < 16); /* at most 16 partitions */
1067 mi += minor;
1068 *pnp++ = ':';
1069 *pnp++ = mi;
1070 *pnp++ = '\0';
1071 bsetprops("fstype", "ufs");
1072 bsetprops("bootpath", vbdpath);
1073
1074 DBG_MSG("VBD bootpath set to ");
1075 DBG_MSG(vbdpath);
1076 DBG_MSG("\n");
1077 }
1078
1079 /*
1080 * parse the xpv-nfsroot property to create properties used by
1081 * nfs_mountroot.
1082 */
1083 static void
1084 xen_nfsroot_props(char *s)
1085 {
1086 char *prop_map[] = {
1087 BP_SERVER_IP, /* server IP address */
1088 BP_SERVER_NAME, /* server hostname */
1089 BP_SERVER_PATH, /* root path */
1090 };
1091 int n_prop = sizeof (prop_map) / sizeof (prop_map[0]);
1092
1093 bsetprops("fstype", "nfs");
1094
1095 xen_parse_props(s, prop_map, n_prop);
1096
1097 /*
1098 * If a server name wasn't specified, use a default.
1099 */
1100 if (do_bsys_getproplen(NULL, BP_SERVER_NAME) == -1)
1101 bsetprops(BP_SERVER_NAME, "unknown");
1102 }
1103
1104 /*
1105 * Extract our IP address, etc. from the "xpv-ip" property.
1106 */
1107 static void
1108 xen_ip_props(char *s)
1109 {
1110 char *prop_map[] = {
1111 BP_HOST_IP, /* IP address */
1112 NULL, /* NFS server IP address (ignored in */
1113 /* favour of xpv-nfsroot) */
1114 BP_ROUTER_IP, /* IP gateway */
1115 BP_SUBNET_MASK, /* IP subnet mask */
1116 "xpv-hostname", /* hostname (ignored) */
1117 BP_NETWORK_INTERFACE, /* interface name */
1118 "xpv-hcp", /* host configuration protocol */
1119 };
1120 int n_prop = sizeof (prop_map) / sizeof (prop_map[0]);
1121 char ifname[IFNAMSIZ];
1122
1123 xen_parse_props(s, prop_map, n_prop);
1124
1125 /*
1126 * A Linux dom0 administrator expects all interfaces to be
1127 * called "ethX", which is not the case here.
1128 *
1129 * If the interface name specified is "eth0", presume that
1130 * this is really intended to be "xnf0" (the first domU ->
1131 * dom0 interface for this domain).
1132 */
1133 if ((do_bsys_getprop(NULL, BP_NETWORK_INTERFACE, ifname) == 0) &&
1134 (strcmp("eth0", ifname) == 0)) {
1135 bsetprops(BP_NETWORK_INTERFACE, "xnf0");
1136 bop_printf(NULL,
1137 "network interface name 'eth0' replaced with 'xnf0'\n");
1138 }
1139 }
1140
1141 #else /* __xpv */
1142
1143 static void
1144 setup_rarp_props(struct sol_netinfo *sip)
1145 {
1146 char buf[BUFLEN]; /* to hold ip/mac addrs */
1147 uint8_t *val;
1148
1149 val = (uint8_t *)&sip->sn_ciaddr;
1150 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1151 val[0], val[1], val[2], val[3]);
1152 bsetprops(BP_HOST_IP, buf);
1153
1154 val = (uint8_t *)&sip->sn_siaddr;
1155 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1156 val[0], val[1], val[2], val[3]);
1157 bsetprops(BP_SERVER_IP, buf);
1158
1159 if (sip->sn_giaddr != 0) {
1160 val = (uint8_t *)&sip->sn_giaddr;
1161 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1162 val[0], val[1], val[2], val[3]);
1163 bsetprops(BP_ROUTER_IP, buf);
1164 }
1165
1166 if (sip->sn_netmask != 0) {
1167 val = (uint8_t *)&sip->sn_netmask;
1168 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1169 val[0], val[1], val[2], val[3]);
1170 bsetprops(BP_SUBNET_MASK, buf);
1171 }
1172
1173 if (sip->sn_mactype != 4 || sip->sn_maclen != 6) {
1174 bop_printf(NULL, "unsupported mac type %d, mac len %d\n",
1175 sip->sn_mactype, sip->sn_maclen);
1176 } else {
1177 val = sip->sn_macaddr;
1178 (void) snprintf(buf, BUFLEN, "%x:%x:%x:%x:%x:%x",
1179 val[0], val[1], val[2], val[3], val[4], val[5]);
1180 bsetprops(BP_BOOT_MAC, buf);
1181 }
1182 }
1183
1184 #endif /* __xpv */
1185
1186 static void
1187 build_panic_cmdline(const char *cmd, int cmdlen)
1188 {
1189 int proplen;
1190 size_t arglen;
1191
1192 arglen = sizeof (fastreboot_onpanic_args);
1193 /*
1194 * If we allready have fastreboot-onpanic set to zero,
1195 * don't add them again.
1196 */
1197 if ((proplen = do_bsys_getproplen(NULL, FASTREBOOT_ONPANIC)) > 0 &&
1198 proplen <= sizeof (fastreboot_onpanic_cmdline)) {
1199 (void) do_bsys_getprop(NULL, FASTREBOOT_ONPANIC,
1200 fastreboot_onpanic_cmdline);
1201 if (FASTREBOOT_ONPANIC_NOTSET(fastreboot_onpanic_cmdline))
1202 arglen = 1;
1203 }
1204
1205 /*
1206 * construct fastreboot_onpanic_cmdline
1207 */
1208 if (cmdlen + arglen > sizeof (fastreboot_onpanic_cmdline)) {
1209 DBG_MSG("Command line too long: clearing "
1210 FASTREBOOT_ONPANIC "\n");
1211 fastreboot_onpanic = 0;
1212 } else {
1213 bcopy(cmd, fastreboot_onpanic_cmdline, cmdlen);
1214 if (arglen != 1)
1215 bcopy(fastreboot_onpanic_args,
1216 fastreboot_onpanic_cmdline + cmdlen, arglen);
1217 else
1218 fastreboot_onpanic_cmdline[cmdlen] = 0;
1219 }
1220 }
1221
1222
1223 #ifndef __xpv
1224 /*
1225 * Construct boot command line for Fast Reboot. The saved_cmdline
1226 * is also reported by "eeprom bootcmd".
1227 */
1228 static void
1229 build_fastboot_cmdline(struct xboot_info *xbp)
1230 {
1231 saved_cmdline_len = strlen(xbp->bi_cmdline) + 1;
1232 if (saved_cmdline_len > FASTBOOT_SAVED_CMDLINE_LEN) {
1233 DBG(saved_cmdline_len);
1234 DBG_MSG("Command line too long: clearing fastreboot_capable\n");
1235 fastreboot_capable = 0;
1236 } else {
1237 bcopy((void *)(xbp->bi_cmdline), (void *)saved_cmdline,
1238 saved_cmdline_len);
1239 saved_cmdline[saved_cmdline_len - 1] = '\0';
1240 build_panic_cmdline(saved_cmdline, saved_cmdline_len - 1);
1241 }
1242 }
1243
1244 /*
1245 * Save memory layout, disk drive information, unix and boot archive sizes for
1246 * Fast Reboot.
1247 */
1248 static void
1249 save_boot_info(struct xboot_info *xbi)
1250 {
1251 multiboot_info_t *mbi = xbi->bi_mb_info;
1252 struct boot_modules *modp;
1253 int i;
1254
1255 bcopy(mbi, &saved_mbi, sizeof (multiboot_info_t));
1256 if (mbi->mmap_length > sizeof (saved_mmap)) {
1257 DBG_MSG("mbi->mmap_length too big: clearing "
1258 "fastreboot_capable\n");
1259 fastreboot_capable = 0;
1260 } else {
1261 bcopy((void *)(uintptr_t)mbi->mmap_addr, (void *)saved_mmap,
1262 mbi->mmap_length);
1263 }
1264
1265 if ((mbi->flags & MB_INFO_DRIVE_INFO) != 0) {
1266 if (mbi->drives_length > sizeof (saved_drives)) {
1267 DBG(mbi->drives_length);
1268 DBG_MSG("mbi->drives_length too big: clearing "
1269 "fastreboot_capable\n");
1270 fastreboot_capable = 0;
1271 } else {
1272 bcopy((void *)(uintptr_t)mbi->drives_addr,
1273 (void *)saved_drives, mbi->drives_length);
1274 }
1275 } else {
1276 saved_mbi.drives_length = 0;
1277 saved_mbi.drives_addr = 0;
1278 }
1279
1280 /*
1281 * Current file sizes. Used by fastboot.c to figure out how much
1282 * memory to reserve for panic reboot.
1283 * Use the module list from the dboot-constructed xboot_info
1284 * instead of the list referenced by the multiboot structure
1285 * because that structure may not be addressable now.
1286 */
1287 saved_file_size[FASTBOOT_NAME_UNIX] = FOUR_MEG - PAGESIZE;
1288 for (i = 0, modp = (struct boot_modules *)(uintptr_t)xbi->bi_modules;
1289 i < xbi->bi_module_cnt; i++, modp++) {
1290 saved_file_size[FASTBOOT_NAME_BOOTARCHIVE] += modp->bm_size;
1291 }
1292 }
1293 #endif /* __xpv */
1294
1295 /*
1296 * Import boot environment module variables as properties, applying
1297 * blacklist filter for variables we know we will not use.
1298 *
1299 * Since the environment can be relatively large, containing many variables
1300 * used only for boot loader purposes, we will use a blacklist based filter.
1301 * To keep the blacklist from growing too large, we use prefix based filtering.
1302 * This is possible because in many cases, the loader variable names are
1303 * using a structured layout.
1304 *
1305 * We will not overwrite already set properties.
1306 *
1307 * Note that the menu items in particular can contain characters not
1308 * well-handled as bootparams, such as spaces, brackets, and the like, so that's
1309 * another reason.
1310 */
1311 static struct bop_blacklist {
1312 const char *bl_name;
1313 int bl_name_len;
1314 } bop_prop_blacklist[] = {
1315 { "ISADIR", sizeof ("ISADIR") },
1316 { "acpi", sizeof ("acpi") },
1317 { "autoboot_delay", sizeof ("autoboot_delay") },
1318 { "beansi_", sizeof ("beansi_") },
1319 { "beastie", sizeof ("beastie") },
1320 { "bemenu", sizeof ("bemenu") },
1321 { "boot.", sizeof ("boot.") },
1322 { "bootenv", sizeof ("bootenv") },
1323 { "currdev", sizeof ("currdev") },
1324 { "dhcp.", sizeof ("dhcp.") },
1325 { "interpret", sizeof ("interpret") },
1326 { "kernel", sizeof ("kernel") },
1327 { "loaddev", sizeof ("loaddev") },
1328 { "loader_", sizeof ("loader_") },
1329 { "mainansi_", sizeof ("mainansi_") },
1330 { "mainmenu_", sizeof ("mainmenu_") },
1331 { "maintoggled_", sizeof ("maintoggled_") },
1332 { "menu_timeout_command", sizeof ("menu_timeout_command") },
1333 { "menuset_", sizeof ("menuset_") },
1334 { "module_path", sizeof ("module_path") },
1335 { "nfs.", sizeof ("nfs.") },
1336 { "optionsansi_", sizeof ("optionsansi_") },
1337 { "optionsmenu_", sizeof ("optionsmenu_") },
1338 { "optionstoggled_", sizeof ("optionstoggled_") },
1339 { "pcibios", sizeof ("pcibios") },
1340 { "prompt", sizeof ("prompt") },
1341 { "smbios", sizeof ("smbios") },
1342 { "tem", sizeof ("tem") },
1343 { "twiddle_divisor", sizeof ("twiddle_divisor") },
1344 { "zfs_be", sizeof ("zfs_be") },
1345 };
1346
1347 /*
1348 * Match the name against prefixes in above blacklist. If the match was
1349 * found, this name is blacklisted.
1350 */
1351 static boolean_t
1352 name_is_blacklisted(const char *name)
1353 {
1354 int i, n;
1355
1356 n = sizeof (bop_prop_blacklist) / sizeof (bop_prop_blacklist[0]);
1357 for (i = 0; i < n; i++) {
1358 if (strncmp(bop_prop_blacklist[i].bl_name, name,
1359 bop_prop_blacklist[i].bl_name_len - 1) == 0) {
1360 return (B_TRUE);
1361 }
1362 }
1363 return (B_FALSE);
1364 }
1365
1366 static void
1367 process_boot_environment(struct boot_modules *benv)
1368 {
1369 char *env, *ptr, *name, *value;
1370 uint32_t size, name_len, value_len;
1371
1372 if (benv == NULL || benv->bm_type != BMT_ENV)
1373 return;
1374 ptr = env = benv->bm_addr;
1375 size = benv->bm_size;
1376 do {
1377 name = ptr;
1378 /* find '=' */
1379 while (*ptr != '=') {
1380 ptr++;
1381 if (ptr > env + size) /* Something is very wrong. */
1382 return;
1383 }
1384 name_len = ptr - name;
1385 if (sizeof (buffer) <= name_len)
1386 continue;
1387
1388 (void) strncpy(buffer, name, sizeof (buffer));
1389 buffer[name_len] = '\0';
1390 name = buffer;
1391
1392 value_len = 0;
1393 value = ++ptr;
1394 while ((uintptr_t)ptr - (uintptr_t)env < size) {
1395 if (*ptr == '\0') {
1396 ptr++;
1397 value_len = (uintptr_t)ptr - (uintptr_t)env;
1398 break;
1399 }
1400 ptr++;
1401 }
1402
1403 /* Did we reach the end of the module? */
1404 if (value_len == 0)
1405 return;
1406
1407 if (*value == '\0')
1408 continue;
1409
1410 /* Is this property already set? */
1411 if (do_bsys_getproplen(NULL, name) >= 0)
1412 continue;
1413
1414 /* Translate netboot variables */
1415 if (strcmp(name, "boot.netif.gateway") == 0) {
1416 bsetprops(BP_ROUTER_IP, value);
1417 continue;
1418 }
1419 if (strcmp(name, "boot.netif.hwaddr") == 0) {
1420 bsetprops(BP_BOOT_MAC, value);
1421 continue;
1422 }
1423 if (strcmp(name, "boot.netif.ip") == 0) {
1424 bsetprops(BP_HOST_IP, value);
1425 continue;
1426 }
1427 if (strcmp(name, "boot.netif.netmask") == 0) {
1428 bsetprops(BP_SUBNET_MASK, value);
1429 continue;
1430 }
1431 if (strcmp(name, "boot.netif.server") == 0) {
1432 bsetprops(BP_SERVER_IP, value);
1433 continue;
1434 }
1435 if (strcmp(name, "boot.netif.server") == 0) {
1436 if (do_bsys_getproplen(NULL, BP_SERVER_IP) < 0)
1437 bsetprops(BP_SERVER_IP, value);
1438 continue;
1439 }
1440 if (strcmp(name, "boot.nfsroot.server") == 0) {
1441 if (do_bsys_getproplen(NULL, BP_SERVER_IP) < 0)
1442 bsetprops(BP_SERVER_IP, value);
1443 continue;
1444 }
1445 if (strcmp(name, "boot.nfsroot.path") == 0) {
1446 bsetprops(BP_SERVER_PATH, value);
1447 continue;
1448 }
1449
1450 if (name_is_blacklisted(name) == B_TRUE)
1451 continue;
1452
1453 /* Create new property. */
1454 bsetprops(name, value);
1455
1456 /* Avoid reading past the module end. */
1457 if (size <= (uintptr_t)ptr - (uintptr_t)env)
1458 return;
1459 } while (*ptr != '\0');
1460 }
1461
1462 /*
1463 * 1st pass at building the table of boot properties. This includes:
1464 * - values set on the command line: -B a=x,b=y,c=z ....
1465 * - known values we just compute (ie. from xbp)
1466 * - values from /boot/solaris/bootenv.rc (ie. eeprom(1m) values)
1467 *
1468 * the grub command line looked like:
1469 * kernel boot-file [-B prop=value[,prop=value]...] [boot-args]
1470 *
1471 * whoami is the same as boot-file
1472 */
1473 static void
1474 build_boot_properties(struct xboot_info *xbp)
1475 {
1476 char *name;
1477 int name_len;
1478 char *value;
1479 int value_len;
1480 struct boot_modules *bm, *rdbm, *benv = NULL;
1481 char *propbuf;
1482 int quoted = 0;
1483 int boot_arg_len;
1484 uint_t i, midx;
1485 char modid[32];
1486 #ifndef __xpv
1487 static int stdout_val = 0;
1488 uchar_t boot_device;
1489 char str[3];
1490 #endif
1491
1492 /*
1493 * These have to be done first, so that kobj_mount_root() works
1494 */
1495 DBG_MSG("Building boot properties\n");
1496 propbuf = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, 0);
1497 DBG((uintptr_t)propbuf);
1498 if (xbp->bi_module_cnt > 0) {
1499 bm = xbp->bi_modules;
1500 rdbm = NULL;
1501 for (midx = i = 0; i < xbp->bi_module_cnt; i++) {
1502 if (bm[i].bm_type == BMT_ROOTFS) {
1503 rdbm = &bm[i];
1504 continue;
1505 }
1506 if (bm[i].bm_type == BMT_HASH ||
1507 bm[i].bm_type == BMT_FONT ||
1508 bm[i].bm_name == NULL)
1509 continue;
1510
1511 if (bm[i].bm_type == BMT_ENV) {
1512 if (benv == NULL)
1513 benv = &bm[i];
1514 else
1515 continue;
1516 }
1517
1518 (void) snprintf(modid, sizeof (modid),
1519 "module-name-%u", midx);
1520 bsetprops(modid, (char *)bm[i].bm_name);
1521 (void) snprintf(modid, sizeof (modid),
1522 "module-addr-%u", midx);
1523 bsetprop64(modid, (uint64_t)(uintptr_t)bm[i].bm_addr);
1524 (void) snprintf(modid, sizeof (modid),
1525 "module-size-%u", midx);
1526 bsetprop64(modid, (uint64_t)bm[i].bm_size);
1527 ++midx;
1528 }
1529 if (rdbm != NULL) {
1530 bsetprop64("ramdisk_start",
1531 (uint64_t)(uintptr_t)rdbm->bm_addr);
1532 bsetprop64("ramdisk_end",
1533 (uint64_t)(uintptr_t)rdbm->bm_addr + rdbm->bm_size);
1534 }
1535 }
1536
1537 /*
1538 * If there are any boot time modules or hashes present, then disable
1539 * fast reboot.
1540 */
1541 if (xbp->bi_module_cnt > 1) {
1542 fastreboot_disable(FBNS_BOOTMOD);
1543 }
1544
1545 #ifndef __xpv
1546 /*
1547 * Disable fast reboot if we're using the Multiboot 2 boot protocol,
1548 * since we don't currently support MB2 info and module relocation.
1549 * Note that fast reboot will have already been disabled if multiple
1550 * modules are present, since the current implementation assumes that
1551 * we only have a single module, the boot_archive.
1552 */
1553 if (xbp->bi_mb_version != 1) {
1554 fastreboot_disable(FBNS_MULTIBOOT2);
1555 }
1556 #endif
1557
1558 DBG_MSG("Parsing command line for boot properties\n");
1559 value = xbp->bi_cmdline;
1560
1561 /*
1562 * allocate memory to collect boot_args into
1563 */
1564 boot_arg_len = strlen(xbp->bi_cmdline) + 1;
1565 boot_args = do_bsys_alloc(NULL, NULL, boot_arg_len, MMU_PAGESIZE);
1566 boot_args[0] = 0;
1567 boot_arg_len = 0;
1568
1569 #ifdef __xpv
1570 /*
1571 * Xen puts a lot of device information in front of the kernel name
1572 * let's grab them and make them boot properties. The first
1573 * string w/o an "=" in it will be the boot-file property.
1574 */
1575 (void) strcpy(namebuf, "xpv-");
1576 for (;;) {
1577 /*
1578 * get to next property
1579 */
1580 while (ISSPACE(*value))
1581 ++value;
1582 name = value;
1583 /*
1584 * look for an "="
1585 */
1586 while (*value && !ISSPACE(*value) && *value != '=') {
1587 value++;
1588 }
1589 if (*value != '=') { /* no "=" in the property */
1590 value = name;
1591 break;
1592 }
1593 name_len = value - name;
1594 value_len = 0;
1595 /*
1596 * skip over the "="
1597 */
1598 value++;
1599 while (value[value_len] && !ISSPACE(value[value_len])) {
1600 ++value_len;
1601 }
1602 /*
1603 * build property name with "xpv-" prefix
1604 */
1605 if (name_len + 4 > 32) { /* skip if name too long */
1606 value += value_len;
1607 continue;
1608 }
1609 bcopy(name, &namebuf[4], name_len);
1610 name_len += 4;
1611 namebuf[name_len] = 0;
1612 bcopy(value, propbuf, value_len);
1613 propbuf[value_len] = 0;
1614 bsetprops(namebuf, propbuf);
1615
1616 /*
1617 * xpv-root is set to the logical disk name of the xen
1618 * VBD when booting from a disk-based filesystem.
1619 */
1620 if (strcmp(namebuf, "xpv-root") == 0)
1621 xen_vbdroot_props(propbuf);
1622 /*
1623 * While we're here, if we have a "xpv-nfsroot" property
1624 * then we need to set "fstype" to "nfs" so we mount
1625 * our root from the nfs server. Also parse the xpv-nfsroot
1626 * property to create the properties that nfs_mountroot will
1627 * need to find the root and mount it.
1628 */
1629 if (strcmp(namebuf, "xpv-nfsroot") == 0)
1630 xen_nfsroot_props(propbuf);
1631
1632 if (strcmp(namebuf, "xpv-ip") == 0)
1633 xen_ip_props(propbuf);
1634 value += value_len;
1635 }
1636 #endif
1637
1638 while (ISSPACE(*value))
1639 ++value;
1640 /*
1641 * value now points at the boot-file
1642 */
1643 value_len = 0;
1644 while (value[value_len] && !ISSPACE(value[value_len]))
1645 ++value_len;
1646 if (value_len > 0) {
1647 whoami = propbuf;
1648 bcopy(value, whoami, value_len);
1649 whoami[value_len] = 0;
1650 bsetprops("boot-file", whoami);
1651 /*
1652 * strip leading path stuff from whoami, so running from
1653 * PXE/miniroot makes sense.
1654 */
1655 if (strstr(whoami, "/platform/") != NULL)
1656 whoami = strstr(whoami, "/platform/");
1657 bsetprops("whoami", whoami);
1658 }
1659
1660 /*
1661 * Values forcibly set boot properties on the command line via -B.
1662 * Allow use of quotes in values. Other stuff goes on kernel
1663 * command line.
1664 */
1665 name = value + value_len;
1666 while (*name != 0) {
1667 /*
1668 * anything not " -B" is copied to the command line
1669 */
1670 if (!ISSPACE(name[0]) || name[1] != '-' || name[2] != 'B') {
1671 boot_args[boot_arg_len++] = *name;
1672 boot_args[boot_arg_len] = 0;
1673 ++name;
1674 continue;
1675 }
1676
1677 /*
1678 * skip the " -B" and following white space
1679 */
1680 name += 3;
1681 while (ISSPACE(*name))
1682 ++name;
1683 while (*name && !ISSPACE(*name)) {
1684 value = strstr(name, "=");
1685 if (value == NULL)
1686 break;
1687 name_len = value - name;
1688 ++value;
1689 value_len = 0;
1690 quoted = 0;
1691 for (; ; ++value_len) {
1692 if (!value[value_len])
1693 break;
1694
1695 /*
1696 * is this value quoted?
1697 */
1698 if (value_len == 0 &&
1699 (value[0] == '\'' || value[0] == '"')) {
1700 quoted = value[0];
1701 ++value_len;
1702 }
1703
1704 /*
1705 * In the quote accept any character,
1706 * but look for ending quote.
1707 */
1708 if (quoted) {
1709 if (value[value_len] == quoted)
1710 quoted = 0;
1711 continue;
1712 }
1713
1714 /*
1715 * a comma or white space ends the value
1716 */
1717 if (value[value_len] == ',' ||
1718 ISSPACE(value[value_len]))
1719 break;
1720 }
1721
1722 if (value_len == 0) {
1723 bsetprop(DDI_PROP_TYPE_ANY, name, name_len,
1724 NULL, 0);
1725 } else {
1726 char *v = value;
1727 int l = value_len;
1728 if (v[0] == v[l - 1] &&
1729 (v[0] == '\'' || v[0] == '"')) {
1730 ++v;
1731 l -= 2;
1732 }
1733 bcopy(v, propbuf, l);
1734 propbuf[l] = '\0';
1735 bsetprop(DDI_PROP_TYPE_STRING, name, name_len,
1736 propbuf, l + 1);
1737 }
1738 name = value + value_len;
1739 while (*name == ',')
1740 ++name;
1741 }
1742 }
1743
1744 /*
1745 * set boot-args property
1746 * 1275 name is bootargs, so set
1747 * that too
1748 */
1749 bsetprops("boot-args", boot_args);
1750 bsetprops("bootargs", boot_args);
1751
1752 process_boot_environment(benv);
1753
1754 #ifndef __xpv
1755 /*
1756 * Build boot command line for Fast Reboot
1757 */
1758 build_fastboot_cmdline(xbp);
1759
1760 if (xbp->bi_mb_version == 1) {
1761 multiboot_info_t *mbi = xbp->bi_mb_info;
1762 int netboot;
1763 struct sol_netinfo *sip;
1764
1765 /*
1766 * set the BIOS boot device from GRUB
1767 */
1768 netboot = 0;
1769
1770 /*
1771 * Save various boot information for Fast Reboot
1772 */
1773 save_boot_info(xbp);
1774
1775 if (mbi != NULL && mbi->flags & MB_INFO_BOOTDEV) {
1776 boot_device = mbi->boot_device >> 24;
1777 if (boot_device == 0x20)
1778 netboot++;
1779 str[0] = (boot_device >> 4) + '0';
1780 str[1] = (boot_device & 0xf) + '0';
1781 str[2] = 0;
1782 bsetprops("bios-boot-device", str);
1783 } else {
1784 netboot = 1;
1785 }
1786
1787 /*
1788 * In the netboot case, drives_info is overloaded with the
1789 * dhcp ack. This is not multiboot compliant and requires
1790 * special pxegrub!
1791 */
1792 if (netboot && mbi->drives_length != 0) {
1793 sip = (struct sol_netinfo *)(uintptr_t)mbi->drives_addr;
1794 if (sip->sn_infotype == SN_TYPE_BOOTP)
1795 bsetprop(DDI_PROP_TYPE_BYTE,
1796 "bootp-response",
1797 sizeof ("bootp-response"),
1798 (void *)(uintptr_t)mbi->drives_addr,
1799 mbi->drives_length);
1800 else if (sip->sn_infotype == SN_TYPE_RARP)
1801 setup_rarp_props(sip);
1802 }
1803 } else {
1804 multiboot2_info_header_t *mbi = xbp->bi_mb_info;
1805 multiboot_tag_bootdev_t *bootdev = NULL;
1806 multiboot_tag_network_t *netdev = NULL;
1807
1808 if (mbi != NULL) {
1809 bootdev = dboot_multiboot2_find_tag(mbi,
1810 MULTIBOOT_TAG_TYPE_BOOTDEV);
1811 netdev = dboot_multiboot2_find_tag(mbi,
1812 MULTIBOOT_TAG_TYPE_NETWORK);
1813 }
1814 if (bootdev != NULL) {
1815 DBG(bootdev->mb_biosdev);
1816 boot_device = bootdev->mb_biosdev;
1817 str[0] = (boot_device >> 4) + '0';
1818 str[1] = (boot_device & 0xf) + '0';
1819 str[2] = 0;
1820 bsetprops("bios-boot-device", str);
1821 }
1822 if (netdev != NULL) {
1823 bsetprop(DDI_PROP_TYPE_BYTE,
1824 "bootp-response", sizeof ("bootp-response"),
1825 (void *)(uintptr_t)netdev->mb_dhcpack,
1826 netdev->mb_size -
1827 sizeof (multiboot_tag_network_t));
1828 }
1829 }
1830
1831 bsetprop32("stdout", stdout_val);
1832 #endif /* __xpv */
1833
1834 /*
1835 * more conjured up values for made up things....
1836 */
1837 #if defined(__xpv)
1838 bsetprops("mfg-name", "i86xpv");
1839 bsetprops("impl-arch-name", "i86xpv");
1840 #else
1841 bsetprops("mfg-name", "i86pc");
1842 bsetprops("impl-arch-name", "i86pc");
1843 #endif
1844
1845 /*
1846 * Build firmware-provided system properties
1847 */
1848 build_firmware_properties(xbp);
1849
1850 /*
1851 * XXPV
1852 *
1853 * Find out what these are:
1854 * - cpuid_feature_ecx_include
1855 * - cpuid_feature_ecx_exclude
1856 * - cpuid_feature_edx_include
1857 * - cpuid_feature_edx_exclude
1858 *
1859 * Find out what these are in multiboot:
1860 * - netdev-path
1861 * - fstype
1862 */
1863 }
1864
1865 #ifdef __xpv
1866 /*
1867 * Under the Hypervisor, memory usable for DMA may be scarce. One
1868 * very likely large pool of DMA friendly memory is occupied by
1869 * the boot_archive, as it was loaded by grub into low MFNs.
1870 *
1871 * Here we free up that memory by copying the boot archive to what are
1872 * likely higher MFN pages and then swapping the mfn/pfn mappings.
1873 */
1874 #define PFN_2GIG 0x80000
1875 static void
1876 relocate_boot_archive(struct xboot_info *xbp)
1877 {
1878 mfn_t max_mfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
1879 struct boot_modules *bm = xbp->bi_modules;
1880 uintptr_t va;
1881 pfn_t va_pfn;
1882 mfn_t va_mfn;
1883 caddr_t copy;
1884 pfn_t copy_pfn;
1885 mfn_t copy_mfn;
1886 size_t len;
1887 int slop;
1888 int total = 0;
1889 int relocated = 0;
1890 int mmu_update_return;
1891 mmu_update_t t[2];
1892 x86pte_t pte;
1893
1894 /*
1895 * If all MFN's are below 2Gig, don't bother doing this.
1896 */
1897 if (max_mfn < PFN_2GIG)
1898 return;
1899 if (xbp->bi_module_cnt < 1) {
1900 DBG_MSG("no boot_archive!");
1901 return;
1902 }
1903
1904 DBG_MSG("moving boot_archive to high MFN memory\n");
1905 va = (uintptr_t)bm->bm_addr;
1906 len = bm->bm_size;
1907 slop = va & MMU_PAGEOFFSET;
1908 if (slop) {
1909 va += MMU_PAGESIZE - slop;
1910 len -= MMU_PAGESIZE - slop;
1911 }
1912 len = P2ALIGN(len, MMU_PAGESIZE);
1913
1914 /*
1915 * Go through all boot_archive pages, swapping any low MFN pages
1916 * with memory at next_phys.
1917 */
1918 while (len != 0) {
1919 ++total;
1920 va_pfn = mmu_btop(va - ONE_GIG);
1921 va_mfn = mfn_list[va_pfn];
1922 if (mfn_list[va_pfn] < PFN_2GIG) {
1923 copy = kbm_remap_window(next_phys, 1);
1924 bcopy((void *)va, copy, MMU_PAGESIZE);
1925 copy_pfn = mmu_btop(next_phys);
1926 copy_mfn = mfn_list[copy_pfn];
1927
1928 pte = mfn_to_ma(copy_mfn) | PT_NOCONSIST | PT_VALID;
1929 if (HYPERVISOR_update_va_mapping(va, pte,
1930 UVMF_INVLPG | UVMF_LOCAL))
1931 bop_panic("relocate_boot_archive(): "
1932 "HYPERVISOR_update_va_mapping() failed");
1933
1934 mfn_list[va_pfn] = copy_mfn;
1935 mfn_list[copy_pfn] = va_mfn;
1936
1937 t[0].ptr = mfn_to_ma(copy_mfn) | MMU_MACHPHYS_UPDATE;
1938 t[0].val = va_pfn;
1939 t[1].ptr = mfn_to_ma(va_mfn) | MMU_MACHPHYS_UPDATE;
1940 t[1].val = copy_pfn;
1941 if (HYPERVISOR_mmu_update(t, 2, &mmu_update_return,
1942 DOMID_SELF) != 0 || mmu_update_return != 2)
1943 bop_panic("relocate_boot_archive(): "
1944 "HYPERVISOR_mmu_update() failed");
1945
1946 next_phys += MMU_PAGESIZE;
1947 ++relocated;
1948 }
1949 len -= MMU_PAGESIZE;
1950 va += MMU_PAGESIZE;
1951 }
1952 DBG_MSG("Relocated pages:\n");
1953 DBG(relocated);
1954 DBG_MSG("Out of total pages:\n");
1955 DBG(total);
1956 }
1957 #endif /* __xpv */
1958
1959 #if !defined(__xpv)
1960 /*
1961 * simple description of a stack frame (args are 32 bit only currently)
1962 */
1963 typedef struct bop_frame {
1964 struct bop_frame *old_frame;
1965 pc_t retaddr;
1966 long arg[1];
1967 } bop_frame_t;
1968
1969 void
1970 bop_traceback(bop_frame_t *frame)
1971 {
1972 pc_t pc;
1973 int cnt;
1974 char *ksym;
1975 ulong_t off;
1976
1977 bop_printf(NULL, "Stack traceback:\n");
1978 for (cnt = 0; cnt < 30; ++cnt) { /* up to 30 frames */
1979 pc = frame->retaddr;
1980 if (pc == 0)
1981 break;
1982 ksym = kobj_getsymname(pc, &off);
1983 if (ksym)
1984 bop_printf(NULL, " %s+%lx", ksym, off);
1985 else
1986 bop_printf(NULL, " 0x%lx", pc);
1987
1988 frame = frame->old_frame;
1989 if (frame == 0) {
1990 bop_printf(NULL, "\n");
1991 break;
1992 }
1993 bop_printf(NULL, "\n");
1994 }
1995 }
1996
1997 struct trapframe {
1998 ulong_t error_code; /* optional */
1999 ulong_t inst_ptr;
2000 ulong_t code_seg;
2001 ulong_t flags_reg;
2002 ulong_t stk_ptr;
2003 ulong_t stk_seg;
2004 };
2005
2006 void
2007 bop_trap(ulong_t *tfp)
2008 {
2009 struct trapframe *tf = (struct trapframe *)tfp;
2010 bop_frame_t fakeframe;
2011 static int depth = 0;
2012
2013 /*
2014 * Check for an infinite loop of traps.
2015 */
2016 if (++depth > 2)
2017 bop_panic("Nested trap");
2018
2019 bop_printf(NULL, "Unexpected trap\n");
2020
2021 /*
2022 * adjust the tf for optional error_code by detecting the code selector
2023 */
2024 if (tf->code_seg != B64CODE_SEL)
2025 tf = (struct trapframe *)(tfp - 1);
2026 else
2027 bop_printf(NULL, "error code 0x%lx\n",
2028 tf->error_code & 0xffffffff);
2029
2030 bop_printf(NULL, "instruction pointer 0x%lx\n", tf->inst_ptr);
2031 bop_printf(NULL, "code segment 0x%lx\n", tf->code_seg & 0xffff);
2032 bop_printf(NULL, "flags register 0x%lx\n", tf->flags_reg);
2033 bop_printf(NULL, "return %%rsp 0x%lx\n", tf->stk_ptr);
2034 bop_printf(NULL, "return %%ss 0x%lx\n", tf->stk_seg & 0xffff);
2035 bop_printf(NULL, "%%cr2 0x%lx\n", getcr2());
2036
2037 /* grab %[er]bp pushed by our code from the stack */
2038 fakeframe.old_frame = (bop_frame_t *)*(tfp - 3);
2039 fakeframe.retaddr = (pc_t)tf->inst_ptr;
2040 bop_printf(NULL, "Attempting stack backtrace:\n");
2041 bop_traceback(&fakeframe);
2042 bop_panic("unexpected trap in early boot");
2043 }
2044
2045 extern void bop_trap_handler(void);
2046
2047 static gate_desc_t *bop_idt;
2048
2049 static desctbr_t bop_idt_info;
2050
2051 /*
2052 * Install a temporary IDT that lets us catch errors in the boot time code.
2053 * We shouldn't get any faults at all while this is installed, so we'll
2054 * just generate a traceback and exit.
2055 */
2056 static void
2057 bop_idt_init(void)
2058 {
2059 int t;
2060
2061 bop_idt = (gate_desc_t *)
2062 do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
2063 bzero(bop_idt, MMU_PAGESIZE);
2064 for (t = 0; t < NIDT; ++t) {
2065 /*
2066 * Note that since boot runs without a TSS, the
2067 * double fault handler cannot use an alternate stack (64-bit).
2068 */
2069 set_gatesegd(&bop_idt[t], &bop_trap_handler, B64CODE_SEL,
2070 SDT_SYSIGT, TRP_KPL, 0);
2071 }
2072 bop_idt_info.dtr_limit = (NIDT * sizeof (gate_desc_t)) - 1;
2073 bop_idt_info.dtr_base = (uintptr_t)bop_idt;
2074 wr_idtr(&bop_idt_info);
2075 }
2076 #endif /* !defined(__xpv) */
2077
2078 /*
2079 * This is where we enter the kernel. It dummies up the boot_ops and
2080 * boot_syscalls vectors and jumps off to _kobj_boot()
2081 */
2082 void
2083 _start(struct xboot_info *xbp)
2084 {
2085 bootops_t *bops = &bootop;
2086 extern void _kobj_boot();
2087
2088 /*
2089 * 1st off - initialize the console for any error messages
2090 */
2091 xbootp = xbp;
2092 #ifdef __xpv
2093 HYPERVISOR_shared_info = (void *)xbp->bi_shared_info;
2094 xen_info = xbp->bi_xen_start_info;
2095 #endif
2096
2097 #ifndef __xpv
2098 if (*((uint32_t *)(FASTBOOT_SWTCH_PA + FASTBOOT_STACK_OFFSET)) ==
2099 FASTBOOT_MAGIC) {
2100 post_fastreboot = 1;
2101 *((uint32_t *)(FASTBOOT_SWTCH_PA + FASTBOOT_STACK_OFFSET)) = 0;
2102 }
2103 #endif
2104
2105 bcons_init(xbp);
2106 have_console = 1;
2107
2108 /*
2109 * enable debugging
2110 */
2111 if (find_boot_prop("kbm_debug") != NULL)
2112 kbm_debug = 1;
2113
2114 DBG_MSG("\n\n*** Entered Solaris in _start() cmdline is: ");
2115 DBG_MSG((char *)xbp->bi_cmdline);
2116 DBG_MSG("\n\n\n");
2117
2118 /*
2119 * physavail is no longer used by startup
2120 */
2121 bm.physinstalled = xbp->bi_phys_install;
2122 bm.pcimem = xbp->bi_pcimem;
2123 bm.rsvdmem = xbp->bi_rsvdmem;
2124 bm.physavail = NULL;
2125
2126 /*
2127 * initialize the boot time allocator
2128 */
2129 next_phys = xbp->bi_next_paddr;
2130 DBG(next_phys);
2131 next_virt = (uintptr_t)xbp->bi_next_vaddr;
2132 DBG(next_virt);
2133 DBG_MSG("Initializing boot time memory management...");
2134 #ifdef __xpv
2135 {
2136 xen_platform_parameters_t p;
2137
2138 /* This call shouldn't fail, dboot already did it once. */
2139 (void) HYPERVISOR_xen_version(XENVER_platform_parameters, &p);
2140 mfn_to_pfn_mapping = (pfn_t *)(xen_virt_start = p.virt_start);
2141 DBG(xen_virt_start);
2142 }
2143 #endif
2144 kbm_init(xbp);
2145 DBG_MSG("done\n");
2146
2147 /*
2148 * Fill in the bootops vector
2149 */
2150 bops->bsys_version = BO_VERSION;
2151 bops->boot_mem = &bm;
2152 bops->bsys_alloc = do_bsys_alloc;
2153 bops->bsys_free = do_bsys_free;
2154 bops->bsys_getproplen = do_bsys_getproplen;
2155 bops->bsys_getprop = do_bsys_getprop;
2156 bops->bsys_nextprop = do_bsys_nextprop;
2157 bops->bsys_printf = bop_printf;
2158 bops->bsys_doint = do_bsys_doint;
2159
2160 /*
2161 * BOP_EALLOC() is no longer needed
2162 */
2163 bops->bsys_ealloc = do_bsys_ealloc;
2164
2165 #ifdef __xpv
2166 /*
2167 * On domain 0 we need to free up some physical memory that is
2168 * usable for DMA. Since GRUB loaded the boot_archive, it is
2169 * sitting in low MFN memory. We'll relocated the boot archive
2170 * pages to high PFN memory.
2171 */
2172 if (DOMAIN_IS_INITDOMAIN(xen_info))
2173 relocate_boot_archive(xbp);
2174 #endif
2175
2176 #ifndef __xpv
2177 /*
2178 * Install an IDT to catch early pagefaults (shouldn't have any).
2179 * Also needed for kmdb.
2180 */
2181 bop_idt_init();
2182 #endif
2183 /* Set up the shadow fb for framebuffer console */
2184 boot_fb_shadow_init(bops);
2185
2186 /*
2187 * Start building the boot properties from the command line
2188 */
2189 DBG_MSG("Initializing boot properties:\n");
2190 build_boot_properties(xbp);
2191
2192 if (find_boot_prop("prom_debug") || kbm_debug) {
2193 char *value;
2194
2195 value = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
2196 boot_prop_display(value);
2197 }
2198
2199 /*
2200 * jump into krtld...
2201 */
2202 _kobj_boot(&bop_sysp, NULL, bops, NULL);
2203 }
2204
2205
2206 /*ARGSUSED*/
2207 static caddr_t
2208 no_more_alloc(bootops_t *bop, caddr_t virthint, size_t size, int align)
2209 {
2210 panic("Attempt to bsys_alloc() too late\n");
2211 return (NULL);
2212 }
2213
2214 /*ARGSUSED*/
2215 static void
2216 no_more_free(bootops_t *bop, caddr_t virt, size_t size)
2217 {
2218 panic("Attempt to bsys_free() too late\n");
2219 }
2220
2221 void
2222 bop_no_more_mem(void)
2223 {
2224 DBG(total_bop_alloc_scratch);
2225 DBG(total_bop_alloc_kernel);
2226 bootops->bsys_alloc = no_more_alloc;
2227 bootops->bsys_free = no_more_free;
2228 }
2229
2230
2231 /*
2232 * Set ACPI firmware properties
2233 */
2234
2235 static caddr_t
2236 vmap_phys(size_t length, paddr_t pa)
2237 {
2238 paddr_t start, end;
2239 caddr_t va;
2240 size_t len, page;
2241
2242 #ifdef __xpv
2243 pa = pfn_to_pa(xen_assign_pfn(mmu_btop(pa))) | (pa & MMU_PAGEOFFSET);
2244 #endif
2245 start = P2ALIGN(pa, MMU_PAGESIZE);
2246 end = P2ROUNDUP(pa + length, MMU_PAGESIZE);
2247 len = end - start;
2248 va = (caddr_t)alloc_vaddr(len, MMU_PAGESIZE);
2249 for (page = 0; page < len; page += MMU_PAGESIZE)
2250 kbm_map((uintptr_t)va + page, start + page, 0, 0);
2251 return (va + (pa & MMU_PAGEOFFSET));
2252 }
2253
2254 static uint8_t
2255 checksum_table(uint8_t *tp, size_t len)
2256 {
2257 uint8_t sum = 0;
2258
2259 while (len-- > 0)
2260 sum += *tp++;
2261
2262 return (sum);
2263 }
2264
2265 static int
2266 valid_rsdp(ACPI_TABLE_RSDP *rp)
2267 {
2268
2269 /* validate the V1.x checksum */
2270 if (checksum_table((uint8_t *)rp, ACPI_RSDP_CHECKSUM_LENGTH) != 0)
2271 return (0);
2272
2273 /* If pre-ACPI 2.0, this is a valid RSDP */
2274 if (rp->Revision < 2)
2275 return (1);
2276
2277 /* validate the V2.x checksum */
2278 if (checksum_table((uint8_t *)rp, ACPI_RSDP_XCHECKSUM_LENGTH) != 0)
2279 return (0);
2280
2281 return (1);
2282 }
2283
2284 /*
2285 * Scan memory range for an RSDP;
2286 * see ACPI 3.0 Spec, 5.2.5.1
2287 */
2288 static ACPI_TABLE_RSDP *
2289 scan_rsdp(paddr_t *paddrp, size_t len)
2290 {
2291 paddr_t paddr = *paddrp;
2292 caddr_t ptr;
2293
2294 ptr = vmap_phys(len, paddr);
2295
2296 while (len > 0) {
2297 if (strncmp(ptr, ACPI_SIG_RSDP, strlen(ACPI_SIG_RSDP)) == 0 &&
2298 valid_rsdp((ACPI_TABLE_RSDP *)ptr)) {
2299 *paddrp = paddr;
2300 return ((ACPI_TABLE_RSDP *)ptr);
2301 }
2302
2303 ptr += ACPI_RSDP_SCAN_STEP;
2304 paddr += ACPI_RSDP_SCAN_STEP;
2305 len -= ACPI_RSDP_SCAN_STEP;
2306 }
2307
2308 return (NULL);
2309 }
2310
2311 /*
2312 * Locate the ACPI RSDP. We search in a particular order:
2313 *
2314 * - If the bootloader told us the location of the RSDP (via the EFI system
2315 * table), try that first.
2316 * - Otherwise, look in the EBDA and BIOS memory as per ACPI 5.2.5.1 (legacy
2317 * case).
2318 * - Finally, our bootloader may have a copy of the RSDP in its info: this might
2319 * get freed after boot, so we always prefer to find the original RSDP first.
2320 *
2321 * Once found, we set acpi-root-tab property (a physical address) for the
2322 * benefit of acpica, acpidump etc.
2323 */
2324
2325 static ACPI_TABLE_RSDP *
2326 find_rsdp(struct xboot_info *xbp)
2327 {
2328 ACPI_TABLE_RSDP *rsdp = NULL;
2329 paddr_t paddr = 0;
2330
2331 if (do_bsys_getproplen(NULL, "acpi-root-tab") == sizeof (uint64_t)) {
2332 (void) do_bsys_getprop(NULL, "acpi-root-tab", &paddr);
2333 rsdp = scan_rsdp(&paddr, sizeof (*rsdp));
2334 }
2335
2336 #ifndef __xpv
2337 if (rsdp == NULL && xbp->bi_acpi_rsdp != NULL) {
2338 paddr = (uintptr_t)xbp->bi_acpi_rsdp;
2339 rsdp = scan_rsdp(&paddr, sizeof (*rsdp));
2340 }
2341 #endif
2342
2343 if (rsdp == NULL) {
2344 uint16_t *ebda_seg = (uint16_t *)vmap_phys(sizeof (uint16_t),
2345 ACPI_EBDA_PTR_LOCATION);
2346 paddr = *ebda_seg << 4;
2347 rsdp = scan_rsdp(&paddr, ACPI_EBDA_WINDOW_SIZE);
2348 }
2349
2350 if (rsdp == NULL) {
2351 paddr = ACPI_HI_RSDP_WINDOW_BASE;
2352 rsdp = scan_rsdp(&paddr, ACPI_HI_RSDP_WINDOW_SIZE);
2353 }
2354
2355 #ifndef __xpv
2356 if (rsdp == NULL && xbp->bi_acpi_rsdp_copy != NULL) {
2357 paddr = (uintptr_t)xbp->bi_acpi_rsdp_copy;
2358 rsdp = scan_rsdp(&paddr, sizeof (*rsdp));
2359 }
2360 #endif
2361
2362 if (rsdp == NULL) {
2363 bop_printf(NULL, "no RSDP found!\n");
2364 return (NULL);
2365 }
2366
2367 if (kbm_debug)
2368 bop_printf(NULL, "RSDP found at physical 0x%lx\n", paddr);
2369
2370 if (do_bsys_getproplen(NULL, "acpi-root-tab") != sizeof (uint64_t))
2371 bsetprop64("acpi-root-tab", paddr);
2372
2373 return (rsdp);
2374 }
2375
2376 static ACPI_TABLE_HEADER *
2377 map_fw_table(paddr_t table_addr)
2378 {
2379 ACPI_TABLE_HEADER *tp;
2380 size_t len = MAX(sizeof (*tp), MMU_PAGESIZE);
2381
2382 /*
2383 * Map at least a page; if the table is larger than this, remap it
2384 */
2385 tp = (ACPI_TABLE_HEADER *)vmap_phys(len, table_addr);
2386 if (tp->Length > len)
2387 tp = (ACPI_TABLE_HEADER *)vmap_phys(tp->Length, table_addr);
2388 return (tp);
2389 }
2390
2391 static ACPI_TABLE_HEADER *
2392 find_fw_table(ACPI_TABLE_RSDP *rsdp, char *signature)
2393 {
2394 static int revision = 0;
2395 static ACPI_TABLE_XSDT *xsdt;
2396 static int len;
2397 paddr_t xsdt_addr;
2398 ACPI_TABLE_HEADER *tp;
2399 paddr_t table_addr;
2400 int n;
2401
2402 if (strlen(signature) != ACPI_NAME_SIZE)
2403 return (NULL);
2404
2405 /*
2406 * Reading the ACPI 3.0 Spec, section 5.2.5.3 will help
2407 * understand this code. If we haven't already found the RSDT/XSDT,
2408 * revision will be 0. Find the RSDP and check the revision
2409 * to find out whether to use the RSDT or XSDT. If revision is
2410 * 0 or 1, use the RSDT and set internal revision to 1; if it is 2,
2411 * use the XSDT. If the XSDT address is 0, though, fall back to
2412 * revision 1 and use the RSDT.
2413 */
2414 if (revision == 0) {
2415 if (rsdp == NULL)
2416 return (NULL);
2417
2418 revision = rsdp->Revision;
2419 /*
2420 * ACPI 6.0 states that current revision is 2
2421 * from acpi_table_rsdp definition:
2422 * Must be (0) for ACPI 1.0 or (2) for ACPI 2.0+
2423 */
2424 if (revision > 2)
2425 revision = 2;
2426 switch (revision) {
2427 case 2:
2428 /*
2429 * Use the XSDT unless BIOS is buggy and
2430 * claims to be rev 2 but has a null XSDT
2431 * address
2432 */
2433 xsdt_addr = rsdp->XsdtPhysicalAddress;
2434 if (xsdt_addr != 0)
2435 break;
2436 /* FALLTHROUGH */
2437 case 0:
2438 /* treat RSDP rev 0 as revision 1 internally */
2439 revision = 1;
2440 /* FALLTHROUGH */
2441 case 1:
2442 /* use the RSDT for rev 0/1 */
2443 xsdt_addr = rsdp->RsdtPhysicalAddress;
2444 break;
2445 default:
2446 /* unknown revision */
2447 revision = 0;
2448 break;
2449 }
2450
2451 if (revision == 0)
2452 return (NULL);
2453
2454 /* cache the XSDT info */
2455 xsdt = (ACPI_TABLE_XSDT *)map_fw_table(xsdt_addr);
2456 len = (xsdt->Header.Length - sizeof (xsdt->Header)) /
2457 ((revision == 1) ? sizeof (uint32_t) : sizeof (uint64_t));
2458 }
2459
2460 /*
2461 * Scan the table headers looking for a signature match
2462 */
2463 for (n = 0; n < len; n++) {
2464 ACPI_TABLE_RSDT *rsdt = (ACPI_TABLE_RSDT *)xsdt;
2465 table_addr = (revision == 1) ? rsdt->TableOffsetEntry[n] :
2466 xsdt->TableOffsetEntry[n];
2467
2468 if (table_addr == 0)
2469 continue;
2470 tp = map_fw_table(table_addr);
2471 if (strncmp(tp->Signature, signature, ACPI_NAME_SIZE) == 0) {
2472 return (tp);
2473 }
2474 }
2475 return (NULL);
2476 }
2477
2478 static void
2479 process_mcfg(ACPI_TABLE_MCFG *tp)
2480 {
2481 ACPI_MCFG_ALLOCATION *cfg_baap;
2482 char *cfg_baa_endp;
2483 int64_t ecfginfo[4];
2484
2485 cfg_baap = (ACPI_MCFG_ALLOCATION *)((uintptr_t)tp + sizeof (*tp));
2486 cfg_baa_endp = ((char *)tp) + tp->Header.Length;
2487 while ((char *)cfg_baap < cfg_baa_endp) {
2488 if (cfg_baap->Address != 0 && cfg_baap->PciSegment == 0) {
2489 ecfginfo[0] = cfg_baap->Address;
2490 ecfginfo[1] = cfg_baap->PciSegment;
2491 ecfginfo[2] = cfg_baap->StartBusNumber;
2492 ecfginfo[3] = cfg_baap->EndBusNumber;
2493 bsetprop(DDI_PROP_TYPE_INT64,
2494 MCFG_PROPNAME, strlen(MCFG_PROPNAME),
2495 ecfginfo, sizeof (ecfginfo));
2496 break;
2497 }
2498 cfg_baap++;
2499 }
2500 }
2501
2502 #ifndef __xpv
2503 static void
2504 process_madt_entries(ACPI_TABLE_MADT *tp, uint32_t *cpu_countp,
2505 uint32_t *cpu_possible_countp, uint32_t *cpu_apicid_array)
2506 {
2507 ACPI_SUBTABLE_HEADER *item, *end;
2508 uint32_t cpu_count = 0;
2509 uint32_t cpu_possible_count = 0;
2510
2511 /*
2512 * Determine number of CPUs and keep track of "final" APIC ID
2513 * for each CPU by walking through ACPI MADT processor list
2514 */
2515 end = (ACPI_SUBTABLE_HEADER *)(tp->Header.Length + (uintptr_t)tp);
2516 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)tp + sizeof (*tp));
2517
2518 while (item < end) {
2519 switch (item->Type) {
2520 case ACPI_MADT_TYPE_LOCAL_APIC: {
2521 ACPI_MADT_LOCAL_APIC *cpu =
2522 (ACPI_MADT_LOCAL_APIC *) item;
2523
2524 if (cpu->LapicFlags & ACPI_MADT_ENABLED) {
2525 if (cpu_apicid_array != NULL)
2526 cpu_apicid_array[cpu_count] = cpu->Id;
2527 cpu_count++;
2528 }
2529 cpu_possible_count++;
2530 break;
2531 }
2532 case ACPI_MADT_TYPE_LOCAL_X2APIC: {
2533 ACPI_MADT_LOCAL_X2APIC *cpu =
2534 (ACPI_MADT_LOCAL_X2APIC *) item;
2535
2536 if (cpu->LapicFlags & ACPI_MADT_ENABLED) {
2537 if (cpu_apicid_array != NULL)
2538 cpu_apicid_array[cpu_count] =
2539 cpu->LocalApicId;
2540 cpu_count++;
2541 }
2542 cpu_possible_count++;
2543 break;
2544 }
2545 default:
2546 if (kbm_debug)
2547 bop_printf(NULL, "MADT type %d\n", item->Type);
2548 break;
2549 }
2550
2551 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)item + item->Length);
2552 }
2553 if (cpu_countp)
2554 *cpu_countp = cpu_count;
2555 if (cpu_possible_countp)
2556 *cpu_possible_countp = cpu_possible_count;
2557 }
2558
2559 static void
2560 process_madt(ACPI_TABLE_MADT *tp)
2561 {
2562 uint32_t cpu_count = 0;
2563 uint32_t cpu_possible_count = 0;
2564 uint32_t *cpu_apicid_array; /* x2APIC ID is 32bit! */
2565
2566 if (tp != NULL) {
2567 /* count cpu's */
2568 process_madt_entries(tp, &cpu_count, &cpu_possible_count, NULL);
2569
2570 cpu_apicid_array = (uint32_t *)do_bsys_alloc(NULL, NULL,
2571 cpu_count * sizeof (*cpu_apicid_array), MMU_PAGESIZE);
2572 if (cpu_apicid_array == NULL)
2573 bop_panic("Not enough memory for APIC ID array");
2574
2575 /* copy IDs */
2576 process_madt_entries(tp, NULL, NULL, cpu_apicid_array);
2577
2578 /*
2579 * Make boot property for array of "final" APIC IDs for each
2580 * CPU
2581 */
2582 bsetprop(DDI_PROP_TYPE_INT,
2583 BP_CPU_APICID_ARRAY, strlen(BP_CPU_APICID_ARRAY),
2584 cpu_apicid_array, cpu_count * sizeof (*cpu_apicid_array));
2585 }
2586
2587 /*
2588 * Check whether property plat-max-ncpus is already set.
2589 */
2590 if (do_bsys_getproplen(NULL, PLAT_MAX_NCPUS_NAME) < 0) {
2591 /*
2592 * Set plat-max-ncpus to number of maximum possible CPUs given
2593 * in MADT if it hasn't been set.
2594 * There's no formal way to detect max possible CPUs supported
2595 * by platform according to ACPI spec3.0b. So current CPU
2596 * hotplug implementation expects that all possible CPUs will
2597 * have an entry in MADT table and set plat-max-ncpus to number
2598 * of entries in MADT.
2599 * With introducing of ACPI4.0, Maximum System Capability Table
2600 * (MSCT) provides maximum number of CPUs supported by platform.
2601 * If MSCT is unavailable, fall back to old way.
2602 */
2603 if (tp != NULL)
2604 bsetpropsi(PLAT_MAX_NCPUS_NAME, cpu_possible_count);
2605 }
2606
2607 /*
2608 * Set boot property boot-max-ncpus to number of CPUs existing at
2609 * boot time. boot-max-ncpus is mainly used for optimization.
2610 */
2611 if (tp != NULL)
2612 bsetpropsi(BOOT_MAX_NCPUS_NAME, cpu_count);
2613
2614 /*
2615 * User-set boot-ncpus overrides firmware count
2616 */
2617 if (do_bsys_getproplen(NULL, BOOT_NCPUS_NAME) >= 0)
2618 return;
2619
2620 /*
2621 * Set boot property boot-ncpus to number of active CPUs given in MADT
2622 * if it hasn't been set yet.
2623 */
2624 if (tp != NULL)
2625 bsetpropsi(BOOT_NCPUS_NAME, cpu_count);
2626 }
2627
2628 static void
2629 process_srat(ACPI_TABLE_SRAT *tp)
2630 {
2631 ACPI_SUBTABLE_HEADER *item, *end;
2632 int i;
2633 int proc_num, mem_num;
2634 #pragma pack(1)
2635 struct {
2636 uint32_t domain;
2637 uint32_t apic_id;
2638 uint32_t sapic_id;
2639 } processor;
2640 struct {
2641 uint32_t domain;
2642 uint32_t x2apic_id;
2643 } x2apic;
2644 struct {
2645 uint32_t domain;
2646 uint64_t addr;
2647 uint64_t length;
2648 uint32_t flags;
2649 } memory;
2650 #pragma pack()
2651 char prop_name[30];
2652 uint64_t maxmem = 0;
2653
2654 if (tp == NULL)
2655 return;
2656
2657 proc_num = mem_num = 0;
2658 end = (ACPI_SUBTABLE_HEADER *)(tp->Header.Length + (uintptr_t)tp);
2659 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)tp + sizeof (*tp));
2660 while (item < end) {
2661 switch (item->Type) {
2662 case ACPI_SRAT_TYPE_CPU_AFFINITY: {
2663 ACPI_SRAT_CPU_AFFINITY *cpu =
2664 (ACPI_SRAT_CPU_AFFINITY *) item;
2665
2666 if (!(cpu->Flags & ACPI_SRAT_CPU_ENABLED))
2667 break;
2668 processor.domain = cpu->ProximityDomainLo;
2669 for (i = 0; i < 3; i++)
2670 processor.domain +=
2671 cpu->ProximityDomainHi[i] << ((i + 1) * 8);
2672 processor.apic_id = cpu->ApicId;
2673 processor.sapic_id = cpu->LocalSapicEid;
2674 (void) snprintf(prop_name, 30, "acpi-srat-processor-%d",
2675 proc_num);
2676 bsetprop(DDI_PROP_TYPE_INT,
2677 prop_name, strlen(prop_name), &processor,
2678 sizeof (processor));
2679 proc_num++;
2680 break;
2681 }
2682 case ACPI_SRAT_TYPE_MEMORY_AFFINITY: {
2683 ACPI_SRAT_MEM_AFFINITY *mem =
2684 (ACPI_SRAT_MEM_AFFINITY *)item;
2685
2686 if (!(mem->Flags & ACPI_SRAT_MEM_ENABLED))
2687 break;
2688 memory.domain = mem->ProximityDomain;
2689 memory.addr = mem->BaseAddress;
2690 memory.length = mem->Length;
2691 memory.flags = mem->Flags;
2692 (void) snprintf(prop_name, 30, "acpi-srat-memory-%d",
2693 mem_num);
2694 bsetprop(DDI_PROP_TYPE_INT,
2695 prop_name, strlen(prop_name), &memory,
2696 sizeof (memory));
2697 if ((mem->Flags & ACPI_SRAT_MEM_HOT_PLUGGABLE) &&
2698 (memory.addr + memory.length > maxmem)) {
2699 maxmem = memory.addr + memory.length;
2700 }
2701 mem_num++;
2702 break;
2703 }
2704 case ACPI_SRAT_TYPE_X2APIC_CPU_AFFINITY: {
2705 ACPI_SRAT_X2APIC_CPU_AFFINITY *x2cpu =
2706 (ACPI_SRAT_X2APIC_CPU_AFFINITY *) item;
2707
2708 if (!(x2cpu->Flags & ACPI_SRAT_CPU_ENABLED))
2709 break;
2710 x2apic.domain = x2cpu->ProximityDomain;
2711 x2apic.x2apic_id = x2cpu->ApicId;
2712 (void) snprintf(prop_name, 30, "acpi-srat-processor-%d",
2713 proc_num);
2714 bsetprop(DDI_PROP_TYPE_INT,
2715 prop_name, strlen(prop_name), &x2apic,
2716 sizeof (x2apic));
2717 proc_num++;
2718 break;
2719 }
2720 default:
2721 if (kbm_debug)
2722 bop_printf(NULL, "SRAT type %d\n", item->Type);
2723 break;
2724 }
2725
2726 item = (ACPI_SUBTABLE_HEADER *)
2727 (item->Length + (uintptr_t)item);
2728 }
2729
2730 /*
2731 * The maximum physical address calculated from the SRAT table is more
2732 * accurate than that calculated from the MSCT table.
2733 */
2734 if (maxmem != 0) {
2735 plat_dr_physmax = btop(maxmem);
2736 }
2737 }
2738
2739 static void
2740 process_slit(ACPI_TABLE_SLIT *tp)
2741 {
2742
2743 /*
2744 * Check the number of localities; if it's too huge, we just
2745 * return and locality enumeration code will handle this later,
2746 * if possible.
2747 *
2748 * Note that the size of the table is the square of the
2749 * number of localities; if the number of localities exceeds
2750 * UINT16_MAX, the table size may overflow an int when being
2751 * passed to bsetprop() below.
2752 */
2753 if (tp->LocalityCount >= SLIT_LOCALITIES_MAX)
2754 return;
2755
2756 bsetprop64(SLIT_NUM_PROPNAME, tp->LocalityCount);
2757 bsetprop(DDI_PROP_TYPE_BYTE,
2758 SLIT_PROPNAME, strlen(SLIT_PROPNAME), &tp->Entry,
2759 tp->LocalityCount * tp->LocalityCount);
2760 }
2761
2762 static ACPI_TABLE_MSCT *
2763 process_msct(ACPI_TABLE_MSCT *tp)
2764 {
2765 int last_seen = 0;
2766 int proc_num = 0;
2767 ACPI_MSCT_PROXIMITY *item, *end;
2768 extern uint64_t plat_dr_options;
2769
2770 ASSERT(tp != NULL);
2771
2772 end = (ACPI_MSCT_PROXIMITY *)(tp->Header.Length + (uintptr_t)tp);
2773 for (item = (void *)((uintptr_t)tp + tp->ProximityOffset);
2774 item < end;
2775 item = (void *)(item->Length + (uintptr_t)item)) {
2776 /*
2777 * Sanity check according to section 5.2.19.1 of ACPI 4.0.
2778 * Revision 1
2779 * Length 22
2780 */
2781 if (item->Revision != 1 || item->Length != 22) {
2782 cmn_err(CE_CONT,
2783 "?boot: unknown proximity domain structure in MSCT "
2784 "with Revision(%d), Length(%d).\n",
2785 (int)item->Revision, (int)item->Length);
2786 return (NULL);
2787 } else if (item->RangeStart > item->RangeEnd) {
2788 cmn_err(CE_CONT,
2789 "?boot: invalid proximity domain structure in MSCT "
2790 "with RangeStart(%u), RangeEnd(%u).\n",
2791 item->RangeStart, item->RangeEnd);
2792 return (NULL);
2793 } else if (item->RangeStart != last_seen) {
2794 /*
2795 * Items must be organized in ascending order of the
2796 * proximity domain enumerations.
2797 */
2798 cmn_err(CE_CONT,
2799 "?boot: invalid proximity domain structure in MSCT,"
2800 " items are not orginized in ascending order.\n");
2801 return (NULL);
2802 }
2803
2804 /*
2805 * If ProcessorCapacity is 0 then there would be no CPUs in this
2806 * domain.
2807 */
2808 if (item->ProcessorCapacity != 0) {
2809 proc_num += (item->RangeEnd - item->RangeStart + 1) *
2810 item->ProcessorCapacity;
2811 }
2812
2813 last_seen = item->RangeEnd - item->RangeStart + 1;
2814 /*
2815 * Break out if all proximity domains have been processed.
2816 * Some BIOSes may have unused items at the end of MSCT table.
2817 */
2818 if (last_seen > tp->MaxProximityDomains) {
2819 break;
2820 }
2821 }
2822 if (last_seen != tp->MaxProximityDomains + 1) {
2823 cmn_err(CE_CONT,
2824 "?boot: invalid proximity domain structure in MSCT, "
2825 "proximity domain count doesn't match.\n");
2826 return (NULL);
2827 }
2828
2829 /*
2830 * Set plat-max-ncpus property if it hasn't been set yet.
2831 */
2832 if (do_bsys_getproplen(NULL, PLAT_MAX_NCPUS_NAME) < 0) {
2833 if (proc_num != 0) {
2834 bsetpropsi(PLAT_MAX_NCPUS_NAME, proc_num);
2835 }
2836 }
2837
2838 /*
2839 * Use Maximum Physical Address from the MSCT table as upper limit for
2840 * memory hot-adding by default. It may be overridden by value from
2841 * the SRAT table or the "plat-dr-physmax" boot option.
2842 */
2843 plat_dr_physmax = btop(tp->MaxAddress + 1);
2844
2845 /*
2846 * Existence of MSCT implies CPU/memory hotplug-capability for the
2847 * platform.
2848 */
2849 plat_dr_options |= PLAT_DR_FEATURE_CPU;
2850 plat_dr_options |= PLAT_DR_FEATURE_MEMORY;
2851
2852 return (tp);
2853 }
2854
2855 #else /* __xpv */
2856 static void
2857 enumerate_xen_cpus()
2858 {
2859 processorid_t id, max_id;
2860
2861 /*
2862 * User-set boot-ncpus overrides enumeration
2863 */
2864 if (do_bsys_getproplen(NULL, BOOT_NCPUS_NAME) >= 0)
2865 return;
2866
2867 /*
2868 * Probe every possible virtual CPU id and remember the
2869 * highest id present; the count of CPUs is one greater
2870 * than this. This tacitly assumes at least cpu 0 is present.
2871 */
2872 max_id = 0;
2873 for (id = 0; id < MAX_VIRT_CPUS; id++)
2874 if (HYPERVISOR_vcpu_op(VCPUOP_is_up, id, NULL) == 0)
2875 max_id = id;
2876
2877 bsetpropsi(BOOT_NCPUS_NAME, max_id+1);
2878
2879 }
2880 #endif /* __xpv */
2881
2882 /*ARGSUSED*/
2883 static void
2884 build_firmware_properties(struct xboot_info *xbp)
2885 {
2886 ACPI_TABLE_HEADER *tp = NULL;
2887 ACPI_TABLE_RSDP *rsdp;
2888
2889 #ifndef __xpv
2890 if (xbp->bi_uefi_arch == XBI_UEFI_ARCH_64) {
2891 bsetprops("efi-systype", "64");
2892 bsetprop64("efi-systab",
2893 (uint64_t)(uintptr_t)xbp->bi_uefi_systab);
2894 if (kbm_debug)
2895 bop_printf(NULL, "64-bit UEFI detected.\n");
2896 } else if (xbp->bi_uefi_arch == XBI_UEFI_ARCH_32) {
2897 bsetprops("efi-systype", "32");
2898 bsetprop64("efi-systab",
2899 (uint64_t)(uintptr_t)xbp->bi_uefi_systab);
2900 if (kbm_debug)
2901 bop_printf(NULL, "32-bit UEFI detected.\n");
2902 }
2903
2904 if (xbp->bi_smbios != NULL) {
2905 bsetprop64("smbios-address",
2906 (uint64_t)(uintptr_t)xbp->bi_smbios);
2907 }
2908
2909 rsdp = find_rsdp(xbp);
2910
2911 if ((tp = find_fw_table(rsdp, ACPI_SIG_MSCT)) != NULL)
2912 msct_ptr = process_msct((ACPI_TABLE_MSCT *)tp);
2913 else
2914 msct_ptr = NULL;
2915
2916 if ((tp = find_fw_table(rsdp, ACPI_SIG_MADT)) != NULL)
2917 process_madt((ACPI_TABLE_MADT *)tp);
2918
2919 if ((srat_ptr = (ACPI_TABLE_SRAT *)
2920 find_fw_table(rsdp, ACPI_SIG_SRAT)) != NULL)
2921 process_srat(srat_ptr);
2922
2923 if (slit_ptr = (ACPI_TABLE_SLIT *)find_fw_table(rsdp, ACPI_SIG_SLIT))
2924 process_slit(slit_ptr);
2925
2926 tp = find_fw_table(rsdp, ACPI_SIG_MCFG);
2927 #else /* __xpv */
2928 enumerate_xen_cpus();
2929 if (DOMAIN_IS_INITDOMAIN(xen_info))
2930 tp = find_fw_table(rsdp, ACPI_SIG_MCFG);
2931 #endif /* __xpv */
2932 if (tp != NULL)
2933 process_mcfg((ACPI_TABLE_MCFG *)tp);
2934 }
2935
2936 /*
2937 * fake up a boot property for deferred early console output
2938 * this is used by both graphical boot and the (developer only)
2939 * USB serial console
2940 */
2941 void *
2942 defcons_init(size_t size)
2943 {
2944 static char *p = NULL;
2945
2946 p = do_bsys_alloc(NULL, NULL, size, MMU_PAGESIZE);
2947 *p = 0;
2948 bsetprop32("deferred-console-buf", (uint32_t)((uintptr_t)&p));
2949 return (p);
2950 }
2951
2952 /*ARGSUSED*/
2953 int
2954 boot_compinfo(int fd, struct compinfo *cbp)
2955 {
2956 cbp->iscmp = 0;
2957 cbp->blksize = MAXBSIZE;
2958 return (0);
2959 }
2960
2961 /*
2962 * Get an integer value for given boot property
2963 */
2964 int
2965 bootprop_getval(const char *prop_name, u_longlong_t *prop_value)
2966 {
2967 int boot_prop_len;
2968 char str[BP_MAX_STRLEN];
2969 u_longlong_t value;
2970
2971 boot_prop_len = BOP_GETPROPLEN(bootops, prop_name);
2972 if (boot_prop_len < 0 || boot_prop_len >= sizeof (str) ||
2973 BOP_GETPROP(bootops, prop_name, str) < 0 ||
2974 kobj_getvalue(str, &value) == -1)
2975 return (-1);
2976
2977 if (prop_value)
2978 *prop_value = value;
2979
2980 return (0);
2981 }
2982
2983 int
2984 bootprop_getstr(const char *prop_name, char *buf, size_t buflen)
2985 {
2986 int boot_prop_len = BOP_GETPROPLEN(bootops, prop_name);
2987
2988 if (boot_prop_len < 0 || boot_prop_len >= buflen ||
2989 BOP_GETPROP(bootops, prop_name, buf) < 0)
2990 return (-1);
2991
2992 return (0);
2993 }