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 2013 Joyent, Inc. All rights reserved.
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/bootvfs.h>
44 #include <sys/bootprops.h>
45 #include <sys/varargs.h>
46 #include <sys/param.h>
47 #include <sys/machparam.h>
48 #include <sys/machsystm.h>
49 #include <sys/archsystm.h>
50 #include <sys/boot_console.h>
51 #include <sys/cmn_err.h>
52 #include <sys/systm.h>
53 #include <sys/promif.h>
54 #include <sys/archsystm.h>
55 #include <sys/x86_archext.h>
56 #include <sys/kobj.h>
57 #include <sys/privregs.h>
58 #include <sys/sysmacros.h>
59 #include <sys/ctype.h>
60 #include <sys/fastboot.h>
61 #ifdef __xpv
62 #include <sys/hypervisor.h>
63 #include <net/if.h>
64 #endif
65 #include <vm/kboot_mmu.h>
66 #include <vm/hat_pte.h>
67 #include <sys/kobj.h>
68 #include <sys/kobj_lex.h>
69 #include <sys/pci_cfgspace_impl.h>
70 #include <sys/fastboot_impl.h>
71 #include <sys/acpi/acconfig.h>
72 #include <sys/acpi/acpi.h>
73
74 static int have_console = 0; /* set once primitive console is initialized */
75 static char *boot_args = "";
76
77 /*
78 * Debugging macros
79 */
80 static uint_t kbm_debug = 0;
81 #define DBG_MSG(s) { if (kbm_debug) bop_printf(NULL, "%s", s); }
82 #define DBG(x) { if (kbm_debug) \
83 bop_printf(NULL, "%s is %" PRIx64 "\n", #x, (uint64_t)(x)); \
84 }
85
86 #define PUT_STRING(s) { \
87 char *cp; \
88 for (cp = (s); *cp; ++cp) \
89 bcons_putchar(*cp); \
90 }
91
92 bootops_t bootop; /* simple bootops we'll pass on to kernel */
93 struct bsys_mem bm;
94
95 /*
96 * Boot info from "glue" code in low memory. xbootp is used by:
97 * do_bop_phys_alloc(), do_bsys_alloc() and boot_prop_finish().
98 */
99 static struct xboot_info *xbootp;
100 static uintptr_t next_virt; /* next available virtual address */
101 static paddr_t next_phys; /* next available physical address from dboot */
102 static paddr_t high_phys = -(paddr_t)1; /* last used physical address */
103
104 /*
105 * buffer for vsnprintf for console I/O
106 */
107 #define BUFFERSIZE 512
108 static char buffer[BUFFERSIZE];
109
110 /*
111 * stuff to store/report/manipulate boot property settings.
112 */
113 typedef struct bootprop {
114 struct bootprop *bp_next;
115 char *bp_name;
116 uint_t bp_vlen;
117 char *bp_value;
118 } bootprop_t;
119
120 static bootprop_t *bprops = NULL;
121 static char *curr_page = NULL; /* ptr to avail bprop memory */
122 static int curr_space = 0; /* amount of memory at curr_page */
123
124 #ifdef __xpv
125 start_info_t *xen_info;
126 shared_info_t *HYPERVISOR_shared_info;
127 #endif
128
129 /*
130 * some allocator statistics
131 */
132 static ulong_t total_bop_alloc_scratch = 0;
133 static ulong_t total_bop_alloc_kernel = 0;
134
135 static void build_firmware_properties(void);
136
137 static int early_allocation = 1;
138
139 int force_fastreboot = 0;
140 volatile int fastreboot_onpanic = 0;
141 int post_fastreboot = 0;
142 #ifdef __xpv
143 volatile int fastreboot_capable = 0;
144 #else
145 volatile int fastreboot_capable = 1;
146 #endif
147
148 /*
149 * Information saved from current boot for fast reboot.
150 * If the information size exceeds what we have allocated, fast reboot
151 * will not be supported.
152 */
153 multiboot_info_t saved_mbi;
154 mb_memory_map_t saved_mmap[FASTBOOT_SAVED_MMAP_COUNT];
155 uint8_t saved_drives[FASTBOOT_SAVED_DRIVES_SIZE];
156 char saved_cmdline[FASTBOOT_SAVED_CMDLINE_LEN];
157 int saved_cmdline_len = 0;
158 size_t saved_file_size[FASTBOOT_MAX_FILES_MAP];
159
160 /*
161 * Turn off fastreboot_onpanic to avoid panic loop.
162 */
163 char fastreboot_onpanic_cmdline[FASTBOOT_SAVED_CMDLINE_LEN];
164 static const char fastreboot_onpanic_args[] = " -B fastreboot_onpanic=0";
165
166 /*
167 * Pointers to where System Resource Affinity Table (SRAT), System Locality
168 * Information Table (SLIT) and Maximum System Capability Table (MSCT)
169 * are mapped into virtual memory
170 */
171 ACPI_TABLE_SRAT *srat_ptr = NULL;
172 ACPI_TABLE_SLIT *slit_ptr = NULL;
173 ACPI_TABLE_MSCT *msct_ptr = NULL;
174
175 /*
176 * Arbitrary limit on number of localities we handle; if
177 * this limit is raised to more than UINT16_MAX, make sure
178 * process_slit() knows how to handle it.
179 */
180 #define SLIT_LOCALITIES_MAX (4096)
181
182 #define SLIT_NUM_PROPNAME "acpi-slit-localities"
183 #define SLIT_PROPNAME "acpi-slit"
184
185 /*
186 * Allocate aligned physical memory at boot time. This allocator allocates
187 * from the highest possible addresses. This avoids exhausting memory that
188 * would be useful for DMA buffers.
189 */
190 paddr_t
191 do_bop_phys_alloc(uint64_t size, uint64_t align)
192 {
193 paddr_t pa = 0;
194 paddr_t start;
195 paddr_t end;
196 struct memlist *ml = (struct memlist *)xbootp->bi_phys_install;
197
198 /*
199 * Be careful if high memory usage is limited in startup.c
200 * Since there are holes in the low part of the physical address
201 * space we can treat physmem as a pfn (not just a pgcnt) and
202 * get a conservative upper limit.
203 */
204 if (physmem != 0 && high_phys > pfn_to_pa(physmem))
205 high_phys = pfn_to_pa(physmem);
206
207 /*
208 * find the lowest or highest available memory in physinstalled
209 * On 32 bit avoid physmem above 4Gig if PAE isn't enabled
210 */
211 #if defined(__i386)
212 if (xbootp->bi_use_pae == 0 && high_phys > FOUR_GIG)
213 high_phys = FOUR_GIG;
214 #endif
215
216 /*
217 * find the highest available memory in physinstalled
218 */
219 size = P2ROUNDUP(size, align);
220 for (; ml; ml = ml->ml_next) {
221 start = P2ROUNDUP(ml->ml_address, align);
222 end = P2ALIGN(ml->ml_address + ml->ml_size, align);
223 if (start < next_phys)
224 start = P2ROUNDUP(next_phys, align);
225 if (end > high_phys)
226 end = P2ALIGN(high_phys, align);
227
228 if (end <= start)
229 continue;
230 if (end - start < size)
231 continue;
232
233 /*
234 * Early allocations need to use low memory, since
235 * physmem might be further limited by bootenv.rc
236 */
237 if (early_allocation) {
238 if (pa == 0 || start < pa)
239 pa = start;
240 } else {
241 if (end - size > pa)
242 pa = end - size;
243 }
244 }
245 if (pa != 0) {
246 if (early_allocation)
247 next_phys = pa + size;
248 else
249 high_phys = pa;
250 return (pa);
251 }
252 bop_panic("do_bop_phys_alloc(0x%" PRIx64 ", 0x%" PRIx64
253 ") Out of memory\n", size, align);
254 /*NOTREACHED*/
255 }
256
257 uintptr_t
258 alloc_vaddr(size_t size, paddr_t align)
259 {
260 uintptr_t rv;
261
262 next_virt = P2ROUNDUP(next_virt, (uintptr_t)align);
263 rv = (uintptr_t)next_virt;
264 next_virt += size;
265 return (rv);
266 }
267
268 /*
269 * Allocate virtual memory. The size is always rounded up to a multiple
270 * of base pagesize.
271 */
272
273 /*ARGSUSED*/
274 static caddr_t
275 do_bsys_alloc(bootops_t *bop, caddr_t virthint, size_t size, int align)
276 {
277 paddr_t a = align; /* same type as pa for masking */
278 uint_t pgsize;
279 paddr_t pa;
280 uintptr_t va;
281 ssize_t s; /* the aligned size */
282 uint_t level;
283 uint_t is_kernel = (virthint != 0);
284
285 if (a < MMU_PAGESIZE)
286 a = MMU_PAGESIZE;
287 else if (!ISP2(a))
288 prom_panic("do_bsys_alloc() incorrect alignment");
289 size = P2ROUNDUP(size, MMU_PAGESIZE);
290
291 /*
292 * Use the next aligned virtual address if we weren't given one.
293 */
294 if (virthint == NULL) {
295 virthint = (caddr_t)alloc_vaddr(size, a);
296 total_bop_alloc_scratch += size;
297 } else {
298 total_bop_alloc_kernel += size;
299 }
300
301 /*
302 * allocate the physical memory
303 */
304 pa = do_bop_phys_alloc(size, a);
305
306 /*
307 * Add the mappings to the page tables, try large pages first.
308 */
309 va = (uintptr_t)virthint;
310 s = size;
311 level = 1;
312 pgsize = xbootp->bi_use_pae ? TWO_MEG : FOUR_MEG;
313 if (xbootp->bi_use_largepage && a == pgsize) {
314 while (IS_P2ALIGNED(pa, pgsize) && IS_P2ALIGNED(va, pgsize) &&
315 s >= pgsize) {
316 kbm_map(va, pa, level, is_kernel);
317 va += pgsize;
318 pa += pgsize;
319 s -= pgsize;
320 }
321 }
322
323 /*
324 * Map remaining pages use small mappings
325 */
326 level = 0;
327 pgsize = MMU_PAGESIZE;
328 while (s > 0) {
329 kbm_map(va, pa, level, is_kernel);
330 va += pgsize;
331 pa += pgsize;
332 s -= pgsize;
333 }
334 return (virthint);
335 }
336
337 /*
338 * Free virtual memory - we'll just ignore these.
339 */
340 /*ARGSUSED*/
341 static void
342 do_bsys_free(bootops_t *bop, caddr_t virt, size_t size)
343 {
344 bop_printf(NULL, "do_bsys_free(virt=0x%p, size=0x%lx) ignored\n",
345 (void *)virt, size);
346 }
347
348 /*
349 * Old interface
350 */
351 /*ARGSUSED*/
352 static caddr_t
353 do_bsys_ealloc(bootops_t *bop, caddr_t virthint, size_t size,
354 int align, int flags)
355 {
356 prom_panic("unsupported call to BOP_EALLOC()\n");
357 return (0);
358 }
359
360
361 static void
362 bsetprop(char *name, int nlen, void *value, int vlen)
363 {
364 uint_t size;
365 uint_t need_size;
366 bootprop_t *b;
367
368 /*
369 * align the size to 16 byte boundary
370 */
371 size = sizeof (bootprop_t) + nlen + 1 + vlen;
372 size = (size + 0xf) & ~0xf;
373 if (size > curr_space) {
374 need_size = (size + (MMU_PAGEOFFSET)) & MMU_PAGEMASK;
375 curr_page = do_bsys_alloc(NULL, 0, need_size, MMU_PAGESIZE);
376 curr_space = need_size;
377 }
378
379 /*
380 * use a bootprop_t at curr_page and link into list
381 */
382 b = (bootprop_t *)curr_page;
383 curr_page += sizeof (bootprop_t);
384 curr_space -= sizeof (bootprop_t);
385 b->bp_next = bprops;
386 bprops = b;
387
388 /*
389 * follow by name and ending zero byte
390 */
391 b->bp_name = curr_page;
392 bcopy(name, curr_page, nlen);
393 curr_page += nlen;
394 *curr_page++ = 0;
395 curr_space -= nlen + 1;
396
397 /*
398 * copy in value, but no ending zero byte
399 */
400 b->bp_value = curr_page;
401 b->bp_vlen = vlen;
402 if (vlen > 0) {
403 bcopy(value, curr_page, vlen);
404 curr_page += vlen;
405 curr_space -= vlen;
406 }
407
408 /*
409 * align new values of curr_page, curr_space
410 */
411 while (curr_space & 0xf) {
412 ++curr_page;
413 --curr_space;
414 }
415 }
416
417 static void
418 bsetprops(char *name, char *value)
419 {
420 bsetprop(name, strlen(name), value, strlen(value) + 1);
421 }
422
423 static void
424 bsetprop64(char *name, uint64_t value)
425 {
426 bsetprop(name, strlen(name), (void *)&value, sizeof (value));
427 }
428
429 static void
430 bsetpropsi(char *name, int value)
431 {
432 char prop_val[32];
433
434 (void) snprintf(prop_val, sizeof (prop_val), "%d", value);
435 bsetprops(name, prop_val);
436 }
437
438 /*
439 * to find the size of the buffer to allocate
440 */
441 /*ARGSUSED*/
442 int
443 do_bsys_getproplen(bootops_t *bop, const char *name)
444 {
445 bootprop_t *b;
446
447 for (b = bprops; b; b = b->bp_next) {
448 if (strcmp(name, b->bp_name) != 0)
449 continue;
450 return (b->bp_vlen);
451 }
452 return (-1);
453 }
454
455 /*
456 * get the value associated with this name
457 */
458 /*ARGSUSED*/
459 int
460 do_bsys_getprop(bootops_t *bop, const char *name, void *value)
461 {
462 bootprop_t *b;
463
464 for (b = bprops; b; b = b->bp_next) {
465 if (strcmp(name, b->bp_name) != 0)
466 continue;
467 bcopy(b->bp_value, value, b->bp_vlen);
468 return (0);
469 }
470 return (-1);
471 }
472
473 /*
474 * get the name of the next property in succession from the standalone
475 */
476 /*ARGSUSED*/
477 static char *
478 do_bsys_nextprop(bootops_t *bop, char *name)
479 {
480 bootprop_t *b;
481
482 /*
483 * A null name is a special signal for the 1st boot property
484 */
485 if (name == NULL || strlen(name) == 0) {
486 if (bprops == NULL)
487 return (NULL);
488 return (bprops->bp_name);
489 }
490
491 for (b = bprops; b; b = b->bp_next) {
492 if (name != b->bp_name)
493 continue;
494 b = b->bp_next;
495 if (b == NULL)
496 return (NULL);
497 return (b->bp_name);
498 }
499 return (NULL);
500 }
501
502 /*
503 * Parse numeric value from a string. Understands decimal, hex, octal, - and ~
504 */
505 static int
506 parse_value(char *p, uint64_t *retval)
507 {
508 int adjust = 0;
509 uint64_t tmp = 0;
510 int digit;
511 int radix = 10;
512
513 *retval = 0;
514 if (*p == '-' || *p == '~')
515 adjust = *p++;
516
517 if (*p == '0') {
518 ++p;
519 if (*p == 0)
520 return (0);
521 if (*p == 'x' || *p == 'X') {
522 radix = 16;
523 ++p;
524 } else {
525 radix = 8;
526 ++p;
527 }
528 }
529 while (*p) {
530 if ('0' <= *p && *p <= '9')
531 digit = *p - '0';
532 else if ('a' <= *p && *p <= 'f')
533 digit = 10 + *p - 'a';
534 else if ('A' <= *p && *p <= 'F')
535 digit = 10 + *p - 'A';
536 else
537 return (-1);
538 if (digit >= radix)
539 return (-1);
540 tmp = tmp * radix + digit;
541 ++p;
542 }
543 if (adjust == '-')
544 tmp = -tmp;
545 else if (adjust == '~')
546 tmp = ~tmp;
547 *retval = tmp;
548 return (0);
549 }
550
551 static boolean_t
552 unprintable(char *value, int size)
553 {
554 int i;
555
556 if (size <= 0 || value[0] == '\0')
557 return (B_TRUE);
558
559 for (i = 0; i < size; i++) {
560 if (value[i] == '\0')
561 return (i != (size - 1));
562
563 if (!isprint(value[i]))
564 return (B_TRUE);
565 }
566 return (B_FALSE);
567 }
568
569 /*
570 * Print out information about all boot properties.
571 * buffer is pointer to pre-allocated space to be used as temporary
572 * space for property values.
573 */
574 static void
575 boot_prop_display(char *buffer)
576 {
577 char *name = "";
578 int i, len;
579
580 bop_printf(NULL, "\nBoot properties:\n");
581
582 while ((name = do_bsys_nextprop(NULL, name)) != NULL) {
583 bop_printf(NULL, "\t0x%p %s = ", (void *)name, name);
584 (void) do_bsys_getprop(NULL, name, buffer);
585 len = do_bsys_getproplen(NULL, name);
586 bop_printf(NULL, "len=%d ", len);
587 if (!unprintable(buffer, len)) {
588 buffer[len] = 0;
589 bop_printf(NULL, "%s\n", buffer);
590 continue;
591 }
592 for (i = 0; i < len; i++) {
593 bop_printf(NULL, "%02x", buffer[i] & 0xff);
594 if (i < len - 1)
595 bop_printf(NULL, ".");
596 }
597 bop_printf(NULL, "\n");
598 }
599 }
600
601 /*
602 * 2nd part of building the table of boot properties. This includes:
603 * - values from /boot/solaris/bootenv.rc (ie. eeprom(1m) values)
604 *
605 * lines look like one of:
606 * ^$
607 * ^# comment till end of line
608 * setprop name 'value'
609 * setprop name value
610 * setprop name "value"
611 *
612 * we do single character I/O since this is really just looking at memory
613 */
614 void
615 boot_prop_finish(void)
616 {
617 int fd;
618 char *line;
619 int c;
620 int bytes_read;
621 char *name;
622 int n_len;
623 char *value;
624 int v_len;
625 char *inputdev; /* these override the command line if serial ports */
626 char *outputdev;
627 char *consoledev;
628 uint64_t lvalue;
629 int use_xencons = 0;
630
631 #ifdef __xpv
632 if (!DOMAIN_IS_INITDOMAIN(xen_info))
633 use_xencons = 1;
634 #endif /* __xpv */
635
636 DBG_MSG("Opening /boot/solaris/bootenv.rc\n");
637 fd = BRD_OPEN(bfs_ops, "/boot/solaris/bootenv.rc", 0);
638 DBG(fd);
639
640 line = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
641 while (fd >= 0) {
642
643 /*
644 * get a line
645 */
646 for (c = 0; ; ++c) {
647 bytes_read = BRD_READ(bfs_ops, fd, line + c, 1);
648 if (bytes_read == 0) {
649 if (c == 0)
650 goto done;
651 break;
652 }
653 if (line[c] == '\n')
654 break;
655 }
656 line[c] = 0;
657
658 /*
659 * ignore comment lines
660 */
661 c = 0;
662 while (ISSPACE(line[c]))
663 ++c;
664 if (line[c] == '#' || line[c] == 0)
665 continue;
666
667 /*
668 * must have "setprop " or "setprop\t"
669 */
670 if (strncmp(line + c, "setprop ", 8) != 0 &&
671 strncmp(line + c, "setprop\t", 8) != 0)
672 continue;
673 c += 8;
674 while (ISSPACE(line[c]))
675 ++c;
676 if (line[c] == 0)
677 continue;
678
679 /*
680 * gather up the property name
681 */
682 name = line + c;
683 n_len = 0;
684 while (line[c] && !ISSPACE(line[c]))
685 ++n_len, ++c;
686
687 /*
688 * gather up the value, if any
689 */
690 value = "";
691 v_len = 0;
692 while (ISSPACE(line[c]))
693 ++c;
694 if (line[c] != 0) {
695 value = line + c;
696 while (line[c] && !ISSPACE(line[c]))
697 ++v_len, ++c;
698 }
699
700 if (v_len >= 2 && value[0] == value[v_len - 1] &&
701 (value[0] == '\'' || value[0] == '"')) {
702 ++value;
703 v_len -= 2;
704 }
705 name[n_len] = 0;
706 if (v_len > 0)
707 value[v_len] = 0;
708 else
709 continue;
710
711 /*
712 * ignore "boot-file" property, it's now meaningless
713 */
714 if (strcmp(name, "boot-file") == 0)
715 continue;
716 if (strcmp(name, "boot-args") == 0 &&
717 strlen(boot_args) > 0)
718 continue;
719
720 /*
721 * If a property was explicitly set on the command line
722 * it will override a setting in bootenv.rc
723 */
724 if (do_bsys_getproplen(NULL, name) > 0)
725 continue;
726
727 bsetprop(name, n_len, value, v_len + 1);
728 }
729 done:
730 if (fd >= 0)
731 (void) BRD_CLOSE(bfs_ops, fd);
732
733 /*
734 * Check if we have to limit the boot time allocator
735 */
736 if (do_bsys_getproplen(NULL, "physmem") != -1 &&
737 do_bsys_getprop(NULL, "physmem", line) >= 0 &&
738 parse_value(line, &lvalue) != -1) {
739 if (0 < lvalue && (lvalue < physmem || physmem == 0)) {
740 physmem = (pgcnt_t)lvalue;
741 DBG(physmem);
742 }
743 }
744 early_allocation = 0;
745
746 /*
747 * check to see if we have to override the default value of the console
748 */
749 if (!use_xencons) {
750 inputdev = line;
751 v_len = do_bsys_getproplen(NULL, "input-device");
752 if (v_len > 0)
753 (void) do_bsys_getprop(NULL, "input-device", inputdev);
754 else
755 v_len = 0;
756 inputdev[v_len] = 0;
757
758 outputdev = inputdev + v_len + 1;
759 v_len = do_bsys_getproplen(NULL, "output-device");
760 if (v_len > 0)
761 (void) do_bsys_getprop(NULL, "output-device",
762 outputdev);
763 else
764 v_len = 0;
765 outputdev[v_len] = 0;
766
767 consoledev = outputdev + v_len + 1;
768 v_len = do_bsys_getproplen(NULL, "console");
769 if (v_len > 0) {
770 (void) do_bsys_getprop(NULL, "console", consoledev);
771 if (post_fastreboot &&
772 strcmp(consoledev, "graphics") == 0) {
773 bsetprops("console", "text");
774 v_len = strlen("text");
775 bcopy("text", consoledev, v_len);
776 }
777 } else {
778 v_len = 0;
779 }
780 consoledev[v_len] = 0;
781 bcons_init2(inputdev, outputdev, consoledev);
782 } else {
783 /*
784 * Ensure console property exists
785 * If not create it as "hypervisor"
786 */
787 v_len = do_bsys_getproplen(NULL, "console");
788 if (v_len < 0)
789 bsetprops("console", "hypervisor");
790 inputdev = outputdev = consoledev = "hypervisor";
791 bcons_init2(inputdev, outputdev, consoledev);
792 }
793
794 if (strstr((char *)xbootp->bi_cmdline, "prom_debug") || kbm_debug)
795 boot_prop_display(line);
796 }
797
798 /*
799 * print formatted output
800 */
801 /*PRINTFLIKE2*/
802 /*ARGSUSED*/
803 void
804 bop_printf(bootops_t *bop, const char *fmt, ...)
805 {
806 va_list ap;
807
808 if (have_console == 0)
809 return;
810
811 va_start(ap, fmt);
812 (void) vsnprintf(buffer, BUFFERSIZE, fmt, ap);
813 va_end(ap);
814 PUT_STRING(buffer);
815 }
816
817 /*
818 * Another panic() variant; this one can be used even earlier during boot than
819 * prom_panic().
820 */
821 /*PRINTFLIKE1*/
822 void
823 bop_panic(const char *fmt, ...)
824 {
825 va_list ap;
826
827 va_start(ap, fmt);
828 bop_printf(NULL, fmt, ap);
829 va_end(ap);
830
831 bop_printf(NULL, "\nPress any key to reboot.\n");
832 (void) bcons_getchar();
833 bop_printf(NULL, "Resetting...\n");
834 pc_reset();
835 }
836
837 /*
838 * Do a real mode interrupt BIOS call
839 */
840 typedef struct bios_regs {
841 unsigned short ax, bx, cx, dx, si, di, bp, es, ds;
842 } bios_regs_t;
843 typedef int (*bios_func_t)(int, bios_regs_t *);
844
845 /*ARGSUSED*/
846 static void
847 do_bsys_doint(bootops_t *bop, int intnum, struct bop_regs *rp)
848 {
849 #if defined(__xpv)
850 prom_panic("unsupported call to BOP_DOINT()\n");
851 #else /* __xpv */
852 static int firsttime = 1;
853 bios_func_t bios_func = (bios_func_t)(void *)(uintptr_t)0x5000;
854 bios_regs_t br;
855
856 /*
857 * The first time we do this, we have to copy the pre-packaged
858 * low memory bios call code image into place.
859 */
860 if (firsttime) {
861 extern char bios_image[];
862 extern uint32_t bios_size;
863
864 bcopy(bios_image, (void *)bios_func, bios_size);
865 firsttime = 0;
866 }
867
868 br.ax = rp->eax.word.ax;
869 br.bx = rp->ebx.word.bx;
870 br.cx = rp->ecx.word.cx;
871 br.dx = rp->edx.word.dx;
872 br.bp = rp->ebp.word.bp;
873 br.si = rp->esi.word.si;
874 br.di = rp->edi.word.di;
875 br.ds = rp->ds;
876 br.es = rp->es;
877
878 DBG_MSG("Doing BIOS call...");
879 DBG(br.ax);
880 DBG(br.bx);
881 DBG(br.dx);
882 rp->eflags = bios_func(intnum, &br);
883 DBG_MSG("done\n");
884
885 rp->eax.word.ax = br.ax;
886 rp->ebx.word.bx = br.bx;
887 rp->ecx.word.cx = br.cx;
888 rp->edx.word.dx = br.dx;
889 rp->ebp.word.bp = br.bp;
890 rp->esi.word.si = br.si;
891 rp->edi.word.di = br.di;
892 rp->ds = br.ds;
893 rp->es = br.es;
894 #endif /* __xpv */
895 }
896
897 static struct boot_syscalls bop_sysp = {
898 bcons_getchar,
899 bcons_putchar,
900 bcons_ischar,
901 };
902
903 static char *whoami;
904
905 #define BUFLEN 64
906
907 #if defined(__xpv)
908
909 static char namebuf[32];
910
911 static void
912 xen_parse_props(char *s, char *prop_map[], int n_prop)
913 {
914 char **prop_name = prop_map;
915 char *cp = s, *scp;
916
917 do {
918 scp = cp;
919 while ((*cp != NULL) && (*cp != ':'))
920 cp++;
921
922 if ((scp != cp) && (*prop_name != NULL)) {
923 *cp = NULL;
924 bsetprops(*prop_name, scp);
925 }
926
927 cp++;
928 prop_name++;
929 n_prop--;
930 } while (n_prop > 0);
931 }
932
933 #define VBDPATHLEN 64
934
935 /*
936 * parse the 'xpv-root' property to create properties used by
937 * ufs_mountroot.
938 */
939 static void
940 xen_vbdroot_props(char *s)
941 {
942 char vbdpath[VBDPATHLEN] = "/xpvd/xdf@";
943 const char lnamefix[] = "/dev/dsk/c0d";
944 char *pnp;
945 char *prop_p;
946 char mi;
947 short minor;
948 long addr = 0;
949
950 pnp = vbdpath + strlen(vbdpath);
951 prop_p = s + strlen(lnamefix);
952 while ((*prop_p != '\0') && (*prop_p != 's') && (*prop_p != 'p'))
953 addr = addr * 10 + *prop_p++ - '0';
954 (void) snprintf(pnp, VBDPATHLEN, "%lx", addr);
955 pnp = vbdpath + strlen(vbdpath);
956 if (*prop_p == 's')
957 mi = 'a';
958 else if (*prop_p == 'p')
959 mi = 'q';
960 else
961 ASSERT(0); /* shouldn't be here */
962 prop_p++;
963 ASSERT(*prop_p != '\0');
964 if (ISDIGIT(*prop_p)) {
965 minor = *prop_p - '0';
966 prop_p++;
967 if (ISDIGIT(*prop_p)) {
968 minor = minor * 10 + *prop_p - '0';
969 }
970 } else {
971 /* malformed root path, use 0 as default */
972 minor = 0;
973 }
974 ASSERT(minor < 16); /* at most 16 partitions */
975 mi += minor;
976 *pnp++ = ':';
977 *pnp++ = mi;
978 *pnp++ = '\0';
979 bsetprops("fstype", "ufs");
980 bsetprops("bootpath", vbdpath);
981
982 DBG_MSG("VBD bootpath set to ");
983 DBG_MSG(vbdpath);
984 DBG_MSG("\n");
985 }
986
987 /*
988 * parse the xpv-nfsroot property to create properties used by
989 * nfs_mountroot.
990 */
991 static void
992 xen_nfsroot_props(char *s)
993 {
994 char *prop_map[] = {
995 BP_SERVER_IP, /* server IP address */
996 BP_SERVER_NAME, /* server hostname */
997 BP_SERVER_PATH, /* root path */
998 };
999 int n_prop = sizeof (prop_map) / sizeof (prop_map[0]);
1000
1001 bsetprop("fstype", 6, "nfs", 4);
1002
1003 xen_parse_props(s, prop_map, n_prop);
1004
1005 /*
1006 * If a server name wasn't specified, use a default.
1007 */
1008 if (do_bsys_getproplen(NULL, BP_SERVER_NAME) == -1)
1009 bsetprops(BP_SERVER_NAME, "unknown");
1010 }
1011
1012 /*
1013 * Extract our IP address, etc. from the "xpv-ip" property.
1014 */
1015 static void
1016 xen_ip_props(char *s)
1017 {
1018 char *prop_map[] = {
1019 BP_HOST_IP, /* IP address */
1020 NULL, /* NFS server IP address (ignored in */
1021 /* favour of xpv-nfsroot) */
1022 BP_ROUTER_IP, /* IP gateway */
1023 BP_SUBNET_MASK, /* IP subnet mask */
1024 "xpv-hostname", /* hostname (ignored) */
1025 BP_NETWORK_INTERFACE, /* interface name */
1026 "xpv-hcp", /* host configuration protocol */
1027 };
1028 int n_prop = sizeof (prop_map) / sizeof (prop_map[0]);
1029 char ifname[IFNAMSIZ];
1030
1031 xen_parse_props(s, prop_map, n_prop);
1032
1033 /*
1034 * A Linux dom0 administrator expects all interfaces to be
1035 * called "ethX", which is not the case here.
1036 *
1037 * If the interface name specified is "eth0", presume that
1038 * this is really intended to be "xnf0" (the first domU ->
1039 * dom0 interface for this domain).
1040 */
1041 if ((do_bsys_getprop(NULL, BP_NETWORK_INTERFACE, ifname) == 0) &&
1042 (strcmp("eth0", ifname) == 0)) {
1043 bsetprops(BP_NETWORK_INTERFACE, "xnf0");
1044 bop_printf(NULL,
1045 "network interface name 'eth0' replaced with 'xnf0'\n");
1046 }
1047 }
1048
1049 #else /* __xpv */
1050
1051 static void
1052 setup_rarp_props(struct sol_netinfo *sip)
1053 {
1054 char buf[BUFLEN]; /* to hold ip/mac addrs */
1055 uint8_t *val;
1056
1057 val = (uint8_t *)&sip->sn_ciaddr;
1058 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1059 val[0], val[1], val[2], val[3]);
1060 bsetprops(BP_HOST_IP, buf);
1061
1062 val = (uint8_t *)&sip->sn_siaddr;
1063 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1064 val[0], val[1], val[2], val[3]);
1065 bsetprops(BP_SERVER_IP, buf);
1066
1067 if (sip->sn_giaddr != 0) {
1068 val = (uint8_t *)&sip->sn_giaddr;
1069 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1070 val[0], val[1], val[2], val[3]);
1071 bsetprops(BP_ROUTER_IP, buf);
1072 }
1073
1074 if (sip->sn_netmask != 0) {
1075 val = (uint8_t *)&sip->sn_netmask;
1076 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1077 val[0], val[1], val[2], val[3]);
1078 bsetprops(BP_SUBNET_MASK, buf);
1079 }
1080
1081 if (sip->sn_mactype != 4 || sip->sn_maclen != 6) {
1082 bop_printf(NULL, "unsupported mac type %d, mac len %d\n",
1083 sip->sn_mactype, sip->sn_maclen);
1084 } else {
1085 val = sip->sn_macaddr;
1086 (void) snprintf(buf, BUFLEN, "%x:%x:%x:%x:%x:%x",
1087 val[0], val[1], val[2], val[3], val[4], val[5]);
1088 bsetprops(BP_BOOT_MAC, buf);
1089 }
1090 }
1091
1092 #endif /* __xpv */
1093
1094 static void
1095 build_panic_cmdline(const char *cmd, int cmdlen)
1096 {
1097 int proplen;
1098 size_t arglen;
1099
1100 arglen = sizeof (fastreboot_onpanic_args);
1101 /*
1102 * If we allready have fastreboot-onpanic set to zero,
1103 * don't add them again.
1104 */
1105 if ((proplen = do_bsys_getproplen(NULL, FASTREBOOT_ONPANIC)) > 0 &&
1106 proplen <= sizeof (fastreboot_onpanic_cmdline)) {
1107 (void) do_bsys_getprop(NULL, FASTREBOOT_ONPANIC,
1108 fastreboot_onpanic_cmdline);
1109 if (FASTREBOOT_ONPANIC_NOTSET(fastreboot_onpanic_cmdline))
1110 arglen = 1;
1111 }
1112
1113 /*
1114 * construct fastreboot_onpanic_cmdline
1115 */
1116 if (cmdlen + arglen > sizeof (fastreboot_onpanic_cmdline)) {
1117 DBG_MSG("Command line too long: clearing "
1118 FASTREBOOT_ONPANIC "\n");
1119 fastreboot_onpanic = 0;
1120 } else {
1121 bcopy(cmd, fastreboot_onpanic_cmdline, cmdlen);
1122 if (arglen != 1)
1123 bcopy(fastreboot_onpanic_args,
1124 fastreboot_onpanic_cmdline + cmdlen, arglen);
1125 else
1126 fastreboot_onpanic_cmdline[cmdlen] = 0;
1127 }
1128 }
1129
1130
1131 #ifndef __xpv
1132 /*
1133 * Construct boot command line for Fast Reboot
1134 */
1135 static void
1136 build_fastboot_cmdline(struct xboot_info *xbp)
1137 {
1138 saved_cmdline_len = strlen(xbp->bi_cmdline) + 1;
1139 if (saved_cmdline_len > FASTBOOT_SAVED_CMDLINE_LEN) {
1140 DBG(saved_cmdline_len);
1141 DBG_MSG("Command line too long: clearing fastreboot_capable\n");
1142 fastreboot_capable = 0;
1143 } else {
1144 bcopy((void *)(xbp->bi_cmdline), (void *)saved_cmdline,
1145 saved_cmdline_len);
1146 saved_cmdline[saved_cmdline_len - 1] = '\0';
1147 build_panic_cmdline(saved_cmdline, saved_cmdline_len - 1);
1148 }
1149 }
1150
1151 /*
1152 * Save memory layout, disk drive information, unix and boot archive sizes for
1153 * Fast Reboot.
1154 */
1155 static void
1156 save_boot_info(struct xboot_info *xbi)
1157 {
1158 multiboot_info_t *mbi = xbi->bi_mb_info;
1159 struct boot_modules *modp;
1160 int i;
1161
1162 bcopy(mbi, &saved_mbi, sizeof (multiboot_info_t));
1163 if (mbi->mmap_length > sizeof (saved_mmap)) {
1164 DBG_MSG("mbi->mmap_length too big: clearing "
1165 "fastreboot_capable\n");
1166 fastreboot_capable = 0;
1167 } else {
1168 bcopy((void *)(uintptr_t)mbi->mmap_addr, (void *)saved_mmap,
1169 mbi->mmap_length);
1170 }
1171
1172 if ((mbi->flags & MB_INFO_DRIVE_INFO) != 0) {
1173 if (mbi->drives_length > sizeof (saved_drives)) {
1174 DBG(mbi->drives_length);
1175 DBG_MSG("mbi->drives_length too big: clearing "
1176 "fastreboot_capable\n");
1177 fastreboot_capable = 0;
1178 } else {
1179 bcopy((void *)(uintptr_t)mbi->drives_addr,
1180 (void *)saved_drives, mbi->drives_length);
1181 }
1182 } else {
1183 saved_mbi.drives_length = 0;
1184 saved_mbi.drives_addr = NULL;
1185 }
1186
1187 /*
1188 * Current file sizes. Used by fastboot.c to figure out how much
1189 * memory to reserve for panic reboot.
1190 * Use the module list from the dboot-constructed xboot_info
1191 * instead of the list referenced by the multiboot structure
1192 * because that structure may not be addressable now.
1193 */
1194 saved_file_size[FASTBOOT_NAME_UNIX] = FOUR_MEG - PAGESIZE;
1195 for (i = 0, modp = (struct boot_modules *)(uintptr_t)xbi->bi_modules;
1196 i < xbi->bi_module_cnt; i++, modp++) {
1197 saved_file_size[FASTBOOT_NAME_BOOTARCHIVE] += modp->bm_size;
1198 }
1199 }
1200 #endif /* __xpv */
1201
1202
1203 /*
1204 * 1st pass at building the table of boot properties. This includes:
1205 * - values set on the command line: -B a=x,b=y,c=z ....
1206 * - known values we just compute (ie. from xbp)
1207 * - values from /boot/solaris/bootenv.rc (ie. eeprom(1m) values)
1208 *
1209 * the grub command line looked like:
1210 * kernel boot-file [-B prop=value[,prop=value]...] [boot-args]
1211 *
1212 * whoami is the same as boot-file
1213 */
1214 static void
1215 build_boot_properties(struct xboot_info *xbp)
1216 {
1217 char *name;
1218 int name_len;
1219 char *value;
1220 int value_len;
1221 struct boot_modules *bm, *rdbm;
1222 char *propbuf;
1223 int quoted = 0;
1224 int boot_arg_len;
1225 uint_t i, midx;
1226 char modid[32];
1227 #ifndef __xpv
1228 static int stdout_val = 0;
1229 uchar_t boot_device;
1230 char str[3];
1231 multiboot_info_t *mbi;
1232 int netboot;
1233 struct sol_netinfo *sip;
1234 #endif
1235
1236 /*
1237 * These have to be done first, so that kobj_mount_root() works
1238 */
1239 DBG_MSG("Building boot properties\n");
1240 propbuf = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, 0);
1241 DBG((uintptr_t)propbuf);
1242 if (xbp->bi_module_cnt > 0) {
1243 bm = xbp->bi_modules;
1244 rdbm = NULL;
1245 for (midx = i = 0; i < xbp->bi_module_cnt; i++) {
1246 if (bm[i].bm_type == BMT_ROOTFS) {
1247 rdbm = &bm[i];
1248 continue;
1249 }
1250 if (bm[i].bm_type == BMT_HASH || bm[i].bm_name == NULL)
1251 continue;
1252
1253 (void) snprintf(modid, sizeof (modid),
1254 "module-name-%u", midx);
1255 bsetprops(modid, (char *)bm[i].bm_name);
1256 (void) snprintf(modid, sizeof (modid),
1257 "module-addr-%u", midx);
1258 bsetprop64(modid, (uint64_t)(uintptr_t)bm[i].bm_addr);
1259 (void) snprintf(modid, sizeof (modid),
1260 "module-size-%u", midx);
1261 bsetprop64(modid, (uint64_t)bm[i].bm_size);
1262 ++midx;
1263 }
1264 if (rdbm != NULL) {
1265 bsetprop64("ramdisk_start",
1266 (uint64_t)(uintptr_t)rdbm->bm_addr);
1267 bsetprop64("ramdisk_end",
1268 (uint64_t)(uintptr_t)rdbm->bm_addr + rdbm->bm_size);
1269 }
1270 }
1271
1272 /*
1273 * If there are any boot time modules or hashes present, then disable
1274 * fast reboot.
1275 */
1276 if (xbp->bi_module_cnt > 1) {
1277 fastreboot_disable(FBNS_BOOTMOD);
1278 }
1279
1280 DBG_MSG("Parsing command line for boot properties\n");
1281 value = xbp->bi_cmdline;
1282
1283 /*
1284 * allocate memory to collect boot_args into
1285 */
1286 boot_arg_len = strlen(xbp->bi_cmdline) + 1;
1287 boot_args = do_bsys_alloc(NULL, NULL, boot_arg_len, MMU_PAGESIZE);
1288 boot_args[0] = 0;
1289 boot_arg_len = 0;
1290
1291 #ifdef __xpv
1292 /*
1293 * Xen puts a lot of device information in front of the kernel name
1294 * let's grab them and make them boot properties. The first
1295 * string w/o an "=" in it will be the boot-file property.
1296 */
1297 (void) strcpy(namebuf, "xpv-");
1298 for (;;) {
1299 /*
1300 * get to next property
1301 */
1302 while (ISSPACE(*value))
1303 ++value;
1304 name = value;
1305 /*
1306 * look for an "="
1307 */
1308 while (*value && !ISSPACE(*value) && *value != '=') {
1309 value++;
1310 }
1311 if (*value != '=') { /* no "=" in the property */
1312 value = name;
1313 break;
1314 }
1315 name_len = value - name;
1316 value_len = 0;
1317 /*
1318 * skip over the "="
1319 */
1320 value++;
1321 while (value[value_len] && !ISSPACE(value[value_len])) {
1322 ++value_len;
1323 }
1324 /*
1325 * build property name with "xpv-" prefix
1326 */
1327 if (name_len + 4 > 32) { /* skip if name too long */
1328 value += value_len;
1329 continue;
1330 }
1331 bcopy(name, &namebuf[4], name_len);
1332 name_len += 4;
1333 namebuf[name_len] = 0;
1334 bcopy(value, propbuf, value_len);
1335 propbuf[value_len] = 0;
1336 bsetprops(namebuf, propbuf);
1337
1338 /*
1339 * xpv-root is set to the logical disk name of the xen
1340 * VBD when booting from a disk-based filesystem.
1341 */
1342 if (strcmp(namebuf, "xpv-root") == 0)
1343 xen_vbdroot_props(propbuf);
1344 /*
1345 * While we're here, if we have a "xpv-nfsroot" property
1346 * then we need to set "fstype" to "nfs" so we mount
1347 * our root from the nfs server. Also parse the xpv-nfsroot
1348 * property to create the properties that nfs_mountroot will
1349 * need to find the root and mount it.
1350 */
1351 if (strcmp(namebuf, "xpv-nfsroot") == 0)
1352 xen_nfsroot_props(propbuf);
1353
1354 if (strcmp(namebuf, "xpv-ip") == 0)
1355 xen_ip_props(propbuf);
1356 value += value_len;
1357 }
1358 #endif
1359
1360 while (ISSPACE(*value))
1361 ++value;
1362 /*
1363 * value now points at the boot-file
1364 */
1365 value_len = 0;
1366 while (value[value_len] && !ISSPACE(value[value_len]))
1367 ++value_len;
1368 if (value_len > 0) {
1369 whoami = propbuf;
1370 bcopy(value, whoami, value_len);
1371 whoami[value_len] = 0;
1372 bsetprops("boot-file", whoami);
1373 /*
1374 * strip leading path stuff from whoami, so running from
1375 * PXE/miniroot makes sense.
1376 */
1377 if (strstr(whoami, "/platform/") != NULL)
1378 whoami = strstr(whoami, "/platform/");
1379 bsetprops("whoami", whoami);
1380 }
1381
1382 /*
1383 * Values forcibly set boot properties on the command line via -B.
1384 * Allow use of quotes in values. Other stuff goes on kernel
1385 * command line.
1386 */
1387 name = value + value_len;
1388 while (*name != 0) {
1389 /*
1390 * anything not " -B" is copied to the command line
1391 */
1392 if (!ISSPACE(name[0]) || name[1] != '-' || name[2] != 'B') {
1393 boot_args[boot_arg_len++] = *name;
1394 boot_args[boot_arg_len] = 0;
1395 ++name;
1396 continue;
1397 }
1398
1399 /*
1400 * skip the " -B" and following white space
1401 */
1402 name += 3;
1403 while (ISSPACE(*name))
1404 ++name;
1405 while (*name && !ISSPACE(*name)) {
1406 value = strstr(name, "=");
1407 if (value == NULL)
1408 break;
1409 name_len = value - name;
1410 ++value;
1411 value_len = 0;
1412 quoted = 0;
1413 for (; ; ++value_len) {
1414 if (!value[value_len])
1415 break;
1416
1417 /*
1418 * is this value quoted?
1419 */
1420 if (value_len == 0 &&
1421 (value[0] == '\'' || value[0] == '"')) {
1422 quoted = value[0];
1423 ++value_len;
1424 }
1425
1426 /*
1427 * In the quote accept any character,
1428 * but look for ending quote.
1429 */
1430 if (quoted) {
1431 if (value[value_len] == quoted)
1432 quoted = 0;
1433 continue;
1434 }
1435
1436 /*
1437 * a comma or white space ends the value
1438 */
1439 if (value[value_len] == ',' ||
1440 ISSPACE(value[value_len]))
1441 break;
1442 }
1443
1444 if (value_len == 0) {
1445 bsetprop(name, name_len, "true", 5);
1446 } else {
1447 char *v = value;
1448 int l = value_len;
1449 if (v[0] == v[l - 1] &&
1450 (v[0] == '\'' || v[0] == '"')) {
1451 ++v;
1452 l -= 2;
1453 }
1454 bcopy(v, propbuf, l);
1455 propbuf[l] = '\0';
1456 bsetprop(name, name_len, propbuf,
1457 l + 1);
1458 }
1459 name = value + value_len;
1460 while (*name == ',')
1461 ++name;
1462 }
1463 }
1464
1465 /*
1466 * set boot-args property
1467 * 1275 name is bootargs, so set
1468 * that too
1469 */
1470 bsetprops("boot-args", boot_args);
1471 bsetprops("bootargs", boot_args);
1472
1473 #ifndef __xpv
1474 /*
1475 * set the BIOS boot device from GRUB
1476 */
1477 netboot = 0;
1478 mbi = xbp->bi_mb_info;
1479
1480 /*
1481 * Build boot command line for Fast Reboot
1482 */
1483 build_fastboot_cmdline(xbp);
1484
1485 /*
1486 * Save various boot information for Fast Reboot
1487 */
1488 save_boot_info(xbp);
1489
1490 if (mbi != NULL && mbi->flags & MB_INFO_BOOTDEV) {
1491 boot_device = mbi->boot_device >> 24;
1492 if (boot_device == 0x20)
1493 netboot++;
1494 str[0] = (boot_device >> 4) + '0';
1495 str[1] = (boot_device & 0xf) + '0';
1496 str[2] = 0;
1497 bsetprops("bios-boot-device", str);
1498 } else {
1499 netboot = 1;
1500 }
1501
1502 /*
1503 * In the netboot case, drives_info is overloaded with the dhcp ack.
1504 * This is not multiboot compliant and requires special pxegrub!
1505 */
1506 if (netboot && mbi->drives_length != 0) {
1507 sip = (struct sol_netinfo *)(uintptr_t)mbi->drives_addr;
1508 if (sip->sn_infotype == SN_TYPE_BOOTP)
1509 bsetprop("bootp-response", sizeof ("bootp-response"),
1510 (void *)(uintptr_t)mbi->drives_addr,
1511 mbi->drives_length);
1512 else if (sip->sn_infotype == SN_TYPE_RARP)
1513 setup_rarp_props(sip);
1514 }
1515 bsetprop("stdout", strlen("stdout"),
1516 &stdout_val, sizeof (stdout_val));
1517 #endif /* __xpv */
1518
1519 /*
1520 * more conjured up values for made up things....
1521 */
1522 #if defined(__xpv)
1523 bsetprops("mfg-name", "i86xpv");
1524 bsetprops("impl-arch-name", "i86xpv");
1525 #else
1526 bsetprops("mfg-name", "i86pc");
1527 bsetprops("impl-arch-name", "i86pc");
1528 #endif
1529
1530 /*
1531 * Build firmware-provided system properties
1532 */
1533 build_firmware_properties();
1534
1535 /*
1536 * XXPV
1537 *
1538 * Find out what these are:
1539 * - cpuid_feature_ecx_include
1540 * - cpuid_feature_ecx_exclude
1541 * - cpuid_feature_edx_include
1542 * - cpuid_feature_edx_exclude
1543 *
1544 * Find out what these are in multiboot:
1545 * - netdev-path
1546 * - fstype
1547 */
1548 }
1549
1550 #ifdef __xpv
1551 /*
1552 * Under the Hypervisor, memory usable for DMA may be scarce. One
1553 * very likely large pool of DMA friendly memory is occupied by
1554 * the boot_archive, as it was loaded by grub into low MFNs.
1555 *
1556 * Here we free up that memory by copying the boot archive to what are
1557 * likely higher MFN pages and then swapping the mfn/pfn mappings.
1558 */
1559 #define PFN_2GIG 0x80000
1560 static void
1561 relocate_boot_archive(struct xboot_info *xbp)
1562 {
1563 mfn_t max_mfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
1564 struct boot_modules *bm = xbp->bi_modules;
1565 uintptr_t va;
1566 pfn_t va_pfn;
1567 mfn_t va_mfn;
1568 caddr_t copy;
1569 pfn_t copy_pfn;
1570 mfn_t copy_mfn;
1571 size_t len;
1572 int slop;
1573 int total = 0;
1574 int relocated = 0;
1575 int mmu_update_return;
1576 mmu_update_t t[2];
1577 x86pte_t pte;
1578
1579 /*
1580 * If all MFN's are below 2Gig, don't bother doing this.
1581 */
1582 if (max_mfn < PFN_2GIG)
1583 return;
1584 if (xbp->bi_module_cnt < 1) {
1585 DBG_MSG("no boot_archive!");
1586 return;
1587 }
1588
1589 DBG_MSG("moving boot_archive to high MFN memory\n");
1590 va = (uintptr_t)bm->bm_addr;
1591 len = bm->bm_size;
1592 slop = va & MMU_PAGEOFFSET;
1593 if (slop) {
1594 va += MMU_PAGESIZE - slop;
1595 len -= MMU_PAGESIZE - slop;
1596 }
1597 len = P2ALIGN(len, MMU_PAGESIZE);
1598
1599 /*
1600 * Go through all boot_archive pages, swapping any low MFN pages
1601 * with memory at next_phys.
1602 */
1603 while (len != 0) {
1604 ++total;
1605 va_pfn = mmu_btop(va - ONE_GIG);
1606 va_mfn = mfn_list[va_pfn];
1607 if (mfn_list[va_pfn] < PFN_2GIG) {
1608 copy = kbm_remap_window(next_phys, 1);
1609 bcopy((void *)va, copy, MMU_PAGESIZE);
1610 copy_pfn = mmu_btop(next_phys);
1611 copy_mfn = mfn_list[copy_pfn];
1612
1613 pte = mfn_to_ma(copy_mfn) | PT_NOCONSIST | PT_VALID;
1614 if (HYPERVISOR_update_va_mapping(va, pte,
1615 UVMF_INVLPG | UVMF_LOCAL))
1616 bop_panic("relocate_boot_archive(): "
1617 "HYPERVISOR_update_va_mapping() failed");
1618
1619 mfn_list[va_pfn] = copy_mfn;
1620 mfn_list[copy_pfn] = va_mfn;
1621
1622 t[0].ptr = mfn_to_ma(copy_mfn) | MMU_MACHPHYS_UPDATE;
1623 t[0].val = va_pfn;
1624 t[1].ptr = mfn_to_ma(va_mfn) | MMU_MACHPHYS_UPDATE;
1625 t[1].val = copy_pfn;
1626 if (HYPERVISOR_mmu_update(t, 2, &mmu_update_return,
1627 DOMID_SELF) != 0 || mmu_update_return != 2)
1628 bop_panic("relocate_boot_archive(): "
1629 "HYPERVISOR_mmu_update() failed");
1630
1631 next_phys += MMU_PAGESIZE;
1632 ++relocated;
1633 }
1634 len -= MMU_PAGESIZE;
1635 va += MMU_PAGESIZE;
1636 }
1637 DBG_MSG("Relocated pages:\n");
1638 DBG(relocated);
1639 DBG_MSG("Out of total pages:\n");
1640 DBG(total);
1641 }
1642 #endif /* __xpv */
1643
1644 #if !defined(__xpv)
1645 /*
1646 * Install a temporary IDT that lets us catch errors in the boot time code.
1647 * We shouldn't get any faults at all while this is installed, so we'll
1648 * just generate a traceback and exit.
1649 */
1650 #ifdef __amd64
1651 static const int bcode_sel = B64CODE_SEL;
1652 #else
1653 static const int bcode_sel = B32CODE_SEL;
1654 #endif
1655
1656 /*
1657 * simple description of a stack frame (args are 32 bit only currently)
1658 */
1659 typedef struct bop_frame {
1660 struct bop_frame *old_frame;
1661 pc_t retaddr;
1662 long arg[1];
1663 } bop_frame_t;
1664
1665 void
1666 bop_traceback(bop_frame_t *frame)
1667 {
1668 pc_t pc;
1669 int cnt;
1670 char *ksym;
1671 ulong_t off;
1672 #if defined(__i386)
1673 int a;
1674 #endif
1675
1676 bop_printf(NULL, "Stack traceback:\n");
1677 for (cnt = 0; cnt < 30; ++cnt) { /* up to 30 frames */
1678 pc = frame->retaddr;
1679 if (pc == 0)
1680 break;
1681 ksym = kobj_getsymname(pc, &off);
1682 if (ksym)
1683 bop_printf(NULL, " %s+%lx", ksym, off);
1684 else
1685 bop_printf(NULL, " 0x%lx", pc);
1686
1687 frame = frame->old_frame;
1688 if (frame == 0) {
1689 bop_printf(NULL, "\n");
1690 break;
1691 }
1692 #if defined(__i386)
1693 for (a = 0; a < 6; ++a) { /* try for 6 args */
1694 if ((void *)&frame->arg[a] == (void *)frame->old_frame)
1695 break;
1696 if (a == 0)
1697 bop_printf(NULL, "(");
1698 else
1699 bop_printf(NULL, ",");
1700 bop_printf(NULL, "0x%lx", frame->arg[a]);
1701 }
1702 bop_printf(NULL, ")");
1703 #endif
1704 bop_printf(NULL, "\n");
1705 }
1706 }
1707
1708 struct trapframe {
1709 ulong_t error_code; /* optional */
1710 ulong_t inst_ptr;
1711 ulong_t code_seg;
1712 ulong_t flags_reg;
1713 #ifdef __amd64
1714 ulong_t stk_ptr;
1715 ulong_t stk_seg;
1716 #endif
1717 };
1718
1719 void
1720 bop_trap(ulong_t *tfp)
1721 {
1722 struct trapframe *tf = (struct trapframe *)tfp;
1723 bop_frame_t fakeframe;
1724 static int depth = 0;
1725
1726 /*
1727 * Check for an infinite loop of traps.
1728 */
1729 if (++depth > 2)
1730 bop_panic("Nested trap");
1731
1732 bop_printf(NULL, "Unexpected trap\n");
1733
1734 /*
1735 * adjust the tf for optional error_code by detecting the code selector
1736 */
1737 if (tf->code_seg != bcode_sel)
1738 tf = (struct trapframe *)(tfp - 1);
1739 else
1740 bop_printf(NULL, "error code 0x%lx\n",
1741 tf->error_code & 0xffffffff);
1742
1743 bop_printf(NULL, "instruction pointer 0x%lx\n", tf->inst_ptr);
1744 bop_printf(NULL, "code segment 0x%lx\n", tf->code_seg & 0xffff);
1745 bop_printf(NULL, "flags register 0x%lx\n", tf->flags_reg);
1746 #ifdef __amd64
1747 bop_printf(NULL, "return %%rsp 0x%lx\n", tf->stk_ptr);
1748 bop_printf(NULL, "return %%ss 0x%lx\n", tf->stk_seg & 0xffff);
1749 #endif
1750
1751 /* grab %[er]bp pushed by our code from the stack */
1752 fakeframe.old_frame = (bop_frame_t *)*(tfp - 3);
1753 fakeframe.retaddr = (pc_t)tf->inst_ptr;
1754 bop_printf(NULL, "Attempting stack backtrace:\n");
1755 bop_traceback(&fakeframe);
1756 bop_panic("unexpected trap in early boot");
1757 }
1758
1759 extern void bop_trap_handler(void);
1760
1761 static gate_desc_t *bop_idt;
1762
1763 static desctbr_t bop_idt_info;
1764
1765 static void
1766 bop_idt_init(void)
1767 {
1768 int t;
1769
1770 bop_idt = (gate_desc_t *)
1771 do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
1772 bzero(bop_idt, MMU_PAGESIZE);
1773 for (t = 0; t < NIDT; ++t) {
1774 /*
1775 * Note that since boot runs without a TSS, the
1776 * double fault handler cannot use an alternate stack
1777 * (64-bit) or a task gate (32-bit).
1778 */
1779 set_gatesegd(&bop_idt[t], &bop_trap_handler, bcode_sel,
1780 SDT_SYSIGT, TRP_KPL, 0);
1781 }
1782 bop_idt_info.dtr_limit = (NIDT * sizeof (gate_desc_t)) - 1;
1783 bop_idt_info.dtr_base = (uintptr_t)bop_idt;
1784 wr_idtr(&bop_idt_info);
1785 }
1786 #endif /* !defined(__xpv) */
1787
1788 /*
1789 * This is where we enter the kernel. It dummies up the boot_ops and
1790 * boot_syscalls vectors and jumps off to _kobj_boot()
1791 */
1792 void
1793 _start(struct xboot_info *xbp)
1794 {
1795 bootops_t *bops = &bootop;
1796 extern void _kobj_boot();
1797
1798 /*
1799 * 1st off - initialize the console for any error messages
1800 */
1801 xbootp = xbp;
1802 #ifdef __xpv
1803 HYPERVISOR_shared_info = (void *)xbp->bi_shared_info;
1804 xen_info = xbp->bi_xen_start_info;
1805 #endif
1806
1807 #ifndef __xpv
1808 if (*((uint32_t *)(FASTBOOT_SWTCH_PA + FASTBOOT_STACK_OFFSET)) ==
1809 FASTBOOT_MAGIC) {
1810 post_fastreboot = 1;
1811 *((uint32_t *)(FASTBOOT_SWTCH_PA + FASTBOOT_STACK_OFFSET)) = 0;
1812 }
1813 #endif
1814
1815 bcons_init((void *)xbp->bi_cmdline);
1816 have_console = 1;
1817
1818 /*
1819 * enable debugging
1820 */
1821 if (strstr((char *)xbp->bi_cmdline, "kbm_debug"))
1822 kbm_debug = 1;
1823
1824 DBG_MSG("\n\n*** Entered Solaris in _start() cmdline is: ");
1825 DBG_MSG((char *)xbp->bi_cmdline);
1826 DBG_MSG("\n\n\n");
1827
1828 /*
1829 * physavail is no longer used by startup
1830 */
1831 bm.physinstalled = xbp->bi_phys_install;
1832 bm.pcimem = xbp->bi_pcimem;
1833 bm.rsvdmem = xbp->bi_rsvdmem;
1834 bm.physavail = NULL;
1835
1836 /*
1837 * initialize the boot time allocator
1838 */
1839 next_phys = xbp->bi_next_paddr;
1840 DBG(next_phys);
1841 next_virt = (uintptr_t)xbp->bi_next_vaddr;
1842 DBG(next_virt);
1843 DBG_MSG("Initializing boot time memory management...");
1844 #ifdef __xpv
1845 {
1846 xen_platform_parameters_t p;
1847
1848 /* This call shouldn't fail, dboot already did it once. */
1849 (void) HYPERVISOR_xen_version(XENVER_platform_parameters, &p);
1850 mfn_to_pfn_mapping = (pfn_t *)(xen_virt_start = p.virt_start);
1851 DBG(xen_virt_start);
1852 }
1853 #endif
1854 kbm_init(xbp);
1855 DBG_MSG("done\n");
1856
1857 /*
1858 * Fill in the bootops vector
1859 */
1860 bops->bsys_version = BO_VERSION;
1861 bops->boot_mem = &bm;
1862 bops->bsys_alloc = do_bsys_alloc;
1863 bops->bsys_free = do_bsys_free;
1864 bops->bsys_getproplen = do_bsys_getproplen;
1865 bops->bsys_getprop = do_bsys_getprop;
1866 bops->bsys_nextprop = do_bsys_nextprop;
1867 bops->bsys_printf = bop_printf;
1868 bops->bsys_doint = do_bsys_doint;
1869
1870 /*
1871 * BOP_EALLOC() is no longer needed
1872 */
1873 bops->bsys_ealloc = do_bsys_ealloc;
1874
1875 #ifdef __xpv
1876 /*
1877 * On domain 0 we need to free up some physical memory that is
1878 * usable for DMA. Since GRUB loaded the boot_archive, it is
1879 * sitting in low MFN memory. We'll relocated the boot archive
1880 * pages to high PFN memory.
1881 */
1882 if (DOMAIN_IS_INITDOMAIN(xen_info))
1883 relocate_boot_archive(xbp);
1884 #endif
1885
1886 #ifndef __xpv
1887 /*
1888 * Install an IDT to catch early pagefaults (shouldn't have any).
1889 * Also needed for kmdb.
1890 */
1891 bop_idt_init();
1892 #endif
1893
1894 /*
1895 * Start building the boot properties from the command line
1896 */
1897 DBG_MSG("Initializing boot properties:\n");
1898 build_boot_properties(xbp);
1899
1900 if (strstr((char *)xbp->bi_cmdline, "prom_debug") || kbm_debug) {
1901 char *value;
1902
1903 value = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
1904 boot_prop_display(value);
1905 }
1906
1907 /*
1908 * jump into krtld...
1909 */
1910 _kobj_boot(&bop_sysp, NULL, bops, NULL);
1911 }
1912
1913
1914 /*ARGSUSED*/
1915 static caddr_t
1916 no_more_alloc(bootops_t *bop, caddr_t virthint, size_t size, int align)
1917 {
1918 panic("Attempt to bsys_alloc() too late\n");
1919 return (NULL);
1920 }
1921
1922 /*ARGSUSED*/
1923 static void
1924 no_more_free(bootops_t *bop, caddr_t virt, size_t size)
1925 {
1926 panic("Attempt to bsys_free() too late\n");
1927 }
1928
1929 void
1930 bop_no_more_mem(void)
1931 {
1932 DBG(total_bop_alloc_scratch);
1933 DBG(total_bop_alloc_kernel);
1934 bootops->bsys_alloc = no_more_alloc;
1935 bootops->bsys_free = no_more_free;
1936 }
1937
1938
1939 /*
1940 * Set ACPI firmware properties
1941 */
1942
1943 static caddr_t
1944 vmap_phys(size_t length, paddr_t pa)
1945 {
1946 paddr_t start, end;
1947 caddr_t va;
1948 size_t len, page;
1949
1950 #ifdef __xpv
1951 pa = pfn_to_pa(xen_assign_pfn(mmu_btop(pa))) | (pa & MMU_PAGEOFFSET);
1952 #endif
1953 start = P2ALIGN(pa, MMU_PAGESIZE);
1954 end = P2ROUNDUP(pa + length, MMU_PAGESIZE);
1955 len = end - start;
1956 va = (caddr_t)alloc_vaddr(len, MMU_PAGESIZE);
1957 for (page = 0; page < len; page += MMU_PAGESIZE)
1958 kbm_map((uintptr_t)va + page, start + page, 0, 0);
1959 return (va + (pa & MMU_PAGEOFFSET));
1960 }
1961
1962 static uint8_t
1963 checksum_table(uint8_t *tp, size_t len)
1964 {
1965 uint8_t sum = 0;
1966
1967 while (len-- > 0)
1968 sum += *tp++;
1969
1970 return (sum);
1971 }
1972
1973 static int
1974 valid_rsdp(ACPI_TABLE_RSDP *rp)
1975 {
1976
1977 /* validate the V1.x checksum */
1978 if (checksum_table((uint8_t *)rp, ACPI_RSDP_CHECKSUM_LENGTH) != 0)
1979 return (0);
1980
1981 /* If pre-ACPI 2.0, this is a valid RSDP */
1982 if (rp->Revision < 2)
1983 return (1);
1984
1985 /* validate the V2.x checksum */
1986 if (checksum_table((uint8_t *)rp, ACPI_RSDP_XCHECKSUM_LENGTH) != 0)
1987 return (0);
1988
1989 return (1);
1990 }
1991
1992 /*
1993 * Scan memory range for an RSDP;
1994 * see ACPI 3.0 Spec, 5.2.5.1
1995 */
1996 static ACPI_TABLE_RSDP *
1997 scan_rsdp(paddr_t start, paddr_t end)
1998 {
1999 ssize_t len = end - start;
2000 caddr_t ptr;
2001
2002 ptr = vmap_phys(len, start);
2003 while (len > 0) {
2004 if (strncmp(ptr, ACPI_SIG_RSDP, strlen(ACPI_SIG_RSDP)) == 0 &&
2005 valid_rsdp((ACPI_TABLE_RSDP *)ptr))
2006 return ((ACPI_TABLE_RSDP *)ptr);
2007
2008 ptr += ACPI_RSDP_SCAN_STEP;
2009 len -= ACPI_RSDP_SCAN_STEP;
2010 }
2011
2012 return (NULL);
2013 }
2014
2015 /*
2016 * Refer to ACPI 3.0 Spec, section 5.2.5.1 to understand this function
2017 */
2018 static ACPI_TABLE_RSDP *
2019 find_rsdp()
2020 {
2021 ACPI_TABLE_RSDP *rsdp;
2022 uint16_t *ebda_seg;
2023 paddr_t ebda_addr;
2024
2025 /*
2026 * Get the EBDA segment and scan the first 1K
2027 */
2028 ebda_seg = (uint16_t *)vmap_phys(sizeof (uint16_t),
2029 ACPI_EBDA_PTR_LOCATION);
2030 ebda_addr = *ebda_seg << 4;
2031 rsdp = scan_rsdp(ebda_addr, ebda_addr + ACPI_EBDA_WINDOW_SIZE);
2032 if (rsdp == NULL)
2033 /* if EBDA doesn't contain RSDP, look in BIOS memory */
2034 rsdp = scan_rsdp(ACPI_HI_RSDP_WINDOW_BASE,
2035 ACPI_HI_RSDP_WINDOW_BASE + ACPI_HI_RSDP_WINDOW_SIZE);
2036 return (rsdp);
2037 }
2038
2039 static ACPI_TABLE_HEADER *
2040 map_fw_table(paddr_t table_addr)
2041 {
2042 ACPI_TABLE_HEADER *tp;
2043 size_t len = MAX(sizeof (*tp), MMU_PAGESIZE);
2044
2045 /*
2046 * Map at least a page; if the table is larger than this, remap it
2047 */
2048 tp = (ACPI_TABLE_HEADER *)vmap_phys(len, table_addr);
2049 if (tp->Length > len)
2050 tp = (ACPI_TABLE_HEADER *)vmap_phys(tp->Length, table_addr);
2051 return (tp);
2052 }
2053
2054 static ACPI_TABLE_HEADER *
2055 find_fw_table(char *signature)
2056 {
2057 static int revision = 0;
2058 static ACPI_TABLE_XSDT *xsdt;
2059 static int len;
2060 paddr_t xsdt_addr;
2061 ACPI_TABLE_RSDP *rsdp;
2062 ACPI_TABLE_HEADER *tp;
2063 paddr_t table_addr;
2064 int n;
2065
2066 if (strlen(signature) != ACPI_NAME_SIZE)
2067 return (NULL);
2068
2069 /*
2070 * Reading the ACPI 3.0 Spec, section 5.2.5.3 will help
2071 * understand this code. If we haven't already found the RSDT/XSDT,
2072 * revision will be 0. Find the RSDP and check the revision
2073 * to find out whether to use the RSDT or XSDT. If revision is
2074 * 0 or 1, use the RSDT and set internal revision to 1; if it is 2,
2075 * use the XSDT. If the XSDT address is 0, though, fall back to
2076 * revision 1 and use the RSDT.
2077 */
2078 if (revision == 0) {
2079 if ((rsdp = find_rsdp()) != NULL) {
2080 revision = rsdp->Revision;
2081 /*
2082 * ACPI 6.0 states that current revision is 2
2083 * from acpi_table_rsdp definition:
2084 * Must be (0) for ACPI 1.0 or (2) for ACPI 2.0+
2085 */
2086 if (revision > 2)
2087 revision = 2;
2088 switch (revision) {
2089 case 2:
2090 /*
2091 * Use the XSDT unless BIOS is buggy and
2092 * claims to be rev 2 but has a null XSDT
2093 * address
2094 */
2095 xsdt_addr = rsdp->XsdtPhysicalAddress;
2096 if (xsdt_addr != 0)
2097 break;
2098 /* FALLTHROUGH */
2099 case 0:
2100 /* treat RSDP rev 0 as revision 1 internally */
2101 revision = 1;
2102 /* FALLTHROUGH */
2103 case 1:
2104 /* use the RSDT for rev 0/1 */
2105 xsdt_addr = rsdp->RsdtPhysicalAddress;
2106 break;
2107 default:
2108 /* unknown revision */
2109 revision = 0;
2110 break;
2111 }
2112 }
2113 if (revision == 0)
2114 return (NULL);
2115
2116 /* cache the XSDT info */
2117 xsdt = (ACPI_TABLE_XSDT *)map_fw_table(xsdt_addr);
2118 len = (xsdt->Header.Length - sizeof (xsdt->Header)) /
2119 ((revision == 1) ? sizeof (uint32_t) : sizeof (uint64_t));
2120 }
2121
2122 /*
2123 * Scan the table headers looking for a signature match
2124 */
2125 for (n = 0; n < len; n++) {
2126 ACPI_TABLE_RSDT *rsdt = (ACPI_TABLE_RSDT *)xsdt;
2127 table_addr = (revision == 1) ? rsdt->TableOffsetEntry[n] :
2128 xsdt->TableOffsetEntry[n];
2129
2130 if (table_addr == 0)
2131 continue;
2132 tp = map_fw_table(table_addr);
2133 if (strncmp(tp->Signature, signature, ACPI_NAME_SIZE) == 0) {
2134 return (tp);
2135 }
2136 }
2137 return (NULL);
2138 }
2139
2140 static void
2141 process_mcfg(ACPI_TABLE_MCFG *tp)
2142 {
2143 ACPI_MCFG_ALLOCATION *cfg_baap;
2144 char *cfg_baa_endp;
2145 int64_t ecfginfo[4];
2146
2147 cfg_baap = (ACPI_MCFG_ALLOCATION *)((uintptr_t)tp + sizeof (*tp));
2148 cfg_baa_endp = ((char *)tp) + tp->Header.Length;
2149 while ((char *)cfg_baap < cfg_baa_endp) {
2150 if (cfg_baap->Address != 0 && cfg_baap->PciSegment == 0) {
2151 ecfginfo[0] = cfg_baap->Address;
2152 ecfginfo[1] = cfg_baap->PciSegment;
2153 ecfginfo[2] = cfg_baap->StartBusNumber;
2154 ecfginfo[3] = cfg_baap->EndBusNumber;
2155 bsetprop(MCFG_PROPNAME, strlen(MCFG_PROPNAME),
2156 ecfginfo, sizeof (ecfginfo));
2157 break;
2158 }
2159 cfg_baap++;
2160 }
2161 }
2162
2163 #ifndef __xpv
2164 static void
2165 process_madt_entries(ACPI_TABLE_MADT *tp, uint32_t *cpu_countp,
2166 uint32_t *cpu_possible_countp, uint32_t *cpu_apicid_array)
2167 {
2168 ACPI_SUBTABLE_HEADER *item, *end;
2169 uint32_t cpu_count = 0;
2170 uint32_t cpu_possible_count = 0;
2171
2172 /*
2173 * Determine number of CPUs and keep track of "final" APIC ID
2174 * for each CPU by walking through ACPI MADT processor list
2175 */
2176 end = (ACPI_SUBTABLE_HEADER *)(tp->Header.Length + (uintptr_t)tp);
2177 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)tp + sizeof (*tp));
2178
2179 while (item < end) {
2180 switch (item->Type) {
2181 case ACPI_MADT_TYPE_LOCAL_APIC: {
2182 ACPI_MADT_LOCAL_APIC *cpu =
2183 (ACPI_MADT_LOCAL_APIC *) item;
2184
2185 if (cpu->LapicFlags & ACPI_MADT_ENABLED) {
2186 if (cpu_apicid_array != NULL)
2187 cpu_apicid_array[cpu_count] = cpu->Id;
2188 cpu_count++;
2189 }
2190 cpu_possible_count++;
2191 break;
2192 }
2193 case ACPI_MADT_TYPE_LOCAL_X2APIC: {
2194 ACPI_MADT_LOCAL_X2APIC *cpu =
2195 (ACPI_MADT_LOCAL_X2APIC *) item;
2196
2197 if (cpu->LapicFlags & ACPI_MADT_ENABLED) {
2198 if (cpu_apicid_array != NULL)
2199 cpu_apicid_array[cpu_count] =
2200 cpu->LocalApicId;
2201 cpu_count++;
2202 }
2203 cpu_possible_count++;
2204 break;
2205 }
2206 default:
2207 if (kbm_debug)
2208 bop_printf(NULL, "MADT type %d\n", item->Type);
2209 break;
2210 }
2211
2212 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)item + item->Length);
2213 }
2214 if (cpu_countp)
2215 *cpu_countp = cpu_count;
2216 if (cpu_possible_countp)
2217 *cpu_possible_countp = cpu_possible_count;
2218 }
2219
2220 static void
2221 process_madt(ACPI_TABLE_MADT *tp)
2222 {
2223 uint32_t cpu_count = 0;
2224 uint32_t cpu_possible_count = 0;
2225 uint32_t *cpu_apicid_array; /* x2APIC ID is 32bit! */
2226
2227 if (tp != NULL) {
2228 /* count cpu's */
2229 process_madt_entries(tp, &cpu_count, &cpu_possible_count, NULL);
2230
2231 cpu_apicid_array = (uint32_t *)do_bsys_alloc(NULL, NULL,
2232 cpu_count * sizeof (*cpu_apicid_array), MMU_PAGESIZE);
2233 if (cpu_apicid_array == NULL)
2234 bop_panic("Not enough memory for APIC ID array");
2235
2236 /* copy IDs */
2237 process_madt_entries(tp, NULL, NULL, cpu_apicid_array);
2238
2239 /*
2240 * Make boot property for array of "final" APIC IDs for each
2241 * CPU
2242 */
2243 bsetprop(BP_CPU_APICID_ARRAY, strlen(BP_CPU_APICID_ARRAY),
2244 cpu_apicid_array, cpu_count * sizeof (*cpu_apicid_array));
2245 }
2246
2247 /*
2248 * Check whether property plat-max-ncpus is already set.
2249 */
2250 if (do_bsys_getproplen(NULL, PLAT_MAX_NCPUS_NAME) < 0) {
2251 /*
2252 * Set plat-max-ncpus to number of maximum possible CPUs given
2253 * in MADT if it hasn't been set.
2254 * There's no formal way to detect max possible CPUs supported
2255 * by platform according to ACPI spec3.0b. So current CPU
2256 * hotplug implementation expects that all possible CPUs will
2257 * have an entry in MADT table and set plat-max-ncpus to number
2258 * of entries in MADT.
2259 * With introducing of ACPI4.0, Maximum System Capability Table
2260 * (MSCT) provides maximum number of CPUs supported by platform.
2261 * If MSCT is unavailable, fall back to old way.
2262 */
2263 if (tp != NULL)
2264 bsetpropsi(PLAT_MAX_NCPUS_NAME, cpu_possible_count);
2265 }
2266
2267 /*
2268 * Set boot property boot-max-ncpus to number of CPUs existing at
2269 * boot time. boot-max-ncpus is mainly used for optimization.
2270 */
2271 if (tp != NULL)
2272 bsetpropsi(BOOT_MAX_NCPUS_NAME, cpu_count);
2273
2274 /*
2275 * User-set boot-ncpus overrides firmware count
2276 */
2277 if (do_bsys_getproplen(NULL, BOOT_NCPUS_NAME) >= 0)
2278 return;
2279
2280 /*
2281 * Set boot property boot-ncpus to number of active CPUs given in MADT
2282 * if it hasn't been set yet.
2283 */
2284 if (tp != NULL)
2285 bsetpropsi(BOOT_NCPUS_NAME, cpu_count);
2286 }
2287
2288 static void
2289 process_srat(ACPI_TABLE_SRAT *tp)
2290 {
2291 ACPI_SUBTABLE_HEADER *item, *end;
2292 int i;
2293 int proc_num, mem_num;
2294 #pragma pack(1)
2295 struct {
2296 uint32_t domain;
2297 uint32_t apic_id;
2298 uint32_t sapic_id;
2299 } processor;
2300 struct {
2301 uint32_t domain;
2302 uint32_t x2apic_id;
2303 } x2apic;
2304 struct {
2305 uint32_t domain;
2306 uint64_t addr;
2307 uint64_t length;
2308 uint32_t flags;
2309 } memory;
2310 #pragma pack()
2311 char prop_name[30];
2312 uint64_t maxmem = 0;
2313
2314 if (tp == NULL)
2315 return;
2316
2317 proc_num = mem_num = 0;
2318 end = (ACPI_SUBTABLE_HEADER *)(tp->Header.Length + (uintptr_t)tp);
2319 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)tp + sizeof (*tp));
2320 while (item < end) {
2321 switch (item->Type) {
2322 case ACPI_SRAT_TYPE_CPU_AFFINITY: {
2323 ACPI_SRAT_CPU_AFFINITY *cpu =
2324 (ACPI_SRAT_CPU_AFFINITY *) item;
2325
2326 if (!(cpu->Flags & ACPI_SRAT_CPU_ENABLED))
2327 break;
2328 processor.domain = cpu->ProximityDomainLo;
2329 for (i = 0; i < 3; i++)
2330 processor.domain +=
2331 cpu->ProximityDomainHi[i] << ((i + 1) * 8);
2332 processor.apic_id = cpu->ApicId;
2333 processor.sapic_id = cpu->LocalSapicEid;
2334 (void) snprintf(prop_name, 30, "acpi-srat-processor-%d",
2335 proc_num);
2336 bsetprop(prop_name, strlen(prop_name), &processor,
2337 sizeof (processor));
2338 proc_num++;
2339 break;
2340 }
2341 case ACPI_SRAT_TYPE_MEMORY_AFFINITY: {
2342 ACPI_SRAT_MEM_AFFINITY *mem =
2343 (ACPI_SRAT_MEM_AFFINITY *)item;
2344
2345 if (!(mem->Flags & ACPI_SRAT_MEM_ENABLED))
2346 break;
2347 memory.domain = mem->ProximityDomain;
2348 memory.addr = mem->BaseAddress;
2349 memory.length = mem->Length;
2350 memory.flags = mem->Flags;
2351 (void) snprintf(prop_name, 30, "acpi-srat-memory-%d",
2352 mem_num);
2353 bsetprop(prop_name, strlen(prop_name), &memory,
2354 sizeof (memory));
2355 if ((mem->Flags & ACPI_SRAT_MEM_HOT_PLUGGABLE) &&
2356 (memory.addr + memory.length > maxmem)) {
2357 maxmem = memory.addr + memory.length;
2358 }
2359 mem_num++;
2360 break;
2361 }
2362 case ACPI_SRAT_TYPE_X2APIC_CPU_AFFINITY: {
2363 ACPI_SRAT_X2APIC_CPU_AFFINITY *x2cpu =
2364 (ACPI_SRAT_X2APIC_CPU_AFFINITY *) item;
2365
2366 if (!(x2cpu->Flags & ACPI_SRAT_CPU_ENABLED))
2367 break;
2368 x2apic.domain = x2cpu->ProximityDomain;
2369 x2apic.x2apic_id = x2cpu->ApicId;
2370 (void) snprintf(prop_name, 30, "acpi-srat-processor-%d",
2371 proc_num);
2372 bsetprop(prop_name, strlen(prop_name), &x2apic,
2373 sizeof (x2apic));
2374 proc_num++;
2375 break;
2376 }
2377 default:
2378 if (kbm_debug)
2379 bop_printf(NULL, "SRAT type %d\n", item->Type);
2380 break;
2381 }
2382
2383 item = (ACPI_SUBTABLE_HEADER *)
2384 (item->Length + (uintptr_t)item);
2385 }
2386
2387 /*
2388 * The maximum physical address calculated from the SRAT table is more
2389 * accurate than that calculated from the MSCT table.
2390 */
2391 if (maxmem != 0) {
2392 plat_dr_physmax = btop(maxmem);
2393 }
2394 }
2395
2396 static void
2397 process_slit(ACPI_TABLE_SLIT *tp)
2398 {
2399
2400 /*
2401 * Check the number of localities; if it's too huge, we just
2402 * return and locality enumeration code will handle this later,
2403 * if possible.
2404 *
2405 * Note that the size of the table is the square of the
2406 * number of localities; if the number of localities exceeds
2407 * UINT16_MAX, the table size may overflow an int when being
2408 * passed to bsetprop() below.
2409 */
2410 if (tp->LocalityCount >= SLIT_LOCALITIES_MAX)
2411 return;
2412
2413 bsetprop(SLIT_NUM_PROPNAME, strlen(SLIT_NUM_PROPNAME),
2414 &tp->LocalityCount, sizeof (tp->LocalityCount));
2415 bsetprop(SLIT_PROPNAME, strlen(SLIT_PROPNAME), &tp->Entry,
2416 tp->LocalityCount * tp->LocalityCount);
2417 }
2418
2419 static ACPI_TABLE_MSCT *
2420 process_msct(ACPI_TABLE_MSCT *tp)
2421 {
2422 int last_seen = 0;
2423 int proc_num = 0;
2424 ACPI_MSCT_PROXIMITY *item, *end;
2425 extern uint64_t plat_dr_options;
2426
2427 ASSERT(tp != NULL);
2428
2429 end = (ACPI_MSCT_PROXIMITY *)(tp->Header.Length + (uintptr_t)tp);
2430 for (item = (void *)((uintptr_t)tp + tp->ProximityOffset);
2431 item < end;
2432 item = (void *)(item->Length + (uintptr_t)item)) {
2433 /*
2434 * Sanity check according to section 5.2.19.1 of ACPI 4.0.
2435 * Revision 1
2436 * Length 22
2437 */
2438 if (item->Revision != 1 || item->Length != 22) {
2439 cmn_err(CE_CONT,
2440 "?boot: unknown proximity domain structure in MSCT "
2441 "with Revision(%d), Length(%d).\n",
2442 (int)item->Revision, (int)item->Length);
2443 return (NULL);
2444 } else if (item->RangeStart > item->RangeEnd) {
2445 cmn_err(CE_CONT,
2446 "?boot: invalid proximity domain structure in MSCT "
2447 "with RangeStart(%u), RangeEnd(%u).\n",
2448 item->RangeStart, item->RangeEnd);
2449 return (NULL);
2450 } else if (item->RangeStart != last_seen) {
2451 /*
2452 * Items must be organized in ascending order of the
2453 * proximity domain enumerations.
2454 */
2455 cmn_err(CE_CONT,
2456 "?boot: invalid proximity domain structure in MSCT,"
2457 " items are not orginized in ascending order.\n");
2458 return (NULL);
2459 }
2460
2461 /*
2462 * If ProcessorCapacity is 0 then there would be no CPUs in this
2463 * domain.
2464 */
2465 if (item->ProcessorCapacity != 0) {
2466 proc_num += (item->RangeEnd - item->RangeStart + 1) *
2467 item->ProcessorCapacity;
2468 }
2469
2470 last_seen = item->RangeEnd - item->RangeStart + 1;
2471 /*
2472 * Break out if all proximity domains have been processed.
2473 * Some BIOSes may have unused items at the end of MSCT table.
2474 */
2475 if (last_seen > tp->MaxProximityDomains) {
2476 break;
2477 }
2478 }
2479 if (last_seen != tp->MaxProximityDomains + 1) {
2480 cmn_err(CE_CONT,
2481 "?boot: invalid proximity domain structure in MSCT, "
2482 "proximity domain count doesn't match.\n");
2483 return (NULL);
2484 }
2485
2486 /*
2487 * Set plat-max-ncpus property if it hasn't been set yet.
2488 */
2489 if (do_bsys_getproplen(NULL, PLAT_MAX_NCPUS_NAME) < 0) {
2490 if (proc_num != 0) {
2491 bsetpropsi(PLAT_MAX_NCPUS_NAME, proc_num);
2492 }
2493 }
2494
2495 /*
2496 * Use Maximum Physical Address from the MSCT table as upper limit for
2497 * memory hot-adding by default. It may be overridden by value from
2498 * the SRAT table or the "plat-dr-physmax" boot option.
2499 */
2500 plat_dr_physmax = btop(tp->MaxAddress + 1);
2501
2502 /*
2503 * Existence of MSCT implies CPU/memory hotplug-capability for the
2504 * platform.
2505 */
2506 plat_dr_options |= PLAT_DR_FEATURE_CPU;
2507 plat_dr_options |= PLAT_DR_FEATURE_MEMORY;
2508
2509 return (tp);
2510 }
2511
2512 #else /* __xpv */
2513 static void
2514 enumerate_xen_cpus()
2515 {
2516 processorid_t id, max_id;
2517
2518 /*
2519 * User-set boot-ncpus overrides enumeration
2520 */
2521 if (do_bsys_getproplen(NULL, BOOT_NCPUS_NAME) >= 0)
2522 return;
2523
2524 /*
2525 * Probe every possible virtual CPU id and remember the
2526 * highest id present; the count of CPUs is one greater
2527 * than this. This tacitly assumes at least cpu 0 is present.
2528 */
2529 max_id = 0;
2530 for (id = 0; id < MAX_VIRT_CPUS; id++)
2531 if (HYPERVISOR_vcpu_op(VCPUOP_is_up, id, NULL) == 0)
2532 max_id = id;
2533
2534 bsetpropsi(BOOT_NCPUS_NAME, max_id+1);
2535
2536 }
2537 #endif /* __xpv */
2538
2539 static void
2540 build_firmware_properties(void)
2541 {
2542 ACPI_TABLE_HEADER *tp = NULL;
2543
2544 #ifndef __xpv
2545 if ((tp = find_fw_table(ACPI_SIG_MSCT)) != NULL)
2546 msct_ptr = process_msct((ACPI_TABLE_MSCT *)tp);
2547 else
2548 msct_ptr = NULL;
2549
2550 if ((tp = find_fw_table(ACPI_SIG_MADT)) != NULL)
2551 process_madt((ACPI_TABLE_MADT *)tp);
2552
2553 if ((srat_ptr = (ACPI_TABLE_SRAT *)
2554 find_fw_table(ACPI_SIG_SRAT)) != NULL)
2555 process_srat(srat_ptr);
2556
2557 if (slit_ptr = (ACPI_TABLE_SLIT *)find_fw_table(ACPI_SIG_SLIT))
2558 process_slit(slit_ptr);
2559
2560 tp = find_fw_table(ACPI_SIG_MCFG);
2561 #else /* __xpv */
2562 enumerate_xen_cpus();
2563 if (DOMAIN_IS_INITDOMAIN(xen_info))
2564 tp = find_fw_table(ACPI_SIG_MCFG);
2565 #endif /* __xpv */
2566 if (tp != NULL)
2567 process_mcfg((ACPI_TABLE_MCFG *)tp);
2568 }
2569
2570 /*
2571 * fake up a boot property for deferred early console output
2572 * this is used by both graphical boot and the (developer only)
2573 * USB serial console
2574 */
2575 void *
2576 defcons_init(size_t size)
2577 {
2578 static char *p = NULL;
2579
2580 p = do_bsys_alloc(NULL, NULL, size, MMU_PAGESIZE);
2581 *p = 0;
2582 bsetprop("deferred-console-buf", strlen("deferred-console-buf") + 1,
2583 &p, sizeof (p));
2584 return (p);
2585 }
2586
2587 /*ARGSUSED*/
2588 int
2589 boot_compinfo(int fd, struct compinfo *cbp)
2590 {
2591 cbp->iscmp = 0;
2592 cbp->blksize = MAXBSIZE;
2593 return (0);
2594 }
2595
2596 #define BP_MAX_STRLEN 32
2597
2598 /*
2599 * Get value for given boot property
2600 */
2601 int
2602 bootprop_getval(const char *prop_name, u_longlong_t *prop_value)
2603 {
2604 int boot_prop_len;
2605 char str[BP_MAX_STRLEN];
2606 u_longlong_t value;
2607
2608 boot_prop_len = BOP_GETPROPLEN(bootops, prop_name);
2609 if (boot_prop_len < 0 || boot_prop_len > sizeof (str) ||
2610 BOP_GETPROP(bootops, prop_name, str) < 0 ||
2611 kobj_getvalue(str, &value) == -1)
2612 return (-1);
2613
2614 if (prop_value)
2615 *prop_value = value;
2616
2617 return (0);
2618 }