8213 uts: get smbios from bootloader
[unleashed.git] / usr / src / uts / i86pc / os / fakebop.c
blobeb29b2d6cb37200a7bf82317b4a75d66b60923af
1 /*
2 * CDDL HEADER START
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.
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.
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]
19 * CDDL HEADER END
23 * Copyright 2010 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2010, Intel Corporation.
27 * All rights reserved.
29 * Copyright 2013 Joyent, Inc. All rights reserved.
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.
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/cmn_err.h>
54 #include <sys/systm.h>
55 #include <sys/promif.h>
56 #include <sys/archsystm.h>
57 #include <sys/x86_archext.h>
58 #include <sys/kobj.h>
59 #include <sys/privregs.h>
60 #include <sys/sysmacros.h>
61 #include <sys/ctype.h>
62 #include <sys/fastboot.h>
63 #ifdef __xpv
64 #include <sys/hypervisor.h>
65 #include <net/if.h>
66 #endif
67 #include <vm/kboot_mmu.h>
68 #include <vm/hat_pte.h>
69 #include <sys/kobj.h>
70 #include <sys/kobj_lex.h>
71 #include <sys/pci_cfgspace_impl.h>
72 #include <sys/fastboot_impl.h>
73 #include <sys/acpi/acconfig.h>
74 #include <sys/acpi/acpi.h>
76 static int have_console = 0; /* set once primitive console is initialized */
77 static char *boot_args = "";
80 * Debugging macros
82 static uint_t kbm_debug = 0;
83 #define DBG_MSG(s) { if (kbm_debug) bop_printf(NULL, "%s", s); }
84 #define DBG(x) { if (kbm_debug) \
85 bop_printf(NULL, "%s is %" PRIx64 "\n", #x, (uint64_t)(x)); \
88 #define PUT_STRING(s) { \
89 char *cp; \
90 for (cp = (s); *cp; ++cp) \
91 bcons_putchar(*cp); \
94 bootops_t bootop; /* simple bootops we'll pass on to kernel */
95 struct bsys_mem bm;
98 * Boot info from "glue" code in low memory. xbootp is used by:
99 * do_bop_phys_alloc(), do_bsys_alloc() and boot_prop_finish().
101 static struct xboot_info *xbootp;
102 static uintptr_t next_virt; /* next available virtual address */
103 static paddr_t next_phys; /* next available physical address from dboot */
104 static paddr_t high_phys = -(paddr_t)1; /* last used physical address */
107 * buffer for vsnprintf for console I/O
109 #define BUFFERSIZE 512
110 static char buffer[BUFFERSIZE];
113 * stuff to store/report/manipulate boot property settings.
115 typedef struct bootprop {
116 struct bootprop *bp_next;
117 char *bp_name;
118 uint_t bp_vlen;
119 char *bp_value;
120 } bootprop_t;
122 static bootprop_t *bprops = NULL;
123 static char *curr_page = NULL; /* ptr to avail bprop memory */
124 static int curr_space = 0; /* amount of memory at curr_page */
126 #ifdef __xpv
127 start_info_t *xen_info;
128 shared_info_t *HYPERVISOR_shared_info;
129 #endif
132 * some allocator statistics
134 static ulong_t total_bop_alloc_scratch = 0;
135 static ulong_t total_bop_alloc_kernel = 0;
137 static void build_firmware_properties(struct xboot_info *);
139 static int early_allocation = 1;
141 int force_fastreboot = 0;
142 volatile int fastreboot_onpanic = 0;
143 int post_fastreboot = 0;
144 #ifdef __xpv
145 volatile int fastreboot_capable = 0;
146 #else
147 volatile int fastreboot_capable = 1;
148 #endif
151 * Information saved from current boot for fast reboot.
152 * If the information size exceeds what we have allocated, fast reboot
153 * will not be supported.
155 multiboot_info_t saved_mbi;
156 mb_memory_map_t saved_mmap[FASTBOOT_SAVED_MMAP_COUNT];
157 uint8_t saved_drives[FASTBOOT_SAVED_DRIVES_SIZE];
158 char saved_cmdline[FASTBOOT_SAVED_CMDLINE_LEN];
159 int saved_cmdline_len = 0;
160 size_t saved_file_size[FASTBOOT_MAX_FILES_MAP];
163 * Turn off fastreboot_onpanic to avoid panic loop.
165 char fastreboot_onpanic_cmdline[FASTBOOT_SAVED_CMDLINE_LEN];
166 static const char fastreboot_onpanic_args[] = " -B fastreboot_onpanic=0";
169 * Pointers to where System Resource Affinity Table (SRAT), System Locality
170 * Information Table (SLIT) and Maximum System Capability Table (MSCT)
171 * are mapped into virtual memory
173 ACPI_TABLE_SRAT *srat_ptr = NULL;
174 ACPI_TABLE_SLIT *slit_ptr = NULL;
175 ACPI_TABLE_MSCT *msct_ptr = NULL;
178 * Arbitrary limit on number of localities we handle; if
179 * this limit is raised to more than UINT16_MAX, make sure
180 * process_slit() knows how to handle it.
182 #define SLIT_LOCALITIES_MAX (4096)
184 #define SLIT_NUM_PROPNAME "acpi-slit-localities"
185 #define SLIT_PROPNAME "acpi-slit"
188 * Allocate aligned physical memory at boot time. This allocator allocates
189 * from the highest possible addresses. This avoids exhausting memory that
190 * would be useful for DMA buffers.
192 paddr_t
193 do_bop_phys_alloc(uint64_t size, uint64_t align)
195 paddr_t pa = 0;
196 paddr_t start;
197 paddr_t end;
198 struct memlist *ml = (struct memlist *)xbootp->bi_phys_install;
201 * Be careful if high memory usage is limited in startup.c
202 * Since there are holes in the low part of the physical address
203 * space we can treat physmem as a pfn (not just a pgcnt) and
204 * get a conservative upper limit.
206 if (physmem != 0 && high_phys > pfn_to_pa(physmem))
207 high_phys = pfn_to_pa(physmem);
210 * find the lowest or highest available memory in physinstalled
211 * On 32 bit avoid physmem above 4Gig if PAE isn't enabled
213 #if defined(__i386)
214 if (xbootp->bi_use_pae == 0 && high_phys > FOUR_GIG)
215 high_phys = FOUR_GIG;
216 #endif
219 * find the highest available memory in physinstalled
221 size = P2ROUNDUP(size, align);
222 for (; ml; ml = ml->ml_next) {
223 start = P2ROUNDUP(ml->ml_address, align);
224 end = P2ALIGN(ml->ml_address + ml->ml_size, align);
225 if (start < next_phys)
226 start = P2ROUNDUP(next_phys, align);
227 if (end > high_phys)
228 end = P2ALIGN(high_phys, align);
230 if (end <= start)
231 continue;
232 if (end - start < size)
233 continue;
236 * Early allocations need to use low memory, since
237 * physmem might be further limited by bootenv.rc
239 if (early_allocation) {
240 if (pa == 0 || start < pa)
241 pa = start;
242 } else {
243 if (end - size > pa)
244 pa = end - size;
247 if (pa != 0) {
248 if (early_allocation)
249 next_phys = pa + size;
250 else
251 high_phys = pa;
252 return (pa);
254 bop_panic("do_bop_phys_alloc(0x%" PRIx64 ", 0x%" PRIx64
255 ") Out of memory\n", size, align);
256 /*NOTREACHED*/
259 uintptr_t
260 alloc_vaddr(size_t size, paddr_t align)
262 uintptr_t rv;
264 next_virt = P2ROUNDUP(next_virt, (uintptr_t)align);
265 rv = (uintptr_t)next_virt;
266 next_virt += size;
267 return (rv);
271 * Allocate virtual memory. The size is always rounded up to a multiple
272 * of base pagesize.
275 /*ARGSUSED*/
276 static caddr_t
277 do_bsys_alloc(bootops_t *bop, caddr_t virthint, size_t size, int align)
279 paddr_t a = align; /* same type as pa for masking */
280 uint_t pgsize;
281 paddr_t pa;
282 uintptr_t va;
283 ssize_t s; /* the aligned size */
284 uint_t level;
285 uint_t is_kernel = (virthint != 0);
287 if (a < MMU_PAGESIZE)
288 a = MMU_PAGESIZE;
289 else if (!ISP2(a))
290 prom_panic("do_bsys_alloc() incorrect alignment");
291 size = P2ROUNDUP(size, MMU_PAGESIZE);
294 * Use the next aligned virtual address if we weren't given one.
296 if (virthint == NULL) {
297 virthint = (caddr_t)alloc_vaddr(size, a);
298 total_bop_alloc_scratch += size;
299 } else {
300 total_bop_alloc_kernel += size;
304 * allocate the physical memory
306 pa = do_bop_phys_alloc(size, a);
309 * Add the mappings to the page tables, try large pages first.
311 va = (uintptr_t)virthint;
312 s = size;
313 level = 1;
314 pgsize = xbootp->bi_use_pae ? TWO_MEG : FOUR_MEG;
315 if (xbootp->bi_use_largepage && a == pgsize) {
316 while (IS_P2ALIGNED(pa, pgsize) && IS_P2ALIGNED(va, pgsize) &&
317 s >= pgsize) {
318 kbm_map(va, pa, level, is_kernel);
319 va += pgsize;
320 pa += pgsize;
321 s -= pgsize;
326 * Map remaining pages use small mappings
328 level = 0;
329 pgsize = MMU_PAGESIZE;
330 while (s > 0) {
331 kbm_map(va, pa, level, is_kernel);
332 va += pgsize;
333 pa += pgsize;
334 s -= pgsize;
336 return (virthint);
340 * Free virtual memory - we'll just ignore these.
342 /*ARGSUSED*/
343 static void
344 do_bsys_free(bootops_t *bop, caddr_t virt, size_t size)
346 bop_printf(NULL, "do_bsys_free(virt=0x%p, size=0x%lx) ignored\n",
347 (void *)virt, size);
351 * Old interface
353 /*ARGSUSED*/
354 static caddr_t
355 do_bsys_ealloc(bootops_t *bop, caddr_t virthint, size_t size,
356 int align, int flags)
358 prom_panic("unsupported call to BOP_EALLOC()\n");
359 return (0);
363 static void
364 bsetprop(char *name, int nlen, void *value, int vlen)
366 uint_t size;
367 uint_t need_size;
368 bootprop_t *b;
371 * align the size to 16 byte boundary
373 size = sizeof (bootprop_t) + nlen + 1 + vlen;
374 size = (size + 0xf) & ~0xf;
375 if (size > curr_space) {
376 need_size = (size + (MMU_PAGEOFFSET)) & MMU_PAGEMASK;
377 curr_page = do_bsys_alloc(NULL, 0, need_size, MMU_PAGESIZE);
378 curr_space = need_size;
382 * use a bootprop_t at curr_page and link into list
384 b = (bootprop_t *)curr_page;
385 curr_page += sizeof (bootprop_t);
386 curr_space -= sizeof (bootprop_t);
387 b->bp_next = bprops;
388 bprops = b;
391 * follow by name and ending zero byte
393 b->bp_name = curr_page;
394 bcopy(name, curr_page, nlen);
395 curr_page += nlen;
396 *curr_page++ = 0;
397 curr_space -= nlen + 1;
400 * copy in value, but no ending zero byte
402 b->bp_value = curr_page;
403 b->bp_vlen = vlen;
404 if (vlen > 0) {
405 bcopy(value, curr_page, vlen);
406 curr_page += vlen;
407 curr_space -= vlen;
411 * align new values of curr_page, curr_space
413 while (curr_space & 0xf) {
414 ++curr_page;
415 --curr_space;
419 static void
420 bsetprops(char *name, char *value)
422 bsetprop(name, strlen(name), value, strlen(value) + 1);
425 static void
426 bsetprop64(char *name, uint64_t value)
428 bsetprop(name, strlen(name), (void *)&value, sizeof (value));
431 static void
432 bsetpropsi(char *name, int value)
434 char prop_val[32];
436 (void) snprintf(prop_val, sizeof (prop_val), "%d", value);
437 bsetprops(name, prop_val);
441 * to find the size of the buffer to allocate
443 /*ARGSUSED*/
445 do_bsys_getproplen(bootops_t *bop, const char *name)
447 bootprop_t *b;
449 for (b = bprops; b; b = b->bp_next) {
450 if (strcmp(name, b->bp_name) != 0)
451 continue;
452 return (b->bp_vlen);
454 return (-1);
458 * get the value associated with this name
460 /*ARGSUSED*/
462 do_bsys_getprop(bootops_t *bop, const char *name, void *value)
464 bootprop_t *b;
466 for (b = bprops; b; b = b->bp_next) {
467 if (strcmp(name, b->bp_name) != 0)
468 continue;
469 bcopy(b->bp_value, value, b->bp_vlen);
470 return (0);
472 return (-1);
476 * get the name of the next property in succession from the standalone
478 /*ARGSUSED*/
479 static char *
480 do_bsys_nextprop(bootops_t *bop, char *name)
482 bootprop_t *b;
485 * A null name is a special signal for the 1st boot property
487 if (name == NULL || strlen(name) == 0) {
488 if (bprops == NULL)
489 return (NULL);
490 return (bprops->bp_name);
493 for (b = bprops; b; b = b->bp_next) {
494 if (name != b->bp_name)
495 continue;
496 b = b->bp_next;
497 if (b == NULL)
498 return (NULL);
499 return (b->bp_name);
501 return (NULL);
505 * Parse numeric value from a string. Understands decimal, hex, octal, - and ~
507 static int
508 parse_value(char *p, uint64_t *retval)
510 int adjust = 0;
511 uint64_t tmp = 0;
512 int digit;
513 int radix = 10;
515 *retval = 0;
516 if (*p == '-' || *p == '~')
517 adjust = *p++;
519 if (*p == '0') {
520 ++p;
521 if (*p == 0)
522 return (0);
523 if (*p == 'x' || *p == 'X') {
524 radix = 16;
525 ++p;
526 } else {
527 radix = 8;
528 ++p;
531 while (*p) {
532 if ('0' <= *p && *p <= '9')
533 digit = *p - '0';
534 else if ('a' <= *p && *p <= 'f')
535 digit = 10 + *p - 'a';
536 else if ('A' <= *p && *p <= 'F')
537 digit = 10 + *p - 'A';
538 else
539 return (-1);
540 if (digit >= radix)
541 return (-1);
542 tmp = tmp * radix + digit;
543 ++p;
545 if (adjust == '-')
546 tmp = -tmp;
547 else if (adjust == '~')
548 tmp = ~tmp;
549 *retval = tmp;
550 return (0);
553 static boolean_t
554 unprintable(char *value, int size)
556 int i;
558 if (size <= 0 || value[0] == '\0')
559 return (B_TRUE);
561 for (i = 0; i < size; i++) {
562 if (value[i] == '\0')
563 return (i != (size - 1));
565 if (!isprint(value[i]))
566 return (B_TRUE);
568 return (B_FALSE);
572 * Print out information about all boot properties.
573 * buffer is pointer to pre-allocated space to be used as temporary
574 * space for property values.
576 static void
577 boot_prop_display(char *buffer)
579 char *name = "";
580 int i, len;
582 bop_printf(NULL, "\nBoot properties:\n");
584 while ((name = do_bsys_nextprop(NULL, name)) != NULL) {
585 bop_printf(NULL, "\t0x%p %s = ", (void *)name, name);
586 (void) do_bsys_getprop(NULL, name, buffer);
587 len = do_bsys_getproplen(NULL, name);
588 bop_printf(NULL, "len=%d ", len);
589 if (!unprintable(buffer, len)) {
590 buffer[len] = 0;
591 bop_printf(NULL, "%s\n", buffer);
592 continue;
594 for (i = 0; i < len; i++) {
595 bop_printf(NULL, "%02x", buffer[i] & 0xff);
596 if (i < len - 1)
597 bop_printf(NULL, ".");
599 bop_printf(NULL, "\n");
604 * 2nd part of building the table of boot properties. This includes:
605 * - values from /boot/solaris/bootenv.rc (ie. eeprom(1m) values)
607 * lines look like one of:
608 * ^$
609 * ^# comment till end of line
610 * setprop name 'value'
611 * setprop name value
612 * setprop name "value"
614 * we do single character I/O since this is really just looking at memory
616 void
617 boot_prop_finish(void)
619 int fd;
620 char *line;
621 int c;
622 int bytes_read;
623 char *name;
624 int n_len;
625 char *value;
626 int v_len;
627 char *inputdev; /* these override the command line if serial ports */
628 char *outputdev;
629 char *consoledev;
630 uint64_t lvalue;
631 int use_xencons = 0;
633 #ifdef __xpv
634 if (!DOMAIN_IS_INITDOMAIN(xen_info))
635 use_xencons = 1;
636 #endif /* __xpv */
638 DBG_MSG("Opening /boot/solaris/bootenv.rc\n");
639 fd = BRD_OPEN(bfs_ops, "/boot/solaris/bootenv.rc", 0);
640 DBG(fd);
642 line = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
643 while (fd >= 0) {
646 * get a line
648 for (c = 0; ; ++c) {
649 bytes_read = BRD_READ(bfs_ops, fd, line + c, 1);
650 if (bytes_read == 0) {
651 if (c == 0)
652 goto done;
653 break;
655 if (line[c] == '\n')
656 break;
658 line[c] = 0;
661 * ignore comment lines
663 c = 0;
664 while (ISSPACE(line[c]))
665 ++c;
666 if (line[c] == '#' || line[c] == 0)
667 continue;
670 * must have "setprop " or "setprop\t"
672 if (strncmp(line + c, "setprop ", 8) != 0 &&
673 strncmp(line + c, "setprop\t", 8) != 0)
674 continue;
675 c += 8;
676 while (ISSPACE(line[c]))
677 ++c;
678 if (line[c] == 0)
679 continue;
682 * gather up the property name
684 name = line + c;
685 n_len = 0;
686 while (line[c] && !ISSPACE(line[c]))
687 ++n_len, ++c;
690 * gather up the value, if any
692 value = "";
693 v_len = 0;
694 while (ISSPACE(line[c]))
695 ++c;
696 if (line[c] != 0) {
697 value = line + c;
698 while (line[c] && !ISSPACE(line[c]))
699 ++v_len, ++c;
702 if (v_len >= 2 && value[0] == value[v_len - 1] &&
703 (value[0] == '\'' || value[0] == '"')) {
704 ++value;
705 v_len -= 2;
707 name[n_len] = 0;
708 if (v_len > 0)
709 value[v_len] = 0;
710 else
711 continue;
714 * ignore "boot-file" property, it's now meaningless
716 if (strcmp(name, "boot-file") == 0)
717 continue;
718 if (strcmp(name, "boot-args") == 0 &&
719 strlen(boot_args) > 0)
720 continue;
723 * If a property was explicitly set on the command line
724 * it will override a setting in bootenv.rc
726 if (do_bsys_getproplen(NULL, name) > 0)
727 continue;
729 bsetprop(name, n_len, value, v_len + 1);
731 done:
732 if (fd >= 0)
733 (void) BRD_CLOSE(bfs_ops, fd);
736 * Check if we have to limit the boot time allocator
738 if (do_bsys_getproplen(NULL, "physmem") != -1 &&
739 do_bsys_getprop(NULL, "physmem", line) >= 0 &&
740 parse_value(line, &lvalue) != -1) {
741 if (0 < lvalue && (lvalue < physmem || physmem == 0)) {
742 physmem = (pgcnt_t)lvalue;
743 DBG(physmem);
746 early_allocation = 0;
749 * check to see if we have to override the default value of the console
751 if (!use_xencons) {
752 inputdev = line;
753 v_len = do_bsys_getproplen(NULL, "input-device");
754 if (v_len > 0)
755 (void) do_bsys_getprop(NULL, "input-device", inputdev);
756 else
757 v_len = 0;
758 inputdev[v_len] = 0;
760 outputdev = inputdev + v_len + 1;
761 v_len = do_bsys_getproplen(NULL, "output-device");
762 if (v_len > 0)
763 (void) do_bsys_getprop(NULL, "output-device",
764 outputdev);
765 else
766 v_len = 0;
767 outputdev[v_len] = 0;
769 consoledev = outputdev + v_len + 1;
770 v_len = do_bsys_getproplen(NULL, "console");
771 if (v_len > 0) {
772 (void) do_bsys_getprop(NULL, "console", consoledev);
773 if (post_fastreboot &&
774 strcmp(consoledev, "graphics") == 0) {
775 bsetprops("console", "text");
776 v_len = strlen("text");
777 bcopy("text", consoledev, v_len);
779 } else {
780 v_len = 0;
782 consoledev[v_len] = 0;
783 bcons_init2(inputdev, outputdev, consoledev);
784 } else {
786 * Ensure console property exists
787 * If not create it as "hypervisor"
789 v_len = do_bsys_getproplen(NULL, "console");
790 if (v_len < 0)
791 bsetprops("console", "hypervisor");
792 inputdev = outputdev = consoledev = "hypervisor";
793 bcons_init2(inputdev, outputdev, consoledev);
796 if (find_boot_prop("prom_debug") || kbm_debug)
797 boot_prop_display(line);
801 * print formatted output
803 /*PRINTFLIKE2*/
804 /*ARGSUSED*/
805 void
806 bop_printf(bootops_t *bop, const char *fmt, ...)
808 va_list ap;
810 if (have_console == 0)
811 return;
813 va_start(ap, fmt);
814 (void) vsnprintf(buffer, BUFFERSIZE, fmt, ap);
815 va_end(ap);
816 PUT_STRING(buffer);
820 * Another panic() variant; this one can be used even earlier during boot than
821 * prom_panic().
823 /*PRINTFLIKE1*/
824 void
825 bop_panic(const char *fmt, ...)
827 va_list ap;
829 va_start(ap, fmt);
830 bop_printf(NULL, fmt, ap);
831 va_end(ap);
833 bop_printf(NULL, "\nPress any key to reboot.\n");
834 (void) bcons_getchar();
835 bop_printf(NULL, "Resetting...\n");
836 pc_reset();
840 * Do a real mode interrupt BIOS call
842 typedef struct bios_regs {
843 unsigned short ax, bx, cx, dx, si, di, bp, es, ds;
844 } bios_regs_t;
845 typedef int (*bios_func_t)(int, bios_regs_t *);
847 /*ARGSUSED*/
848 static void
849 do_bsys_doint(bootops_t *bop, int intnum, struct bop_regs *rp)
851 #if defined(__xpv)
852 prom_panic("unsupported call to BOP_DOINT()\n");
853 #else /* __xpv */
854 static int firsttime = 1;
855 bios_func_t bios_func = (bios_func_t)(void *)(uintptr_t)0x5000;
856 bios_regs_t br;
859 * The first time we do this, we have to copy the pre-packaged
860 * low memory bios call code image into place.
862 if (firsttime) {
863 extern char bios_image[];
864 extern uint32_t bios_size;
866 bcopy(bios_image, (void *)bios_func, bios_size);
867 firsttime = 0;
870 br.ax = rp->eax.word.ax;
871 br.bx = rp->ebx.word.bx;
872 br.cx = rp->ecx.word.cx;
873 br.dx = rp->edx.word.dx;
874 br.bp = rp->ebp.word.bp;
875 br.si = rp->esi.word.si;
876 br.di = rp->edi.word.di;
877 br.ds = rp->ds;
878 br.es = rp->es;
880 DBG_MSG("Doing BIOS call...");
881 DBG(br.ax);
882 DBG(br.bx);
883 DBG(br.dx);
884 rp->eflags = bios_func(intnum, &br);
885 DBG_MSG("done\n");
887 rp->eax.word.ax = br.ax;
888 rp->ebx.word.bx = br.bx;
889 rp->ecx.word.cx = br.cx;
890 rp->edx.word.dx = br.dx;
891 rp->ebp.word.bp = br.bp;
892 rp->esi.word.si = br.si;
893 rp->edi.word.di = br.di;
894 rp->ds = br.ds;
895 rp->es = br.es;
896 #endif /* __xpv */
899 static struct boot_syscalls bop_sysp = {
900 bcons_getchar,
901 bcons_putchar,
902 bcons_ischar,
905 static char *whoami;
907 #define BUFLEN 64
909 #if defined(__xpv)
911 static char namebuf[32];
913 static void
914 xen_parse_props(char *s, char *prop_map[], int n_prop)
916 char **prop_name = prop_map;
917 char *cp = s, *scp;
919 do {
920 scp = cp;
921 while ((*cp != NULL) && (*cp != ':'))
922 cp++;
924 if ((scp != cp) && (*prop_name != NULL)) {
925 *cp = NULL;
926 bsetprops(*prop_name, scp);
929 cp++;
930 prop_name++;
931 n_prop--;
932 } while (n_prop > 0);
935 #define VBDPATHLEN 64
938 * parse the 'xpv-root' property to create properties used by
939 * ufs_mountroot.
941 static void
942 xen_vbdroot_props(char *s)
944 char vbdpath[VBDPATHLEN] = "/xpvd/xdf@";
945 const char lnamefix[] = "/dev/dsk/c0d";
946 char *pnp;
947 char *prop_p;
948 char mi;
949 short minor;
950 long addr = 0;
952 pnp = vbdpath + strlen(vbdpath);
953 prop_p = s + strlen(lnamefix);
954 while ((*prop_p != '\0') && (*prop_p != 's') && (*prop_p != 'p'))
955 addr = addr * 10 + *prop_p++ - '0';
956 (void) snprintf(pnp, VBDPATHLEN, "%lx", addr);
957 pnp = vbdpath + strlen(vbdpath);
958 if (*prop_p == 's')
959 mi = 'a';
960 else if (*prop_p == 'p')
961 mi = 'q';
962 else
963 ASSERT(0); /* shouldn't be here */
964 prop_p++;
965 ASSERT(*prop_p != '\0');
966 if (ISDIGIT(*prop_p)) {
967 minor = *prop_p - '0';
968 prop_p++;
969 if (ISDIGIT(*prop_p)) {
970 minor = minor * 10 + *prop_p - '0';
972 } else {
973 /* malformed root path, use 0 as default */
974 minor = 0;
976 ASSERT(minor < 16); /* at most 16 partitions */
977 mi += minor;
978 *pnp++ = ':';
979 *pnp++ = mi;
980 *pnp++ = '\0';
981 bsetprops("fstype", "ufs");
982 bsetprops("bootpath", vbdpath);
984 DBG_MSG("VBD bootpath set to ");
985 DBG_MSG(vbdpath);
986 DBG_MSG("\n");
990 * parse the xpv-nfsroot property to create properties used by
991 * nfs_mountroot.
993 static void
994 xen_nfsroot_props(char *s)
996 char *prop_map[] = {
997 BP_SERVER_IP, /* server IP address */
998 BP_SERVER_NAME, /* server hostname */
999 BP_SERVER_PATH, /* root path */
1001 int n_prop = sizeof (prop_map) / sizeof (prop_map[0]);
1003 bsetprop("fstype", 6, "nfs", 4);
1005 xen_parse_props(s, prop_map, n_prop);
1008 * If a server name wasn't specified, use a default.
1010 if (do_bsys_getproplen(NULL, BP_SERVER_NAME) == -1)
1011 bsetprops(BP_SERVER_NAME, "unknown");
1015 * Extract our IP address, etc. from the "xpv-ip" property.
1017 static void
1018 xen_ip_props(char *s)
1020 char *prop_map[] = {
1021 BP_HOST_IP, /* IP address */
1022 NULL, /* NFS server IP address (ignored in */
1023 /* favour of xpv-nfsroot) */
1024 BP_ROUTER_IP, /* IP gateway */
1025 BP_SUBNET_MASK, /* IP subnet mask */
1026 "xpv-hostname", /* hostname (ignored) */
1027 BP_NETWORK_INTERFACE, /* interface name */
1028 "xpv-hcp", /* host configuration protocol */
1030 int n_prop = sizeof (prop_map) / sizeof (prop_map[0]);
1031 char ifname[IFNAMSIZ];
1033 xen_parse_props(s, prop_map, n_prop);
1036 * A Linux dom0 administrator expects all interfaces to be
1037 * called "ethX", which is not the case here.
1039 * If the interface name specified is "eth0", presume that
1040 * this is really intended to be "xnf0" (the first domU ->
1041 * dom0 interface for this domain).
1043 if ((do_bsys_getprop(NULL, BP_NETWORK_INTERFACE, ifname) == 0) &&
1044 (strcmp("eth0", ifname) == 0)) {
1045 bsetprops(BP_NETWORK_INTERFACE, "xnf0");
1046 bop_printf(NULL,
1047 "network interface name 'eth0' replaced with 'xnf0'\n");
1051 #else /* __xpv */
1053 static void
1054 setup_rarp_props(struct sol_netinfo *sip)
1056 char buf[BUFLEN]; /* to hold ip/mac addrs */
1057 uint8_t *val;
1059 val = (uint8_t *)&sip->sn_ciaddr;
1060 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1061 val[0], val[1], val[2], val[3]);
1062 bsetprops(BP_HOST_IP, buf);
1064 val = (uint8_t *)&sip->sn_siaddr;
1065 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1066 val[0], val[1], val[2], val[3]);
1067 bsetprops(BP_SERVER_IP, buf);
1069 if (sip->sn_giaddr != 0) {
1070 val = (uint8_t *)&sip->sn_giaddr;
1071 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1072 val[0], val[1], val[2], val[3]);
1073 bsetprops(BP_ROUTER_IP, buf);
1076 if (sip->sn_netmask != 0) {
1077 val = (uint8_t *)&sip->sn_netmask;
1078 (void) snprintf(buf, BUFLEN, "%d.%d.%d.%d",
1079 val[0], val[1], val[2], val[3]);
1080 bsetprops(BP_SUBNET_MASK, buf);
1083 if (sip->sn_mactype != 4 || sip->sn_maclen != 6) {
1084 bop_printf(NULL, "unsupported mac type %d, mac len %d\n",
1085 sip->sn_mactype, sip->sn_maclen);
1086 } else {
1087 val = sip->sn_macaddr;
1088 (void) snprintf(buf, BUFLEN, "%x:%x:%x:%x:%x:%x",
1089 val[0], val[1], val[2], val[3], val[4], val[5]);
1090 bsetprops(BP_BOOT_MAC, buf);
1094 #endif /* __xpv */
1096 static void
1097 build_panic_cmdline(const char *cmd, int cmdlen)
1099 int proplen;
1100 size_t arglen;
1102 arglen = sizeof (fastreboot_onpanic_args);
1104 * If we allready have fastreboot-onpanic set to zero,
1105 * don't add them again.
1107 if ((proplen = do_bsys_getproplen(NULL, FASTREBOOT_ONPANIC)) > 0 &&
1108 proplen <= sizeof (fastreboot_onpanic_cmdline)) {
1109 (void) do_bsys_getprop(NULL, FASTREBOOT_ONPANIC,
1110 fastreboot_onpanic_cmdline);
1111 if (FASTREBOOT_ONPANIC_NOTSET(fastreboot_onpanic_cmdline))
1112 arglen = 1;
1116 * construct fastreboot_onpanic_cmdline
1118 if (cmdlen + arglen > sizeof (fastreboot_onpanic_cmdline)) {
1119 DBG_MSG("Command line too long: clearing "
1120 FASTREBOOT_ONPANIC "\n");
1121 fastreboot_onpanic = 0;
1122 } else {
1123 bcopy(cmd, fastreboot_onpanic_cmdline, cmdlen);
1124 if (arglen != 1)
1125 bcopy(fastreboot_onpanic_args,
1126 fastreboot_onpanic_cmdline + cmdlen, arglen);
1127 else
1128 fastreboot_onpanic_cmdline[cmdlen] = 0;
1133 #ifndef __xpv
1135 * Construct boot command line for Fast Reboot. The saved_cmdline
1136 * is also reported by "eeprom bootcmd".
1138 static void
1139 build_fastboot_cmdline(struct xboot_info *xbp)
1141 saved_cmdline_len = strlen(xbp->bi_cmdline) + 1;
1142 if (saved_cmdline_len > FASTBOOT_SAVED_CMDLINE_LEN) {
1143 DBG(saved_cmdline_len);
1144 DBG_MSG("Command line too long: clearing fastreboot_capable\n");
1145 fastreboot_capable = 0;
1146 } else {
1147 bcopy((void *)(xbp->bi_cmdline), (void *)saved_cmdline,
1148 saved_cmdline_len);
1149 saved_cmdline[saved_cmdline_len - 1] = '\0';
1150 build_panic_cmdline(saved_cmdline, saved_cmdline_len - 1);
1155 * Save memory layout, disk drive information, unix and boot archive sizes for
1156 * Fast Reboot.
1158 static void
1159 save_boot_info(struct xboot_info *xbi)
1161 multiboot_info_t *mbi = xbi->bi_mb_info;
1162 struct boot_modules *modp;
1163 int i;
1165 bcopy(mbi, &saved_mbi, sizeof (multiboot_info_t));
1166 if (mbi->mmap_length > sizeof (saved_mmap)) {
1167 DBG_MSG("mbi->mmap_length too big: clearing "
1168 "fastreboot_capable\n");
1169 fastreboot_capable = 0;
1170 } else {
1171 bcopy((void *)(uintptr_t)mbi->mmap_addr, (void *)saved_mmap,
1172 mbi->mmap_length);
1175 if ((mbi->flags & MB_INFO_DRIVE_INFO) != 0) {
1176 if (mbi->drives_length > sizeof (saved_drives)) {
1177 DBG(mbi->drives_length);
1178 DBG_MSG("mbi->drives_length too big: clearing "
1179 "fastreboot_capable\n");
1180 fastreboot_capable = 0;
1181 } else {
1182 bcopy((void *)(uintptr_t)mbi->drives_addr,
1183 (void *)saved_drives, mbi->drives_length);
1185 } else {
1186 saved_mbi.drives_length = 0;
1187 saved_mbi.drives_addr = NULL;
1191 * Current file sizes. Used by fastboot.c to figure out how much
1192 * memory to reserve for panic reboot.
1193 * Use the module list from the dboot-constructed xboot_info
1194 * instead of the list referenced by the multiboot structure
1195 * because that structure may not be addressable now.
1197 saved_file_size[FASTBOOT_NAME_UNIX] = FOUR_MEG - PAGESIZE;
1198 for (i = 0, modp = (struct boot_modules *)(uintptr_t)xbi->bi_modules;
1199 i < xbi->bi_module_cnt; i++, modp++) {
1200 saved_file_size[FASTBOOT_NAME_BOOTARCHIVE] += modp->bm_size;
1203 #endif /* __xpv */
1206 * Import boot environment module variables as properties, applying
1207 * blacklist filter for variables we know we will not use.
1209 * Since the environment can be relatively large, containing many variables
1210 * used only for boot loader purposes, we will use a blacklist based filter.
1211 * To keep the blacklist from growing too large, we use prefix based filtering.
1212 * This is possible because in many cases, the loader variable names are
1213 * using a structured layout.
1215 * We will not overwrite already set properties.
1217 static struct bop_blacklist {
1218 const char *bl_name;
1219 int bl_name_len;
1220 } bop_prop_blacklist[] = {
1221 { "ISADIR", sizeof ("ISADIR") },
1222 { "acpi", sizeof ("acpi") },
1223 { "autoboot_delay", sizeof ("autoboot_delay") },
1224 { "autoboot_delay", sizeof ("autoboot_delay") },
1225 { "beansi_", sizeof ("beansi_") },
1226 { "beastie", sizeof ("beastie") },
1227 { "bemenu", sizeof ("bemenu") },
1228 { "boot.", sizeof ("boot.") },
1229 { "bootenv", sizeof ("bootenv") },
1230 { "currdev", sizeof ("currdev") },
1231 { "dhcp.", sizeof ("dhcp.") },
1232 { "interpret", sizeof ("interpret") },
1233 { "kernel", sizeof ("kernel") },
1234 { "loaddev", sizeof ("loaddev") },
1235 { "loader_", sizeof ("loader_") },
1236 { "module_path", sizeof ("module_path") },
1237 { "nfs.", sizeof ("nfs.") },
1238 { "pcibios", sizeof ("pcibios") },
1239 { "prompt", sizeof ("prompt") },
1240 { "smbios", sizeof ("smbios") },
1241 { "tem", sizeof ("tem") },
1242 { "twiddle_divisor", sizeof ("twiddle_divisor") },
1243 { "zfs_be", sizeof ("zfs_be") },
1247 * Match the name against prefixes in above blacklist. If the match was
1248 * found, this name is blacklisted.
1250 static boolean_t
1251 name_is_blacklisted(const char *name)
1253 int i, n;
1255 n = sizeof (bop_prop_blacklist) / sizeof (bop_prop_blacklist[0]);
1256 for (i = 0; i < n; i++) {
1257 if (strncmp(bop_prop_blacklist[i].bl_name, name,
1258 bop_prop_blacklist[i].bl_name_len - 1) == 0) {
1259 return (B_TRUE);
1262 return (B_FALSE);
1265 static void
1266 process_boot_environment(struct boot_modules *benv)
1268 char *env, *ptr, *name, *value;
1269 uint32_t size, name_len, value_len;
1271 if (benv == NULL || benv->bm_type != BMT_ENV)
1272 return;
1273 ptr = env = benv->bm_addr;
1274 size = benv->bm_size;
1275 do {
1276 name = ptr;
1277 /* find '=' */
1278 while (*ptr != '=') {
1279 ptr++;
1280 if (ptr > env + size) /* Something is very wrong. */
1281 return;
1283 name_len = ptr - name;
1284 if (sizeof (buffer) <= name_len)
1285 continue;
1287 (void) strncpy(buffer, name, sizeof (buffer));
1288 buffer[name_len] = '\0';
1289 name = buffer;
1291 value_len = 0;
1292 value = ++ptr;
1293 while ((uintptr_t)ptr - (uintptr_t)env < size) {
1294 if (*ptr == '\0') {
1295 ptr++;
1296 value_len = (uintptr_t)ptr - (uintptr_t)env;
1297 break;
1299 ptr++;
1302 /* Did we reach the end of the module? */
1303 if (value_len == 0)
1304 return;
1306 if (*value == '\0')
1307 continue;
1309 /* Is this property already set? */
1310 if (do_bsys_getproplen(NULL, name) >= 0)
1311 continue;
1313 if (name_is_blacklisted(name) == B_TRUE)
1314 continue;
1316 /* Create new property. */
1317 bsetprops(name, value);
1319 /* Avoid reading past the module end. */
1320 if (size <= (uintptr_t)ptr - (uintptr_t)env)
1321 return;
1322 } while (*ptr != '\0');
1326 * 1st pass at building the table of boot properties. This includes:
1327 * - values set on the command line: -B a=x,b=y,c=z ....
1328 * - known values we just compute (ie. from xbp)
1329 * - values from /boot/solaris/bootenv.rc (ie. eeprom(1m) values)
1331 * the grub command line looked like:
1332 * kernel boot-file [-B prop=value[,prop=value]...] [boot-args]
1334 * whoami is the same as boot-file
1336 static void
1337 build_boot_properties(struct xboot_info *xbp)
1339 char *name;
1340 int name_len;
1341 char *value;
1342 int value_len;
1343 struct boot_modules *bm, *rdbm, *benv = NULL;
1344 char *propbuf;
1345 int quoted = 0;
1346 int boot_arg_len;
1347 uint_t i, midx;
1348 char modid[32];
1349 #ifndef __xpv
1350 static int stdout_val = 0;
1351 uchar_t boot_device;
1352 char str[3];
1353 #endif
1356 * These have to be done first, so that kobj_mount_root() works
1358 DBG_MSG("Building boot properties\n");
1359 propbuf = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, 0);
1360 DBG((uintptr_t)propbuf);
1361 if (xbp->bi_module_cnt > 0) {
1362 bm = xbp->bi_modules;
1363 rdbm = NULL;
1364 for (midx = i = 0; i < xbp->bi_module_cnt; i++) {
1365 if (bm[i].bm_type == BMT_ROOTFS) {
1366 rdbm = &bm[i];
1367 continue;
1369 if (bm[i].bm_type == BMT_HASH || bm[i].bm_name == NULL)
1370 continue;
1372 if (bm[i].bm_type == BMT_ENV) {
1373 if (benv == NULL)
1374 benv = &bm[i];
1375 else
1376 continue;
1379 (void) snprintf(modid, sizeof (modid),
1380 "module-name-%u", midx);
1381 bsetprops(modid, (char *)bm[i].bm_name);
1382 (void) snprintf(modid, sizeof (modid),
1383 "module-addr-%u", midx);
1384 bsetprop64(modid, (uint64_t)(uintptr_t)bm[i].bm_addr);
1385 (void) snprintf(modid, sizeof (modid),
1386 "module-size-%u", midx);
1387 bsetprop64(modid, (uint64_t)bm[i].bm_size);
1388 ++midx;
1390 if (rdbm != NULL) {
1391 bsetprop64("ramdisk_start",
1392 (uint64_t)(uintptr_t)rdbm->bm_addr);
1393 bsetprop64("ramdisk_end",
1394 (uint64_t)(uintptr_t)rdbm->bm_addr + rdbm->bm_size);
1399 * If there are any boot time modules or hashes present, then disable
1400 * fast reboot.
1402 if (xbp->bi_module_cnt > 1) {
1403 fastreboot_disable(FBNS_BOOTMOD);
1406 #ifndef __xpv
1408 * Disable fast reboot if we're using the Multiboot 2 boot protocol,
1409 * since we don't currently support MB2 info and module relocation.
1410 * Note that fast reboot will have already been disabled if multiple
1411 * modules are present, since the current implementation assumes that
1412 * we only have a single module, the boot_archive.
1414 if (xbp->bi_mb_version != 1) {
1415 fastreboot_disable(FBNS_MULTIBOOT2);
1417 #endif
1419 DBG_MSG("Parsing command line for boot properties\n");
1420 value = xbp->bi_cmdline;
1423 * allocate memory to collect boot_args into
1425 boot_arg_len = strlen(xbp->bi_cmdline) + 1;
1426 boot_args = do_bsys_alloc(NULL, NULL, boot_arg_len, MMU_PAGESIZE);
1427 boot_args[0] = 0;
1428 boot_arg_len = 0;
1430 #ifdef __xpv
1432 * Xen puts a lot of device information in front of the kernel name
1433 * let's grab them and make them boot properties. The first
1434 * string w/o an "=" in it will be the boot-file property.
1436 (void) strcpy(namebuf, "xpv-");
1437 for (;;) {
1439 * get to next property
1441 while (ISSPACE(*value))
1442 ++value;
1443 name = value;
1445 * look for an "="
1447 while (*value && !ISSPACE(*value) && *value != '=') {
1448 value++;
1450 if (*value != '=') { /* no "=" in the property */
1451 value = name;
1452 break;
1454 name_len = value - name;
1455 value_len = 0;
1457 * skip over the "="
1459 value++;
1460 while (value[value_len] && !ISSPACE(value[value_len])) {
1461 ++value_len;
1464 * build property name with "xpv-" prefix
1466 if (name_len + 4 > 32) { /* skip if name too long */
1467 value += value_len;
1468 continue;
1470 bcopy(name, &namebuf[4], name_len);
1471 name_len += 4;
1472 namebuf[name_len] = 0;
1473 bcopy(value, propbuf, value_len);
1474 propbuf[value_len] = 0;
1475 bsetprops(namebuf, propbuf);
1478 * xpv-root is set to the logical disk name of the xen
1479 * VBD when booting from a disk-based filesystem.
1481 if (strcmp(namebuf, "xpv-root") == 0)
1482 xen_vbdroot_props(propbuf);
1484 * While we're here, if we have a "xpv-nfsroot" property
1485 * then we need to set "fstype" to "nfs" so we mount
1486 * our root from the nfs server. Also parse the xpv-nfsroot
1487 * property to create the properties that nfs_mountroot will
1488 * need to find the root and mount it.
1490 if (strcmp(namebuf, "xpv-nfsroot") == 0)
1491 xen_nfsroot_props(propbuf);
1493 if (strcmp(namebuf, "xpv-ip") == 0)
1494 xen_ip_props(propbuf);
1495 value += value_len;
1497 #endif
1499 while (ISSPACE(*value))
1500 ++value;
1502 * value now points at the boot-file
1504 value_len = 0;
1505 while (value[value_len] && !ISSPACE(value[value_len]))
1506 ++value_len;
1507 if (value_len > 0) {
1508 whoami = propbuf;
1509 bcopy(value, whoami, value_len);
1510 whoami[value_len] = 0;
1511 bsetprops("boot-file", whoami);
1513 * strip leading path stuff from whoami, so running from
1514 * PXE/miniroot makes sense.
1516 if (strstr(whoami, "/platform/") != NULL)
1517 whoami = strstr(whoami, "/platform/");
1518 bsetprops("whoami", whoami);
1522 * Values forcibly set boot properties on the command line via -B.
1523 * Allow use of quotes in values. Other stuff goes on kernel
1524 * command line.
1526 name = value + value_len;
1527 while (*name != 0) {
1529 * anything not " -B" is copied to the command line
1531 if (!ISSPACE(name[0]) || name[1] != '-' || name[2] != 'B') {
1532 boot_args[boot_arg_len++] = *name;
1533 boot_args[boot_arg_len] = 0;
1534 ++name;
1535 continue;
1539 * skip the " -B" and following white space
1541 name += 3;
1542 while (ISSPACE(*name))
1543 ++name;
1544 while (*name && !ISSPACE(*name)) {
1545 value = strstr(name, "=");
1546 if (value == NULL)
1547 break;
1548 name_len = value - name;
1549 ++value;
1550 value_len = 0;
1551 quoted = 0;
1552 for (; ; ++value_len) {
1553 if (!value[value_len])
1554 break;
1557 * is this value quoted?
1559 if (value_len == 0 &&
1560 (value[0] == '\'' || value[0] == '"')) {
1561 quoted = value[0];
1562 ++value_len;
1566 * In the quote accept any character,
1567 * but look for ending quote.
1569 if (quoted) {
1570 if (value[value_len] == quoted)
1571 quoted = 0;
1572 continue;
1576 * a comma or white space ends the value
1578 if (value[value_len] == ',' ||
1579 ISSPACE(value[value_len]))
1580 break;
1583 if (value_len == 0) {
1584 bsetprop(name, name_len, "true", 5);
1585 } else {
1586 char *v = value;
1587 int l = value_len;
1588 if (v[0] == v[l - 1] &&
1589 (v[0] == '\'' || v[0] == '"')) {
1590 ++v;
1591 l -= 2;
1593 bcopy(v, propbuf, l);
1594 propbuf[l] = '\0';
1595 bsetprop(name, name_len, propbuf,
1596 l + 1);
1598 name = value + value_len;
1599 while (*name == ',')
1600 ++name;
1605 * set boot-args property
1606 * 1275 name is bootargs, so set
1607 * that too
1609 bsetprops("boot-args", boot_args);
1610 bsetprops("bootargs", boot_args);
1612 process_boot_environment(benv);
1614 #ifndef __xpv
1616 * Build boot command line for Fast Reboot
1618 build_fastboot_cmdline(xbp);
1620 if (xbp->bi_mb_version == 1) {
1621 multiboot_info_t *mbi = xbp->bi_mb_info;
1622 int netboot;
1623 struct sol_netinfo *sip;
1626 * set the BIOS boot device from GRUB
1628 netboot = 0;
1631 * Save various boot information for Fast Reboot
1633 save_boot_info(xbp);
1635 if (mbi != NULL && mbi->flags & MB_INFO_BOOTDEV) {
1636 boot_device = mbi->boot_device >> 24;
1637 if (boot_device == 0x20)
1638 netboot++;
1639 str[0] = (boot_device >> 4) + '0';
1640 str[1] = (boot_device & 0xf) + '0';
1641 str[2] = 0;
1642 bsetprops("bios-boot-device", str);
1643 } else {
1644 netboot = 1;
1648 * In the netboot case, drives_info is overloaded with the
1649 * dhcp ack. This is not multiboot compliant and requires
1650 * special pxegrub!
1652 if (netboot && mbi->drives_length != 0) {
1653 sip = (struct sol_netinfo *)(uintptr_t)mbi->drives_addr;
1654 if (sip->sn_infotype == SN_TYPE_BOOTP)
1655 bsetprop("bootp-response",
1656 sizeof ("bootp-response"),
1657 (void *)(uintptr_t)mbi->drives_addr,
1658 mbi->drives_length);
1659 else if (sip->sn_infotype == SN_TYPE_RARP)
1660 setup_rarp_props(sip);
1662 } else {
1663 multiboot2_info_header_t *mbi = xbp->bi_mb_info;
1664 multiboot_tag_bootdev_t *bootdev = NULL;
1665 multiboot_tag_network_t *netdev = NULL;
1667 if (mbi != NULL) {
1668 bootdev = dboot_multiboot2_find_tag(mbi,
1669 MULTIBOOT_TAG_TYPE_BOOTDEV);
1670 netdev = dboot_multiboot2_find_tag(mbi,
1671 MULTIBOOT_TAG_TYPE_NETWORK);
1673 if (bootdev != NULL) {
1674 DBG(bootdev->mb_biosdev);
1675 boot_device = bootdev->mb_biosdev;
1676 str[0] = (boot_device >> 4) + '0';
1677 str[1] = (boot_device & 0xf) + '0';
1678 str[2] = 0;
1679 bsetprops("bios-boot-device", str);
1681 if (netdev != NULL) {
1682 bsetprop("bootp-response", sizeof ("bootp-response"),
1683 (void *)(uintptr_t)netdev->mb_dhcpack,
1684 netdev->mb_size -
1685 sizeof (multiboot_tag_network_t));
1689 bsetprop("stdout", strlen("stdout"),
1690 &stdout_val, sizeof (stdout_val));
1691 #endif /* __xpv */
1694 * more conjured up values for made up things....
1696 #if defined(__xpv)
1697 bsetprops("mfg-name", "i86xpv");
1698 bsetprops("impl-arch-name", "i86xpv");
1699 #else
1700 bsetprops("mfg-name", "i86pc");
1701 bsetprops("impl-arch-name", "i86pc");
1702 #endif
1705 * Build firmware-provided system properties
1707 build_firmware_properties(xbp);
1710 * XXPV
1712 * Find out what these are:
1713 * - cpuid_feature_ecx_include
1714 * - cpuid_feature_ecx_exclude
1715 * - cpuid_feature_edx_include
1716 * - cpuid_feature_edx_exclude
1718 * Find out what these are in multiboot:
1719 * - netdev-path
1720 * - fstype
1724 #ifdef __xpv
1726 * Under the Hypervisor, memory usable for DMA may be scarce. One
1727 * very likely large pool of DMA friendly memory is occupied by
1728 * the boot_archive, as it was loaded by grub into low MFNs.
1730 * Here we free up that memory by copying the boot archive to what are
1731 * likely higher MFN pages and then swapping the mfn/pfn mappings.
1733 #define PFN_2GIG 0x80000
1734 static void
1735 relocate_boot_archive(struct xboot_info *xbp)
1737 mfn_t max_mfn = HYPERVISOR_memory_op(XENMEM_maximum_ram_page, NULL);
1738 struct boot_modules *bm = xbp->bi_modules;
1739 uintptr_t va;
1740 pfn_t va_pfn;
1741 mfn_t va_mfn;
1742 caddr_t copy;
1743 pfn_t copy_pfn;
1744 mfn_t copy_mfn;
1745 size_t len;
1746 int slop;
1747 int total = 0;
1748 int relocated = 0;
1749 int mmu_update_return;
1750 mmu_update_t t[2];
1751 x86pte_t pte;
1754 * If all MFN's are below 2Gig, don't bother doing this.
1756 if (max_mfn < PFN_2GIG)
1757 return;
1758 if (xbp->bi_module_cnt < 1) {
1759 DBG_MSG("no boot_archive!");
1760 return;
1763 DBG_MSG("moving boot_archive to high MFN memory\n");
1764 va = (uintptr_t)bm->bm_addr;
1765 len = bm->bm_size;
1766 slop = va & MMU_PAGEOFFSET;
1767 if (slop) {
1768 va += MMU_PAGESIZE - slop;
1769 len -= MMU_PAGESIZE - slop;
1771 len = P2ALIGN(len, MMU_PAGESIZE);
1774 * Go through all boot_archive pages, swapping any low MFN pages
1775 * with memory at next_phys.
1777 while (len != 0) {
1778 ++total;
1779 va_pfn = mmu_btop(va - ONE_GIG);
1780 va_mfn = mfn_list[va_pfn];
1781 if (mfn_list[va_pfn] < PFN_2GIG) {
1782 copy = kbm_remap_window(next_phys, 1);
1783 bcopy((void *)va, copy, MMU_PAGESIZE);
1784 copy_pfn = mmu_btop(next_phys);
1785 copy_mfn = mfn_list[copy_pfn];
1787 pte = mfn_to_ma(copy_mfn) | PT_NOCONSIST | PT_VALID;
1788 if (HYPERVISOR_update_va_mapping(va, pte,
1789 UVMF_INVLPG | UVMF_LOCAL))
1790 bop_panic("relocate_boot_archive(): "
1791 "HYPERVISOR_update_va_mapping() failed");
1793 mfn_list[va_pfn] = copy_mfn;
1794 mfn_list[copy_pfn] = va_mfn;
1796 t[0].ptr = mfn_to_ma(copy_mfn) | MMU_MACHPHYS_UPDATE;
1797 t[0].val = va_pfn;
1798 t[1].ptr = mfn_to_ma(va_mfn) | MMU_MACHPHYS_UPDATE;
1799 t[1].val = copy_pfn;
1800 if (HYPERVISOR_mmu_update(t, 2, &mmu_update_return,
1801 DOMID_SELF) != 0 || mmu_update_return != 2)
1802 bop_panic("relocate_boot_archive(): "
1803 "HYPERVISOR_mmu_update() failed");
1805 next_phys += MMU_PAGESIZE;
1806 ++relocated;
1808 len -= MMU_PAGESIZE;
1809 va += MMU_PAGESIZE;
1811 DBG_MSG("Relocated pages:\n");
1812 DBG(relocated);
1813 DBG_MSG("Out of total pages:\n");
1814 DBG(total);
1816 #endif /* __xpv */
1818 #if !defined(__xpv)
1820 * Install a temporary IDT that lets us catch errors in the boot time code.
1821 * We shouldn't get any faults at all while this is installed, so we'll
1822 * just generate a traceback and exit.
1824 #ifdef __amd64
1825 static const int bcode_sel = B64CODE_SEL;
1826 #else
1827 static const int bcode_sel = B32CODE_SEL;
1828 #endif
1831 * simple description of a stack frame (args are 32 bit only currently)
1833 typedef struct bop_frame {
1834 struct bop_frame *old_frame;
1835 pc_t retaddr;
1836 long arg[1];
1837 } bop_frame_t;
1839 void
1840 bop_traceback(bop_frame_t *frame)
1842 pc_t pc;
1843 int cnt;
1844 char *ksym;
1845 ulong_t off;
1846 #if defined(__i386)
1847 int a;
1848 #endif
1850 bop_printf(NULL, "Stack traceback:\n");
1851 for (cnt = 0; cnt < 30; ++cnt) { /* up to 30 frames */
1852 pc = frame->retaddr;
1853 if (pc == 0)
1854 break;
1855 ksym = kobj_getsymname(pc, &off);
1856 if (ksym)
1857 bop_printf(NULL, " %s+%lx", ksym, off);
1858 else
1859 bop_printf(NULL, " 0x%lx", pc);
1861 frame = frame->old_frame;
1862 if (frame == 0) {
1863 bop_printf(NULL, "\n");
1864 break;
1866 #if defined(__i386)
1867 for (a = 0; a < 6; ++a) { /* try for 6 args */
1868 if ((void *)&frame->arg[a] == (void *)frame->old_frame)
1869 break;
1870 if (a == 0)
1871 bop_printf(NULL, "(");
1872 else
1873 bop_printf(NULL, ",");
1874 bop_printf(NULL, "0x%lx", frame->arg[a]);
1876 bop_printf(NULL, ")");
1877 #endif
1878 bop_printf(NULL, "\n");
1882 struct trapframe {
1883 ulong_t error_code; /* optional */
1884 ulong_t inst_ptr;
1885 ulong_t code_seg;
1886 ulong_t flags_reg;
1887 #ifdef __amd64
1888 ulong_t stk_ptr;
1889 ulong_t stk_seg;
1890 #endif
1893 void
1894 bop_trap(ulong_t *tfp)
1896 struct trapframe *tf = (struct trapframe *)tfp;
1897 bop_frame_t fakeframe;
1898 static int depth = 0;
1901 * Check for an infinite loop of traps.
1903 if (++depth > 2)
1904 bop_panic("Nested trap");
1906 bop_printf(NULL, "Unexpected trap\n");
1909 * adjust the tf for optional error_code by detecting the code selector
1911 if (tf->code_seg != bcode_sel)
1912 tf = (struct trapframe *)(tfp - 1);
1913 else
1914 bop_printf(NULL, "error code 0x%lx\n",
1915 tf->error_code & 0xffffffff);
1917 bop_printf(NULL, "instruction pointer 0x%lx\n", tf->inst_ptr);
1918 bop_printf(NULL, "code segment 0x%lx\n", tf->code_seg & 0xffff);
1919 bop_printf(NULL, "flags register 0x%lx\n", tf->flags_reg);
1920 #ifdef __amd64
1921 bop_printf(NULL, "return %%rsp 0x%lx\n", tf->stk_ptr);
1922 bop_printf(NULL, "return %%ss 0x%lx\n", tf->stk_seg & 0xffff);
1923 #endif
1925 /* grab %[er]bp pushed by our code from the stack */
1926 fakeframe.old_frame = (bop_frame_t *)*(tfp - 3);
1927 fakeframe.retaddr = (pc_t)tf->inst_ptr;
1928 bop_printf(NULL, "Attempting stack backtrace:\n");
1929 bop_traceback(&fakeframe);
1930 bop_panic("unexpected trap in early boot");
1933 extern void bop_trap_handler(void);
1935 static gate_desc_t *bop_idt;
1937 static desctbr_t bop_idt_info;
1939 static void
1940 bop_idt_init(void)
1942 int t;
1944 bop_idt = (gate_desc_t *)
1945 do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
1946 bzero(bop_idt, MMU_PAGESIZE);
1947 for (t = 0; t < NIDT; ++t) {
1949 * Note that since boot runs without a TSS, the
1950 * double fault handler cannot use an alternate stack
1951 * (64-bit) or a task gate (32-bit).
1953 set_gatesegd(&bop_idt[t], &bop_trap_handler, bcode_sel,
1954 SDT_SYSIGT, TRP_KPL, 0);
1956 bop_idt_info.dtr_limit = (NIDT * sizeof (gate_desc_t)) - 1;
1957 bop_idt_info.dtr_base = (uintptr_t)bop_idt;
1958 wr_idtr(&bop_idt_info);
1960 #endif /* !defined(__xpv) */
1963 * This is where we enter the kernel. It dummies up the boot_ops and
1964 * boot_syscalls vectors and jumps off to _kobj_boot()
1966 void
1967 _start(struct xboot_info *xbp)
1969 bootops_t *bops = &bootop;
1970 extern void _kobj_boot();
1973 * 1st off - initialize the console for any error messages
1975 xbootp = xbp;
1976 #ifdef __xpv
1977 HYPERVISOR_shared_info = (void *)xbp->bi_shared_info;
1978 xen_info = xbp->bi_xen_start_info;
1979 #endif
1981 #ifndef __xpv
1982 if (*((uint32_t *)(FASTBOOT_SWTCH_PA + FASTBOOT_STACK_OFFSET)) ==
1983 FASTBOOT_MAGIC) {
1984 post_fastreboot = 1;
1985 *((uint32_t *)(FASTBOOT_SWTCH_PA + FASTBOOT_STACK_OFFSET)) = 0;
1987 #endif
1989 bcons_init(xbp);
1990 have_console = 1;
1993 * enable debugging
1995 if (find_boot_prop("kbm_debug") != NULL)
1996 kbm_debug = 1;
1998 DBG_MSG("\n\n*** Entered Solaris in _start() cmdline is: ");
1999 DBG_MSG((char *)xbp->bi_cmdline);
2000 DBG_MSG("\n\n\n");
2003 * physavail is no longer used by startup
2005 bm.physinstalled = xbp->bi_phys_install;
2006 bm.pcimem = xbp->bi_pcimem;
2007 bm.rsvdmem = xbp->bi_rsvdmem;
2008 bm.physavail = NULL;
2011 * initialize the boot time allocator
2013 next_phys = xbp->bi_next_paddr;
2014 DBG(next_phys);
2015 next_virt = (uintptr_t)xbp->bi_next_vaddr;
2016 DBG(next_virt);
2017 DBG_MSG("Initializing boot time memory management...");
2018 #ifdef __xpv
2020 xen_platform_parameters_t p;
2022 /* This call shouldn't fail, dboot already did it once. */
2023 (void) HYPERVISOR_xen_version(XENVER_platform_parameters, &p);
2024 mfn_to_pfn_mapping = (pfn_t *)(xen_virt_start = p.virt_start);
2025 DBG(xen_virt_start);
2027 #endif
2028 kbm_init(xbp);
2029 DBG_MSG("done\n");
2032 * Fill in the bootops vector
2034 bops->bsys_version = BO_VERSION;
2035 bops->boot_mem = &bm;
2036 bops->bsys_alloc = do_bsys_alloc;
2037 bops->bsys_free = do_bsys_free;
2038 bops->bsys_getproplen = do_bsys_getproplen;
2039 bops->bsys_getprop = do_bsys_getprop;
2040 bops->bsys_nextprop = do_bsys_nextprop;
2041 bops->bsys_printf = bop_printf;
2042 bops->bsys_doint = do_bsys_doint;
2045 * BOP_EALLOC() is no longer needed
2047 bops->bsys_ealloc = do_bsys_ealloc;
2049 #ifdef __xpv
2051 * On domain 0 we need to free up some physical memory that is
2052 * usable for DMA. Since GRUB loaded the boot_archive, it is
2053 * sitting in low MFN memory. We'll relocated the boot archive
2054 * pages to high PFN memory.
2056 if (DOMAIN_IS_INITDOMAIN(xen_info))
2057 relocate_boot_archive(xbp);
2058 #endif
2060 #ifndef __xpv
2062 * Install an IDT to catch early pagefaults (shouldn't have any).
2063 * Also needed for kmdb.
2065 bop_idt_init();
2066 #endif
2069 * Start building the boot properties from the command line
2071 DBG_MSG("Initializing boot properties:\n");
2072 build_boot_properties(xbp);
2074 if (find_boot_prop("prom_debug") || kbm_debug) {
2075 char *value;
2077 value = do_bsys_alloc(NULL, NULL, MMU_PAGESIZE, MMU_PAGESIZE);
2078 boot_prop_display(value);
2082 * jump into krtld...
2084 _kobj_boot(&bop_sysp, NULL, bops, NULL);
2088 /*ARGSUSED*/
2089 static caddr_t
2090 no_more_alloc(bootops_t *bop, caddr_t virthint, size_t size, int align)
2092 panic("Attempt to bsys_alloc() too late\n");
2093 return (NULL);
2096 /*ARGSUSED*/
2097 static void
2098 no_more_free(bootops_t *bop, caddr_t virt, size_t size)
2100 panic("Attempt to bsys_free() too late\n");
2103 void
2104 bop_no_more_mem(void)
2106 DBG(total_bop_alloc_scratch);
2107 DBG(total_bop_alloc_kernel);
2108 bootops->bsys_alloc = no_more_alloc;
2109 bootops->bsys_free = no_more_free;
2114 * Set ACPI firmware properties
2117 static caddr_t
2118 vmap_phys(size_t length, paddr_t pa)
2120 paddr_t start, end;
2121 caddr_t va;
2122 size_t len, page;
2124 #ifdef __xpv
2125 pa = pfn_to_pa(xen_assign_pfn(mmu_btop(pa))) | (pa & MMU_PAGEOFFSET);
2126 #endif
2127 start = P2ALIGN(pa, MMU_PAGESIZE);
2128 end = P2ROUNDUP(pa + length, MMU_PAGESIZE);
2129 len = end - start;
2130 va = (caddr_t)alloc_vaddr(len, MMU_PAGESIZE);
2131 for (page = 0; page < len; page += MMU_PAGESIZE)
2132 kbm_map((uintptr_t)va + page, start + page, 0, 0);
2133 return (va + (pa & MMU_PAGEOFFSET));
2136 static uint8_t
2137 checksum_table(uint8_t *tp, size_t len)
2139 uint8_t sum = 0;
2141 while (len-- > 0)
2142 sum += *tp++;
2144 return (sum);
2147 static int
2148 valid_rsdp(ACPI_TABLE_RSDP *rp)
2151 /* validate the V1.x checksum */
2152 if (checksum_table((uint8_t *)rp, ACPI_RSDP_CHECKSUM_LENGTH) != 0)
2153 return (0);
2155 /* If pre-ACPI 2.0, this is a valid RSDP */
2156 if (rp->Revision < 2)
2157 return (1);
2159 /* validate the V2.x checksum */
2160 if (checksum_table((uint8_t *)rp, ACPI_RSDP_XCHECKSUM_LENGTH) != 0)
2161 return (0);
2163 return (1);
2167 * Scan memory range for an RSDP;
2168 * see ACPI 3.0 Spec, 5.2.5.1
2170 static ACPI_TABLE_RSDP *
2171 scan_rsdp(paddr_t start, paddr_t end)
2173 ssize_t len = end - start;
2174 caddr_t ptr;
2176 ptr = vmap_phys(len, start);
2177 while (len > 0) {
2178 if (strncmp(ptr, ACPI_SIG_RSDP, strlen(ACPI_SIG_RSDP)) == 0 &&
2179 valid_rsdp((ACPI_TABLE_RSDP *)ptr))
2180 return ((ACPI_TABLE_RSDP *)ptr);
2182 ptr += ACPI_RSDP_SCAN_STEP;
2183 len -= ACPI_RSDP_SCAN_STEP;
2186 return (NULL);
2190 * Refer to ACPI 3.0 Spec, section 5.2.5.1 to understand this function
2192 static ACPI_TABLE_RSDP *
2193 find_rsdp()
2195 ACPI_TABLE_RSDP *rsdp;
2196 uint64_t rsdp_val = 0;
2197 uint16_t *ebda_seg;
2198 paddr_t ebda_addr;
2200 /* check for "acpi-root-tab" property */
2201 if (do_bsys_getproplen(NULL, "acpi-root-tab") == sizeof (uint64_t)) {
2202 (void) do_bsys_getprop(NULL, "acpi-root-tab", &rsdp_val);
2203 if (rsdp_val != 0) {
2204 rsdp = scan_rsdp(rsdp_val, rsdp_val + sizeof (*rsdp));
2205 if (rsdp != NULL) {
2206 if (kbm_debug) {
2207 bop_printf(NULL,
2208 "Using RSDP from bootloader: "
2209 "0x%p\n", (void *)rsdp);
2211 return (rsdp);
2217 * Get the EBDA segment and scan the first 1K
2219 ebda_seg = (uint16_t *)vmap_phys(sizeof (uint16_t),
2220 ACPI_EBDA_PTR_LOCATION);
2221 ebda_addr = *ebda_seg << 4;
2222 rsdp = scan_rsdp(ebda_addr, ebda_addr + ACPI_EBDA_WINDOW_SIZE);
2223 if (rsdp == NULL)
2224 /* if EBDA doesn't contain RSDP, look in BIOS memory */
2225 rsdp = scan_rsdp(ACPI_HI_RSDP_WINDOW_BASE,
2226 ACPI_HI_RSDP_WINDOW_BASE + ACPI_HI_RSDP_WINDOW_SIZE);
2227 return (rsdp);
2230 static ACPI_TABLE_HEADER *
2231 map_fw_table(paddr_t table_addr)
2233 ACPI_TABLE_HEADER *tp;
2234 size_t len = MAX(sizeof (*tp), MMU_PAGESIZE);
2237 * Map at least a page; if the table is larger than this, remap it
2239 tp = (ACPI_TABLE_HEADER *)vmap_phys(len, table_addr);
2240 if (tp->Length > len)
2241 tp = (ACPI_TABLE_HEADER *)vmap_phys(tp->Length, table_addr);
2242 return (tp);
2245 static ACPI_TABLE_HEADER *
2246 find_fw_table(char *signature)
2248 static int revision = 0;
2249 static ACPI_TABLE_XSDT *xsdt;
2250 static int len;
2251 paddr_t xsdt_addr;
2252 ACPI_TABLE_RSDP *rsdp;
2253 ACPI_TABLE_HEADER *tp;
2254 paddr_t table_addr;
2255 int n;
2257 if (strlen(signature) != ACPI_NAME_SIZE)
2258 return (NULL);
2261 * Reading the ACPI 3.0 Spec, section 5.2.5.3 will help
2262 * understand this code. If we haven't already found the RSDT/XSDT,
2263 * revision will be 0. Find the RSDP and check the revision
2264 * to find out whether to use the RSDT or XSDT. If revision is
2265 * 0 or 1, use the RSDT and set internal revision to 1; if it is 2,
2266 * use the XSDT. If the XSDT address is 0, though, fall back to
2267 * revision 1 and use the RSDT.
2269 if (revision == 0) {
2270 if ((rsdp = find_rsdp()) != NULL) {
2271 revision = rsdp->Revision;
2273 * ACPI 6.0 states that current revision is 2
2274 * from acpi_table_rsdp definition:
2275 * Must be (0) for ACPI 1.0 or (2) for ACPI 2.0+
2277 if (revision > 2)
2278 revision = 2;
2279 switch (revision) {
2280 case 2:
2282 * Use the XSDT unless BIOS is buggy and
2283 * claims to be rev 2 but has a null XSDT
2284 * address
2286 xsdt_addr = rsdp->XsdtPhysicalAddress;
2287 if (xsdt_addr != 0)
2288 break;
2289 /* FALLTHROUGH */
2290 case 0:
2291 /* treat RSDP rev 0 as revision 1 internally */
2292 revision = 1;
2293 /* FALLTHROUGH */
2294 case 1:
2295 /* use the RSDT for rev 0/1 */
2296 xsdt_addr = rsdp->RsdtPhysicalAddress;
2297 break;
2298 default:
2299 /* unknown revision */
2300 revision = 0;
2301 break;
2304 if (revision == 0)
2305 return (NULL);
2307 /* cache the XSDT info */
2308 xsdt = (ACPI_TABLE_XSDT *)map_fw_table(xsdt_addr);
2309 len = (xsdt->Header.Length - sizeof (xsdt->Header)) /
2310 ((revision == 1) ? sizeof (uint32_t) : sizeof (uint64_t));
2314 * Scan the table headers looking for a signature match
2316 for (n = 0; n < len; n++) {
2317 ACPI_TABLE_RSDT *rsdt = (ACPI_TABLE_RSDT *)xsdt;
2318 table_addr = (revision == 1) ? rsdt->TableOffsetEntry[n] :
2319 xsdt->TableOffsetEntry[n];
2321 if (table_addr == 0)
2322 continue;
2323 tp = map_fw_table(table_addr);
2324 if (strncmp(tp->Signature, signature, ACPI_NAME_SIZE) == 0) {
2325 return (tp);
2328 return (NULL);
2331 static void
2332 process_mcfg(ACPI_TABLE_MCFG *tp)
2334 ACPI_MCFG_ALLOCATION *cfg_baap;
2335 char *cfg_baa_endp;
2336 int64_t ecfginfo[4];
2338 cfg_baap = (ACPI_MCFG_ALLOCATION *)((uintptr_t)tp + sizeof (*tp));
2339 cfg_baa_endp = ((char *)tp) + tp->Header.Length;
2340 while ((char *)cfg_baap < cfg_baa_endp) {
2341 if (cfg_baap->Address != 0 && cfg_baap->PciSegment == 0) {
2342 ecfginfo[0] = cfg_baap->Address;
2343 ecfginfo[1] = cfg_baap->PciSegment;
2344 ecfginfo[2] = cfg_baap->StartBusNumber;
2345 ecfginfo[3] = cfg_baap->EndBusNumber;
2346 bsetprop(MCFG_PROPNAME, strlen(MCFG_PROPNAME),
2347 ecfginfo, sizeof (ecfginfo));
2348 break;
2350 cfg_baap++;
2354 #ifndef __xpv
2355 static void
2356 process_madt_entries(ACPI_TABLE_MADT *tp, uint32_t *cpu_countp,
2357 uint32_t *cpu_possible_countp, uint32_t *cpu_apicid_array)
2359 ACPI_SUBTABLE_HEADER *item, *end;
2360 uint32_t cpu_count = 0;
2361 uint32_t cpu_possible_count = 0;
2364 * Determine number of CPUs and keep track of "final" APIC ID
2365 * for each CPU by walking through ACPI MADT processor list
2367 end = (ACPI_SUBTABLE_HEADER *)(tp->Header.Length + (uintptr_t)tp);
2368 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)tp + sizeof (*tp));
2370 while (item < end) {
2371 switch (item->Type) {
2372 case ACPI_MADT_TYPE_LOCAL_APIC: {
2373 ACPI_MADT_LOCAL_APIC *cpu =
2374 (ACPI_MADT_LOCAL_APIC *) item;
2376 if (cpu->LapicFlags & ACPI_MADT_ENABLED) {
2377 if (cpu_apicid_array != NULL)
2378 cpu_apicid_array[cpu_count] = cpu->Id;
2379 cpu_count++;
2381 cpu_possible_count++;
2382 break;
2384 case ACPI_MADT_TYPE_LOCAL_X2APIC: {
2385 ACPI_MADT_LOCAL_X2APIC *cpu =
2386 (ACPI_MADT_LOCAL_X2APIC *) item;
2388 if (cpu->LapicFlags & ACPI_MADT_ENABLED) {
2389 if (cpu_apicid_array != NULL)
2390 cpu_apicid_array[cpu_count] =
2391 cpu->LocalApicId;
2392 cpu_count++;
2394 cpu_possible_count++;
2395 break;
2397 default:
2398 if (kbm_debug)
2399 bop_printf(NULL, "MADT type %d\n", item->Type);
2400 break;
2403 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)item + item->Length);
2405 if (cpu_countp)
2406 *cpu_countp = cpu_count;
2407 if (cpu_possible_countp)
2408 *cpu_possible_countp = cpu_possible_count;
2411 static void
2412 process_madt(ACPI_TABLE_MADT *tp)
2414 uint32_t cpu_count = 0;
2415 uint32_t cpu_possible_count = 0;
2416 uint32_t *cpu_apicid_array; /* x2APIC ID is 32bit! */
2418 if (tp != NULL) {
2419 /* count cpu's */
2420 process_madt_entries(tp, &cpu_count, &cpu_possible_count, NULL);
2422 cpu_apicid_array = (uint32_t *)do_bsys_alloc(NULL, NULL,
2423 cpu_count * sizeof (*cpu_apicid_array), MMU_PAGESIZE);
2424 if (cpu_apicid_array == NULL)
2425 bop_panic("Not enough memory for APIC ID array");
2427 /* copy IDs */
2428 process_madt_entries(tp, NULL, NULL, cpu_apicid_array);
2431 * Make boot property for array of "final" APIC IDs for each
2432 * CPU
2434 bsetprop(BP_CPU_APICID_ARRAY, strlen(BP_CPU_APICID_ARRAY),
2435 cpu_apicid_array, cpu_count * sizeof (*cpu_apicid_array));
2439 * Check whether property plat-max-ncpus is already set.
2441 if (do_bsys_getproplen(NULL, PLAT_MAX_NCPUS_NAME) < 0) {
2443 * Set plat-max-ncpus to number of maximum possible CPUs given
2444 * in MADT if it hasn't been set.
2445 * There's no formal way to detect max possible CPUs supported
2446 * by platform according to ACPI spec3.0b. So current CPU
2447 * hotplug implementation expects that all possible CPUs will
2448 * have an entry in MADT table and set plat-max-ncpus to number
2449 * of entries in MADT.
2450 * With introducing of ACPI4.0, Maximum System Capability Table
2451 * (MSCT) provides maximum number of CPUs supported by platform.
2452 * If MSCT is unavailable, fall back to old way.
2454 if (tp != NULL)
2455 bsetpropsi(PLAT_MAX_NCPUS_NAME, cpu_possible_count);
2459 * Set boot property boot-max-ncpus to number of CPUs existing at
2460 * boot time. boot-max-ncpus is mainly used for optimization.
2462 if (tp != NULL)
2463 bsetpropsi(BOOT_MAX_NCPUS_NAME, cpu_count);
2466 * User-set boot-ncpus overrides firmware count
2468 if (do_bsys_getproplen(NULL, BOOT_NCPUS_NAME) >= 0)
2469 return;
2472 * Set boot property boot-ncpus to number of active CPUs given in MADT
2473 * if it hasn't been set yet.
2475 if (tp != NULL)
2476 bsetpropsi(BOOT_NCPUS_NAME, cpu_count);
2479 static void
2480 process_srat(ACPI_TABLE_SRAT *tp)
2482 ACPI_SUBTABLE_HEADER *item, *end;
2483 int i;
2484 int proc_num, mem_num;
2485 #pragma pack(1)
2486 struct {
2487 uint32_t domain;
2488 uint32_t apic_id;
2489 uint32_t sapic_id;
2490 } processor;
2491 struct {
2492 uint32_t domain;
2493 uint32_t x2apic_id;
2494 } x2apic;
2495 struct {
2496 uint32_t domain;
2497 uint64_t addr;
2498 uint64_t length;
2499 uint32_t flags;
2500 } memory;
2501 #pragma pack()
2502 char prop_name[30];
2503 uint64_t maxmem = 0;
2505 if (tp == NULL)
2506 return;
2508 proc_num = mem_num = 0;
2509 end = (ACPI_SUBTABLE_HEADER *)(tp->Header.Length + (uintptr_t)tp);
2510 item = (ACPI_SUBTABLE_HEADER *)((uintptr_t)tp + sizeof (*tp));
2511 while (item < end) {
2512 switch (item->Type) {
2513 case ACPI_SRAT_TYPE_CPU_AFFINITY: {
2514 ACPI_SRAT_CPU_AFFINITY *cpu =
2515 (ACPI_SRAT_CPU_AFFINITY *) item;
2517 if (!(cpu->Flags & ACPI_SRAT_CPU_ENABLED))
2518 break;
2519 processor.domain = cpu->ProximityDomainLo;
2520 for (i = 0; i < 3; i++)
2521 processor.domain +=
2522 cpu->ProximityDomainHi[i] << ((i + 1) * 8);
2523 processor.apic_id = cpu->ApicId;
2524 processor.sapic_id = cpu->LocalSapicEid;
2525 (void) snprintf(prop_name, 30, "acpi-srat-processor-%d",
2526 proc_num);
2527 bsetprop(prop_name, strlen(prop_name), &processor,
2528 sizeof (processor));
2529 proc_num++;
2530 break;
2532 case ACPI_SRAT_TYPE_MEMORY_AFFINITY: {
2533 ACPI_SRAT_MEM_AFFINITY *mem =
2534 (ACPI_SRAT_MEM_AFFINITY *)item;
2536 if (!(mem->Flags & ACPI_SRAT_MEM_ENABLED))
2537 break;
2538 memory.domain = mem->ProximityDomain;
2539 memory.addr = mem->BaseAddress;
2540 memory.length = mem->Length;
2541 memory.flags = mem->Flags;
2542 (void) snprintf(prop_name, 30, "acpi-srat-memory-%d",
2543 mem_num);
2544 bsetprop(prop_name, strlen(prop_name), &memory,
2545 sizeof (memory));
2546 if ((mem->Flags & ACPI_SRAT_MEM_HOT_PLUGGABLE) &&
2547 (memory.addr + memory.length > maxmem)) {
2548 maxmem = memory.addr + memory.length;
2550 mem_num++;
2551 break;
2553 case ACPI_SRAT_TYPE_X2APIC_CPU_AFFINITY: {
2554 ACPI_SRAT_X2APIC_CPU_AFFINITY *x2cpu =
2555 (ACPI_SRAT_X2APIC_CPU_AFFINITY *) item;
2557 if (!(x2cpu->Flags & ACPI_SRAT_CPU_ENABLED))
2558 break;
2559 x2apic.domain = x2cpu->ProximityDomain;
2560 x2apic.x2apic_id = x2cpu->ApicId;
2561 (void) snprintf(prop_name, 30, "acpi-srat-processor-%d",
2562 proc_num);
2563 bsetprop(prop_name, strlen(prop_name), &x2apic,
2564 sizeof (x2apic));
2565 proc_num++;
2566 break;
2568 default:
2569 if (kbm_debug)
2570 bop_printf(NULL, "SRAT type %d\n", item->Type);
2571 break;
2574 item = (ACPI_SUBTABLE_HEADER *)
2575 (item->Length + (uintptr_t)item);
2579 * The maximum physical address calculated from the SRAT table is more
2580 * accurate than that calculated from the MSCT table.
2582 if (maxmem != 0) {
2583 plat_dr_physmax = btop(maxmem);
2587 static void
2588 process_slit(ACPI_TABLE_SLIT *tp)
2592 * Check the number of localities; if it's too huge, we just
2593 * return and locality enumeration code will handle this later,
2594 * if possible.
2596 * Note that the size of the table is the square of the
2597 * number of localities; if the number of localities exceeds
2598 * UINT16_MAX, the table size may overflow an int when being
2599 * passed to bsetprop() below.
2601 if (tp->LocalityCount >= SLIT_LOCALITIES_MAX)
2602 return;
2604 bsetprop(SLIT_NUM_PROPNAME, strlen(SLIT_NUM_PROPNAME),
2605 &tp->LocalityCount, sizeof (tp->LocalityCount));
2606 bsetprop(SLIT_PROPNAME, strlen(SLIT_PROPNAME), &tp->Entry,
2607 tp->LocalityCount * tp->LocalityCount);
2610 static ACPI_TABLE_MSCT *
2611 process_msct(ACPI_TABLE_MSCT *tp)
2613 int last_seen = 0;
2614 int proc_num = 0;
2615 ACPI_MSCT_PROXIMITY *item, *end;
2616 extern uint64_t plat_dr_options;
2618 ASSERT(tp != NULL);
2620 end = (ACPI_MSCT_PROXIMITY *)(tp->Header.Length + (uintptr_t)tp);
2621 for (item = (void *)((uintptr_t)tp + tp->ProximityOffset);
2622 item < end;
2623 item = (void *)(item->Length + (uintptr_t)item)) {
2625 * Sanity check according to section 5.2.19.1 of ACPI 4.0.
2626 * Revision 1
2627 * Length 22
2629 if (item->Revision != 1 || item->Length != 22) {
2630 cmn_err(CE_CONT,
2631 "?boot: unknown proximity domain structure in MSCT "
2632 "with Revision(%d), Length(%d).\n",
2633 (int)item->Revision, (int)item->Length);
2634 return (NULL);
2635 } else if (item->RangeStart > item->RangeEnd) {
2636 cmn_err(CE_CONT,
2637 "?boot: invalid proximity domain structure in MSCT "
2638 "with RangeStart(%u), RangeEnd(%u).\n",
2639 item->RangeStart, item->RangeEnd);
2640 return (NULL);
2641 } else if (item->RangeStart != last_seen) {
2643 * Items must be organized in ascending order of the
2644 * proximity domain enumerations.
2646 cmn_err(CE_CONT,
2647 "?boot: invalid proximity domain structure in MSCT,"
2648 " items are not orginized in ascending order.\n");
2649 return (NULL);
2653 * If ProcessorCapacity is 0 then there would be no CPUs in this
2654 * domain.
2656 if (item->ProcessorCapacity != 0) {
2657 proc_num += (item->RangeEnd - item->RangeStart + 1) *
2658 item->ProcessorCapacity;
2661 last_seen = item->RangeEnd - item->RangeStart + 1;
2663 * Break out if all proximity domains have been processed.
2664 * Some BIOSes may have unused items at the end of MSCT table.
2666 if (last_seen > tp->MaxProximityDomains) {
2667 break;
2670 if (last_seen != tp->MaxProximityDomains + 1) {
2671 cmn_err(CE_CONT,
2672 "?boot: invalid proximity domain structure in MSCT, "
2673 "proximity domain count doesn't match.\n");
2674 return (NULL);
2678 * Set plat-max-ncpus property if it hasn't been set yet.
2680 if (do_bsys_getproplen(NULL, PLAT_MAX_NCPUS_NAME) < 0) {
2681 if (proc_num != 0) {
2682 bsetpropsi(PLAT_MAX_NCPUS_NAME, proc_num);
2687 * Use Maximum Physical Address from the MSCT table as upper limit for
2688 * memory hot-adding by default. It may be overridden by value from
2689 * the SRAT table or the "plat-dr-physmax" boot option.
2691 plat_dr_physmax = btop(tp->MaxAddress + 1);
2694 * Existence of MSCT implies CPU/memory hotplug-capability for the
2695 * platform.
2697 plat_dr_options |= PLAT_DR_FEATURE_CPU;
2698 plat_dr_options |= PLAT_DR_FEATURE_MEMORY;
2700 return (tp);
2703 #else /* __xpv */
2704 static void
2705 enumerate_xen_cpus()
2707 processorid_t id, max_id;
2710 * User-set boot-ncpus overrides enumeration
2712 if (do_bsys_getproplen(NULL, BOOT_NCPUS_NAME) >= 0)
2713 return;
2716 * Probe every possible virtual CPU id and remember the
2717 * highest id present; the count of CPUs is one greater
2718 * than this. This tacitly assumes at least cpu 0 is present.
2720 max_id = 0;
2721 for (id = 0; id < MAX_VIRT_CPUS; id++)
2722 if (HYPERVISOR_vcpu_op(VCPUOP_is_up, id, NULL) == 0)
2723 max_id = id;
2725 bsetpropsi(BOOT_NCPUS_NAME, max_id+1);
2728 #endif /* __xpv */
2730 /*ARGSUSED*/
2731 static void
2732 build_firmware_properties(struct xboot_info *xbp)
2734 ACPI_TABLE_HEADER *tp = NULL;
2736 #ifndef __xpv
2737 if (xbp->bi_acpi_rsdp != NULL) {
2738 bsetprop64("acpi-root-tab",
2739 (uint64_t)(uintptr_t)xbp->bi_acpi_rsdp);
2742 if (xbp->bi_smbios != NULL) {
2743 bsetprop64("smbios-address",
2744 (uint64_t)(uintptr_t)xbp->bi_smbios);
2747 if ((tp = find_fw_table(ACPI_SIG_MSCT)) != NULL)
2748 msct_ptr = process_msct((ACPI_TABLE_MSCT *)tp);
2749 else
2750 msct_ptr = NULL;
2752 if ((tp = find_fw_table(ACPI_SIG_MADT)) != NULL)
2753 process_madt((ACPI_TABLE_MADT *)tp);
2755 if ((srat_ptr = (ACPI_TABLE_SRAT *)
2756 find_fw_table(ACPI_SIG_SRAT)) != NULL)
2757 process_srat(srat_ptr);
2759 if (slit_ptr = (ACPI_TABLE_SLIT *)find_fw_table(ACPI_SIG_SLIT))
2760 process_slit(slit_ptr);
2762 tp = find_fw_table(ACPI_SIG_MCFG);
2763 #else /* __xpv */
2764 enumerate_xen_cpus();
2765 if (DOMAIN_IS_INITDOMAIN(xen_info))
2766 tp = find_fw_table(ACPI_SIG_MCFG);
2767 #endif /* __xpv */
2768 if (tp != NULL)
2769 process_mcfg((ACPI_TABLE_MCFG *)tp);
2773 * fake up a boot property for deferred early console output
2774 * this is used by both graphical boot and the (developer only)
2775 * USB serial console
2777 void *
2778 defcons_init(size_t size)
2780 static char *p = NULL;
2782 p = do_bsys_alloc(NULL, NULL, size, MMU_PAGESIZE);
2783 *p = 0;
2784 bsetprop("deferred-console-buf", strlen("deferred-console-buf") + 1,
2785 &p, sizeof (p));
2786 return (p);
2789 /*ARGSUSED*/
2791 boot_compinfo(int fd, struct compinfo *cbp)
2793 cbp->iscmp = 0;
2794 cbp->blksize = MAXBSIZE;
2795 return (0);
2798 #define BP_MAX_STRLEN 32
2801 * Get value for given boot property
2804 bootprop_getval(const char *prop_name, u_longlong_t *prop_value)
2806 int boot_prop_len;
2807 char str[BP_MAX_STRLEN];
2808 u_longlong_t value;
2810 boot_prop_len = BOP_GETPROPLEN(bootops, prop_name);
2811 if (boot_prop_len < 0 || boot_prop_len > sizeof (str) ||
2812 BOP_GETPROP(bootops, prop_name, str) < 0 ||
2813 kobj_getvalue(str, &value) == -1)
2814 return (-1);
2816 if (prop_value)
2817 *prop_value = value;
2819 return (0);