Annotation of src/sys/arch/i386/i386/machdep.c, Revision 1.586
1.586 ! christos 1: /* $NetBSD: machdep.c,v 1.585 2006/10/25 13:56:15 jmmv Exp $ */
1.231 thorpej 2:
3: /*-
1.571 jmmv 4: * Copyright (c) 1996, 1997, 1998, 2000, 2004, 2006 The NetBSD Foundation, Inc.
1.231 thorpej 5: * All rights reserved.
6: *
7: * This code is derived from software contributed to The NetBSD Foundation
1.571 jmmv 8: * by Charles M. Hannum, by Jason R. Thorpe of the Numerical Aerospace
9: * Simulation Facility, NASA Ames Research Center and by Julio M. Merino Vidal.
1.231 thorpej 10: *
11: * Redistribution and use in source and binary forms, with or without
12: * modification, are permitted provided that the following conditions
13: * are met:
14: * 1. Redistributions of source code must retain the above copyright
15: * notice, this list of conditions and the following disclaimer.
16: * 2. Redistributions in binary form must reproduce the above copyright
17: * notice, this list of conditions and the following disclaimer in the
18: * documentation and/or other materials provided with the distribution.
19: * 3. All advertising materials mentioning features or use of this software
20: * must display the following acknowledgement:
21: * This product includes software developed by the NetBSD
22: * Foundation, Inc. and its contributors.
23: * 4. Neither the name of The NetBSD Foundation nor the names of its
24: * contributors may be used to endorse or promote products derived
25: * from this software without specific prior written permission.
26: *
27: * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28: * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29: * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30: * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31: * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32: * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33: * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34: * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35: * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36: * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37: * POSSIBILITY OF SUCH DAMAGE.
38: */
1.125 cgd 39:
1.1 cgd 40: /*-
41: * Copyright (c) 1982, 1987, 1990 The Regents of the University of California.
42: * All rights reserved.
43: *
44: * This code is derived from software contributed to Berkeley by
45: * William Jolitz.
46: *
47: * Redistribution and use in source and binary forms, with or without
48: * modification, are permitted provided that the following conditions
49: * are met:
50: * 1. Redistributions of source code must retain the above copyright
51: * notice, this list of conditions and the following disclaimer.
52: * 2. Redistributions in binary form must reproduce the above copyright
53: * notice, this list of conditions and the following disclaimer in the
54: * documentation and/or other materials provided with the distribution.
1.528 agc 55: * 3. Neither the name of the University nor the names of its contributors
1.1 cgd 56: * may be used to endorse or promote products derived from this software
57: * without specific prior written permission.
58: *
59: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69: * SUCH DAMAGE.
70: *
1.125 cgd 71: * @(#)machdep.c 7.4 (Berkeley) 6/3/91
1.1 cgd 72: */
1.460 lukem 73:
74: #include <sys/cdefs.h>
1.586 ! christos 75: __KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.585 2006/10/25 13:56:15 jmmv Exp $");
1.271 thorpej 76:
1.538 christos 77: #include "opt_beep.h"
78: #include "opt_compat_ibcs2.h"
79: #include "opt_compat_mach.h" /* need to get the right segment def */
80: #include "opt_compat_netbsd.h"
81: #include "opt_compat_svr4.h"
82: #include "opt_cpureset_delay.h"
1.271 thorpej 83: #include "opt_cputype.h"
1.309 jonathan 84: #include "opt_ddb.h"
1.377 ws 85: #include "opt_ipkdb.h"
1.443 lukem 86: #include "opt_kgdb.h"
1.455 fvdl 87: #include "opt_mtrr.h"
1.523 martin 88: #include "opt_multiprocessor.h"
1.538 christos 89: #include "opt_realmem.h"
90: #include "opt_user_ldt.h"
91: #include "opt_vm86.h"
1.1 cgd 92:
1.59 mycroft 93: #include <sys/param.h>
94: #include <sys/systm.h>
1.508 thorpej 95: #include <sys/signal.h>
1.59 mycroft 96: #include <sys/signalvar.h>
97: #include <sys/kernel.h>
98: #include <sys/proc.h>
99: #include <sys/user.h>
100: #include <sys/exec.h>
101: #include <sys/buf.h>
102: #include <sys/reboot.h>
103: #include <sys/conf.h>
104: #include <sys/file.h>
105: #include <sys/malloc.h>
106: #include <sys/mbuf.h>
107: #include <sys/msgbuf.h>
108: #include <sys/mount.h>
109: #include <sys/vnode.h>
1.204 thorpej 110: #include <sys/extent.h>
1.123 cgd 111: #include <sys/syscallargs.h>
1.291 thorpej 112: #include <sys/core.h>
113: #include <sys/kcore.h>
1.508 thorpej 114: #include <sys/ucontext.h>
1.519 nathanw 115: #include <sys/ras.h>
1.508 thorpej 116: #include <sys/sa.h>
117: #include <sys/savar.h>
1.522 ragge 118: #include <sys/ksyms.h>
1.57 cgd 119:
1.377 ws 120: #ifdef IPKDB
121: #include <ipkdb/ipkdb.h>
122: #endif
123:
1.235 thorpej 124: #ifdef KGDB
125: #include <sys/kgdb.h>
126: #endif
127:
1.104 cgd 128: #include <dev/cons.h>
1.390 mrg 129:
130: #include <uvm/uvm_extern.h>
1.393 fvdl 131: #include <uvm/uvm_page.h>
1.284 mrg 132:
1.200 christos 133: #include <sys/sysctl.h>
134:
1.59 mycroft 135: #include <machine/cpu.h>
136: #include <machine/cpufunc.h>
1.484 fvdl 137: #include <machine/cpuvar.h>
1.178 mycroft 138: #include <machine/gdt.h>
1.574 kardel 139: #include <machine/kcore.h>
1.149 mycroft 140: #include <machine/pio.h>
1.59 mycroft 141: #include <machine/psl.h>
142: #include <machine/reg.h>
143: #include <machine/specialreg.h>
1.255 drochner 144: #include <machine/bootinfo.h>
1.455 fvdl 145: #include <machine/mtrr.h>
1.574 kardel 146: #include <x86/x86/tsc.h>
147:
1.570 jmmv 148: #include <machine/multiboot.h>
1.43 brezak 149:
1.146 cgd 150: #include <dev/isa/isareg.h>
1.372 drochner 151: #include <machine/isa_machdep.h>
1.164 cgd 152: #include <dev/ic/i8042reg.h>
1.43 brezak 153:
1.200 christos 154: #ifdef DDB
155: #include <machine/db_machdep.h>
156: #include <ddb/db_extern.h>
157: #endif
158:
1.184 mycroft 159: #ifdef VM86
160: #include <machine/vm86.h>
161: #endif
162:
1.473 tshiozak 163: #include "acpi.h"
1.576 christos 164: #include "apmbios.h"
1.258 jtk 165: #include "bioscall.h"
1.207 jtk 166:
1.259 jtk 167: #if NBIOSCALL > 0
168: #include <machine/bioscall.h>
169: #endif
170:
1.473 tshiozak 171: #if NACPI > 0
172: #include <dev/acpi/acpivar.h>
173: #define ACPI_MACHDEP_PRIVATE
174: #include <machine/acpi_machdep.h>
175: #endif
176:
1.576 christos 177: #if NAPMBIOS > 0
1.207 jtk 178: #include <machine/apmvar.h>
1.258 jtk 179: #endif
180:
1.59 mycroft 181: #include "isa.h"
1.231 thorpej 182: #include "isadma.h"
1.59 mycroft 183: #include "npx.h"
1.522 ragge 184: #include "ksyms.h"
1.2 cgd 185:
1.384 jdolecek 186: #include "mca.h"
187: #if NMCA > 0
188: #include <machine/mca_machdep.h> /* for mca_busprobe() */
189: #endif
190:
1.484 fvdl 191: #ifdef MULTIPROCESSOR /* XXX */
192: #include <machine/mpbiosvar.h> /* XXX */
193: #endif /* XXX */
194:
1.517 jmmv 195: #ifndef BEEP_ONHALT_COUNT
196: #define BEEP_ONHALT_COUNT 3
197: #endif
198: #ifndef BEEP_ONHALT_PITCH
199: #define BEEP_ONHALT_PITCH 1500
200: #endif
201: #ifndef BEEP_ONHALT_PERIOD
202: #define BEEP_ONHALT_PERIOD 250
203: #endif
204:
1.104 cgd 205: /* the following is used externally (sysctl_hw) */
1.549 wiz 206: char machine[] = "i386"; /* CPU "architecture" */
1.232 veego 207: char machine_arch[] = "i386"; /* machine == machine_arch */
1.104 cgd 208:
1.559 thorpej 209: extern struct bi_devmatch *x86_alldisks;
210: extern int x86_ndisks;
1.342 fvdl 211:
1.328 bouyer 212: #ifdef CPURESET_DELAY
213: int cpureset_delay = CPURESET_DELAY;
214: #else
215: int cpureset_delay = 2000; /* default to 2s */
216: #endif
217:
1.455 fvdl 218: #ifdef MTRR
1.513 fvdl 219: struct mtrr_funcs *mtrr_funcs;
1.455 fvdl 220: #endif
221:
1.515 fvdl 222: #ifdef COMPAT_NOMID
1.566 christos 223: static int exec_nomid(struct lwp *, struct exec_package *);
1.550 junyoung 224: #endif
1.515 fvdl 225:
1.59 mycroft 226: int physmem;
1.163 cgd 227: int dumpmem_low;
228: int dumpmem_high;
1.524 drochner 229: unsigned int cpu_feature;
1.553 lukem 230: unsigned int cpu_feature2;
1.59 mycroft 231: int cpu_class;
1.428 fvdl 232: int i386_fpu_present;
233: int i386_fpu_exception;
234: int i386_fpu_fdivbug;
1.59 mycroft 235:
1.451 thorpej 236: int i386_use_fxsave;
237: int i386_has_sse;
238: int i386_has_sse2;
1.450 thorpej 239:
1.461 christos 240: int tmx86_has_longrun;
241:
1.314 thorpej 242: vaddr_t msgbuf_vaddr;
243: paddr_t msgbuf_paddr;
244:
245: vaddr_t idt_vaddr;
246: paddr_t idt_paddr;
247:
1.264 mycroft 248: #ifdef I586_CPU
1.314 thorpej 249: vaddr_t pentium_idt_vaddr;
1.264 mycroft 250: #endif
1.59 mycroft 251:
1.444 chs 252: struct vm_map *exec_map = NULL;
253: struct vm_map *mb_map = NULL;
254: struct vm_map *phys_map = NULL;
1.48 brezak 255:
1.314 thorpej 256: extern paddr_t avail_start, avail_end;
1.1 cgd 257:
1.542 junyoung 258: void (*delay_func)(int) = i8254_delay;
259: void (*initclock_func)(void) = i8254_initclocks;
1.484 fvdl 260:
1.204 thorpej 261: /*
1.291 thorpej 262: * Size of memory segments, before any memory is stolen.
263: */
264: phys_ram_seg_t mem_clusters[VM_PHYSSEG_MAX];
265: int mem_cluster_cnt;
266:
1.542 junyoung 267: int cpu_dump(void);
268: int cpu_dumpsize(void);
269: u_long cpu_dump_mempagecnt(void);
270: void dumpsys(void);
271: void init386(paddr_t);
272: void initgdt(union descriptor *);
1.255 drochner 273:
1.568 perry 274: void add_mem_cluster(uint64_t, uint64_t, uint32_t);
1.433 kanaoka 275:
1.521 dsl 276: extern int time_adjusted;
1.484 fvdl 277:
1.571 jmmv 278: struct bootinfo bootinfo;
279: int *esym;
280: extern int boothowto;
281:
282: /* Base memory reported by BIOS. */
283: #ifndef REALBASEMEM
284: int biosbasemem = 0;
285: #else
286: int biosbasemem = REALBASEMEM;
287: #endif
288:
289: /* Extended memory reported by BIOS. */
290: #ifndef REALEXTMEM
291: int biosextmem = 0;
292: #else
293: int biosextmem = REALEXTMEM;
294: #endif
295:
1.577 mrg 296: /* Set if any boot-loader set biosbasemem/biosextmem. */
297: int biosmem_implicit;
298:
1.571 jmmv 299: /* Representation of the bootinfo structure constructed by a NetBSD native
300: * boot loader. Only be used by native_loader(). */
301: struct bootinfo_source {
302: uint32_t bs_naddrs;
303: paddr_t bs_addrs[1]; /* Actually longer. */
304: };
305:
306: /* Only called by locore.h; no need to be in a header file. */
307: void native_loader(int, int, struct bootinfo_source *, paddr_t, int, int);
308:
309: /*
310: * Called as one of the very first things during system startup (just after
311: * the boot loader gave control to the kernel image), this routine is in
312: * charge of retrieving the parameters passed in by the boot loader and
313: * storing them in the appropriate kernel variables.
314: *
315: * WARNING: Because the kernel has not yet relocated itself to KERNBASE,
316: * special care has to be taken when accessing memory because absolute
317: * addresses (referring to kernel symbols) do not work. So:
318: *
319: * 1) Avoid jumps to absolute addresses (such as gotos and switches).
320: * 2) To access global variables use their physical address, which
321: * can be obtained using the RELOC macro.
322: */
323: void
1.586 ! christos 324: native_loader(int bl_boothowto, int bl_bootdev,
1.571 jmmv 325: struct bootinfo_source *bl_bootinfo, paddr_t bl_esym,
1.572 jmmv 326: int bl_biosextmem, int bl_biosbasemem)
1.571 jmmv 327: {
328: #define RELOC(type, x) ((type)((vaddr_t)(x) - KERNBASE))
329:
330: *RELOC(int *, &boothowto) = bl_boothowto;
331:
332: #ifdef COMPAT_OLDBOOT
333: /*
334: * Pre-1.3 boot loaders gave the boot device as a parameter
335: * (instead of a bootinfo entry).
336: */
337: *RELOC(int *, &bootdev) = bl_bootdev;
338: #endif
339:
340: /*
341: * The boot loader provides a physical, non-relocated address
342: * for the symbols table's end. We need to convert it to a
343: * virtual address.
344: */
345: if (bl_esym != 0)
346: *RELOC(int **, &esym) = (int *)((vaddr_t)bl_esym + KERNBASE);
347: else
348: *RELOC(int **, &esym) = 0;
349:
350: /*
351: * Copy bootinfo entries (if any) from the boot loader's
352: * representation to the kernel's bootinfo space.
353: */
354: if (bl_bootinfo != NULL) {
355: size_t i;
356: uint8_t *data;
357: struct bootinfo *bidest;
358:
359: bidest = RELOC(struct bootinfo *, &bootinfo);
360:
361: data = &bidest->bi_data[0];
362:
363: for (i = 0; i < bl_bootinfo->bs_naddrs; i++) {
364: struct btinfo_common *bc;
365:
366: bc = (struct btinfo_common *)(bl_bootinfo->bs_addrs[i]);
367:
368: if ((paddr_t)(data + bc->len) >
369: (paddr_t)(&bidest->bi_data[0] + BOOTINFO_MAXSIZE))
370: break;
371:
372: memcpy(data, bc, bc->len);
373: data += bc->len;
374: }
375: bidest->bi_nentries = i;
376: }
377:
378: /*
379: * Configure biosbasemem and biosextmem only if they were not
380: * explicitly given during the kernel's build.
381: */
1.577 mrg 382: if (*RELOC(int *, &biosbasemem) == 0) {
1.571 jmmv 383: *RELOC(int *, &biosbasemem) = bl_biosbasemem;
1.577 mrg 384: *RELOC(int *, &biosmem_implicit) = 1;
385: }
386: if (*RELOC(int *, &biosextmem) == 0) {
1.571 jmmv 387: *RELOC(int *, &biosextmem) = bl_biosextmem;
1.577 mrg 388: *RELOC(int *, &biosmem_implicit) = 1;
389: }
1.571 jmmv 390: #undef RELOC
391: }
392:
1.59 mycroft 393: /*
394: * Machine-dependent startup code
395: */
1.32 andrew 396: void
1.1 cgd 397: cpu_startup()
398: {
1.547 pk 399: int x;
1.314 thorpej 400: vaddr_t minaddr, maxaddr;
1.354 lukem 401: char pbuf[9];
1.1 cgd 402:
1.284 mrg 403: /*
404: * Initialize error message buffer (et end of core).
405: */
1.561 yamt 406: msgbuf_vaddr = uvm_km_alloc(kernel_map, x86_round_page(MSGBUFSIZE), 0,
407: UVM_KMF_VAONLY);
1.385 thorpej 408: if (msgbuf_vaddr == 0)
1.284 mrg 409: panic("failed to valloc msgbuf_vaddr");
1.359 thorpej 410:
1.284 mrg 411: /* msgbuf_paddr was init'd in pmap */
412: for (x = 0; x < btoc(MSGBUFSIZE); x++)
1.414 thorpej 413: pmap_kenter_pa((vaddr_t)msgbuf_vaddr + x * PAGE_SIZE,
414: msgbuf_paddr + x * PAGE_SIZE, VM_PROT_READ|VM_PROT_WRITE);
1.456 chris 415: pmap_update(pmap_kernel());
1.359 thorpej 416:
1.284 mrg 417: initmsgbuf((caddr_t)msgbuf_vaddr, round_page(MSGBUFSIZE));
418:
1.562 lukem 419: printf("%s%s", copyright, version);
1.484 fvdl 420:
1.570 jmmv 421: #ifdef MULTIBOOT
422: multiboot_print_info();
423: #endif
424:
1.484 fvdl 425: #ifdef TRAPLOG
426: /*
427: * Enable recording of branch from/to in MSR's
428: */
429: wrmsr(MSR_DEBUGCTLMSR, 0x1);
430: #endif
431:
1.382 mycroft 432: format_bytes(pbuf, sizeof(pbuf), ptoa(physmem));
1.354 lukem 433: printf("total memory = %s\n", pbuf);
1.1 cgd 434:
1.547 pk 435: minaddr = 0;
1.41 cgd 436:
1.1 cgd 437: /*
1.36 cgd 438: * Allocate a submap for exec arguments. This map effectively
439: * limits the number of processes exec'ing at any time.
1.1 cgd 440: */
1.284 mrg 441: exec_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 442: 16*NCARGS, VM_MAP_PAGEABLE, FALSE, NULL);
1.59 mycroft 443:
1.1 cgd 444: /*
445: * Allocate a submap for physio
446: */
1.284 mrg 447: phys_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 448: VM_PHYS_SIZE, 0, FALSE, NULL);
1.1 cgd 449:
450: /*
1.229 thorpej 451: * Finally, allocate mbuf cluster submap.
1.1 cgd 452: */
1.334 thorpej 453: mb_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 454: nmbclusters * mclbytes, VM_MAP_INTRSAFE, FALSE, NULL);
1.1 cgd 455:
1.547 pk 456: format_bytes(pbuf, sizeof(pbuf), ptoa(uvmexp.free));
1.354 lukem 457: printf("avail memory = %s\n", pbuf);
1.1 cgd 458:
1.375 drochner 459: /* Safe for i/o port / memory space allocation to use malloc now. */
1.514 fvdl 460: x86_bus_space_mallocok();
1.349 thorpej 461: }
462:
463: /*
464: * Set up proc0's TSS and LDT.
465: */
466: void
467: i386_proc0_tss_ldt_init()
468: {
469: struct pcb *pcb;
470: int x;
1.268 thorpej 471:
1.326 thorpej 472: gdt_init();
1.484 fvdl 473:
1.508 thorpej 474: cpu_info_primary.ci_curpcb = pcb = &lwp0.l_addr->u_pcb;
1.484 fvdl 475:
1.326 thorpej 476: pcb->pcb_tss.tss_ioopt =
477: ((caddr_t)pcb->pcb_iomap - (caddr_t)&pcb->pcb_tss) << 16;
1.484 fvdl 478:
1.326 thorpej 479: for (x = 0; x < sizeof(pcb->pcb_iomap) / 4; x++)
480: pcb->pcb_iomap[x] = 0xffffffff;
481:
1.394 thorpej 482: pcb->pcb_ldt_sel = pmap_kernel()->pm_ldt_sel = GSEL(GLDT_SEL, SEL_KPL);
1.326 thorpej 483: pcb->pcb_cr0 = rcr0();
484: pcb->pcb_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
1.573 yamt 485: pcb->pcb_tss.tss_esp0 = USER_TO_UAREA(lwp0.l_addr) + KSTACK_SIZE - 16;
1.508 thorpej 486: lwp0.l_md.md_regs = (struct trapframe *)pcb->pcb_tss.tss_esp0 - 1;
487: lwp0.l_md.md_tss_sel = tss_alloc(pcb);
1.326 thorpej 488:
1.508 thorpej 489: ltr(lwp0.l_md.md_tss_sel);
1.326 thorpej 490: lldt(pcb->pcb_ldt_sel);
1.484 fvdl 491: }
492:
493: /*
494: * Set up TSS and LDT for a new PCB.
495: */
496:
497: void
1.551 junyoung 498: i386_init_pcb_tss_ldt(struct cpu_info *ci)
1.484 fvdl 499: {
500: int x;
501: struct pcb *pcb = ci->ci_idle_pcb;
1.326 thorpej 502:
1.484 fvdl 503: pcb->pcb_tss.tss_ioopt =
504: ((caddr_t)pcb->pcb_iomap - (caddr_t)&pcb->pcb_tss) << 16;
505: for (x = 0; x < sizeof(pcb->pcb_iomap) / 4; x++)
506: pcb->pcb_iomap[x] = 0xffffffff;
507:
508: pcb->pcb_ldt_sel = pmap_kernel()->pm_ldt_sel = GSEL(GLDT_SEL, SEL_KPL);
509: pcb->pcb_cr0 = rcr0();
510:
511: ci->ci_idle_tss_sel = tss_alloc(pcb);
1.326 thorpej 512: }
513:
514: /*
1.545 atatat 515: * sysctl helper routine for machdep.tm* nodes.
1.484 fvdl 516: */
1.545 atatat 517: static int
518: sysctl_machdep_tm_longrun(SYSCTLFN_ARGS)
1.104 cgd 519: {
1.545 atatat 520: struct sysctlnode node;
521: int io, error;
522:
523: if (!tmx86_has_longrun)
524: return (EOPNOTSUPP);
525:
526: node = *rnode;
527: node.sysctl_data = &io;
1.104 cgd 528:
1.545 atatat 529: switch (rnode->sysctl_num) {
1.461 christos 530: case CPU_TMLR_MODE:
1.545 atatat 531: io = (int)(crusoe_longrun = tmx86_get_longrun_mode());
532: break;
1.461 christos 533: case CPU_TMLR_FREQUENCY:
534: tmx86_get_longrun_status_all();
1.545 atatat 535: io = crusoe_frequency;
536: break;
1.461 christos 537: case CPU_TMLR_VOLTAGE:
538: tmx86_get_longrun_status_all();
1.545 atatat 539: io = crusoe_voltage;
540: break;
1.461 christos 541: case CPU_TMLR_PERCENTAGE:
542: tmx86_get_longrun_status_all();
1.545 atatat 543: io = crusoe_percentage;
544: break;
1.104 cgd 545: default:
546: return (EOPNOTSUPP);
547: }
1.545 atatat 548:
549: error = sysctl_lookup(SYSCTLFN_CALL(&node));
550: if (error || newp == NULL)
551: return (error);
552:
553: if (rnode->sysctl_num == CPU_TMLR_MODE) {
554: if (tmx86_set_longrun_mode(io))
555: crusoe_longrun = (u_int)io;
556: else
557: return (EINVAL);
558: }
559:
560: return (0);
561: }
562:
563: /*
564: * sysctl helper routine for machdep.booted_kernel
565: */
566: static int
567: sysctl_machdep_booted_kernel(SYSCTLFN_ARGS)
568: {
569: struct btinfo_bootpath *bibp;
570: struct sysctlnode node;
571:
572: bibp = lookup_bootinfo(BTINFO_BOOTPATH);
573: if(!bibp)
574: return(ENOENT); /* ??? */
575:
576: node = *rnode;
577: node.sysctl_data = bibp->bootpath;
578: node.sysctl_size = sizeof(bibp->bootpath);
579: return (sysctl_lookup(SYSCTLFN_CALL(&node)));
580: }
581:
582: /*
583: * sysctl helper routine for machdep.diskinfo
584: */
585: static int
586: sysctl_machdep_diskinfo(SYSCTLFN_ARGS)
587: {
588: struct sysctlnode node;
589:
590: node = *rnode;
1.559 thorpej 591: if (x86_alldisks == NULL)
1.557 dbj 592: return(EOPNOTSUPP);
1.559 thorpej 593: node.sysctl_data = x86_alldisks;
1.545 atatat 594: node.sysctl_size = sizeof(struct disklist) +
1.559 thorpej 595: (x86_ndisks - 1) * sizeof(struct nativedisk_info);
1.557 dbj 596: return (sysctl_lookup(SYSCTLFN_CALL(&node)));
1.545 atatat 597: }
598:
599: /*
600: * machine dependent system variables.
601: */
602: SYSCTL_SETUP(sysctl_machdep_setup, "sysctl machdep subtree setup")
603: {
604:
1.552 atatat 605: sysctl_createv(clog, 0, NULL, NULL,
606: CTLFLAG_PERMANENT,
1.545 atatat 607: CTLTYPE_NODE, "machdep", NULL,
608: NULL, 0, NULL, 0,
609: CTL_MACHDEP, CTL_EOL);
610:
1.552 atatat 611: sysctl_createv(clog, 0, NULL, NULL,
612: CTLFLAG_PERMANENT,
1.545 atatat 613: CTLTYPE_STRUCT, "console_device", NULL,
614: sysctl_consdev, 0, NULL, sizeof(dev_t),
615: CTL_MACHDEP, CPU_CONSDEV, CTL_EOL);
1.552 atatat 616: sysctl_createv(clog, 0, NULL, NULL,
617: CTLFLAG_PERMANENT,
1.545 atatat 618: CTLTYPE_INT, "biosbasemem", NULL,
619: NULL, 0, &biosbasemem, 0,
620: CTL_MACHDEP, CPU_BIOSBASEMEM, CTL_EOL);
1.552 atatat 621: sysctl_createv(clog, 0, NULL, NULL,
622: CTLFLAG_PERMANENT,
1.545 atatat 623: CTLTYPE_INT, "biosextmem", NULL,
624: NULL, 0, &biosextmem, 0,
625: CTL_MACHDEP, CPU_BIOSEXTMEM, CTL_EOL);
1.552 atatat 626: sysctl_createv(clog, 0, NULL, NULL,
627: CTLFLAG_PERMANENT,
1.545 atatat 628: CTLTYPE_INT, "nkpde", NULL,
629: NULL, 0, &nkpde, 0,
630: CTL_MACHDEP, CPU_NKPDE, CTL_EOL);
1.552 atatat 631: sysctl_createv(clog, 0, NULL, NULL,
632: CTLFLAG_PERMANENT,
1.545 atatat 633: CTLTYPE_STRING, "booted_kernel", NULL,
634: sysctl_machdep_booted_kernel, 0, NULL, 0,
635: CTL_MACHDEP, CPU_BOOTED_KERNEL, CTL_EOL);
1.552 atatat 636: sysctl_createv(clog, 0, NULL, NULL,
637: CTLFLAG_PERMANENT,
1.545 atatat 638: CTLTYPE_STRUCT, "diskinfo", NULL,
639: sysctl_machdep_diskinfo, 0, NULL, 0,
640: CTL_MACHDEP, CPU_DISKINFO, CTL_EOL);
1.552 atatat 641: sysctl_createv(clog, 0, NULL, NULL,
642: CTLFLAG_PERMANENT,
1.545 atatat 643: CTLTYPE_INT, "fpu_present", NULL,
644: NULL, 0, &i386_fpu_present, 0,
645: CTL_MACHDEP, CPU_FPU_PRESENT, CTL_EOL);
1.552 atatat 646: sysctl_createv(clog, 0, NULL, NULL,
647: CTLFLAG_PERMANENT,
1.545 atatat 648: CTLTYPE_INT, "osfxsr", NULL,
649: NULL, 0, &i386_use_fxsave, 0,
650: CTL_MACHDEP, CPU_OSFXSR, CTL_EOL);
1.552 atatat 651: sysctl_createv(clog, 0, NULL, NULL,
652: CTLFLAG_PERMANENT,
1.545 atatat 653: CTLTYPE_INT, "sse", NULL,
654: NULL, 0, &i386_has_sse, 0,
655: CTL_MACHDEP, CPU_SSE, CTL_EOL);
1.552 atatat 656: sysctl_createv(clog, 0, NULL, NULL,
657: CTLFLAG_PERMANENT,
1.545 atatat 658: CTLTYPE_INT, "sse2", NULL,
659: NULL, 0, &i386_has_sse2, 0,
660: CTL_MACHDEP, CPU_SSE2, CTL_EOL);
1.554 lukem 661: sysctl_createv(clog, 0, NULL, NULL,
662: CTLFLAG_PERMANENT,
663: CTLTYPE_STRING, "cpu_brand", NULL,
664: NULL, 0, &cpu_brand_string, 0,
665: CTL_MACHDEP, CTL_CREATE, CTL_EOL);
1.552 atatat 666: sysctl_createv(clog, 0, NULL, NULL,
667: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.545 atatat 668: CTLTYPE_INT, "tm_longrun_mode", NULL,
669: sysctl_machdep_tm_longrun, 0, NULL, 0,
670: CTL_MACHDEP, CPU_TMLR_MODE, CTL_EOL);
1.552 atatat 671: sysctl_createv(clog, 0, NULL, NULL,
672: CTLFLAG_PERMANENT,
1.545 atatat 673: CTLTYPE_INT, "tm_longrun_frequency", NULL,
674: sysctl_machdep_tm_longrun, 0, NULL, 0,
675: CTL_MACHDEP, CPU_TMLR_FREQUENCY, CTL_EOL);
1.552 atatat 676: sysctl_createv(clog, 0, NULL, NULL,
677: CTLFLAG_PERMANENT,
1.545 atatat 678: CTLTYPE_INT, "tm_longrun_voltage", NULL,
679: sysctl_machdep_tm_longrun, 0, NULL, 0,
680: CTL_MACHDEP, CPU_TMLR_VOLTAGE, CTL_EOL);
1.552 atatat 681: sysctl_createv(clog, 0, NULL, NULL,
682: CTLFLAG_PERMANENT,
1.545 atatat 683: CTLTYPE_INT, "tm_longrun_percentage", NULL,
684: sysctl_machdep_tm_longrun, 0, NULL, 0,
685: CTL_MACHDEP, CPU_TMLR_PERCENTAGE, CTL_EOL);
1.104 cgd 686: }
1.151 christos 687:
1.534 christos 688: void *
689: getframe(struct lwp *l, int sig, int *onstack)
690: {
691: struct proc *p = l->l_proc;
692: struct sigctx *ctx = &p->p_sigctx;
693: struct trapframe *tf = l->l_md.md_regs;
694:
695: /* Do we need to jump onto the signal stack? */
696: *onstack = (ctx->ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0
697: && (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0;
698: if (*onstack)
699: return (char *)ctx->ps_sigstk.ss_sp + ctx->ps_sigstk.ss_size;
700: #ifdef VM86
701: if (tf->tf_eflags & PSL_VM)
702: return (void *)(tf->tf_esp + (tf->tf_ss << 4));
703: else
704: #endif
705: return (void *)tf->tf_esp;
706: }
707:
708: /*
709: * Build context to run handler in. We invoke the handler
710: * directly, only returning via the trampoline. Note the
711: * trampoline version numbers are coordinated with machine-
712: * dependent code in libc.
713: */
714: void
715: buildcontext(struct lwp *l, int sel, void *catcher, void *fp)
716: {
717: struct trapframe *tf = l->l_md.md_regs;
718:
719: tf->tf_gs = GSEL(GUDATA_SEL, SEL_UPL);
720: tf->tf_fs = GSEL(GUDATA_SEL, SEL_UPL);
721: tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
722: tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
723: tf->tf_eip = (int)catcher;
724: tf->tf_cs = GSEL(sel, SEL_UPL);
725: tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
726: tf->tf_esp = (int)fp;
727: tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
728: }
729:
730: static void
1.539 christos 731: sendsig_siginfo(const ksiginfo_t *ksi, const sigset_t *mask)
1.534 christos 732: {
733: struct lwp *l = curlwp;
734: struct proc *p = l->l_proc;
735: struct pmap *pmap = vm_map_pmap(&p->p_vmspace->vm_map);
736: int sel = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
737: GUCODEBIG_SEL : GUCODE_SEL;
738: struct sigacts *ps = p->p_sigacts;
739: int onstack;
740: int sig = ksi->ksi_signo;
741: struct sigframe_siginfo *fp = getframe(l, sig, &onstack), frame;
742: sig_t catcher = SIGACTION(p, sig).sa_handler;
743: struct trapframe *tf = l->l_md.md_regs;
744:
745: fp--;
746:
747: /* Build stack frame for signal trampoline. */
748: switch (ps->sa_sigdesc[sig].sd_vers) {
1.537 christos 749: case 0: /* handled by sendsig_sigcontext */
750: case 1: /* handled by sendsig_sigcontext */
1.534 christos 751: default: /* unknown version */
752: printf("nsendsig: bad version %d\n",
753: ps->sa_sigdesc[sig].sd_vers);
754: sigexit(l, SIGILL);
755: case 2:
756: break;
757: }
758:
759: frame.sf_ra = (int)ps->sa_sigdesc[sig].sd_tramp;
760: frame.sf_signum = sig;
761: frame.sf_sip = &fp->sf_si;
762: frame.sf_ucp = &fp->sf_uc;
1.540 thorpej 763: frame.sf_si._info = ksi->ksi_info;
1.534 christos 764: frame.sf_uc.uc_flags = _UC_SIGMASK|_UC_VM;
765: frame.sf_uc.uc_sigmask = *mask;
766: frame.sf_uc.uc_link = NULL;
767: frame.sf_uc.uc_flags |= (p->p_sigctx.ps_sigstk.ss_flags & SS_ONSTACK)
768: ? _UC_SETSTACK : _UC_CLRSTACK;
769: memset(&frame.sf_uc.uc_stack, 0, sizeof(frame.sf_uc.uc_stack));
770: cpu_getmcontext(l, &frame.sf_uc.uc_mcontext, &frame.sf_uc.uc_flags);
771:
772: if (tf->tf_eflags & PSL_VM)
773: (*p->p_emul->e_syscall_intern)(p);
774:
775: if (copyout(&frame, fp, sizeof(frame)) != 0) {
776: /*
777: * Process has trashed its stack; give it an illegal
778: * instruction to halt it in its tracks.
779: */
780: sigexit(l, SIGILL);
781: /* NOTREACHED */
782: }
783:
784: buildcontext(l, sel, catcher, fp);
1.319 mycroft 785:
786: /* Remember that we're now on the signal stack. */
787: if (onstack)
1.425 jdolecek 788: p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
1.1 cgd 789: }
790:
1.534 christos 791: void
1.539 christos 792: sendsig(const ksiginfo_t *ksi, const sigset_t *mask)
1.534 christos 793: {
794: #ifdef COMPAT_16
795: if (curproc->p_sigacts->sa_sigdesc[ksi->ksi_signo].sd_vers < 2)
796: sendsig_sigcontext(ksi, mask);
797: else
798: #endif
799: sendsig_siginfo(ksi, mask);
800: }
1.508 thorpej 801:
1.550 junyoung 802: void
1.534 christos 803: cpu_upcall(struct lwp *l, int type, int nevents, int ninterrupted, void *sas,
804: void *ap, void *sp, sa_upcall_t upcall)
1.508 thorpej 805: {
1.533 chs 806: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 807: struct saframe *sf, frame;
808: struct trapframe *tf;
809:
810: tf = l->l_md.md_regs;
811:
812: /* Finally, copy out the rest of the frame. */
813: frame.sa_type = type;
814: frame.sa_sas = sas;
815: frame.sa_events = nevents;
816: frame.sa_interrupted = ninterrupted;
817: frame.sa_arg = ap;
818: frame.sa_ra = 0;
1.550 junyoung 819:
1.508 thorpej 820: sf = (struct saframe *)sp - 1;
821: if (copyout(&frame, sf, sizeof(frame)) != 0) {
822: /* Copying onto the stack didn't work. Die. */
823: sigexit(l, SIGILL);
824: /* NOTREACHED */
825: }
826:
827: tf->tf_eip = (int) upcall;
828: tf->tf_esp = (int) sf;
829: tf->tf_ebp = 0; /* indicate call-frame-top to debuggers */
830: tf->tf_gs = GSEL(GUDATA_SEL, SEL_UPL);
831: tf->tf_fs = GSEL(GUDATA_SEL, SEL_UPL);
832: tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
833: tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
1.533 chs 834: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
835: GSEL(GUCODEBIG_SEL, SEL_UPL) : GSEL(GUCODE_SEL, SEL_UPL);
1.508 thorpej 836: tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
837: tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
838: }
1.37 cgd 839:
1.1 cgd 840: int waittime = -1;
841: struct pcb dumppcb;
842:
1.32 andrew 843: void
1.586 ! christos 844: cpu_reboot(int howto, char *bootstr)
1.1 cgd 845: {
846:
1.106 mycroft 847: if (cold) {
1.193 mycroft 848: howto |= RB_HALT;
849: goto haltsys;
1.1 cgd 850: }
1.193 mycroft 851:
1.106 mycroft 852: boothowto = howto;
1.193 mycroft 853: if ((howto & RB_NOSYNC) == 0 && waittime < 0) {
1.1 cgd 854: waittime = 0;
1.150 mycroft 855: vfs_shutdown();
1.521 dsl 856: /*
857: * If we've been adjusting the clock, the todr
858: * will be out of synch; adjust it now.
859: */
860: if (time_adjusted != 0)
861: resettodr();
1.1 cgd 862: }
1.193 mycroft 863:
864: /* Disable interrupts. */
1.1 cgd 865: splhigh();
1.193 mycroft 866:
867: /* Do a dump if requested. */
868: if ((howto & (RB_DUMP | RB_HALT)) == RB_DUMP)
869: dumpsys();
870:
871: haltsys:
872: doshutdownhooks();
873:
1.484 fvdl 874: #ifdef MULTIPROCESSOR
1.514 fvdl 875: x86_broadcast_ipi(X86_IPI_HALT);
1.484 fvdl 876: #endif
877:
1.307 thorpej 878: if ((howto & RB_POWERDOWN) == RB_POWERDOWN) {
1.473 tshiozak 879: #if NACPI > 0
1.530 kochi 880: if (acpi_softc != NULL) {
881: delay(500000);
882: acpi_enter_sleep_state(acpi_softc, ACPI_STATE_S5);
883: printf("WARNING: ACPI powerdown failed!\n");
884: }
1.473 tshiozak 885: #endif
1.576 christos 886: #if NAPMBIOS > 0 && !defined(APM_NO_POWEROFF)
1.208 jtk 887: /* turn off, if we can. But try to turn disk off and
888: * wait a bit first--some disk drives are slow to clean up
889: * and users have reported disk corruption.
890: */
891: delay(500000);
1.578 xtraeme 892: apm_set_powstate(NULL, APM_DEV_DISK(APM_DEV_ALLUNITS), APM_SYS_OFF);
1.208 jtk 893: delay(500000);
1.576 christos 894: apm_set_powstate(NULL, APM_DEV_ALLDEVS, APM_SYS_OFF);
1.531 kochi 895: printf("WARNING: APM powerdown failed!\n");
1.307 thorpej 896: /*
897: * RB_POWERDOWN implies RB_HALT... fall into it...
898: */
1.208 jtk 899: #endif
1.307 thorpej 900: }
901:
902: if (howto & RB_HALT) {
1.210 christos 903: printf("\n");
904: printf("The operating system has halted.\n");
905: printf("Please press any key to reboot.\n\n");
1.517 jmmv 906:
907: #ifdef BEEP_ONHALT
908: {
909: int c;
910: for (c = BEEP_ONHALT_COUNT; c > 0; c--) {
911: sysbeep(BEEP_ONHALT_PITCH,
1.565 simonb 912: BEEP_ONHALT_PERIOD * hz / 1000);
1.517 jmmv 913: delay(BEEP_ONHALT_PERIOD * 1000);
914: sysbeep(0, BEEP_ONHALT_PERIOD * hz / 1000);
915: delay(BEEP_ONHALT_PERIOD * 1000);
916: }
917: }
918: #endif
919:
1.300 drochner 920: cnpollc(1); /* for proper keyboard command handling */
1.512 perry 921: if (cngetc() == 0) {
922: /* no console attached, so just hlt */
923: for(;;) {
1.567 perry 924: __asm volatile("hlt");
1.512 perry 925: }
926: }
1.300 drochner 927: cnpollc(0);
1.1 cgd 928: }
1.193 mycroft 929:
1.210 christos 930: printf("rebooting...\n");
1.328 bouyer 931: if (cpureset_delay > 0)
932: delay(cpureset_delay * 1000);
1.1 cgd 933: cpu_reset();
934: for(;;) ;
935: /*NOTREACHED*/
936: }
937:
1.116 gwr 938: /*
939: * These variables are needed by /sbin/savecore
940: */
1.568 perry 941: uint32_t dumpmag = 0x8fca0101; /* magic number */
1.116 gwr 942: int dumpsize = 0; /* pages */
943: long dumplo = 0; /* blocks */
944:
945: /*
1.291 thorpej 946: * cpu_dumpsize: calculate size of machine-dependent kernel core dump headers.
947: */
948: int
949: cpu_dumpsize()
950: {
951: int size;
952:
953: size = ALIGN(sizeof(kcore_seg_t)) + ALIGN(sizeof(cpu_kcore_hdr_t)) +
954: ALIGN(mem_cluster_cnt * sizeof(phys_ram_seg_t));
955: if (roundup(size, dbtob(1)) != dbtob(1))
956: return (-1);
957:
958: return (1);
959: }
960:
961: /*
962: * cpu_dump_mempagecnt: calculate the size of RAM (in pages) to be dumped.
963: */
964: u_long
965: cpu_dump_mempagecnt()
966: {
967: u_long i, n;
968:
969: n = 0;
970: for (i = 0; i < mem_cluster_cnt; i++)
971: n += atop(mem_clusters[i].size);
972: return (n);
973: }
974:
975: /*
976: * cpu_dump: dump the machine-dependent kernel core dump headers.
977: */
978: int
979: cpu_dump()
980: {
1.542 junyoung 981: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.563 christos 982: char bf[dbtob(1)];
1.291 thorpej 983: kcore_seg_t *segp;
984: cpu_kcore_hdr_t *cpuhdrp;
985: phys_ram_seg_t *memsegp;
1.481 gehenna 986: const struct bdevsw *bdev;
1.291 thorpej 987: int i;
988:
1.481 gehenna 989: bdev = bdevsw_lookup(dumpdev);
990: if (bdev == NULL)
991: return (ENXIO);
992: dump = bdev->d_dump;
1.291 thorpej 993:
1.563 christos 994: memset(bf, 0, sizeof bf);
995: segp = (kcore_seg_t *)bf;
996: cpuhdrp = (cpu_kcore_hdr_t *)&bf[ALIGN(sizeof(*segp))];
997: memsegp = (phys_ram_seg_t *)&bf[ ALIGN(sizeof(*segp)) +
1.291 thorpej 998: ALIGN(sizeof(*cpuhdrp))];
999:
1000: /*
1001: * Generate a segment header.
1002: */
1003: CORE_SETMAGIC(*segp, KCORE_MAGIC, MID_MACHINE, CORE_CPU);
1004: segp->c_size = dbtob(1) - ALIGN(sizeof(*segp));
1005:
1006: /*
1007: * Add the machine-dependent header info.
1008: */
1.558 junyoung 1009: cpuhdrp->pdppaddr = PDPpaddr;
1.291 thorpej 1010: cpuhdrp->nmemsegs = mem_cluster_cnt;
1011:
1012: /*
1013: * Fill in the memory segment descriptors.
1014: */
1015: for (i = 0; i < mem_cluster_cnt; i++) {
1016: memsegp[i].start = mem_clusters[i].start;
1017: memsegp[i].size = mem_clusters[i].size;
1018: }
1019:
1.563 christos 1020: return (dump(dumpdev, dumplo, (caddr_t)bf, dbtob(1)));
1.291 thorpej 1021: }
1022:
1023: /*
1.228 gwr 1024: * This is called by main to set dumplo and dumpsize.
1.414 thorpej 1025: * Dumps always skip the first PAGE_SIZE of disk space
1.116 gwr 1026: * in case there might be a disk label stored there.
1027: * If there is extra space, put dump at the end to
1028: * reduce the chance that swapping trashes it.
1029: */
1030: void
1.228 gwr 1031: cpu_dumpconf()
1.116 gwr 1032: {
1.481 gehenna 1033: const struct bdevsw *bdev;
1.291 thorpej 1034: int nblks, dumpblks; /* size of dump area */
1.116 gwr 1035:
1036: if (dumpdev == NODEV)
1.291 thorpej 1037: goto bad;
1.481 gehenna 1038: bdev = bdevsw_lookup(dumpdev);
1.582 mrg 1039: if (bdev == NULL) {
1040: dumpdev = NODEV;
1.583 skrll 1041: goto bad;
1.582 mrg 1042: }
1.481 gehenna 1043: if (bdev->d_psize == NULL)
1.291 thorpej 1044: goto bad;
1.481 gehenna 1045: nblks = (*bdev->d_psize)(dumpdev);
1.116 gwr 1046: if (nblks <= ctod(1))
1.291 thorpej 1047: goto bad;
1.116 gwr 1048:
1.291 thorpej 1049: dumpblks = cpu_dumpsize();
1050: if (dumpblks < 0)
1051: goto bad;
1052: dumpblks += ctod(cpu_dump_mempagecnt());
1053:
1054: /* If dump won't fit (incl. room for possible label), punt. */
1055: if (dumpblks > (nblks - ctod(1)))
1056: goto bad;
1057:
1058: /* Put dump at end of partition */
1059: dumplo = nblks - dumpblks;
1060:
1061: /* dumpsize is in page units, and doesn't include headers. */
1062: dumpsize = cpu_dump_mempagecnt();
1063: return;
1.116 gwr 1064:
1.291 thorpej 1065: bad:
1066: dumpsize = 0;
1.116 gwr 1067: }
1068:
1.1 cgd 1069: /*
1070: * Doadump comes here after turning off memory management and
1071: * getting on the dump stack, either when called above, or by
1072: * the auto-restart code.
1073: */
1.414 thorpej 1074: #define BYTES_PER_DUMP PAGE_SIZE /* must be a multiple of pagesize XXX small */
1.314 thorpej 1075: static vaddr_t dumpspace;
1.163 cgd 1076:
1.314 thorpej 1077: vaddr_t
1.551 junyoung 1078: reserve_dumppages(vaddr_t p)
1.163 cgd 1079: {
1080:
1081: dumpspace = p;
1082: return (p + BYTES_PER_DUMP);
1083: }
1084:
1.32 andrew 1085: void
1.1 cgd 1086: dumpsys()
1087: {
1.556 mycroft 1088: u_long totalbytesleft, bytes, i, n, m, memseg;
1.291 thorpej 1089: u_long maddr;
1090: int psize;
1.163 cgd 1091: daddr_t blkno;
1.481 gehenna 1092: const struct bdevsw *bdev;
1.542 junyoung 1093: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.200 christos 1094: int error;
1.193 mycroft 1095:
1096: /* Save registers. */
1097: savectx(&dumppcb);
1.1 cgd 1098:
1099: if (dumpdev == NODEV)
1100: return;
1.484 fvdl 1101:
1.481 gehenna 1102: bdev = bdevsw_lookup(dumpdev);
1103: if (bdev == NULL || bdev->d_psize == NULL)
1104: return;
1.163 cgd 1105:
1106: /*
1107: * For dumps during autoconfiguration,
1108: * if dump device has already configured...
1109: */
1110: if (dumpsize == 0)
1.228 gwr 1111: cpu_dumpconf();
1.330 jtk 1112: if (dumplo <= 0 || dumpsize == 0) {
1.275 mycroft 1113: printf("\ndump to dev %u,%u not possible\n", major(dumpdev),
1114: minor(dumpdev));
1.163 cgd 1115: return;
1.275 mycroft 1116: }
1117: printf("\ndumping to dev %u,%u offset %ld\n", major(dumpdev),
1118: minor(dumpdev), dumplo);
1.134 mycroft 1119:
1.481 gehenna 1120: psize = (*bdev->d_psize)(dumpdev);
1.210 christos 1121: printf("dump ");
1.163 cgd 1122: if (psize == -1) {
1.210 christos 1123: printf("area unavailable\n");
1.163 cgd 1124: return;
1125: }
1126:
1127: #if 0 /* XXX this doesn't work. grr. */
1.565 simonb 1128: /* toss any characters present prior to dump */
1.163 cgd 1129: while (sget() != NULL); /*syscons and pccons differ */
1130: #endif
1131:
1.291 thorpej 1132: if ((error = cpu_dump()) != 0)
1133: goto err;
1134:
1135: totalbytesleft = ptoa(cpu_dump_mempagecnt());
1136: blkno = dumplo + cpu_dumpsize();
1.481 gehenna 1137: dump = bdev->d_dump;
1.200 christos 1138: error = 0;
1.291 thorpej 1139:
1140: for (memseg = 0; memseg < mem_cluster_cnt; memseg++) {
1141: maddr = mem_clusters[memseg].start;
1142: bytes = mem_clusters[memseg].size;
1143:
1144: for (i = 0; i < bytes; i += n, totalbytesleft -= n) {
1145: /* Print out how many MBs we have left to go. */
1146: if ((totalbytesleft % (1024*1024)) == 0)
1147: printf("%ld ", totalbytesleft / (1024 * 1024));
1148:
1149: /* Limit size for next transfer. */
1150: n = bytes - i;
1151: if (n > BYTES_PER_DUMP)
1152: n = BYTES_PER_DUMP;
1153:
1.556 mycroft 1154: for (m = 0; m < n; m += NBPG)
1155: pmap_kenter_pa(dumpspace + m, maddr + m,
1156: VM_PROT_READ);
1.291 thorpej 1157:
1158: error = (*dump)(dumpdev, blkno, (caddr_t)dumpspace, n);
1159: if (error)
1160: goto err;
1.163 cgd 1161: maddr += n;
1.291 thorpej 1162: blkno += btodb(n); /* XXX? */
1.163 cgd 1163:
1164: #if 0 /* XXX this doesn't work. grr. */
1.291 thorpej 1165: /* operator aborting dump? */
1166: if (sget() != NULL) {
1167: error = EINTR;
1168: break;
1169: }
1170: #endif
1.163 cgd 1171: }
1172: }
1173:
1.291 thorpej 1174: err:
1.163 cgd 1175: switch (error) {
1.1 cgd 1176:
1177: case ENXIO:
1.210 christos 1178: printf("device bad\n");
1.1 cgd 1179: break;
1180:
1181: case EFAULT:
1.210 christos 1182: printf("device not ready\n");
1.1 cgd 1183: break;
1184:
1185: case EINVAL:
1.210 christos 1186: printf("area improper\n");
1.1 cgd 1187: break;
1188:
1189: case EIO:
1.210 christos 1190: printf("i/o error\n");
1.1 cgd 1191: break;
1192:
1193: case EINTR:
1.210 christos 1194: printf("aborted from console\n");
1.1 cgd 1195: break;
1196:
1.163 cgd 1197: case 0:
1.210 christos 1198: printf("succeeded\n");
1.163 cgd 1199: break;
1200:
1.1 cgd 1201: default:
1.210 christos 1202: printf("error %d\n", error);
1.1 cgd 1203: break;
1204: }
1.210 christos 1205: printf("\n\n");
1.163 cgd 1206: delay(5000000); /* 5 seconds */
1.1 cgd 1207: }
1208:
1209: /*
1210: * Clear registers on exec
1211: */
1.33 cgd 1212: void
1.551 junyoung 1213: setregs(struct lwp *l, struct exec_package *pack, u_long stack)
1.1 cgd 1214: {
1.533 chs 1215: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 1216: struct pcb *pcb = &l->l_addr->u_pcb;
1.298 mycroft 1217: struct trapframe *tf;
1.1 cgd 1218:
1.161 mycroft 1219: #if NNPX > 0
1220: /* If we were using the FPU, forget about it. */
1.508 thorpej 1221: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
1222: npxsave_lwp(l, 0);
1.161 mycroft 1223: #endif
1.166 mycroft 1224:
1.178 mycroft 1225: #ifdef USER_LDT
1.508 thorpej 1226: pmap_ldt_cleanup(l);
1.178 mycroft 1227: #endif
1228:
1.548 yamt 1229: l->l_md.md_flags &= ~MDL_USEDFPU;
1.452 thorpej 1230: if (i386_use_fxsave) {
1.450 thorpej 1231: pcb->pcb_savefpu.sv_xmm.sv_env.en_cw = __NetBSD_NPXCW__;
1.452 thorpej 1232: pcb->pcb_savefpu.sv_xmm.sv_env.en_mxcsr = __INITIAL_MXCSR__;
1233: } else
1.450 thorpej 1234: pcb->pcb_savefpu.sv_87.sv_env.en_cw = __NetBSD_NPXCW__;
1.59 mycroft 1235:
1.508 thorpej 1236: tf = l->l_md.md_regs;
1.445 sommerfe 1237: tf->tf_gs = LSEL(LUDATA_SEL, SEL_UPL);
1238: tf->tf_fs = LSEL(LUDATA_SEL, SEL_UPL);
1.154 mycroft 1239: tf->tf_es = LSEL(LUDATA_SEL, SEL_UPL);
1240: tf->tf_ds = LSEL(LUDATA_SEL, SEL_UPL);
1.252 mycroft 1241: tf->tf_edi = 0;
1242: tf->tf_esi = 0;
1.154 mycroft 1243: tf->tf_ebp = 0;
1.508 thorpej 1244: tf->tf_ebx = (int)l->l_proc->p_psstr;
1.252 mycroft 1245: tf->tf_edx = 0;
1246: tf->tf_ecx = 0;
1247: tf->tf_eax = 0;
1.154 mycroft 1248: tf->tf_eip = pack->ep_entry;
1.533 chs 1249: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
1250: LSEL(LUCODEBIG_SEL, SEL_UPL) : LSEL(LUCODE_SEL, SEL_UPL);
1.154 mycroft 1251: tf->tf_eflags = PSL_USERSET;
1252: tf->tf_esp = stack;
1253: tf->tf_ss = LSEL(LUDATA_SEL, SEL_UPL);
1.1 cgd 1254: }
1255:
1256: /*
1.55 brezak 1257: * Initialize segments and descriptor tables
1.1 cgd 1258: */
1259:
1.516 fvdl 1260: union descriptor *gdt, *ldt;
1261: struct gate_descriptor *idt;
1.498 fvdl 1262: char idt_allocmap[NIDT];
1263: struct simplelock idt_lock = SIMPLELOCK_INITIALIZER;
1.275 mycroft 1264: #ifdef I586_CPU
1265: union descriptor *pentium_idt;
1266: #endif
1.573 yamt 1267: struct user *proc0paddr;
1268: extern vaddr_t proc0uarea;
1.49 brezak 1269:
1.178 mycroft 1270: void
1.551 junyoung 1271: setgate(struct gate_descriptor *gd, void *func, int args, int type, int dpl,
1272: int sel)
1.178 mycroft 1273: {
1.1 cgd 1274:
1.178 mycroft 1275: gd->gd_looffset = (int)func;
1.489 fvdl 1276: gd->gd_selector = sel;
1.178 mycroft 1277: gd->gd_stkcpy = args;
1278: gd->gd_xx = 0;
1279: gd->gd_type = type;
1280: gd->gd_dpl = dpl;
1281: gd->gd_p = 1;
1282: gd->gd_hioffset = (int)func >> 16;
1283: }
1284:
1285: void
1.551 junyoung 1286: unsetgate(struct gate_descriptor *gd)
1.484 fvdl 1287: {
1288: gd->gd_p = 0;
1289: gd->gd_hioffset = 0;
1290: gd->gd_looffset = 0;
1291: gd->gd_selector = 0;
1292: gd->gd_xx = 0;
1293: gd->gd_stkcpy = 0;
1294: gd->gd_type = 0;
1295: gd->gd_dpl = 0;
1296: }
1297:
1298:
1299: void
1.551 junyoung 1300: setregion(struct region_descriptor *rd, void *base, size_t limit)
1.178 mycroft 1301: {
1302:
1303: rd->rd_limit = (int)limit;
1304: rd->rd_base = (int)base;
1305: }
1.1 cgd 1306:
1.174 mycroft 1307: void
1.551 junyoung 1308: setsegment(struct segment_descriptor *sd, void *base, size_t limit, int type,
1309: int dpl, int def32, int gran)
1.174 mycroft 1310: {
1.1 cgd 1311:
1.174 mycroft 1312: sd->sd_lolimit = (int)limit;
1313: sd->sd_lobase = (int)base;
1314: sd->sd_type = type;
1315: sd->sd_dpl = dpl;
1316: sd->sd_p = 1;
1317: sd->sd_hilimit = (int)limit >> 16;
1318: sd->sd_xx = 0;
1319: sd->sd_def32 = def32;
1320: sd->sd_gran = gran;
1321: sd->sd_hibase = (int)base >> 24;
1322: }
1.1 cgd 1323:
1324: #define IDTVEC(name) __CONCAT(X, name)
1.542 junyoung 1325: typedef void (vector)(void);
1.299 mycroft 1326: extern vector IDTVEC(syscall);
1327: extern vector IDTVEC(osyscall);
1328: extern vector *IDTVEC(exceptions)[];
1.333 christos 1329: #ifdef COMPAT_SVR4
1330: extern vector IDTVEC(svr4_fasttrap);
1331: #endif /* COMPAT_SVR4 */
1.447 christos 1332: #ifdef COMPAT_MACH
1333: extern vector IDTVEC(mach_trap);
1334: #endif
1.1 cgd 1335:
1.381 thorpej 1336: #define KBTOB(x) ((size_t)(x) * 1024UL)
1337:
1.484 fvdl 1338: void cpu_init_idt()
1339: {
1340: struct region_descriptor region;
1341: #ifdef I586_CPU
1342: setregion(®ion, pentium_idt, NIDT * sizeof(idt[0]) - 1);
1343: #else
1344: setregion(®ion, idt, NIDT * sizeof(idt[0]) - 1);
1345: #endif
1.565 simonb 1346: lidt(®ion);
1.484 fvdl 1347: }
1348:
1.433 kanaoka 1349: void
1.568 perry 1350: add_mem_cluster(uint64_t seg_start, uint64_t seg_end, uint32_t type)
1.433 kanaoka 1351: {
1352: extern struct extent *iomem_ex;
1.492 kanaoka 1353: int i;
1.433 kanaoka 1354:
1355: if (seg_end > 0x100000000ULL) {
1356: printf("WARNING: skipping large "
1357: "memory map entry: "
1358: "0x%qx/0x%qx/0x%x\n",
1359: seg_start,
1360: (seg_end - seg_start),
1361: type);
1362: return;
1363: }
1364:
1365: /*
1366: * XXX Chop the last page off the size so that
1367: * XXX it can fit in avail_end.
1368: */
1369: if (seg_end == 0x100000000ULL)
1370: seg_end -= PAGE_SIZE;
1371:
1372: if (seg_end <= seg_start)
1373: return;
1374:
1.492 kanaoka 1375: for (i = 0; i < mem_cluster_cnt; i++) {
1376: if ((mem_clusters[i].start == round_page(seg_start))
1377: && (mem_clusters[i].size
1378: == trunc_page(seg_end) - mem_clusters[i].start)) {
1379: #ifdef DEBUG_MEMLOAD
1380: printf("WARNING: skipping duplicate segment entry\n");
1381: #endif
1382: return;
1383: }
1384: }
1385:
1.433 kanaoka 1386: /*
1387: * Allocate the physical addresses used by RAM
1388: * from the iomem extent map. This is done before
1389: * the addresses are page rounded just to make
1390: * sure we get them all.
1391: */
1392: if (extent_alloc_region(iomem_ex, seg_start,
1393: seg_end - seg_start, EX_NOWAIT)) {
1394: /* XXX What should we do? */
1395: printf("WARNING: CAN'T ALLOCATE "
1396: "MEMORY SEGMENT "
1397: "(0x%qx/0x%qx/0x%x) FROM "
1398: "IOMEM EXTENT MAP!\n",
1399: seg_start, seg_end - seg_start, type);
1.492 kanaoka 1400: return;
1.433 kanaoka 1401: }
1402:
1403: /*
1404: * If it's not free memory, skip it.
1405: */
1406: if (type != BIM_Memory)
1407: return;
1408:
1409: /* XXX XXX XXX */
1410: if (mem_cluster_cnt >= VM_PHYSSEG_MAX)
1.579 cube 1411: panic("init386: too many memory segments "
1412: "(increase VM_PHYSSEG_MAX)");
1.433 kanaoka 1413:
1414: seg_start = round_page(seg_start);
1415: seg_end = trunc_page(seg_end);
1416:
1417: if (seg_start == seg_end)
1418: return;
1419:
1420: mem_clusters[mem_cluster_cnt].start = seg_start;
1421: mem_clusters[mem_cluster_cnt].size =
1422: seg_end - seg_start;
1423:
1424: if (avail_end < seg_end)
1425: avail_end = seg_end;
1426: physmem += atop(mem_clusters[mem_cluster_cnt].size);
1427: mem_cluster_cnt++;
1428: }
1429:
1.59 mycroft 1430: void
1.484 fvdl 1431: initgdt(union descriptor *tgdt)
1432: {
1433: struct region_descriptor region;
1434: gdt = tgdt;
1435: memset(gdt, 0, NGDT*sizeof(*gdt));
1436: /* make gdt gates and memory segments */
1437: setsegment(&gdt[GCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 1, 1);
1438: setsegment(&gdt[GDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 1, 1);
1.533 chs 1439: setsegment(&gdt[GUCODE_SEL].sd, 0, x86_btop(I386_MAX_EXE_ADDR) - 1,
1440: SDT_MEMERA, SEL_UPL, 1, 1);
1441: setsegment(&gdt[GUCODEBIG_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1442: SDT_MEMERA, SEL_UPL, 1, 1);
1.514 fvdl 1443: setsegment(&gdt[GUDATA_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1444: SDT_MEMRWA, SEL_UPL, 1, 1);
1445: #ifdef COMPAT_MACH
1446: setgate(&gdt[GMACHCALLS_SEL].gd, &IDTVEC(mach_trap), 1,
1.489 fvdl 1447: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 1448: #endif
1449: #if NBIOSCALL > 0
1450: /* bios trampoline GDT entries */
1451: setsegment(&gdt[GBIOSCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 0,
1452: 0);
1453: setsegment(&gdt[GBIOSDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 0,
1454: 0);
1455: #endif
1456: setsegment(&gdt[GCPU_SEL].sd, &cpu_info_primary,
1457: sizeof(struct cpu_info)-1, SDT_MEMRWA, SEL_KPL, 1, 1);
1458:
1459: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
1460: lgdt(®ion);
1461: }
1462:
1463: void
1.551 junyoung 1464: init386(paddr_t first_avail)
1.2 cgd 1465: {
1.484 fvdl 1466: union descriptor *tgdt;
1.542 junyoung 1467: extern void consinit(void);
1.375 drochner 1468: extern struct extent *iomem_ex;
1.401 thorpej 1469: struct btinfo_memmap *bim;
1.381 thorpej 1470: struct region_descriptor region;
1.401 thorpej 1471: int x, first16q;
1.568 perry 1472: uint64_t seg_start, seg_end;
1473: uint64_t seg_start1, seg_end1;
1.473 tshiozak 1474: paddr_t realmode_reserved_start;
1475: psize_t realmode_reserved_size;
1476: int needs_earlier_install_pte0;
1.436 jdolecek 1477: #if NBIOSCALL > 0
1478: extern int biostramp_image_size;
1479: extern u_char biostramp_image[];
1480: #endif
1.1 cgd 1481:
1.484 fvdl 1482: cpu_probe_features(&cpu_info_primary);
1483: cpu_feature = cpu_info_primary.ci_feature_flags;
1.553 lukem 1484: cpu_feature2 = cpu_info_primary.ci_feature2_flags;
1.484 fvdl 1485:
1.573 yamt 1486: proc0paddr = UAREA_TO_USER(proc0uarea);
1.508 thorpej 1487: lwp0.l_addr = proc0paddr;
1488: cpu_info_primary.ci_curpcb = &lwp0.l_addr->u_pcb;
1.275 mycroft 1489:
1.514 fvdl 1490: x86_bus_space_init();
1.84 cgd 1491: consinit(); /* XXX SHOULD NOT BE DONE HERE */
1.401 thorpej 1492: /*
1493: * Initailize PAGE_SIZE-dependent variables.
1494: */
1495: uvm_setpagesize();
1.450 thorpej 1496:
1497: /*
1498: * Saving SSE registers won't work if the save area isn't
1499: * 16-byte aligned.
1500: */
1501: if (offsetof(struct user, u_pcb.pcb_savefpu) & 0xf)
1502: panic("init386: pcb_savefpu not 16-byte aligned");
1.440 thorpej 1503:
1504: /*
1505: * Start with 2 color bins -- this is just a guess to get us
1506: * started. We'll recolor when we determine the largest cache
1507: * sizes on the system.
1508: */
1509: uvmexp.ncolors = 2;
1.401 thorpej 1510:
1.473 tshiozak 1511: /*
1.484 fvdl 1512: * BIOS leaves data in physical page 0
1513: * Even if it didn't, our VM system doesn't like using zero as a
1514: * physical page number.
1515: * We may also need pages in low memory (one each) for secondary CPU
1516: * startup, for BIOS calls, and for ACPI, plus a page table page to map
1517: * them into the first few pages of the kernel's pmap.
1.473 tshiozak 1518: */
1519: avail_start = PAGE_SIZE;
1520:
1521: /*
1522: * reserve memory for real-mode call
1523: */
1524: needs_earlier_install_pte0 = 0;
1525: realmode_reserved_start = 0;
1526: realmode_reserved_size = 0;
1.414 thorpej 1527: #if NBIOSCALL > 0
1.473 tshiozak 1528: /* save us a page for trampoline code */
1529: realmode_reserved_size += PAGE_SIZE;
1530: needs_earlier_install_pte0 = 1;
1531: #endif
1.484 fvdl 1532: #ifdef MULTIPROCESSOR /* XXX */
1533: KASSERT(avail_start == PAGE_SIZE); /* XXX */
1534: if (realmode_reserved_size < MP_TRAMPOLINE) /* XXX */
1535: realmode_reserved_size = MP_TRAMPOLINE; /* XXX */
1536: needs_earlier_install_pte0 = 1; /* XXX */
1537: #endif /* XXX */
1.473 tshiozak 1538: #if NACPI > 0
1539: /* trampoline code for wake handler */
1540: realmode_reserved_size += ptoa(acpi_md_get_npages_of_wakecode()+1);
1541: needs_earlier_install_pte0 = 1;
1542: #endif
1543: if (needs_earlier_install_pte0) {
1544: /* page table for directory entry 0 */
1545: realmode_reserved_size += PAGE_SIZE;
1546: }
1547: if (realmode_reserved_size>0) {
1548: realmode_reserved_start = avail_start;
1549: avail_start += realmode_reserved_size;
1550: }
1.414 thorpej 1551:
1.492 kanaoka 1552: #ifdef DEBUG_MEMLOAD
1553: printf("mem_cluster_count: %d\n", mem_cluster_cnt);
1554: #endif
1555:
1.401 thorpej 1556: /*
1557: * Call pmap initialization to make new kernel address space.
1558: * We must do this before loading pages into the VM system.
1559: */
1.314 thorpej 1560: pmap_bootstrap((vaddr_t)atdevbase + IOM_SIZE);
1.2 cgd 1561:
1.401 thorpej 1562: /*
1563: * Check to see if we have a memory map from the BIOS (passed
1564: * to us by the boot program.
1565: */
1.577 mrg 1566: if ((biosmem_implicit || (biosbasemem == 0 && biosextmem == 0)) &&
1.575 mrg 1567: (bim = lookup_bootinfo(BTINFO_MEMMAP)) != NULL && bim->num > 0) {
1.463 enami 1568: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1569: printf("BIOS MEMORY MAP (%d ENTRIES):\n", bim->num);
1570: #endif
1571: for (x = 0; x < bim->num; x++) {
1.463 enami 1572: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1573: printf(" addr 0x%qx size 0x%qx type 0x%x\n",
1574: bim->entry[x].addr,
1575: bim->entry[x].size,
1576: bim->entry[x].type);
1577: #endif
1578:
1579: /*
1580: * If the segment is not memory, skip it.
1581: */
1582: switch (bim->entry[x].type) {
1583: case BIM_Memory:
1584: case BIM_ACPI:
1585: case BIM_NVS:
1586: break;
1587: default:
1588: continue;
1589: }
1590:
1591: /*
1592: * Sanity check the entry.
1593: * XXX Need to handle uint64_t in extent code
1594: * XXX and 64-bit physical addresses in i386
1595: * XXX port.
1596: */
1597: seg_start = bim->entry[x].addr;
1598: seg_end = bim->entry[x].addr + bim->entry[x].size;
1599:
1.432 kanaoka 1600: /*
1.433 kanaoka 1601: * Avoid Compatibility Holes.
1602: * XXX Holes within memory space that allow access
1603: * XXX to be directed to the PC-compatible frame buffer
1.484 fvdl 1604: * XXX (0xa0000-0xbffff),to adapter ROM space
1.433 kanaoka 1605: * XXX (0xc0000-0xdffff), and to system BIOS space
1606: * XXX (0xe0000-0xfffff).
1607: * XXX Some laptop(for example,Toshiba Satellite2550X)
1608: * XXX report this area and occurred problems,
1609: * XXX so we avoid this area.
1.432 kanaoka 1610: */
1.433 kanaoka 1611: if (seg_start < 0x100000 && seg_end > 0xa0000) {
1612: printf("WARNING: memory map entry overlaps "
1613: "with ``Compatibility Holes'': "
1614: "0x%qx/0x%qx/0x%x\n", seg_start,
1615: seg_end - seg_start, bim->entry[x].type);
1616: add_mem_cluster(seg_start, 0xa0000,
1617: bim->entry[x].type);
1618: add_mem_cluster(0x100000, seg_end,
1.401 thorpej 1619: bim->entry[x].type);
1.433 kanaoka 1620: } else
1621: add_mem_cluster(seg_start, seg_end,
1.401 thorpej 1622: bim->entry[x].type);
1623: }
1.421 aymeric 1624: }
1.575 mrg 1625:
1.421 aymeric 1626: /*
1627: * If the loop above didn't find any valid segment, fall back to
1628: * former code.
1629: */
1630: if (mem_cluster_cnt == 0) {
1.401 thorpej 1631: /*
1632: * Allocate the physical addresses used by RAM from the iomem
1633: * extent map. This is done before the addresses are
1634: * page rounded just to make sure we get them all.
1635: */
1636: if (extent_alloc_region(iomem_ex, 0, KBTOB(biosbasemem),
1637: EX_NOWAIT)) {
1638: /* XXX What should we do? */
1639: printf("WARNING: CAN'T ALLOCATE BASE MEMORY FROM "
1640: "IOMEM EXTENT MAP!\n");
1641: }
1642: mem_clusters[0].start = 0;
1643: mem_clusters[0].size = trunc_page(KBTOB(biosbasemem));
1644: physmem += atop(mem_clusters[0].size);
1645: if (extent_alloc_region(iomem_ex, IOM_END, KBTOB(biosextmem),
1646: EX_NOWAIT)) {
1647: /* XXX What should we do? */
1648: printf("WARNING: CAN'T ALLOCATE EXTENDED MEMORY FROM "
1649: "IOMEM EXTENT MAP!\n");
1650: }
1651: #if NISADMA > 0
1652: /*
1653: * Some motherboards/BIOSes remap the 384K of RAM that would
1654: * normally be covered by the ISA hole to the end of memory
1655: * so that it can be used. However, on a 16M system, this
1656: * would cause bounce buffers to be allocated and used.
1657: * This is not desirable behaviour, as more than 384K of
1658: * bounce buffers might be allocated. As a work-around,
1659: * we round memory down to the nearest 1M boundary if
1660: * we're using any isadma devices and the remapped memory
1661: * is what puts us over 16M.
1662: */
1663: if (biosextmem > (15*1024) && biosextmem < (16*1024)) {
1664: char pbuf[9];
1665:
1666: format_bytes(pbuf, sizeof(pbuf),
1667: biosextmem - (15*1024));
1668: printf("Warning: ignoring %s of remapped memory\n",
1669: pbuf);
1670: biosextmem = (15*1024);
1671: }
1672: #endif
1673: mem_clusters[1].start = IOM_END;
1674: mem_clusters[1].size = trunc_page(KBTOB(biosextmem));
1675: physmem += atop(mem_clusters[1].size);
1676:
1677: mem_cluster_cnt = 2;
1678:
1679: avail_end = IOM_END + trunc_page(KBTOB(biosextmem));
1680: }
1681: /*
1682: * If we have 16M of RAM or less, just put it all on
1683: * the default free list. Otherwise, put the first
1684: * 16M of RAM on a lower priority free list (so that
1685: * all of the ISA DMA'able memory won't be eaten up
1686: * first-off).
1687: */
1688: if (avail_end <= (16 * 1024 * 1024))
1689: first16q = VM_FREELIST_DEFAULT;
1690: else
1691: first16q = VM_FREELIST_FIRST16;
1692:
1693: /* Make sure the end of the space used by the kernel is rounded. */
1694: first_avail = round_page(first_avail);
1695:
1696: /*
1697: * Now, load the memory clusters (which have already been
1698: * rounded and truncated) into the VM system.
1699: *
1700: * NOTE: WE ASSUME THAT MEMORY STARTS AT 0 AND THAT THE KERNEL
1701: * IS LOADED AT IOM_END (1M).
1702: */
1703: for (x = 0; x < mem_cluster_cnt; x++) {
1704: seg_start = mem_clusters[x].start;
1705: seg_end = mem_clusters[x].start + mem_clusters[x].size;
1706: seg_start1 = 0;
1707: seg_end1 = 0;
1708:
1709: /*
1710: * Skip memory before our available starting point.
1711: */
1712: if (seg_end <= avail_start)
1713: continue;
1714:
1715: if (avail_start >= seg_start && avail_start < seg_end) {
1716: if (seg_start != 0)
1717: panic("init386: memory doesn't start at 0");
1718: seg_start = avail_start;
1719: if (seg_start == seg_end)
1720: continue;
1721: }
1722:
1723: /*
1724: * If this segment contains the kernel, split it
1725: * in two, around the kernel.
1726: */
1727: if (seg_start <= IOM_END && first_avail <= seg_end) {
1728: seg_start1 = first_avail;
1729: seg_end1 = seg_end;
1730: seg_end = IOM_END;
1731: }
1732:
1733: /* First hunk */
1734: if (seg_start != seg_end) {
1.480 erh 1735: if (seg_start < (16 * 1024 * 1024) &&
1.401 thorpej 1736: first16q != VM_FREELIST_DEFAULT) {
1.568 perry 1737: uint64_t tmp;
1.401 thorpej 1738:
1739: if (seg_end > (16 * 1024 * 1024))
1740: tmp = (16 * 1024 * 1024);
1741: else
1742: tmp = seg_end;
1.492 kanaoka 1743:
1744: if (tmp != seg_start) {
1.463 enami 1745: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1746: printf("loading 0x%qx-0x%qx "
1747: "(0x%lx-0x%lx)\n",
1748: seg_start, tmp,
1749: atop(seg_start), atop(tmp));
1750: #endif
1751: uvm_page_physload(atop(seg_start),
1752: atop(tmp), atop(seg_start),
1753: atop(tmp), first16q);
1754: }
1.401 thorpej 1755: seg_start = tmp;
1756: }
1.411 enami 1757:
1758: if (seg_start != seg_end) {
1.463 enami 1759: #ifdef DEBUG_MEMLOAD
1.411 enami 1760: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1761: seg_start, seg_end,
1762: atop(seg_start), atop(seg_end));
1763: #endif
1764: uvm_page_physload(atop(seg_start),
1765: atop(seg_end), atop(seg_start),
1766: atop(seg_end), VM_FREELIST_DEFAULT);
1767: }
1.401 thorpej 1768: }
1769:
1770: /* Second hunk */
1771: if (seg_start1 != seg_end1) {
1.480 erh 1772: if (seg_start1 < (16 * 1024 * 1024) &&
1.401 thorpej 1773: first16q != VM_FREELIST_DEFAULT) {
1.568 perry 1774: uint64_t tmp;
1.401 thorpej 1775:
1776: if (seg_end1 > (16 * 1024 * 1024))
1777: tmp = (16 * 1024 * 1024);
1778: else
1779: tmp = seg_end1;
1.492 kanaoka 1780:
1781: if (tmp != seg_start1) {
1.463 enami 1782: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1783: printf("loading 0x%qx-0x%qx "
1784: "(0x%lx-0x%lx)\n",
1785: seg_start1, tmp,
1786: atop(seg_start1), atop(tmp));
1787: #endif
1788: uvm_page_physload(atop(seg_start1),
1789: atop(tmp), atop(seg_start1),
1790: atop(tmp), first16q);
1791: }
1.401 thorpej 1792: seg_start1 = tmp;
1793: }
1.412 enami 1794:
1795: if (seg_start1 != seg_end1) {
1.463 enami 1796: #ifdef DEBUG_MEMLOAD
1.412 enami 1797: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1798: seg_start1, seg_end1,
1799: atop(seg_start1), atop(seg_end1));
1800: #endif
1801: uvm_page_physload(atop(seg_start1),
1802: atop(seg_end1), atop(seg_start1),
1803: atop(seg_end1), VM_FREELIST_DEFAULT);
1804: }
1.401 thorpej 1805: }
1806: }
1807:
1808: /*
1809: * Steal memory for the message buffer (at end of core).
1810: */
1811: {
1.543 mycroft 1812: struct vm_physseg *vps;
1.401 thorpej 1813: psize_t sz = round_page(MSGBUFSIZE);
1814: psize_t reqsz = sz;
1815:
1816: for (x = 0; x < vm_nphysseg; x++) {
1817: vps = &vm_physmem[x];
1818: if (ptoa(vps->avail_end) == avail_end)
1.543 mycroft 1819: goto found;
1.401 thorpej 1820: }
1.543 mycroft 1821: panic("init386: can't find end of memory");
1.401 thorpej 1822:
1.543 mycroft 1823: found:
1.401 thorpej 1824: /* Shrink so it'll fit in the last segment. */
1825: if ((vps->avail_end - vps->avail_start) < atop(sz))
1826: sz = ptoa(vps->avail_end - vps->avail_start);
1827:
1828: vps->avail_end -= atop(sz);
1829: vps->end -= atop(sz);
1830: msgbuf_paddr = ptoa(vps->avail_end);
1831:
1832: /* Remove the last segment if it now has no pages. */
1833: if (vps->start == vps->end) {
1834: for (vm_nphysseg--; x < vm_nphysseg; x++)
1835: vm_physmem[x] = vm_physmem[x + 1];
1836: }
1837:
1838: /* Now find where the new avail_end is. */
1839: for (avail_end = 0, x = 0; x < vm_nphysseg; x++)
1840: if (vm_physmem[x].avail_end > avail_end)
1841: avail_end = vm_physmem[x].avail_end;
1842: avail_end = ptoa(avail_end);
1843:
1844: /* Warn if the message buffer had to be shrunk. */
1845: if (sz != reqsz)
1846: printf("WARNING: %ld bytes not available for msgbuf "
1847: "in last cluster (%ld used)\n", reqsz, sz);
1848: }
1849:
1.473 tshiozak 1850: /*
1851: * install PT page for the first 4M if needed.
1852: */
1853: if (needs_earlier_install_pte0) {
1854: paddr_t paddr;
1855: #ifdef DIAGNOSTIC
1856: if (realmode_reserved_size < PAGE_SIZE) {
1857: panic("cannot steal memory for first 4M PT page.");
1858: }
1859: #endif
1860: paddr=realmode_reserved_start+realmode_reserved_size-PAGE_SIZE;
1.561 yamt 1861: pmap_kenter_pa((vaddr_t)vtopte(0), paddr,
1862: VM_PROT_READ|VM_PROT_WRITE);
1.473 tshiozak 1863: pmap_update(pmap_kernel());
1864: /* make sure it is clean before using */
1865: memset(vtopte(0), 0, PAGE_SIZE);
1866: realmode_reserved_size -= PAGE_SIZE;
1867: }
1868:
1.295 drochner 1869: #if NBIOSCALL > 0
1.436 jdolecek 1870: /*
1871: * this should be caught at kernel build time, but put it here
1872: * in case someone tries to fake it out...
1873: */
1874: #ifdef DIAGNOSTIC
1.473 tshiozak 1875: if (realmode_reserved_start > BIOSTRAMP_BASE ||
1.474 tron 1876: (realmode_reserved_start+realmode_reserved_size) < (BIOSTRAMP_BASE+
1.473 tshiozak 1877: PAGE_SIZE)) {
1878: panic("cannot steal memory for PT page of bioscall.");
1879: }
1.436 jdolecek 1880: if (biostramp_image_size > PAGE_SIZE)
1.483 provos 1881: panic("biostramp_image_size too big: %x vs. %x",
1.436 jdolecek 1882: biostramp_image_size, PAGE_SIZE);
1883: #endif
1884: pmap_kenter_pa((vaddr_t)BIOSTRAMP_BASE, /* virtual */
1885: (paddr_t)BIOSTRAMP_BASE, /* physical */
1886: VM_PROT_ALL); /* protection */
1.456 chris 1887: pmap_update(pmap_kernel());
1.436 jdolecek 1888: memcpy((caddr_t)BIOSTRAMP_BASE, biostramp_image, biostramp_image_size);
1889: #ifdef DEBUG_BIOSCALL
1890: printf("biostramp installed @ %x\n", BIOSTRAMP_BASE);
1891: #endif
1.473 tshiozak 1892: realmode_reserved_size -= PAGE_SIZE;
1893: realmode_reserved_start += PAGE_SIZE;
1894: #endif
1895:
1896: #if NACPI > 0
1897: /*
1898: * Steal memory for the acpi wake code
1899: */
1900: {
1901: paddr_t paddr, p;
1902: psize_t sz;
1903: int npg;
1904:
1905: paddr = realmode_reserved_start;
1906: npg = acpi_md_get_npages_of_wakecode();
1907: sz = ptoa(npg);
1908: #ifdef DIAGNOSTIC
1909: if (realmode_reserved_size < sz) {
1910: panic("cannot steal memory for ACPI wake code.");
1911: }
1912: #endif
1913:
1914: /* identical mapping */
1915: p = paddr;
1916: for (x=0; x<npg; x++) {
1917: printf("kenter: 0x%08X\n", (unsigned)p);
1918: pmap_kenter_pa((vaddr_t)p, p, VM_PROT_ALL);
1919: p += PAGE_SIZE;
1920: }
1921: pmap_update(pmap_kernel());
1922:
1923: acpi_md_install_wakecode(paddr);
1924:
1925: realmode_reserved_size -= sz;
1926: realmode_reserved_start += sz;
1927: }
1.295 drochner 1928: #endif
1.59 mycroft 1929:
1.561 yamt 1930: pmap_kenter_pa(idt_vaddr, idt_paddr, VM_PROT_READ|VM_PROT_WRITE);
1.456 chris 1931: pmap_update(pmap_kernel());
1.484 fvdl 1932: memset((void *)idt_vaddr, 0, PAGE_SIZE);
1933:
1.516 fvdl 1934: idt = (struct gate_descriptor *)idt_vaddr;
1.275 mycroft 1935: #ifdef I586_CPU
1.561 yamt 1936: pmap_kenter_pa(pentium_idt_vaddr, idt_paddr, VM_PROT_READ);
1.275 mycroft 1937: pentium_idt = (union descriptor *)pentium_idt_vaddr;
1938: #endif
1.484 fvdl 1939: pmap_update(pmap_kernel());
1940:
1941: tgdt = gdt;
1.516 fvdl 1942: gdt = (union descriptor *)
1943: ((char *)idt + NIDT * sizeof (struct gate_descriptor));
1.275 mycroft 1944: ldt = gdt + NGDT;
1945:
1.484 fvdl 1946: memcpy(gdt, tgdt, NGDT*sizeof(*gdt));
1.275 mycroft 1947:
1948: setsegment(&gdt[GLDT_SEL].sd, ldt, NLDT * sizeof(ldt[0]) - 1,
1949: SDT_SYSLDT, SEL_KPL, 0, 0);
1950:
1951: /* make ldt gates and memory segments */
1952: setgate(&ldt[LSYS5CALLS_SEL].gd, &IDTVEC(osyscall), 1,
1.489 fvdl 1953: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.447 christos 1954:
1.275 mycroft 1955: ldt[LUCODE_SEL] = gdt[GUCODE_SEL];
1.533 chs 1956: ldt[LUCODEBIG_SEL] = gdt[GUCODEBIG_SEL];
1.275 mycroft 1957: ldt[LUDATA_SEL] = gdt[GUDATA_SEL];
1.324 christos 1958: ldt[LSOL26CALLS_SEL] = ldt[LBSDICALLS_SEL] = ldt[LSYS5CALLS_SEL];
1.275 mycroft 1959:
1960: /* exceptions */
1.498 fvdl 1961: for (x = 0; x < 32; x++) {
1.516 fvdl 1962: setgate(&idt[x], IDTVEC(exceptions)[x], 0, SDT_SYS386TGT,
1.489 fvdl 1963: (x == 3 || x == 4) ? SEL_UPL : SEL_KPL,
1964: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 1965: idt_allocmap[x] = 1;
1966: }
1.257 thorpej 1967:
1.275 mycroft 1968: /* new-style interrupt gate for syscalls */
1.516 fvdl 1969: setgate(&idt[128], &IDTVEC(syscall), 0, SDT_SYS386TGT, SEL_UPL,
1.489 fvdl 1970: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 1971: idt_allocmap[128] = 1;
1.333 christos 1972: #ifdef COMPAT_SVR4
1.516 fvdl 1973: setgate(&idt[0xd2], &IDTVEC(svr4_fasttrap), 0, SDT_SYS386TGT,
1.489 fvdl 1974: SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 1975: idt_allocmap[0xd2] = 1;
1.333 christos 1976: #endif /* COMPAT_SVR4 */
1.264 mycroft 1977:
1.275 mycroft 1978: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
1979: lgdt(®ion);
1.484 fvdl 1980:
1981: cpu_init_idt();
1.264 mycroft 1982:
1.522 ragge 1983: #if NKSYMS || defined(DDB) || defined(LKM)
1.308 tv 1984: {
1985: extern int end;
1.585 jmmv 1986: boolean_t loaded;
1.336 christos 1987: struct btinfo_symtab *symtab;
1.308 tv 1988:
1.522 ragge 1989: #ifdef DDB
1.484 fvdl 1990: db_machine_init();
1.522 ragge 1991: #endif
1.484 fvdl 1992:
1.585 jmmv 1993: #if defined(MULTIBOOT)
1994: loaded = multiboot_ksyms_init();
1995: #else
1996: loaded = FALSE;
1997: #endif
1998: if (!loaded) {
1999: symtab = lookup_bootinfo(BTINFO_SYMTAB);
2000: if (symtab) {
2001: symtab->ssym += KERNBASE;
2002: symtab->esym += KERNBASE;
2003: ksyms_init(symtab->nsym, (int *)symtab->ssym,
2004: (int *)symtab->esym);
2005: } else
2006: ksyms_init(*(int *)&end, ((int *)&end) + 1, esym);
1.336 christos 2007: }
1.308 tv 2008: }
1.522 ragge 2009: #endif
2010: #ifdef DDB
1.190 mycroft 2011: if (boothowto & RB_KDB)
2012: Debugger();
1.377 ws 2013: #endif
2014: #ifdef IPKDB
2015: ipkdb_init();
2016: if (boothowto & RB_KDB)
2017: ipkdb_connect(0);
1.190 mycroft 2018: #endif
2019: #ifdef KGDB
1.243 drochner 2020: kgdb_port_init();
1.235 thorpej 2021: if (boothowto & RB_KDB) {
2022: kgdb_debug_init = 1;
1.242 drochner 2023: kgdb_connect(1);
1.235 thorpej 2024: }
1.384 jdolecek 2025: #endif
2026:
2027: #if NMCA > 0
2028: /* check for MCA bus, needed to be done before ISA stuff - if
2029: * MCA is detected, ISA needs to use level triggered interrupts
2030: * by default */
2031: mca_busprobe();
1.190 mycroft 2032: #endif
1.275 mycroft 2033:
1.498 fvdl 2034: intr_default_setup();
1.431 thorpej 2035:
2036: /* Initialize software interrupts. */
2037: softintr_init();
1.275 mycroft 2038:
1.498 fvdl 2039: splraise(IPL_IPI);
1.275 mycroft 2040: enable_intr();
2041:
2042: if (physmem < btoc(2 * 1024 * 1024)) {
2043: printf("warning: too little memory available; "
1.383 mycroft 2044: "have %lu bytes, want %lu bytes\n"
1.275 mycroft 2045: "running in degraded mode\n"
2046: "press a key to confirm\n\n",
1.383 mycroft 2047: ptoa(physmem), 2*1024*1024UL);
1.275 mycroft 2048: cngetc();
2049: }
1.507 jdolecek 2050:
2051: #ifdef __HAVE_CPU_MAXPROC
2052: /* Make sure maxproc is sane */
2053: if (maxproc > cpu_maxproc())
2054: maxproc = cpu_maxproc();
2055: #endif
1.1 cgd 2056: }
2057:
1.107 deraadt 2058: #ifdef COMPAT_NOMID
2059: static int
1.566 christos 2060: exec_nomid(struct lwp *l, struct exec_package *epp)
1.31 cgd 2061: {
1.59 mycroft 2062: int error;
2063: u_long midmag, magic;
2064: u_short mid;
1.80 cgd 2065: struct exec *execp = epp->ep_hdr;
1.31 cgd 2066:
1.80 cgd 2067: /* check on validity of epp->ep_hdr performed by exec_out_makecmds */
2068:
2069: midmag = ntohl(execp->a_midmag);
1.59 mycroft 2070: mid = (midmag >> 16) & 0xffff;
2071: magic = midmag & 0xffff;
2072:
2073: if (magic == 0) {
1.80 cgd 2074: magic = (execp->a_midmag & 0xffff);
1.59 mycroft 2075: mid = MID_ZERO;
2076: }
2077:
2078: midmag = mid << 16 | magic;
2079:
2080: switch (midmag) {
2081: case (MID_ZERO << 16) | ZMAGIC:
2082: /*
2083: * 386BSD's ZMAGIC format:
2084: */
1.566 christos 2085: error = exec_aout_prep_oldzmagic(l, epp);
1.59 mycroft 2086: break;
2087:
2088: case (MID_ZERO << 16) | QMAGIC:
2089: /*
2090: * BSDI's QMAGIC format:
2091: * same as new ZMAGIC format, but with different magic number
2092: */
1.566 christos 2093: error = exec_aout_prep_zmagic(l, epp);
1.59 mycroft 2094: break;
2095:
1.202 christos 2096: case (MID_ZERO << 16) | NMAGIC:
2097: /*
2098: * BSDI's NMAGIC format:
2099: * same as NMAGIC format, but with different magic number
2100: * and with text starting at 0.
2101: */
1.566 christos 2102: error = exec_aout_prep_oldnmagic(l, epp);
1.202 christos 2103: break;
2104:
2105: case (MID_ZERO << 16) | OMAGIC:
2106: /*
2107: * BSDI's OMAGIC format:
2108: * same as OMAGIC format, but with different magic number
2109: * and with text starting at 0.
2110: */
1.566 christos 2111: error = exec_aout_prep_oldomagic(l, epp);
1.202 christos 2112: break;
2113:
1.59 mycroft 2114: default:
2115: error = ENOEXEC;
2116: }
2117:
2118: return error;
1.107 deraadt 2119: }
1.31 cgd 2120: #endif
1.107 deraadt 2121:
2122: /*
2123: * cpu_exec_aout_makecmds():
1.549 wiz 2124: * CPU-dependent a.out format hook for execve().
1.107 deraadt 2125: *
2126: * Determine of the given exec package refers to something which we
2127: * understand and, if so, set up the vmcmds for it.
2128: *
2129: * On the i386, old (386bsd) ZMAGIC binaries and BSDI QMAGIC binaries
2130: * if COMPAT_NOMID is given as a kernel option.
2131: */
2132: int
1.566 christos 2133: cpu_exec_aout_makecmds(struct lwp *l, struct exec_package *epp)
1.107 deraadt 2134: {
2135: int error = ENOEXEC;
2136:
2137: #ifdef COMPAT_NOMID
1.566 christos 2138: if ((error = exec_nomid(l, epp)) == 0)
1.107 deraadt 2139: return error;
1.581 thorpej 2140: #else
2141: (void) l;
2142: (void) epp;
1.107 deraadt 2143: #endif /* ! COMPAT_NOMID */
2144:
2145: return error;
1.31 cgd 2146: }
1.84 cgd 2147:
1.484 fvdl 2148: #include <dev/ic/mc146818reg.h> /* for NVRAM POST */
2149: #include <i386/isa/nvram.h> /* for NVRAM POST */
2150:
1.149 mycroft 2151: void
2152: cpu_reset()
2153: {
1.555 mycroft 2154: struct region_descriptor region;
1.149 mycroft 2155:
1.224 mycroft 2156: disable_intr();
2157:
1.227 mycroft 2158: /*
1.484 fvdl 2159: * Ensure the NVRAM reset byte contains something vaguely sane.
2160: */
2161:
2162: outb(IO_RTC, NVRAM_RESET);
2163: outb(IO_RTC+1, NVRAM_RESET_RST);
2164:
2165: /*
1.564 dyoung 2166: * Reset AMD Geode SC1100.
2167: *
1.565 simonb 2168: * 1) Write PCI Configuration Address Register (0xcf8) to
2169: * select Function 0, Register 0x44: Bridge Configuration,
2170: * GPIO and LPC Configuration Register Space, Reset
2171: * Control Register.
1.564 dyoung 2172: *
1.565 simonb 2173: * 2) Write 0xf to PCI Configuration Data Register (0xcfc)
2174: * to reset IDE controller, IDE bus, and PCI bus, and
2175: * to trigger a system-wide reset.
1.564 dyoung 2176: *
2177: * See AMD Geode SC1100 Processor Data Book, Revision 2.0,
2178: * sections 6.3.1, 6.3.2, and 6.4.1.
2179: */
2180: if (cpu_info_primary.ci_signature == 0x540) {
2181: outl(0xcf8, 0x80009044ul);
2182: outl(0xcfc, 0xf);
1.565 simonb 2183: }
1.564 dyoung 2184:
2185: /*
1.227 mycroft 2186: * The keyboard controller has 4 random output pins, one of which is
2187: * connected to the RESET pin on the CPU in many PCs. We tell the
2188: * keyboard controller to pulse this line a couple of times.
2189: */
1.273 drochner 2190: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2191: delay(100000);
1.273 drochner 2192: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2193: delay(100000);
1.149 mycroft 2194:
2195: /*
1.224 mycroft 2196: * Try to cause a triple fault and watchdog reset by making the IDT
2197: * invalid and causing a fault.
1.149 mycroft 2198: */
1.313 perry 2199: memset((caddr_t)idt, 0, NIDT * sizeof(idt[0]));
1.555 mycroft 2200: setregion(®ion, idt, NIDT * sizeof(idt[0]) - 1);
1.565 simonb 2201: lidt(®ion);
1.567 perry 2202: __asm volatile("divl %0,%1" : : "q" (0), "a" (0));
1.149 mycroft 2203:
1.224 mycroft 2204: #if 0
1.149 mycroft 2205: /*
2206: * Try to cause a triple fault and watchdog reset by unmapping the
1.224 mycroft 2207: * entire address space and doing a TLB flush.
1.149 mycroft 2208: */
1.414 thorpej 2209: memset((caddr_t)PTD, 0, PAGE_SIZE);
1.484 fvdl 2210: tlbflush();
1.224 mycroft 2211: #endif
1.149 mycroft 2212:
2213: for (;;);
1.45 cgd 2214: }
1.484 fvdl 2215:
1.499 jdolecek 2216: void
1.551 junyoung 2217: cpu_getmcontext(struct lwp *l, mcontext_t *mcp, unsigned int *flags)
1.508 thorpej 2218: {
2219: const struct trapframe *tf = l->l_md.md_regs;
2220: __greg_t *gr = mcp->__gregs;
1.519 nathanw 2221: __greg_t ras_eip;
1.508 thorpej 2222:
2223: /* Save register context. */
2224: #ifdef VM86
2225: if (tf->tf_eflags & PSL_VM) {
2226: gr[_REG_GS] = tf->tf_vm86_gs;
2227: gr[_REG_FS] = tf->tf_vm86_fs;
2228: gr[_REG_ES] = tf->tf_vm86_es;
2229: gr[_REG_DS] = tf->tf_vm86_ds;
2230: gr[_REG_EFL] = get_vflags(l);
2231: } else
2232: #endif
2233: {
2234: gr[_REG_GS] = tf->tf_gs;
2235: gr[_REG_FS] = tf->tf_fs;
2236: gr[_REG_ES] = tf->tf_es;
2237: gr[_REG_DS] = tf->tf_ds;
2238: gr[_REG_EFL] = tf->tf_eflags;
2239: }
2240: gr[_REG_EDI] = tf->tf_edi;
2241: gr[_REG_ESI] = tf->tf_esi;
2242: gr[_REG_EBP] = tf->tf_ebp;
2243: gr[_REG_EBX] = tf->tf_ebx;
2244: gr[_REG_EDX] = tf->tf_edx;
2245: gr[_REG_ECX] = tf->tf_ecx;
2246: gr[_REG_EAX] = tf->tf_eax;
2247: gr[_REG_EIP] = tf->tf_eip;
2248: gr[_REG_CS] = tf->tf_cs;
2249: gr[_REG_ESP] = tf->tf_esp;
2250: gr[_REG_UESP] = tf->tf_esp;
2251: gr[_REG_SS] = tf->tf_ss;
2252: gr[_REG_TRAPNO] = tf->tf_trapno;
2253: gr[_REG_ERR] = tf->tf_err;
1.519 nathanw 2254:
2255: if ((ras_eip = (__greg_t)ras_lookup(l->l_proc,
2256: (caddr_t) gr[_REG_EIP])) != -1)
2257: gr[_REG_EIP] = ras_eip;
2258:
1.508 thorpej 2259: *flags |= _UC_CPU;
2260:
2261: /* Save floating point register context, if any. */
1.548 yamt 2262: if ((l->l_md.md_flags & MDL_USEDFPU) != 0) {
1.508 thorpej 2263: #if NNPX > 0
2264: /*
2265: * If this process is the current FP owner, dump its
2266: * context to the PCB first.
2267: * XXX npxsave() also clears the FPU state; depending on the
2268: * XXX application this might be a penalty.
2269: */
2270: if (l->l_addr->u_pcb.pcb_fpcpu) {
2271: npxsave_lwp(l, 1);
2272: }
2273: #endif
2274: if (i386_use_fxsave) {
2275: memcpy(&mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2276: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2277: sizeof (mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm));
2278: *flags |= _UC_FXSAVE;
2279: } else {
2280: memcpy(&mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2281: &l->l_addr->u_pcb.pcb_savefpu.sv_87,
2282: sizeof (mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state));
2283: }
2284: #if 0
2285: /* Apparently nothing ever touches this. */
2286: ucp->mcp.mc_fp.fp_emcsts = l->l_addr->u_pcb.pcb_saveemc;
2287: #endif
2288: *flags |= _UC_FPU;
2289: }
2290: }
2291:
2292: int
1.551 junyoung 2293: cpu_setmcontext(struct lwp *l, const mcontext_t *mcp, unsigned int flags)
1.508 thorpej 2294: {
2295: struct trapframe *tf = l->l_md.md_regs;
1.584 pooka 2296: const __greg_t *gr = mcp->__gregs;
1.508 thorpej 2297:
2298: /* Restore register context, if any. */
2299: if ((flags & _UC_CPU) != 0) {
2300: #ifdef VM86
1.535 drochner 2301: if (gr[_REG_EFL] & PSL_VM) {
1.508 thorpej 2302: tf->tf_vm86_gs = gr[_REG_GS];
2303: tf->tf_vm86_fs = gr[_REG_FS];
2304: tf->tf_vm86_es = gr[_REG_ES];
2305: tf->tf_vm86_ds = gr[_REG_DS];
2306: set_vflags(l, gr[_REG_EFL]);
1.534 christos 2307: if (flags & _UC_VM) {
1.542 junyoung 2308: void syscall_vm86(struct trapframe *);
1.534 christos 2309: l->l_proc->p_md.md_syscall = syscall_vm86;
2310: }
1.508 thorpej 2311: } else
2312: #endif
2313: {
2314: /*
2315: * Check for security violations. If we're returning
2316: * to protected mode, the CPU will validate the segment
2317: * registers automatically and generate a trap on
2318: * violations. We handle the trap, rather than doing
2319: * all of the checking here.
2320: */
1.534 christos 2321: if (((gr[_REG_EFL] ^ tf->tf_eflags) & PSL_USERSTATIC) ||
2322: !USERMODE(gr[_REG_CS], gr[_REG_EFL])) {
2323: printf("cpu_setmcontext error: uc EFL: 0x%08x"
1.535 drochner 2324: " tf EFL: 0x%08x uc CS: 0x%x\n",
1.508 thorpej 2325: gr[_REG_EFL], tf->tf_eflags, gr[_REG_CS]);
2326: return (EINVAL);
2327: }
2328: tf->tf_gs = gr[_REG_GS];
2329: tf->tf_fs = gr[_REG_FS];
2330: tf->tf_es = gr[_REG_ES];
2331: tf->tf_ds = gr[_REG_DS];
2332: /* Only change the user-alterable part of eflags */
2333: tf->tf_eflags &= ~PSL_USER;
2334: tf->tf_eflags |= (gr[_REG_EFL] & PSL_USER);
2335: }
2336: tf->tf_edi = gr[_REG_EDI];
2337: tf->tf_esi = gr[_REG_ESI];
2338: tf->tf_ebp = gr[_REG_EBP];
2339: tf->tf_ebx = gr[_REG_EBX];
2340: tf->tf_edx = gr[_REG_EDX];
2341: tf->tf_ecx = gr[_REG_ECX];
2342: tf->tf_eax = gr[_REG_EAX];
2343: tf->tf_eip = gr[_REG_EIP];
2344: tf->tf_cs = gr[_REG_CS];
2345: tf->tf_esp = gr[_REG_UESP];
2346: tf->tf_ss = gr[_REG_SS];
2347: }
2348:
2349: /* Restore floating point register context, if any. */
2350: if ((flags & _UC_FPU) != 0) {
2351: #if NNPX > 0
2352: /*
2353: * If we were using the FPU, forget that we were.
2354: */
2355: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
2356: npxsave_lwp(l, 0);
2357: #endif
2358: if (flags & _UC_FXSAVE) {
2359: if (i386_use_fxsave) {
2360: memcpy(
2361: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2362: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2363: sizeof (&l->l_addr->u_pcb.pcb_savefpu.sv_xmm));
2364: } else {
2365: /* This is a weird corner case */
2366: process_xmm_to_s87((struct savexmm *)
2367: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2368: &l->l_addr->u_pcb.pcb_savefpu.sv_87);
2369: }
2370: } else {
2371: if (i386_use_fxsave) {
2372: process_s87_to_xmm((struct save87 *)
2373: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2374: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm);
2375: } else {
2376: memcpy(&l->l_addr->u_pcb.pcb_savefpu.sv_87,
2377: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2378: sizeof (l->l_addr->u_pcb.pcb_savefpu.sv_87));
2379: }
2380: }
2381: /* If not set already. */
1.548 yamt 2382: l->l_md.md_flags |= MDL_USEDFPU;
1.508 thorpej 2383: #if 0
2384: /* Apparently unused. */
2385: l->l_addr->u_pcb.pcb_saveemc = mcp->mc_fp.fp_emcsts;
2386: #endif
2387: }
1.534 christos 2388: if (flags & _UC_SETSTACK)
2389: l->l_proc->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
2390: if (flags & _UC_CLRSTACK)
2391: l->l_proc->p_sigctx.ps_sigstk.ss_flags &= ~SS_ONSTACK;
1.508 thorpej 2392: return (0);
2393: }
2394:
2395: void
1.499 jdolecek 2396: cpu_initclocks()
1.484 fvdl 2397: {
1.574 kardel 2398:
1.484 fvdl 2399: (*initclock_func)();
2400: }
2401:
2402: #ifdef MULTIPROCESSOR
1.499 jdolecek 2403: void
2404: need_resched(struct cpu_info *ci)
1.484 fvdl 2405: {
1.546 yamt 2406:
2407: if (ci->ci_want_resched)
2408: return;
2409:
1.484 fvdl 2410: ci->ci_want_resched = 1;
1.508 thorpej 2411: if ((ci)->ci_curlwp != NULL)
2412: aston((ci)->ci_curlwp->l_proc);
1.546 yamt 2413: else if (ci != curcpu())
2414: x86_send_ipi(ci, 0);
1.484 fvdl 2415: }
2416: #endif
2417:
2418: /*
2419: * Allocate an IDT vector slot within the given range.
2420: * XXX needs locking to avoid MP allocation races.
2421: */
2422:
2423: int
1.551 junyoung 2424: idt_vec_alloc(int low, int high)
1.484 fvdl 2425: {
2426: int vec;
2427:
1.498 fvdl 2428: simple_lock(&idt_lock);
2429: for (vec = low; vec <= high; vec++) {
2430: if (idt_allocmap[vec] == 0) {
2431: idt_allocmap[vec] = 1;
2432: simple_unlock(&idt_lock);
1.484 fvdl 2433: return vec;
1.498 fvdl 2434: }
2435: }
2436: simple_unlock(&idt_lock);
1.484 fvdl 2437: return 0;
2438: }
2439:
1.498 fvdl 2440: void
1.551 junyoung 2441: idt_vec_set(int vec, void (*function)(void))
1.484 fvdl 2442: {
1.498 fvdl 2443: /*
2444: * Vector should be allocated, so no locking needed.
2445: */
2446: KASSERT(idt_allocmap[vec] == 1);
1.516 fvdl 2447: setgate(&idt[vec], function, 0, SDT_SYS386IGT, SEL_KPL,
1.489 fvdl 2448: GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 2449: }
2450:
2451: void
1.551 junyoung 2452: idt_vec_free(int vec)
1.484 fvdl 2453: {
1.498 fvdl 2454: simple_lock(&idt_lock);
1.516 fvdl 2455: unsetgate(&idt[vec]);
1.498 fvdl 2456: idt_allocmap[vec] = 0;
2457: simple_unlock(&idt_lock);
1.507 jdolecek 2458: }
2459:
2460: /*
2461: * Number of processes is limited by number of available GDT slots.
2462: */
2463: int
2464: cpu_maxproc(void)
2465: {
2466: #ifdef USER_LDT
2467: return ((MAXGDTSIZ - NGDT) / 2);
2468: #else
2469: return (MAXGDTSIZ - NGDT);
2470: #endif
1.484 fvdl 2471: }
CVSweb <webmaster@jp.NetBSD.org>