Annotation of src/sys/arch/i386/i386/machdep.c, Revision 1.586.2.3
1.586.2.3! jdc 1: /* $NetBSD: machdep.c,v 1.586.2.2 2007/02/07 12:53:10 tron 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>
1.586.2.3! jdc 75: __KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.586.2.2 2007/02/07 12:53:10 tron 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:
1.586.2.3! jdc 186: #include "cardbus.h"
! 187: #if NCARDBUS > 0
! 188: /* For rbus_min_start hint. */
! 189: #include <machine/bus.h>
! 190: #include <dev/cardbus/rbus.h>
! 191: #include <machine/rbus_machdep.h>
! 192: #endif
! 193:
1.384 jdolecek 194: #include "mca.h"
195: #if NMCA > 0
196: #include <machine/mca_machdep.h> /* for mca_busprobe() */
197: #endif
198:
1.484 fvdl 199: #ifdef MULTIPROCESSOR /* XXX */
200: #include <machine/mpbiosvar.h> /* XXX */
201: #endif /* XXX */
202:
1.517 jmmv 203: #ifndef BEEP_ONHALT_COUNT
204: #define BEEP_ONHALT_COUNT 3
205: #endif
206: #ifndef BEEP_ONHALT_PITCH
207: #define BEEP_ONHALT_PITCH 1500
208: #endif
209: #ifndef BEEP_ONHALT_PERIOD
210: #define BEEP_ONHALT_PERIOD 250
211: #endif
212:
1.104 cgd 213: /* the following is used externally (sysctl_hw) */
1.549 wiz 214: char machine[] = "i386"; /* CPU "architecture" */
1.232 veego 215: char machine_arch[] = "i386"; /* machine == machine_arch */
1.104 cgd 216:
1.559 thorpej 217: extern struct bi_devmatch *x86_alldisks;
218: extern int x86_ndisks;
1.342 fvdl 219:
1.328 bouyer 220: #ifdef CPURESET_DELAY
221: int cpureset_delay = CPURESET_DELAY;
222: #else
223: int cpureset_delay = 2000; /* default to 2s */
224: #endif
225:
1.455 fvdl 226: #ifdef MTRR
1.513 fvdl 227: struct mtrr_funcs *mtrr_funcs;
1.455 fvdl 228: #endif
229:
1.515 fvdl 230: #ifdef COMPAT_NOMID
1.566 christos 231: static int exec_nomid(struct lwp *, struct exec_package *);
1.550 junyoung 232: #endif
1.515 fvdl 233:
1.59 mycroft 234: int physmem;
1.163 cgd 235: int dumpmem_low;
236: int dumpmem_high;
1.524 drochner 237: unsigned int cpu_feature;
1.553 lukem 238: unsigned int cpu_feature2;
1.59 mycroft 239: int cpu_class;
1.428 fvdl 240: int i386_fpu_present;
241: int i386_fpu_exception;
242: int i386_fpu_fdivbug;
1.59 mycroft 243:
1.451 thorpej 244: int i386_use_fxsave;
245: int i386_has_sse;
246: int i386_has_sse2;
1.450 thorpej 247:
1.461 christos 248: int tmx86_has_longrun;
249:
1.314 thorpej 250: vaddr_t msgbuf_vaddr;
1.586.2.1 tron 251: struct {
252: paddr_t paddr;
253: psize_t sz;
254: } msgbuf_p_seg[VM_PHYSSEG_MAX];
255: unsigned int msgbuf_p_cnt = 0;
1.314 thorpej 256:
257: vaddr_t idt_vaddr;
258: paddr_t idt_paddr;
259:
1.264 mycroft 260: #ifdef I586_CPU
1.314 thorpej 261: vaddr_t pentium_idt_vaddr;
1.264 mycroft 262: #endif
1.59 mycroft 263:
1.444 chs 264: struct vm_map *exec_map = NULL;
265: struct vm_map *mb_map = NULL;
266: struct vm_map *phys_map = NULL;
1.48 brezak 267:
1.314 thorpej 268: extern paddr_t avail_start, avail_end;
1.1 cgd 269:
1.542 junyoung 270: void (*delay_func)(int) = i8254_delay;
271: void (*initclock_func)(void) = i8254_initclocks;
1.484 fvdl 272:
1.204 thorpej 273: /*
1.291 thorpej 274: * Size of memory segments, before any memory is stolen.
275: */
276: phys_ram_seg_t mem_clusters[VM_PHYSSEG_MAX];
277: int mem_cluster_cnt;
278:
1.542 junyoung 279: int cpu_dump(void);
280: int cpu_dumpsize(void);
281: u_long cpu_dump_mempagecnt(void);
282: void dumpsys(void);
283: void init386(paddr_t);
284: void initgdt(union descriptor *);
1.255 drochner 285:
1.568 perry 286: void add_mem_cluster(uint64_t, uint64_t, uint32_t);
1.433 kanaoka 287:
1.521 dsl 288: extern int time_adjusted;
1.484 fvdl 289:
1.571 jmmv 290: struct bootinfo bootinfo;
291: int *esym;
292: extern int boothowto;
293:
294: /* Base memory reported by BIOS. */
295: #ifndef REALBASEMEM
296: int biosbasemem = 0;
297: #else
298: int biosbasemem = REALBASEMEM;
299: #endif
300:
301: /* Extended memory reported by BIOS. */
302: #ifndef REALEXTMEM
303: int biosextmem = 0;
304: #else
305: int biosextmem = REALEXTMEM;
306: #endif
307:
1.577 mrg 308: /* Set if any boot-loader set biosbasemem/biosextmem. */
309: int biosmem_implicit;
310:
1.571 jmmv 311: /* Representation of the bootinfo structure constructed by a NetBSD native
312: * boot loader. Only be used by native_loader(). */
313: struct bootinfo_source {
314: uint32_t bs_naddrs;
315: paddr_t bs_addrs[1]; /* Actually longer. */
316: };
317:
318: /* Only called by locore.h; no need to be in a header file. */
319: void native_loader(int, int, struct bootinfo_source *, paddr_t, int, int);
320:
321: /*
322: * Called as one of the very first things during system startup (just after
323: * the boot loader gave control to the kernel image), this routine is in
324: * charge of retrieving the parameters passed in by the boot loader and
325: * storing them in the appropriate kernel variables.
326: *
327: * WARNING: Because the kernel has not yet relocated itself to KERNBASE,
328: * special care has to be taken when accessing memory because absolute
329: * addresses (referring to kernel symbols) do not work. So:
330: *
331: * 1) Avoid jumps to absolute addresses (such as gotos and switches).
332: * 2) To access global variables use their physical address, which
333: * can be obtained using the RELOC macro.
334: */
335: void
1.586 christos 336: native_loader(int bl_boothowto, int bl_bootdev,
1.571 jmmv 337: struct bootinfo_source *bl_bootinfo, paddr_t bl_esym,
1.572 jmmv 338: int bl_biosextmem, int bl_biosbasemem)
1.571 jmmv 339: {
340: #define RELOC(type, x) ((type)((vaddr_t)(x) - KERNBASE))
341:
342: *RELOC(int *, &boothowto) = bl_boothowto;
343:
344: #ifdef COMPAT_OLDBOOT
345: /*
346: * Pre-1.3 boot loaders gave the boot device as a parameter
347: * (instead of a bootinfo entry).
348: */
349: *RELOC(int *, &bootdev) = bl_bootdev;
350: #endif
351:
352: /*
353: * The boot loader provides a physical, non-relocated address
354: * for the symbols table's end. We need to convert it to a
355: * virtual address.
356: */
357: if (bl_esym != 0)
358: *RELOC(int **, &esym) = (int *)((vaddr_t)bl_esym + KERNBASE);
359: else
360: *RELOC(int **, &esym) = 0;
361:
362: /*
363: * Copy bootinfo entries (if any) from the boot loader's
364: * representation to the kernel's bootinfo space.
365: */
366: if (bl_bootinfo != NULL) {
367: size_t i;
368: uint8_t *data;
369: struct bootinfo *bidest;
370:
371: bidest = RELOC(struct bootinfo *, &bootinfo);
372:
373: data = &bidest->bi_data[0];
374:
375: for (i = 0; i < bl_bootinfo->bs_naddrs; i++) {
376: struct btinfo_common *bc;
377:
378: bc = (struct btinfo_common *)(bl_bootinfo->bs_addrs[i]);
379:
380: if ((paddr_t)(data + bc->len) >
381: (paddr_t)(&bidest->bi_data[0] + BOOTINFO_MAXSIZE))
382: break;
383:
384: memcpy(data, bc, bc->len);
385: data += bc->len;
386: }
387: bidest->bi_nentries = i;
388: }
389:
390: /*
391: * Configure biosbasemem and biosextmem only if they were not
392: * explicitly given during the kernel's build.
393: */
1.577 mrg 394: if (*RELOC(int *, &biosbasemem) == 0) {
1.571 jmmv 395: *RELOC(int *, &biosbasemem) = bl_biosbasemem;
1.577 mrg 396: *RELOC(int *, &biosmem_implicit) = 1;
397: }
398: if (*RELOC(int *, &biosextmem) == 0) {
1.571 jmmv 399: *RELOC(int *, &biosextmem) = bl_biosextmem;
1.577 mrg 400: *RELOC(int *, &biosmem_implicit) = 1;
401: }
1.571 jmmv 402: #undef RELOC
403: }
404:
1.59 mycroft 405: /*
406: * Machine-dependent startup code
407: */
1.32 andrew 408: void
1.1 cgd 409: cpu_startup()
410: {
1.586.2.1 tron 411: int x, y;
1.314 thorpej 412: vaddr_t minaddr, maxaddr;
1.586.2.1 tron 413: psize_t sz;
1.354 lukem 414: char pbuf[9];
1.1 cgd 415:
1.284 mrg 416: /*
417: * Initialize error message buffer (et end of core).
418: */
1.586.2.1 tron 419: if (msgbuf_p_cnt == 0)
420: panic("msgbuf paddr map has not been set up");
421: for (x = 0, sz = 0; x < msgbuf_p_cnt; sz += msgbuf_p_seg[x++].sz)
422: continue;
423: msgbuf_vaddr = uvm_km_alloc(kernel_map, sz, 0, UVM_KMF_VAONLY);
1.385 thorpej 424: if (msgbuf_vaddr == 0)
1.284 mrg 425: panic("failed to valloc msgbuf_vaddr");
1.359 thorpej 426:
1.284 mrg 427: /* msgbuf_paddr was init'd in pmap */
1.586.2.1 tron 428: for (y = 0, sz = 0; y < msgbuf_p_cnt; y++) {
429: for (x = 0; x < btoc(msgbuf_p_seg[y].sz); x++, sz += PAGE_SIZE)
430: pmap_kenter_pa((vaddr_t)msgbuf_vaddr + sz,
431: msgbuf_p_seg[y].paddr + x * PAGE_SIZE, VM_PROT_READ|VM_PROT_WRITE);
432: }
1.456 chris 433: pmap_update(pmap_kernel());
1.359 thorpej 434:
1.586.2.1 tron 435: initmsgbuf((caddr_t)msgbuf_vaddr, sz);
1.284 mrg 436:
1.562 lukem 437: printf("%s%s", copyright, version);
1.484 fvdl 438:
1.570 jmmv 439: #ifdef MULTIBOOT
440: multiboot_print_info();
441: #endif
442:
1.484 fvdl 443: #ifdef TRAPLOG
444: /*
445: * Enable recording of branch from/to in MSR's
446: */
447: wrmsr(MSR_DEBUGCTLMSR, 0x1);
448: #endif
449:
1.382 mycroft 450: format_bytes(pbuf, sizeof(pbuf), ptoa(physmem));
1.354 lukem 451: printf("total memory = %s\n", pbuf);
1.1 cgd 452:
1.586.2.3! jdc 453: #if NCARDBUS > 0
! 454: /* Tell RBUS how much RAM we have, so it can use heuristics. */
! 455: rbus_min_start_hint(ptoa(physmem));
! 456: #endif
! 457:
1.547 pk 458: minaddr = 0;
1.41 cgd 459:
1.1 cgd 460: /*
1.36 cgd 461: * Allocate a submap for exec arguments. This map effectively
462: * limits the number of processes exec'ing at any time.
1.1 cgd 463: */
1.284 mrg 464: exec_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 465: 16*NCARGS, VM_MAP_PAGEABLE, FALSE, NULL);
1.59 mycroft 466:
1.1 cgd 467: /*
468: * Allocate a submap for physio
469: */
1.284 mrg 470: phys_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 471: VM_PHYS_SIZE, 0, FALSE, NULL);
1.1 cgd 472:
473: /*
1.229 thorpej 474: * Finally, allocate mbuf cluster submap.
1.1 cgd 475: */
1.334 thorpej 476: mb_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 477: nmbclusters * mclbytes, VM_MAP_INTRSAFE, FALSE, NULL);
1.1 cgd 478:
1.547 pk 479: format_bytes(pbuf, sizeof(pbuf), ptoa(uvmexp.free));
1.354 lukem 480: printf("avail memory = %s\n", pbuf);
1.1 cgd 481:
1.375 drochner 482: /* Safe for i/o port / memory space allocation to use malloc now. */
1.514 fvdl 483: x86_bus_space_mallocok();
1.349 thorpej 484: }
485:
486: /*
487: * Set up proc0's TSS and LDT.
488: */
489: void
490: i386_proc0_tss_ldt_init()
491: {
492: struct pcb *pcb;
493: int x;
1.268 thorpej 494:
1.326 thorpej 495: gdt_init();
1.484 fvdl 496:
1.508 thorpej 497: cpu_info_primary.ci_curpcb = pcb = &lwp0.l_addr->u_pcb;
1.484 fvdl 498:
1.326 thorpej 499: pcb->pcb_tss.tss_ioopt =
500: ((caddr_t)pcb->pcb_iomap - (caddr_t)&pcb->pcb_tss) << 16;
1.484 fvdl 501:
1.326 thorpej 502: for (x = 0; x < sizeof(pcb->pcb_iomap) / 4; x++)
503: pcb->pcb_iomap[x] = 0xffffffff;
504:
1.394 thorpej 505: pcb->pcb_ldt_sel = pmap_kernel()->pm_ldt_sel = GSEL(GLDT_SEL, SEL_KPL);
1.326 thorpej 506: pcb->pcb_cr0 = rcr0();
507: pcb->pcb_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
1.573 yamt 508: pcb->pcb_tss.tss_esp0 = USER_TO_UAREA(lwp0.l_addr) + KSTACK_SIZE - 16;
1.508 thorpej 509: lwp0.l_md.md_regs = (struct trapframe *)pcb->pcb_tss.tss_esp0 - 1;
510: lwp0.l_md.md_tss_sel = tss_alloc(pcb);
1.326 thorpej 511:
1.508 thorpej 512: ltr(lwp0.l_md.md_tss_sel);
1.326 thorpej 513: lldt(pcb->pcb_ldt_sel);
1.484 fvdl 514: }
515:
516: /*
517: * Set up TSS and LDT for a new PCB.
518: */
519:
520: void
1.551 junyoung 521: i386_init_pcb_tss_ldt(struct cpu_info *ci)
1.484 fvdl 522: {
523: int x;
524: struct pcb *pcb = ci->ci_idle_pcb;
1.326 thorpej 525:
1.484 fvdl 526: pcb->pcb_tss.tss_ioopt =
527: ((caddr_t)pcb->pcb_iomap - (caddr_t)&pcb->pcb_tss) << 16;
528: for (x = 0; x < sizeof(pcb->pcb_iomap) / 4; x++)
529: pcb->pcb_iomap[x] = 0xffffffff;
530:
531: pcb->pcb_ldt_sel = pmap_kernel()->pm_ldt_sel = GSEL(GLDT_SEL, SEL_KPL);
532: pcb->pcb_cr0 = rcr0();
533:
534: ci->ci_idle_tss_sel = tss_alloc(pcb);
1.326 thorpej 535: }
536:
537: /*
1.545 atatat 538: * sysctl helper routine for machdep.tm* nodes.
1.484 fvdl 539: */
1.545 atatat 540: static int
541: sysctl_machdep_tm_longrun(SYSCTLFN_ARGS)
1.104 cgd 542: {
1.545 atatat 543: struct sysctlnode node;
544: int io, error;
545:
546: if (!tmx86_has_longrun)
547: return (EOPNOTSUPP);
548:
549: node = *rnode;
550: node.sysctl_data = &io;
1.104 cgd 551:
1.545 atatat 552: switch (rnode->sysctl_num) {
1.461 christos 553: case CPU_TMLR_MODE:
1.545 atatat 554: io = (int)(crusoe_longrun = tmx86_get_longrun_mode());
555: break;
1.461 christos 556: case CPU_TMLR_FREQUENCY:
557: tmx86_get_longrun_status_all();
1.545 atatat 558: io = crusoe_frequency;
559: break;
1.461 christos 560: case CPU_TMLR_VOLTAGE:
561: tmx86_get_longrun_status_all();
1.545 atatat 562: io = crusoe_voltage;
563: break;
1.461 christos 564: case CPU_TMLR_PERCENTAGE:
565: tmx86_get_longrun_status_all();
1.545 atatat 566: io = crusoe_percentage;
567: break;
1.104 cgd 568: default:
569: return (EOPNOTSUPP);
570: }
1.545 atatat 571:
572: error = sysctl_lookup(SYSCTLFN_CALL(&node));
573: if (error || newp == NULL)
574: return (error);
575:
576: if (rnode->sysctl_num == CPU_TMLR_MODE) {
577: if (tmx86_set_longrun_mode(io))
578: crusoe_longrun = (u_int)io;
579: else
580: return (EINVAL);
581: }
582:
583: return (0);
584: }
585:
586: /*
587: * sysctl helper routine for machdep.booted_kernel
588: */
589: static int
590: sysctl_machdep_booted_kernel(SYSCTLFN_ARGS)
591: {
592: struct btinfo_bootpath *bibp;
593: struct sysctlnode node;
594:
595: bibp = lookup_bootinfo(BTINFO_BOOTPATH);
596: if(!bibp)
597: return(ENOENT); /* ??? */
598:
599: node = *rnode;
600: node.sysctl_data = bibp->bootpath;
601: node.sysctl_size = sizeof(bibp->bootpath);
602: return (sysctl_lookup(SYSCTLFN_CALL(&node)));
603: }
604:
605: /*
606: * sysctl helper routine for machdep.diskinfo
607: */
608: static int
609: sysctl_machdep_diskinfo(SYSCTLFN_ARGS)
610: {
611: struct sysctlnode node;
612:
613: node = *rnode;
1.559 thorpej 614: if (x86_alldisks == NULL)
1.557 dbj 615: return(EOPNOTSUPP);
1.559 thorpej 616: node.sysctl_data = x86_alldisks;
1.545 atatat 617: node.sysctl_size = sizeof(struct disklist) +
1.559 thorpej 618: (x86_ndisks - 1) * sizeof(struct nativedisk_info);
1.557 dbj 619: return (sysctl_lookup(SYSCTLFN_CALL(&node)));
1.545 atatat 620: }
621:
622: /*
623: * machine dependent system variables.
624: */
625: SYSCTL_SETUP(sysctl_machdep_setup, "sysctl machdep subtree setup")
626: {
627:
1.552 atatat 628: sysctl_createv(clog, 0, NULL, NULL,
629: CTLFLAG_PERMANENT,
1.545 atatat 630: CTLTYPE_NODE, "machdep", NULL,
631: NULL, 0, NULL, 0,
632: CTL_MACHDEP, CTL_EOL);
633:
1.552 atatat 634: sysctl_createv(clog, 0, NULL, NULL,
635: CTLFLAG_PERMANENT,
1.545 atatat 636: CTLTYPE_STRUCT, "console_device", NULL,
637: sysctl_consdev, 0, NULL, sizeof(dev_t),
638: CTL_MACHDEP, CPU_CONSDEV, CTL_EOL);
1.552 atatat 639: sysctl_createv(clog, 0, NULL, NULL,
640: CTLFLAG_PERMANENT,
1.545 atatat 641: CTLTYPE_INT, "biosbasemem", NULL,
642: NULL, 0, &biosbasemem, 0,
643: CTL_MACHDEP, CPU_BIOSBASEMEM, CTL_EOL);
1.552 atatat 644: sysctl_createv(clog, 0, NULL, NULL,
645: CTLFLAG_PERMANENT,
1.545 atatat 646: CTLTYPE_INT, "biosextmem", NULL,
647: NULL, 0, &biosextmem, 0,
648: CTL_MACHDEP, CPU_BIOSEXTMEM, CTL_EOL);
1.552 atatat 649: sysctl_createv(clog, 0, NULL, NULL,
650: CTLFLAG_PERMANENT,
1.545 atatat 651: CTLTYPE_INT, "nkpde", NULL,
652: NULL, 0, &nkpde, 0,
653: CTL_MACHDEP, CPU_NKPDE, CTL_EOL);
1.552 atatat 654: sysctl_createv(clog, 0, NULL, NULL,
655: CTLFLAG_PERMANENT,
1.545 atatat 656: CTLTYPE_STRING, "booted_kernel", NULL,
657: sysctl_machdep_booted_kernel, 0, NULL, 0,
658: CTL_MACHDEP, CPU_BOOTED_KERNEL, CTL_EOL);
1.552 atatat 659: sysctl_createv(clog, 0, NULL, NULL,
660: CTLFLAG_PERMANENT,
1.545 atatat 661: CTLTYPE_STRUCT, "diskinfo", NULL,
662: sysctl_machdep_diskinfo, 0, NULL, 0,
663: CTL_MACHDEP, CPU_DISKINFO, CTL_EOL);
1.552 atatat 664: sysctl_createv(clog, 0, NULL, NULL,
665: CTLFLAG_PERMANENT,
1.545 atatat 666: CTLTYPE_INT, "fpu_present", NULL,
667: NULL, 0, &i386_fpu_present, 0,
668: CTL_MACHDEP, CPU_FPU_PRESENT, CTL_EOL);
1.552 atatat 669: sysctl_createv(clog, 0, NULL, NULL,
670: CTLFLAG_PERMANENT,
1.545 atatat 671: CTLTYPE_INT, "osfxsr", NULL,
672: NULL, 0, &i386_use_fxsave, 0,
673: CTL_MACHDEP, CPU_OSFXSR, CTL_EOL);
1.552 atatat 674: sysctl_createv(clog, 0, NULL, NULL,
675: CTLFLAG_PERMANENT,
1.545 atatat 676: CTLTYPE_INT, "sse", NULL,
677: NULL, 0, &i386_has_sse, 0,
678: CTL_MACHDEP, CPU_SSE, CTL_EOL);
1.552 atatat 679: sysctl_createv(clog, 0, NULL, NULL,
680: CTLFLAG_PERMANENT,
1.545 atatat 681: CTLTYPE_INT, "sse2", NULL,
682: NULL, 0, &i386_has_sse2, 0,
683: CTL_MACHDEP, CPU_SSE2, CTL_EOL);
1.554 lukem 684: sysctl_createv(clog, 0, NULL, NULL,
685: CTLFLAG_PERMANENT,
686: CTLTYPE_STRING, "cpu_brand", NULL,
687: NULL, 0, &cpu_brand_string, 0,
688: CTL_MACHDEP, CTL_CREATE, CTL_EOL);
1.552 atatat 689: sysctl_createv(clog, 0, NULL, NULL,
690: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.545 atatat 691: CTLTYPE_INT, "tm_longrun_mode", NULL,
692: sysctl_machdep_tm_longrun, 0, NULL, 0,
693: CTL_MACHDEP, CPU_TMLR_MODE, CTL_EOL);
1.552 atatat 694: sysctl_createv(clog, 0, NULL, NULL,
695: CTLFLAG_PERMANENT,
1.545 atatat 696: CTLTYPE_INT, "tm_longrun_frequency", NULL,
697: sysctl_machdep_tm_longrun, 0, NULL, 0,
698: CTL_MACHDEP, CPU_TMLR_FREQUENCY, CTL_EOL);
1.552 atatat 699: sysctl_createv(clog, 0, NULL, NULL,
700: CTLFLAG_PERMANENT,
1.545 atatat 701: CTLTYPE_INT, "tm_longrun_voltage", NULL,
702: sysctl_machdep_tm_longrun, 0, NULL, 0,
703: CTL_MACHDEP, CPU_TMLR_VOLTAGE, CTL_EOL);
1.552 atatat 704: sysctl_createv(clog, 0, NULL, NULL,
705: CTLFLAG_PERMANENT,
1.545 atatat 706: CTLTYPE_INT, "tm_longrun_percentage", NULL,
707: sysctl_machdep_tm_longrun, 0, NULL, 0,
708: CTL_MACHDEP, CPU_TMLR_PERCENTAGE, CTL_EOL);
1.104 cgd 709: }
1.151 christos 710:
1.534 christos 711: void *
712: getframe(struct lwp *l, int sig, int *onstack)
713: {
714: struct proc *p = l->l_proc;
715: struct sigctx *ctx = &p->p_sigctx;
716: struct trapframe *tf = l->l_md.md_regs;
717:
718: /* Do we need to jump onto the signal stack? */
719: *onstack = (ctx->ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0
720: && (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0;
721: if (*onstack)
722: return (char *)ctx->ps_sigstk.ss_sp + ctx->ps_sigstk.ss_size;
723: #ifdef VM86
724: if (tf->tf_eflags & PSL_VM)
725: return (void *)(tf->tf_esp + (tf->tf_ss << 4));
726: else
727: #endif
728: return (void *)tf->tf_esp;
729: }
730:
731: /*
732: * Build context to run handler in. We invoke the handler
733: * directly, only returning via the trampoline. Note the
734: * trampoline version numbers are coordinated with machine-
735: * dependent code in libc.
736: */
737: void
738: buildcontext(struct lwp *l, int sel, void *catcher, void *fp)
739: {
740: struct trapframe *tf = l->l_md.md_regs;
741:
742: tf->tf_gs = GSEL(GUDATA_SEL, SEL_UPL);
743: tf->tf_fs = GSEL(GUDATA_SEL, SEL_UPL);
744: tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
745: tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
746: tf->tf_eip = (int)catcher;
747: tf->tf_cs = GSEL(sel, SEL_UPL);
748: tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
749: tf->tf_esp = (int)fp;
750: tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
751: }
752:
753: static void
1.539 christos 754: sendsig_siginfo(const ksiginfo_t *ksi, const sigset_t *mask)
1.534 christos 755: {
756: struct lwp *l = curlwp;
757: struct proc *p = l->l_proc;
758: struct pmap *pmap = vm_map_pmap(&p->p_vmspace->vm_map);
759: int sel = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
760: GUCODEBIG_SEL : GUCODE_SEL;
761: struct sigacts *ps = p->p_sigacts;
762: int onstack;
763: int sig = ksi->ksi_signo;
764: struct sigframe_siginfo *fp = getframe(l, sig, &onstack), frame;
765: sig_t catcher = SIGACTION(p, sig).sa_handler;
766: struct trapframe *tf = l->l_md.md_regs;
767:
768: fp--;
769:
770: /* Build stack frame for signal trampoline. */
771: switch (ps->sa_sigdesc[sig].sd_vers) {
1.537 christos 772: case 0: /* handled by sendsig_sigcontext */
773: case 1: /* handled by sendsig_sigcontext */
1.534 christos 774: default: /* unknown version */
775: printf("nsendsig: bad version %d\n",
776: ps->sa_sigdesc[sig].sd_vers);
777: sigexit(l, SIGILL);
778: case 2:
779: break;
780: }
781:
782: frame.sf_ra = (int)ps->sa_sigdesc[sig].sd_tramp;
783: frame.sf_signum = sig;
784: frame.sf_sip = &fp->sf_si;
785: frame.sf_ucp = &fp->sf_uc;
1.540 thorpej 786: frame.sf_si._info = ksi->ksi_info;
1.534 christos 787: frame.sf_uc.uc_flags = _UC_SIGMASK|_UC_VM;
788: frame.sf_uc.uc_sigmask = *mask;
789: frame.sf_uc.uc_link = NULL;
790: frame.sf_uc.uc_flags |= (p->p_sigctx.ps_sigstk.ss_flags & SS_ONSTACK)
791: ? _UC_SETSTACK : _UC_CLRSTACK;
792: memset(&frame.sf_uc.uc_stack, 0, sizeof(frame.sf_uc.uc_stack));
793: cpu_getmcontext(l, &frame.sf_uc.uc_mcontext, &frame.sf_uc.uc_flags);
794:
795: if (tf->tf_eflags & PSL_VM)
796: (*p->p_emul->e_syscall_intern)(p);
797:
798: if (copyout(&frame, fp, sizeof(frame)) != 0) {
799: /*
800: * Process has trashed its stack; give it an illegal
801: * instruction to halt it in its tracks.
802: */
803: sigexit(l, SIGILL);
804: /* NOTREACHED */
805: }
806:
807: buildcontext(l, sel, catcher, fp);
1.319 mycroft 808:
809: /* Remember that we're now on the signal stack. */
810: if (onstack)
1.425 jdolecek 811: p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
1.1 cgd 812: }
813:
1.534 christos 814: void
1.539 christos 815: sendsig(const ksiginfo_t *ksi, const sigset_t *mask)
1.534 christos 816: {
817: #ifdef COMPAT_16
818: if (curproc->p_sigacts->sa_sigdesc[ksi->ksi_signo].sd_vers < 2)
819: sendsig_sigcontext(ksi, mask);
820: else
821: #endif
822: sendsig_siginfo(ksi, mask);
823: }
1.508 thorpej 824:
1.550 junyoung 825: void
1.534 christos 826: cpu_upcall(struct lwp *l, int type, int nevents, int ninterrupted, void *sas,
827: void *ap, void *sp, sa_upcall_t upcall)
1.508 thorpej 828: {
1.533 chs 829: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 830: struct saframe *sf, frame;
831: struct trapframe *tf;
832:
833: tf = l->l_md.md_regs;
834:
835: /* Finally, copy out the rest of the frame. */
836: frame.sa_type = type;
837: frame.sa_sas = sas;
838: frame.sa_events = nevents;
839: frame.sa_interrupted = ninterrupted;
840: frame.sa_arg = ap;
841: frame.sa_ra = 0;
1.550 junyoung 842:
1.508 thorpej 843: sf = (struct saframe *)sp - 1;
844: if (copyout(&frame, sf, sizeof(frame)) != 0) {
845: /* Copying onto the stack didn't work. Die. */
846: sigexit(l, SIGILL);
847: /* NOTREACHED */
848: }
849:
850: tf->tf_eip = (int) upcall;
851: tf->tf_esp = (int) sf;
852: tf->tf_ebp = 0; /* indicate call-frame-top to debuggers */
853: tf->tf_gs = GSEL(GUDATA_SEL, SEL_UPL);
854: tf->tf_fs = GSEL(GUDATA_SEL, SEL_UPL);
855: tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
856: tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
1.533 chs 857: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
858: GSEL(GUCODEBIG_SEL, SEL_UPL) : GSEL(GUCODE_SEL, SEL_UPL);
1.508 thorpej 859: tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
860: tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
861: }
1.37 cgd 862:
1.1 cgd 863: int waittime = -1;
864: struct pcb dumppcb;
865:
1.32 andrew 866: void
1.586 christos 867: cpu_reboot(int howto, char *bootstr)
1.1 cgd 868: {
869:
1.106 mycroft 870: if (cold) {
1.193 mycroft 871: howto |= RB_HALT;
872: goto haltsys;
1.1 cgd 873: }
1.193 mycroft 874:
1.106 mycroft 875: boothowto = howto;
1.193 mycroft 876: if ((howto & RB_NOSYNC) == 0 && waittime < 0) {
1.1 cgd 877: waittime = 0;
1.150 mycroft 878: vfs_shutdown();
1.521 dsl 879: /*
880: * If we've been adjusting the clock, the todr
881: * will be out of synch; adjust it now.
882: */
883: if (time_adjusted != 0)
884: resettodr();
1.1 cgd 885: }
1.193 mycroft 886:
887: /* Disable interrupts. */
1.1 cgd 888: splhigh();
1.193 mycroft 889:
890: /* Do a dump if requested. */
891: if ((howto & (RB_DUMP | RB_HALT)) == RB_DUMP)
892: dumpsys();
893:
894: haltsys:
895: doshutdownhooks();
896:
1.484 fvdl 897: #ifdef MULTIPROCESSOR
1.514 fvdl 898: x86_broadcast_ipi(X86_IPI_HALT);
1.484 fvdl 899: #endif
900:
1.307 thorpej 901: if ((howto & RB_POWERDOWN) == RB_POWERDOWN) {
1.473 tshiozak 902: #if NACPI > 0
1.530 kochi 903: if (acpi_softc != NULL) {
904: delay(500000);
905: acpi_enter_sleep_state(acpi_softc, ACPI_STATE_S5);
906: printf("WARNING: ACPI powerdown failed!\n");
907: }
1.473 tshiozak 908: #endif
1.576 christos 909: #if NAPMBIOS > 0 && !defined(APM_NO_POWEROFF)
1.208 jtk 910: /* turn off, if we can. But try to turn disk off and
911: * wait a bit first--some disk drives are slow to clean up
912: * and users have reported disk corruption.
913: */
914: delay(500000);
1.578 xtraeme 915: apm_set_powstate(NULL, APM_DEV_DISK(APM_DEV_ALLUNITS), APM_SYS_OFF);
1.208 jtk 916: delay(500000);
1.576 christos 917: apm_set_powstate(NULL, APM_DEV_ALLDEVS, APM_SYS_OFF);
1.531 kochi 918: printf("WARNING: APM powerdown failed!\n");
1.307 thorpej 919: /*
920: * RB_POWERDOWN implies RB_HALT... fall into it...
921: */
1.208 jtk 922: #endif
1.307 thorpej 923: }
924:
925: if (howto & RB_HALT) {
1.210 christos 926: printf("\n");
927: printf("The operating system has halted.\n");
928: printf("Please press any key to reboot.\n\n");
1.517 jmmv 929:
930: #ifdef BEEP_ONHALT
931: {
932: int c;
933: for (c = BEEP_ONHALT_COUNT; c > 0; c--) {
934: sysbeep(BEEP_ONHALT_PITCH,
1.565 simonb 935: BEEP_ONHALT_PERIOD * hz / 1000);
1.517 jmmv 936: delay(BEEP_ONHALT_PERIOD * 1000);
937: sysbeep(0, BEEP_ONHALT_PERIOD * hz / 1000);
938: delay(BEEP_ONHALT_PERIOD * 1000);
939: }
940: }
941: #endif
942:
1.300 drochner 943: cnpollc(1); /* for proper keyboard command handling */
1.512 perry 944: if (cngetc() == 0) {
945: /* no console attached, so just hlt */
946: for(;;) {
1.567 perry 947: __asm volatile("hlt");
1.512 perry 948: }
949: }
1.300 drochner 950: cnpollc(0);
1.1 cgd 951: }
1.193 mycroft 952:
1.210 christos 953: printf("rebooting...\n");
1.328 bouyer 954: if (cpureset_delay > 0)
955: delay(cpureset_delay * 1000);
1.1 cgd 956: cpu_reset();
957: for(;;) ;
958: /*NOTREACHED*/
959: }
960:
1.116 gwr 961: /*
962: * These variables are needed by /sbin/savecore
963: */
1.568 perry 964: uint32_t dumpmag = 0x8fca0101; /* magic number */
1.116 gwr 965: int dumpsize = 0; /* pages */
966: long dumplo = 0; /* blocks */
967:
968: /*
1.291 thorpej 969: * cpu_dumpsize: calculate size of machine-dependent kernel core dump headers.
970: */
971: int
972: cpu_dumpsize()
973: {
974: int size;
975:
976: size = ALIGN(sizeof(kcore_seg_t)) + ALIGN(sizeof(cpu_kcore_hdr_t)) +
977: ALIGN(mem_cluster_cnt * sizeof(phys_ram_seg_t));
978: if (roundup(size, dbtob(1)) != dbtob(1))
979: return (-1);
980:
981: return (1);
982: }
983:
984: /*
985: * cpu_dump_mempagecnt: calculate the size of RAM (in pages) to be dumped.
986: */
987: u_long
988: cpu_dump_mempagecnt()
989: {
990: u_long i, n;
991:
992: n = 0;
993: for (i = 0; i < mem_cluster_cnt; i++)
994: n += atop(mem_clusters[i].size);
995: return (n);
996: }
997:
998: /*
999: * cpu_dump: dump the machine-dependent kernel core dump headers.
1000: */
1001: int
1002: cpu_dump()
1003: {
1.542 junyoung 1004: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.563 christos 1005: char bf[dbtob(1)];
1.291 thorpej 1006: kcore_seg_t *segp;
1007: cpu_kcore_hdr_t *cpuhdrp;
1008: phys_ram_seg_t *memsegp;
1.481 gehenna 1009: const struct bdevsw *bdev;
1.291 thorpej 1010: int i;
1011:
1.481 gehenna 1012: bdev = bdevsw_lookup(dumpdev);
1013: if (bdev == NULL)
1014: return (ENXIO);
1015: dump = bdev->d_dump;
1.291 thorpej 1016:
1.563 christos 1017: memset(bf, 0, sizeof bf);
1018: segp = (kcore_seg_t *)bf;
1019: cpuhdrp = (cpu_kcore_hdr_t *)&bf[ALIGN(sizeof(*segp))];
1020: memsegp = (phys_ram_seg_t *)&bf[ ALIGN(sizeof(*segp)) +
1.291 thorpej 1021: ALIGN(sizeof(*cpuhdrp))];
1022:
1023: /*
1024: * Generate a segment header.
1025: */
1026: CORE_SETMAGIC(*segp, KCORE_MAGIC, MID_MACHINE, CORE_CPU);
1027: segp->c_size = dbtob(1) - ALIGN(sizeof(*segp));
1028:
1029: /*
1030: * Add the machine-dependent header info.
1031: */
1.558 junyoung 1032: cpuhdrp->pdppaddr = PDPpaddr;
1.291 thorpej 1033: cpuhdrp->nmemsegs = mem_cluster_cnt;
1034:
1035: /*
1036: * Fill in the memory segment descriptors.
1037: */
1038: for (i = 0; i < mem_cluster_cnt; i++) {
1039: memsegp[i].start = mem_clusters[i].start;
1040: memsegp[i].size = mem_clusters[i].size;
1041: }
1042:
1.563 christos 1043: return (dump(dumpdev, dumplo, (caddr_t)bf, dbtob(1)));
1.291 thorpej 1044: }
1045:
1046: /*
1.228 gwr 1047: * This is called by main to set dumplo and dumpsize.
1.414 thorpej 1048: * Dumps always skip the first PAGE_SIZE of disk space
1.116 gwr 1049: * in case there might be a disk label stored there.
1050: * If there is extra space, put dump at the end to
1051: * reduce the chance that swapping trashes it.
1052: */
1053: void
1.228 gwr 1054: cpu_dumpconf()
1.116 gwr 1055: {
1.481 gehenna 1056: const struct bdevsw *bdev;
1.291 thorpej 1057: int nblks, dumpblks; /* size of dump area */
1.116 gwr 1058:
1059: if (dumpdev == NODEV)
1.291 thorpej 1060: goto bad;
1.481 gehenna 1061: bdev = bdevsw_lookup(dumpdev);
1.582 mrg 1062: if (bdev == NULL) {
1063: dumpdev = NODEV;
1.583 skrll 1064: goto bad;
1.582 mrg 1065: }
1.481 gehenna 1066: if (bdev->d_psize == NULL)
1.291 thorpej 1067: goto bad;
1.481 gehenna 1068: nblks = (*bdev->d_psize)(dumpdev);
1.116 gwr 1069: if (nblks <= ctod(1))
1.291 thorpej 1070: goto bad;
1.116 gwr 1071:
1.291 thorpej 1072: dumpblks = cpu_dumpsize();
1073: if (dumpblks < 0)
1074: goto bad;
1075: dumpblks += ctod(cpu_dump_mempagecnt());
1076:
1077: /* If dump won't fit (incl. room for possible label), punt. */
1078: if (dumpblks > (nblks - ctod(1)))
1079: goto bad;
1080:
1081: /* Put dump at end of partition */
1082: dumplo = nblks - dumpblks;
1083:
1084: /* dumpsize is in page units, and doesn't include headers. */
1085: dumpsize = cpu_dump_mempagecnt();
1086: return;
1.116 gwr 1087:
1.291 thorpej 1088: bad:
1089: dumpsize = 0;
1.116 gwr 1090: }
1091:
1.1 cgd 1092: /*
1093: * Doadump comes here after turning off memory management and
1094: * getting on the dump stack, either when called above, or by
1095: * the auto-restart code.
1096: */
1.414 thorpej 1097: #define BYTES_PER_DUMP PAGE_SIZE /* must be a multiple of pagesize XXX small */
1.314 thorpej 1098: static vaddr_t dumpspace;
1.163 cgd 1099:
1.314 thorpej 1100: vaddr_t
1.551 junyoung 1101: reserve_dumppages(vaddr_t p)
1.163 cgd 1102: {
1103:
1104: dumpspace = p;
1105: return (p + BYTES_PER_DUMP);
1106: }
1107:
1.32 andrew 1108: void
1.1 cgd 1109: dumpsys()
1110: {
1.556 mycroft 1111: u_long totalbytesleft, bytes, i, n, m, memseg;
1.291 thorpej 1112: u_long maddr;
1113: int psize;
1.163 cgd 1114: daddr_t blkno;
1.481 gehenna 1115: const struct bdevsw *bdev;
1.542 junyoung 1116: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.200 christos 1117: int error;
1.193 mycroft 1118:
1119: /* Save registers. */
1120: savectx(&dumppcb);
1.1 cgd 1121:
1122: if (dumpdev == NODEV)
1123: return;
1.484 fvdl 1124:
1.481 gehenna 1125: bdev = bdevsw_lookup(dumpdev);
1126: if (bdev == NULL || bdev->d_psize == NULL)
1127: return;
1.163 cgd 1128:
1129: /*
1130: * For dumps during autoconfiguration,
1131: * if dump device has already configured...
1132: */
1133: if (dumpsize == 0)
1.228 gwr 1134: cpu_dumpconf();
1.330 jtk 1135: if (dumplo <= 0 || dumpsize == 0) {
1.275 mycroft 1136: printf("\ndump to dev %u,%u not possible\n", major(dumpdev),
1137: minor(dumpdev));
1.163 cgd 1138: return;
1.275 mycroft 1139: }
1140: printf("\ndumping to dev %u,%u offset %ld\n", major(dumpdev),
1141: minor(dumpdev), dumplo);
1.134 mycroft 1142:
1.481 gehenna 1143: psize = (*bdev->d_psize)(dumpdev);
1.210 christos 1144: printf("dump ");
1.163 cgd 1145: if (psize == -1) {
1.210 christos 1146: printf("area unavailable\n");
1.163 cgd 1147: return;
1148: }
1149:
1150: #if 0 /* XXX this doesn't work. grr. */
1.565 simonb 1151: /* toss any characters present prior to dump */
1.163 cgd 1152: while (sget() != NULL); /*syscons and pccons differ */
1153: #endif
1154:
1.291 thorpej 1155: if ((error = cpu_dump()) != 0)
1156: goto err;
1157:
1158: totalbytesleft = ptoa(cpu_dump_mempagecnt());
1159: blkno = dumplo + cpu_dumpsize();
1.481 gehenna 1160: dump = bdev->d_dump;
1.200 christos 1161: error = 0;
1.291 thorpej 1162:
1163: for (memseg = 0; memseg < mem_cluster_cnt; memseg++) {
1164: maddr = mem_clusters[memseg].start;
1165: bytes = mem_clusters[memseg].size;
1166:
1167: for (i = 0; i < bytes; i += n, totalbytesleft -= n) {
1168: /* Print out how many MBs we have left to go. */
1169: if ((totalbytesleft % (1024*1024)) == 0)
1170: printf("%ld ", totalbytesleft / (1024 * 1024));
1171:
1172: /* Limit size for next transfer. */
1173: n = bytes - i;
1174: if (n > BYTES_PER_DUMP)
1175: n = BYTES_PER_DUMP;
1176:
1.556 mycroft 1177: for (m = 0; m < n; m += NBPG)
1178: pmap_kenter_pa(dumpspace + m, maddr + m,
1179: VM_PROT_READ);
1.291 thorpej 1180:
1181: error = (*dump)(dumpdev, blkno, (caddr_t)dumpspace, n);
1182: if (error)
1183: goto err;
1.163 cgd 1184: maddr += n;
1.291 thorpej 1185: blkno += btodb(n); /* XXX? */
1.163 cgd 1186:
1187: #if 0 /* XXX this doesn't work. grr. */
1.291 thorpej 1188: /* operator aborting dump? */
1189: if (sget() != NULL) {
1190: error = EINTR;
1191: break;
1192: }
1193: #endif
1.163 cgd 1194: }
1195: }
1196:
1.291 thorpej 1197: err:
1.163 cgd 1198: switch (error) {
1.1 cgd 1199:
1200: case ENXIO:
1.210 christos 1201: printf("device bad\n");
1.1 cgd 1202: break;
1203:
1204: case EFAULT:
1.210 christos 1205: printf("device not ready\n");
1.1 cgd 1206: break;
1207:
1208: case EINVAL:
1.210 christos 1209: printf("area improper\n");
1.1 cgd 1210: break;
1211:
1212: case EIO:
1.210 christos 1213: printf("i/o error\n");
1.1 cgd 1214: break;
1215:
1216: case EINTR:
1.210 christos 1217: printf("aborted from console\n");
1.1 cgd 1218: break;
1219:
1.163 cgd 1220: case 0:
1.210 christos 1221: printf("succeeded\n");
1.163 cgd 1222: break;
1223:
1.1 cgd 1224: default:
1.210 christos 1225: printf("error %d\n", error);
1.1 cgd 1226: break;
1227: }
1.210 christos 1228: printf("\n\n");
1.163 cgd 1229: delay(5000000); /* 5 seconds */
1.1 cgd 1230: }
1231:
1232: /*
1233: * Clear registers on exec
1234: */
1.33 cgd 1235: void
1.551 junyoung 1236: setregs(struct lwp *l, struct exec_package *pack, u_long stack)
1.1 cgd 1237: {
1.533 chs 1238: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 1239: struct pcb *pcb = &l->l_addr->u_pcb;
1.298 mycroft 1240: struct trapframe *tf;
1.1 cgd 1241:
1.161 mycroft 1242: #if NNPX > 0
1243: /* If we were using the FPU, forget about it. */
1.508 thorpej 1244: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
1245: npxsave_lwp(l, 0);
1.161 mycroft 1246: #endif
1.166 mycroft 1247:
1.178 mycroft 1248: #ifdef USER_LDT
1.508 thorpej 1249: pmap_ldt_cleanup(l);
1.178 mycroft 1250: #endif
1251:
1.548 yamt 1252: l->l_md.md_flags &= ~MDL_USEDFPU;
1.452 thorpej 1253: if (i386_use_fxsave) {
1.450 thorpej 1254: pcb->pcb_savefpu.sv_xmm.sv_env.en_cw = __NetBSD_NPXCW__;
1.452 thorpej 1255: pcb->pcb_savefpu.sv_xmm.sv_env.en_mxcsr = __INITIAL_MXCSR__;
1256: } else
1.450 thorpej 1257: pcb->pcb_savefpu.sv_87.sv_env.en_cw = __NetBSD_NPXCW__;
1.59 mycroft 1258:
1.508 thorpej 1259: tf = l->l_md.md_regs;
1.445 sommerfe 1260: tf->tf_gs = LSEL(LUDATA_SEL, SEL_UPL);
1261: tf->tf_fs = LSEL(LUDATA_SEL, SEL_UPL);
1.154 mycroft 1262: tf->tf_es = LSEL(LUDATA_SEL, SEL_UPL);
1263: tf->tf_ds = LSEL(LUDATA_SEL, SEL_UPL);
1.252 mycroft 1264: tf->tf_edi = 0;
1265: tf->tf_esi = 0;
1.154 mycroft 1266: tf->tf_ebp = 0;
1.508 thorpej 1267: tf->tf_ebx = (int)l->l_proc->p_psstr;
1.252 mycroft 1268: tf->tf_edx = 0;
1269: tf->tf_ecx = 0;
1270: tf->tf_eax = 0;
1.154 mycroft 1271: tf->tf_eip = pack->ep_entry;
1.533 chs 1272: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
1273: LSEL(LUCODEBIG_SEL, SEL_UPL) : LSEL(LUCODE_SEL, SEL_UPL);
1.154 mycroft 1274: tf->tf_eflags = PSL_USERSET;
1275: tf->tf_esp = stack;
1276: tf->tf_ss = LSEL(LUDATA_SEL, SEL_UPL);
1.1 cgd 1277: }
1278:
1279: /*
1.55 brezak 1280: * Initialize segments and descriptor tables
1.1 cgd 1281: */
1282:
1.516 fvdl 1283: union descriptor *gdt, *ldt;
1284: struct gate_descriptor *idt;
1.498 fvdl 1285: char idt_allocmap[NIDT];
1286: struct simplelock idt_lock = SIMPLELOCK_INITIALIZER;
1.275 mycroft 1287: #ifdef I586_CPU
1288: union descriptor *pentium_idt;
1289: #endif
1.573 yamt 1290: struct user *proc0paddr;
1291: extern vaddr_t proc0uarea;
1.49 brezak 1292:
1.178 mycroft 1293: void
1.551 junyoung 1294: setgate(struct gate_descriptor *gd, void *func, int args, int type, int dpl,
1295: int sel)
1.178 mycroft 1296: {
1.1 cgd 1297:
1.178 mycroft 1298: gd->gd_looffset = (int)func;
1.489 fvdl 1299: gd->gd_selector = sel;
1.178 mycroft 1300: gd->gd_stkcpy = args;
1301: gd->gd_xx = 0;
1302: gd->gd_type = type;
1303: gd->gd_dpl = dpl;
1304: gd->gd_p = 1;
1305: gd->gd_hioffset = (int)func >> 16;
1306: }
1307:
1308: void
1.551 junyoung 1309: unsetgate(struct gate_descriptor *gd)
1.484 fvdl 1310: {
1311: gd->gd_p = 0;
1312: gd->gd_hioffset = 0;
1313: gd->gd_looffset = 0;
1314: gd->gd_selector = 0;
1315: gd->gd_xx = 0;
1316: gd->gd_stkcpy = 0;
1317: gd->gd_type = 0;
1318: gd->gd_dpl = 0;
1319: }
1320:
1321:
1322: void
1.551 junyoung 1323: setregion(struct region_descriptor *rd, void *base, size_t limit)
1.178 mycroft 1324: {
1325:
1326: rd->rd_limit = (int)limit;
1327: rd->rd_base = (int)base;
1328: }
1.1 cgd 1329:
1.174 mycroft 1330: void
1.551 junyoung 1331: setsegment(struct segment_descriptor *sd, void *base, size_t limit, int type,
1332: int dpl, int def32, int gran)
1.174 mycroft 1333: {
1.1 cgd 1334:
1.174 mycroft 1335: sd->sd_lolimit = (int)limit;
1336: sd->sd_lobase = (int)base;
1337: sd->sd_type = type;
1338: sd->sd_dpl = dpl;
1339: sd->sd_p = 1;
1340: sd->sd_hilimit = (int)limit >> 16;
1341: sd->sd_xx = 0;
1342: sd->sd_def32 = def32;
1343: sd->sd_gran = gran;
1344: sd->sd_hibase = (int)base >> 24;
1345: }
1.1 cgd 1346:
1347: #define IDTVEC(name) __CONCAT(X, name)
1.542 junyoung 1348: typedef void (vector)(void);
1.299 mycroft 1349: extern vector IDTVEC(syscall);
1350: extern vector IDTVEC(osyscall);
1351: extern vector *IDTVEC(exceptions)[];
1.333 christos 1352: #ifdef COMPAT_SVR4
1353: extern vector IDTVEC(svr4_fasttrap);
1354: #endif /* COMPAT_SVR4 */
1.447 christos 1355: #ifdef COMPAT_MACH
1356: extern vector IDTVEC(mach_trap);
1357: #endif
1.1 cgd 1358:
1.381 thorpej 1359: #define KBTOB(x) ((size_t)(x) * 1024UL)
1360:
1.484 fvdl 1361: void cpu_init_idt()
1362: {
1363: struct region_descriptor region;
1364: #ifdef I586_CPU
1365: setregion(®ion, pentium_idt, NIDT * sizeof(idt[0]) - 1);
1366: #else
1367: setregion(®ion, idt, NIDT * sizeof(idt[0]) - 1);
1368: #endif
1.565 simonb 1369: lidt(®ion);
1.484 fvdl 1370: }
1371:
1.433 kanaoka 1372: void
1.568 perry 1373: add_mem_cluster(uint64_t seg_start, uint64_t seg_end, uint32_t type)
1.433 kanaoka 1374: {
1375: extern struct extent *iomem_ex;
1.492 kanaoka 1376: int i;
1.433 kanaoka 1377:
1378: if (seg_end > 0x100000000ULL) {
1379: printf("WARNING: skipping large "
1380: "memory map entry: "
1381: "0x%qx/0x%qx/0x%x\n",
1382: seg_start,
1383: (seg_end - seg_start),
1384: type);
1385: return;
1386: }
1387:
1388: /*
1389: * XXX Chop the last page off the size so that
1390: * XXX it can fit in avail_end.
1391: */
1392: if (seg_end == 0x100000000ULL)
1393: seg_end -= PAGE_SIZE;
1394:
1395: if (seg_end <= seg_start)
1396: return;
1397:
1.492 kanaoka 1398: for (i = 0; i < mem_cluster_cnt; i++) {
1399: if ((mem_clusters[i].start == round_page(seg_start))
1400: && (mem_clusters[i].size
1401: == trunc_page(seg_end) - mem_clusters[i].start)) {
1402: #ifdef DEBUG_MEMLOAD
1403: printf("WARNING: skipping duplicate segment entry\n");
1404: #endif
1405: return;
1406: }
1407: }
1408:
1.433 kanaoka 1409: /*
1410: * Allocate the physical addresses used by RAM
1411: * from the iomem extent map. This is done before
1412: * the addresses are page rounded just to make
1413: * sure we get them all.
1414: */
1415: if (extent_alloc_region(iomem_ex, seg_start,
1416: seg_end - seg_start, EX_NOWAIT)) {
1417: /* XXX What should we do? */
1418: printf("WARNING: CAN'T ALLOCATE "
1419: "MEMORY SEGMENT "
1420: "(0x%qx/0x%qx/0x%x) FROM "
1421: "IOMEM EXTENT MAP!\n",
1422: seg_start, seg_end - seg_start, type);
1.492 kanaoka 1423: return;
1.433 kanaoka 1424: }
1425:
1426: /*
1427: * If it's not free memory, skip it.
1428: */
1429: if (type != BIM_Memory)
1430: return;
1431:
1432: /* XXX XXX XXX */
1433: if (mem_cluster_cnt >= VM_PHYSSEG_MAX)
1.579 cube 1434: panic("init386: too many memory segments "
1435: "(increase VM_PHYSSEG_MAX)");
1.433 kanaoka 1436:
1437: seg_start = round_page(seg_start);
1438: seg_end = trunc_page(seg_end);
1439:
1440: if (seg_start == seg_end)
1441: return;
1442:
1443: mem_clusters[mem_cluster_cnt].start = seg_start;
1444: mem_clusters[mem_cluster_cnt].size =
1445: seg_end - seg_start;
1446:
1447: if (avail_end < seg_end)
1448: avail_end = seg_end;
1449: physmem += atop(mem_clusters[mem_cluster_cnt].size);
1450: mem_cluster_cnt++;
1451: }
1452:
1.59 mycroft 1453: void
1.484 fvdl 1454: initgdt(union descriptor *tgdt)
1455: {
1456: struct region_descriptor region;
1457: gdt = tgdt;
1458: memset(gdt, 0, NGDT*sizeof(*gdt));
1459: /* make gdt gates and memory segments */
1460: setsegment(&gdt[GCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 1, 1);
1461: setsegment(&gdt[GDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 1, 1);
1.533 chs 1462: setsegment(&gdt[GUCODE_SEL].sd, 0, x86_btop(I386_MAX_EXE_ADDR) - 1,
1463: SDT_MEMERA, SEL_UPL, 1, 1);
1464: setsegment(&gdt[GUCODEBIG_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1465: SDT_MEMERA, SEL_UPL, 1, 1);
1.514 fvdl 1466: setsegment(&gdt[GUDATA_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1467: SDT_MEMRWA, SEL_UPL, 1, 1);
1468: #ifdef COMPAT_MACH
1469: setgate(&gdt[GMACHCALLS_SEL].gd, &IDTVEC(mach_trap), 1,
1.489 fvdl 1470: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 1471: #endif
1472: #if NBIOSCALL > 0
1473: /* bios trampoline GDT entries */
1474: setsegment(&gdt[GBIOSCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 0,
1475: 0);
1476: setsegment(&gdt[GBIOSDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 0,
1477: 0);
1478: #endif
1479: setsegment(&gdt[GCPU_SEL].sd, &cpu_info_primary,
1480: sizeof(struct cpu_info)-1, SDT_MEMRWA, SEL_KPL, 1, 1);
1481:
1482: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
1483: lgdt(®ion);
1484: }
1485:
1486: void
1.551 junyoung 1487: init386(paddr_t first_avail)
1.2 cgd 1488: {
1.484 fvdl 1489: union descriptor *tgdt;
1.542 junyoung 1490: extern void consinit(void);
1.375 drochner 1491: extern struct extent *iomem_ex;
1.401 thorpej 1492: struct btinfo_memmap *bim;
1.381 thorpej 1493: struct region_descriptor region;
1.401 thorpej 1494: int x, first16q;
1.568 perry 1495: uint64_t seg_start, seg_end;
1496: uint64_t seg_start1, seg_end1;
1.473 tshiozak 1497: paddr_t realmode_reserved_start;
1498: psize_t realmode_reserved_size;
1499: int needs_earlier_install_pte0;
1.436 jdolecek 1500: #if NBIOSCALL > 0
1501: extern int biostramp_image_size;
1502: extern u_char biostramp_image[];
1503: #endif
1.1 cgd 1504:
1.484 fvdl 1505: cpu_probe_features(&cpu_info_primary);
1506: cpu_feature = cpu_info_primary.ci_feature_flags;
1.553 lukem 1507: cpu_feature2 = cpu_info_primary.ci_feature2_flags;
1.484 fvdl 1508:
1.573 yamt 1509: proc0paddr = UAREA_TO_USER(proc0uarea);
1.508 thorpej 1510: lwp0.l_addr = proc0paddr;
1511: cpu_info_primary.ci_curpcb = &lwp0.l_addr->u_pcb;
1.275 mycroft 1512:
1.514 fvdl 1513: x86_bus_space_init();
1.84 cgd 1514: consinit(); /* XXX SHOULD NOT BE DONE HERE */
1.401 thorpej 1515: /*
1516: * Initailize PAGE_SIZE-dependent variables.
1517: */
1518: uvm_setpagesize();
1.450 thorpej 1519:
1520: /*
1521: * Saving SSE registers won't work if the save area isn't
1522: * 16-byte aligned.
1523: */
1524: if (offsetof(struct user, u_pcb.pcb_savefpu) & 0xf)
1525: panic("init386: pcb_savefpu not 16-byte aligned");
1.440 thorpej 1526:
1527: /*
1528: * Start with 2 color bins -- this is just a guess to get us
1529: * started. We'll recolor when we determine the largest cache
1530: * sizes on the system.
1531: */
1532: uvmexp.ncolors = 2;
1.401 thorpej 1533:
1.473 tshiozak 1534: /*
1.484 fvdl 1535: * BIOS leaves data in physical page 0
1536: * Even if it didn't, our VM system doesn't like using zero as a
1537: * physical page number.
1538: * We may also need pages in low memory (one each) for secondary CPU
1539: * startup, for BIOS calls, and for ACPI, plus a page table page to map
1540: * them into the first few pages of the kernel's pmap.
1.473 tshiozak 1541: */
1542: avail_start = PAGE_SIZE;
1543:
1544: /*
1545: * reserve memory for real-mode call
1546: */
1547: needs_earlier_install_pte0 = 0;
1548: realmode_reserved_start = 0;
1549: realmode_reserved_size = 0;
1.414 thorpej 1550: #if NBIOSCALL > 0
1.473 tshiozak 1551: /* save us a page for trampoline code */
1552: realmode_reserved_size += PAGE_SIZE;
1553: needs_earlier_install_pte0 = 1;
1554: #endif
1.484 fvdl 1555: #ifdef MULTIPROCESSOR /* XXX */
1556: KASSERT(avail_start == PAGE_SIZE); /* XXX */
1557: if (realmode_reserved_size < MP_TRAMPOLINE) /* XXX */
1558: realmode_reserved_size = MP_TRAMPOLINE; /* XXX */
1559: needs_earlier_install_pte0 = 1; /* XXX */
1560: #endif /* XXX */
1.473 tshiozak 1561: #if NACPI > 0
1562: /* trampoline code for wake handler */
1563: realmode_reserved_size += ptoa(acpi_md_get_npages_of_wakecode()+1);
1564: needs_earlier_install_pte0 = 1;
1565: #endif
1566: if (needs_earlier_install_pte0) {
1567: /* page table for directory entry 0 */
1568: realmode_reserved_size += PAGE_SIZE;
1569: }
1570: if (realmode_reserved_size>0) {
1571: realmode_reserved_start = avail_start;
1572: avail_start += realmode_reserved_size;
1573: }
1.414 thorpej 1574:
1.492 kanaoka 1575: #ifdef DEBUG_MEMLOAD
1576: printf("mem_cluster_count: %d\n", mem_cluster_cnt);
1577: #endif
1578:
1.401 thorpej 1579: /*
1580: * Call pmap initialization to make new kernel address space.
1581: * We must do this before loading pages into the VM system.
1582: */
1.314 thorpej 1583: pmap_bootstrap((vaddr_t)atdevbase + IOM_SIZE);
1.2 cgd 1584:
1.401 thorpej 1585: /*
1586: * Check to see if we have a memory map from the BIOS (passed
1587: * to us by the boot program.
1588: */
1.577 mrg 1589: if ((biosmem_implicit || (biosbasemem == 0 && biosextmem == 0)) &&
1.575 mrg 1590: (bim = lookup_bootinfo(BTINFO_MEMMAP)) != NULL && bim->num > 0) {
1.463 enami 1591: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1592: printf("BIOS MEMORY MAP (%d ENTRIES):\n", bim->num);
1593: #endif
1594: for (x = 0; x < bim->num; x++) {
1.463 enami 1595: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1596: printf(" addr 0x%qx size 0x%qx type 0x%x\n",
1597: bim->entry[x].addr,
1598: bim->entry[x].size,
1599: bim->entry[x].type);
1600: #endif
1601:
1602: /*
1603: * If the segment is not memory, skip it.
1604: */
1605: switch (bim->entry[x].type) {
1606: case BIM_Memory:
1607: case BIM_ACPI:
1608: case BIM_NVS:
1609: break;
1610: default:
1611: continue;
1612: }
1613:
1614: /*
1615: * Sanity check the entry.
1616: * XXX Need to handle uint64_t in extent code
1617: * XXX and 64-bit physical addresses in i386
1618: * XXX port.
1619: */
1620: seg_start = bim->entry[x].addr;
1621: seg_end = bim->entry[x].addr + bim->entry[x].size;
1622:
1.432 kanaoka 1623: /*
1.433 kanaoka 1624: * Avoid Compatibility Holes.
1625: * XXX Holes within memory space that allow access
1626: * XXX to be directed to the PC-compatible frame buffer
1.484 fvdl 1627: * XXX (0xa0000-0xbffff),to adapter ROM space
1.433 kanaoka 1628: * XXX (0xc0000-0xdffff), and to system BIOS space
1629: * XXX (0xe0000-0xfffff).
1630: * XXX Some laptop(for example,Toshiba Satellite2550X)
1631: * XXX report this area and occurred problems,
1632: * XXX so we avoid this area.
1.432 kanaoka 1633: */
1.433 kanaoka 1634: if (seg_start < 0x100000 && seg_end > 0xa0000) {
1635: printf("WARNING: memory map entry overlaps "
1636: "with ``Compatibility Holes'': "
1637: "0x%qx/0x%qx/0x%x\n", seg_start,
1638: seg_end - seg_start, bim->entry[x].type);
1639: add_mem_cluster(seg_start, 0xa0000,
1640: bim->entry[x].type);
1641: add_mem_cluster(0x100000, seg_end,
1.401 thorpej 1642: bim->entry[x].type);
1.433 kanaoka 1643: } else
1644: add_mem_cluster(seg_start, seg_end,
1.401 thorpej 1645: bim->entry[x].type);
1646: }
1.421 aymeric 1647: }
1.575 mrg 1648:
1.421 aymeric 1649: /*
1650: * If the loop above didn't find any valid segment, fall back to
1651: * former code.
1652: */
1653: if (mem_cluster_cnt == 0) {
1.401 thorpej 1654: /*
1655: * Allocate the physical addresses used by RAM from the iomem
1656: * extent map. This is done before the addresses are
1657: * page rounded just to make sure we get them all.
1658: */
1659: if (extent_alloc_region(iomem_ex, 0, KBTOB(biosbasemem),
1660: EX_NOWAIT)) {
1661: /* XXX What should we do? */
1662: printf("WARNING: CAN'T ALLOCATE BASE MEMORY FROM "
1663: "IOMEM EXTENT MAP!\n");
1664: }
1665: mem_clusters[0].start = 0;
1666: mem_clusters[0].size = trunc_page(KBTOB(biosbasemem));
1667: physmem += atop(mem_clusters[0].size);
1668: if (extent_alloc_region(iomem_ex, IOM_END, KBTOB(biosextmem),
1669: EX_NOWAIT)) {
1670: /* XXX What should we do? */
1671: printf("WARNING: CAN'T ALLOCATE EXTENDED MEMORY FROM "
1672: "IOMEM EXTENT MAP!\n");
1673: }
1674: #if NISADMA > 0
1675: /*
1676: * Some motherboards/BIOSes remap the 384K of RAM that would
1677: * normally be covered by the ISA hole to the end of memory
1678: * so that it can be used. However, on a 16M system, this
1679: * would cause bounce buffers to be allocated and used.
1680: * This is not desirable behaviour, as more than 384K of
1681: * bounce buffers might be allocated. As a work-around,
1682: * we round memory down to the nearest 1M boundary if
1683: * we're using any isadma devices and the remapped memory
1684: * is what puts us over 16M.
1685: */
1686: if (biosextmem > (15*1024) && biosextmem < (16*1024)) {
1687: char pbuf[9];
1688:
1689: format_bytes(pbuf, sizeof(pbuf),
1690: biosextmem - (15*1024));
1691: printf("Warning: ignoring %s of remapped memory\n",
1692: pbuf);
1693: biosextmem = (15*1024);
1694: }
1695: #endif
1696: mem_clusters[1].start = IOM_END;
1697: mem_clusters[1].size = trunc_page(KBTOB(biosextmem));
1698: physmem += atop(mem_clusters[1].size);
1699:
1700: mem_cluster_cnt = 2;
1701:
1702: avail_end = IOM_END + trunc_page(KBTOB(biosextmem));
1703: }
1704: /*
1705: * If we have 16M of RAM or less, just put it all on
1706: * the default free list. Otherwise, put the first
1707: * 16M of RAM on a lower priority free list (so that
1708: * all of the ISA DMA'able memory won't be eaten up
1709: * first-off).
1710: */
1711: if (avail_end <= (16 * 1024 * 1024))
1712: first16q = VM_FREELIST_DEFAULT;
1713: else
1714: first16q = VM_FREELIST_FIRST16;
1715:
1716: /* Make sure the end of the space used by the kernel is rounded. */
1717: first_avail = round_page(first_avail);
1718:
1719: /*
1720: * Now, load the memory clusters (which have already been
1721: * rounded and truncated) into the VM system.
1722: *
1723: * NOTE: WE ASSUME THAT MEMORY STARTS AT 0 AND THAT THE KERNEL
1724: * IS LOADED AT IOM_END (1M).
1725: */
1726: for (x = 0; x < mem_cluster_cnt; x++) {
1727: seg_start = mem_clusters[x].start;
1728: seg_end = mem_clusters[x].start + mem_clusters[x].size;
1729: seg_start1 = 0;
1730: seg_end1 = 0;
1731:
1732: /*
1733: * Skip memory before our available starting point.
1734: */
1735: if (seg_end <= avail_start)
1736: continue;
1737:
1738: if (avail_start >= seg_start && avail_start < seg_end) {
1739: if (seg_start != 0)
1740: panic("init386: memory doesn't start at 0");
1741: seg_start = avail_start;
1742: if (seg_start == seg_end)
1743: continue;
1744: }
1745:
1746: /*
1747: * If this segment contains the kernel, split it
1748: * in two, around the kernel.
1749: */
1750: if (seg_start <= IOM_END && first_avail <= seg_end) {
1751: seg_start1 = first_avail;
1752: seg_end1 = seg_end;
1753: seg_end = IOM_END;
1754: }
1755:
1756: /* First hunk */
1757: if (seg_start != seg_end) {
1.480 erh 1758: if (seg_start < (16 * 1024 * 1024) &&
1.401 thorpej 1759: first16q != VM_FREELIST_DEFAULT) {
1.568 perry 1760: uint64_t tmp;
1.401 thorpej 1761:
1762: if (seg_end > (16 * 1024 * 1024))
1763: tmp = (16 * 1024 * 1024);
1764: else
1765: tmp = seg_end;
1.492 kanaoka 1766:
1767: if (tmp != seg_start) {
1.463 enami 1768: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1769: printf("loading 0x%qx-0x%qx "
1770: "(0x%lx-0x%lx)\n",
1771: seg_start, tmp,
1772: atop(seg_start), atop(tmp));
1773: #endif
1774: uvm_page_physload(atop(seg_start),
1775: atop(tmp), atop(seg_start),
1776: atop(tmp), first16q);
1777: }
1.401 thorpej 1778: seg_start = tmp;
1779: }
1.411 enami 1780:
1781: if (seg_start != seg_end) {
1.463 enami 1782: #ifdef DEBUG_MEMLOAD
1.411 enami 1783: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1784: seg_start, seg_end,
1785: atop(seg_start), atop(seg_end));
1786: #endif
1787: uvm_page_physload(atop(seg_start),
1788: atop(seg_end), atop(seg_start),
1789: atop(seg_end), VM_FREELIST_DEFAULT);
1790: }
1.401 thorpej 1791: }
1792:
1793: /* Second hunk */
1794: if (seg_start1 != seg_end1) {
1.480 erh 1795: if (seg_start1 < (16 * 1024 * 1024) &&
1.401 thorpej 1796: first16q != VM_FREELIST_DEFAULT) {
1.568 perry 1797: uint64_t tmp;
1.401 thorpej 1798:
1799: if (seg_end1 > (16 * 1024 * 1024))
1800: tmp = (16 * 1024 * 1024);
1801: else
1802: tmp = seg_end1;
1.492 kanaoka 1803:
1804: if (tmp != seg_start1) {
1.463 enami 1805: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1806: printf("loading 0x%qx-0x%qx "
1807: "(0x%lx-0x%lx)\n",
1808: seg_start1, tmp,
1809: atop(seg_start1), atop(tmp));
1810: #endif
1811: uvm_page_physload(atop(seg_start1),
1812: atop(tmp), atop(seg_start1),
1813: atop(tmp), first16q);
1814: }
1.401 thorpej 1815: seg_start1 = tmp;
1816: }
1.412 enami 1817:
1818: if (seg_start1 != seg_end1) {
1.463 enami 1819: #ifdef DEBUG_MEMLOAD
1.412 enami 1820: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1821: seg_start1, seg_end1,
1822: atop(seg_start1), atop(seg_end1));
1823: #endif
1824: uvm_page_physload(atop(seg_start1),
1825: atop(seg_end1), atop(seg_start1),
1826: atop(seg_end1), VM_FREELIST_DEFAULT);
1827: }
1.401 thorpej 1828: }
1829: }
1830:
1831: /*
1832: * Steal memory for the message buffer (at end of core).
1833: */
1834: {
1.543 mycroft 1835: struct vm_physseg *vps;
1.401 thorpej 1836: psize_t sz = round_page(MSGBUFSIZE);
1837: psize_t reqsz = sz;
1838:
1.586.2.1 tron 1839: search_again:
1.401 thorpej 1840: for (x = 0; x < vm_nphysseg; x++) {
1841: vps = &vm_physmem[x];
1842: if (ptoa(vps->avail_end) == avail_end)
1.543 mycroft 1843: goto found;
1.401 thorpej 1844: }
1.543 mycroft 1845: panic("init386: can't find end of memory");
1.401 thorpej 1846:
1.543 mycroft 1847: found:
1.401 thorpej 1848: /* Shrink so it'll fit in the last segment. */
1849: if ((vps->avail_end - vps->avail_start) < atop(sz))
1850: sz = ptoa(vps->avail_end - vps->avail_start);
1851:
1852: vps->avail_end -= atop(sz);
1853: vps->end -= atop(sz);
1.586.2.1 tron 1854: msgbuf_p_seg[msgbuf_p_cnt].sz = sz;
1855: msgbuf_p_seg[msgbuf_p_cnt++].paddr = ptoa(vps->avail_end);
1.401 thorpej 1856:
1857: /* Remove the last segment if it now has no pages. */
1858: if (vps->start == vps->end) {
1859: for (vm_nphysseg--; x < vm_nphysseg; x++)
1860: vm_physmem[x] = vm_physmem[x + 1];
1861: }
1862:
1863: /* Now find where the new avail_end is. */
1864: for (avail_end = 0, x = 0; x < vm_nphysseg; x++)
1865: if (vm_physmem[x].avail_end > avail_end)
1866: avail_end = vm_physmem[x].avail_end;
1867: avail_end = ptoa(avail_end);
1868:
1.586.2.1 tron 1869: if (sz != reqsz) {
1870: reqsz -= sz;
1871: if (msgbuf_p_cnt != VM_PHYSSEG_MAX) {
1872: /* if still segments available, get memory from next one ... */
1.586.2.2 tron 1873: sz = reqsz;
1874: goto search_again;
1.586.2.1 tron 1875: }
1.401 thorpej 1876: /* Warn if the message buffer had to be shrunk. */
1877: printf("WARNING: %ld bytes not available for msgbuf "
1.586.2.2 tron 1878: "in last cluster (%ld used)\n", (long)MSGBUFSIZE, MSGBUFSIZE - reqsz);
1.586.2.1 tron 1879: }
1.401 thorpej 1880: }
1881:
1.473 tshiozak 1882: /*
1883: * install PT page for the first 4M if needed.
1884: */
1885: if (needs_earlier_install_pte0) {
1886: paddr_t paddr;
1887: #ifdef DIAGNOSTIC
1888: if (realmode_reserved_size < PAGE_SIZE) {
1889: panic("cannot steal memory for first 4M PT page.");
1890: }
1891: #endif
1892: paddr=realmode_reserved_start+realmode_reserved_size-PAGE_SIZE;
1.561 yamt 1893: pmap_kenter_pa((vaddr_t)vtopte(0), paddr,
1894: VM_PROT_READ|VM_PROT_WRITE);
1.473 tshiozak 1895: pmap_update(pmap_kernel());
1896: /* make sure it is clean before using */
1897: memset(vtopte(0), 0, PAGE_SIZE);
1898: realmode_reserved_size -= PAGE_SIZE;
1899: }
1900:
1.295 drochner 1901: #if NBIOSCALL > 0
1.436 jdolecek 1902: /*
1903: * this should be caught at kernel build time, but put it here
1904: * in case someone tries to fake it out...
1905: */
1906: #ifdef DIAGNOSTIC
1.473 tshiozak 1907: if (realmode_reserved_start > BIOSTRAMP_BASE ||
1.474 tron 1908: (realmode_reserved_start+realmode_reserved_size) < (BIOSTRAMP_BASE+
1.473 tshiozak 1909: PAGE_SIZE)) {
1910: panic("cannot steal memory for PT page of bioscall.");
1911: }
1.436 jdolecek 1912: if (biostramp_image_size > PAGE_SIZE)
1.483 provos 1913: panic("biostramp_image_size too big: %x vs. %x",
1.436 jdolecek 1914: biostramp_image_size, PAGE_SIZE);
1915: #endif
1916: pmap_kenter_pa((vaddr_t)BIOSTRAMP_BASE, /* virtual */
1917: (paddr_t)BIOSTRAMP_BASE, /* physical */
1918: VM_PROT_ALL); /* protection */
1.456 chris 1919: pmap_update(pmap_kernel());
1.436 jdolecek 1920: memcpy((caddr_t)BIOSTRAMP_BASE, biostramp_image, biostramp_image_size);
1921: #ifdef DEBUG_BIOSCALL
1922: printf("biostramp installed @ %x\n", BIOSTRAMP_BASE);
1923: #endif
1.473 tshiozak 1924: realmode_reserved_size -= PAGE_SIZE;
1925: realmode_reserved_start += PAGE_SIZE;
1926: #endif
1927:
1928: #if NACPI > 0
1929: /*
1930: * Steal memory for the acpi wake code
1931: */
1932: {
1933: paddr_t paddr, p;
1934: psize_t sz;
1935: int npg;
1936:
1937: paddr = realmode_reserved_start;
1938: npg = acpi_md_get_npages_of_wakecode();
1939: sz = ptoa(npg);
1940: #ifdef DIAGNOSTIC
1941: if (realmode_reserved_size < sz) {
1942: panic("cannot steal memory for ACPI wake code.");
1943: }
1944: #endif
1945:
1946: /* identical mapping */
1947: p = paddr;
1948: for (x=0; x<npg; x++) {
1949: printf("kenter: 0x%08X\n", (unsigned)p);
1950: pmap_kenter_pa((vaddr_t)p, p, VM_PROT_ALL);
1951: p += PAGE_SIZE;
1952: }
1953: pmap_update(pmap_kernel());
1954:
1955: acpi_md_install_wakecode(paddr);
1956:
1957: realmode_reserved_size -= sz;
1958: realmode_reserved_start += sz;
1959: }
1.295 drochner 1960: #endif
1.59 mycroft 1961:
1.561 yamt 1962: pmap_kenter_pa(idt_vaddr, idt_paddr, VM_PROT_READ|VM_PROT_WRITE);
1.456 chris 1963: pmap_update(pmap_kernel());
1.484 fvdl 1964: memset((void *)idt_vaddr, 0, PAGE_SIZE);
1965:
1.516 fvdl 1966: idt = (struct gate_descriptor *)idt_vaddr;
1.275 mycroft 1967: #ifdef I586_CPU
1.561 yamt 1968: pmap_kenter_pa(pentium_idt_vaddr, idt_paddr, VM_PROT_READ);
1.275 mycroft 1969: pentium_idt = (union descriptor *)pentium_idt_vaddr;
1970: #endif
1.484 fvdl 1971: pmap_update(pmap_kernel());
1972:
1973: tgdt = gdt;
1.516 fvdl 1974: gdt = (union descriptor *)
1975: ((char *)idt + NIDT * sizeof (struct gate_descriptor));
1.275 mycroft 1976: ldt = gdt + NGDT;
1977:
1.484 fvdl 1978: memcpy(gdt, tgdt, NGDT*sizeof(*gdt));
1.275 mycroft 1979:
1980: setsegment(&gdt[GLDT_SEL].sd, ldt, NLDT * sizeof(ldt[0]) - 1,
1981: SDT_SYSLDT, SEL_KPL, 0, 0);
1982:
1983: /* make ldt gates and memory segments */
1984: setgate(&ldt[LSYS5CALLS_SEL].gd, &IDTVEC(osyscall), 1,
1.489 fvdl 1985: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.447 christos 1986:
1.275 mycroft 1987: ldt[LUCODE_SEL] = gdt[GUCODE_SEL];
1.533 chs 1988: ldt[LUCODEBIG_SEL] = gdt[GUCODEBIG_SEL];
1.275 mycroft 1989: ldt[LUDATA_SEL] = gdt[GUDATA_SEL];
1.324 christos 1990: ldt[LSOL26CALLS_SEL] = ldt[LBSDICALLS_SEL] = ldt[LSYS5CALLS_SEL];
1.275 mycroft 1991:
1992: /* exceptions */
1.498 fvdl 1993: for (x = 0; x < 32; x++) {
1.516 fvdl 1994: setgate(&idt[x], IDTVEC(exceptions)[x], 0, SDT_SYS386TGT,
1.489 fvdl 1995: (x == 3 || x == 4) ? SEL_UPL : SEL_KPL,
1996: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 1997: idt_allocmap[x] = 1;
1998: }
1.257 thorpej 1999:
1.275 mycroft 2000: /* new-style interrupt gate for syscalls */
1.516 fvdl 2001: setgate(&idt[128], &IDTVEC(syscall), 0, SDT_SYS386TGT, SEL_UPL,
1.489 fvdl 2002: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 2003: idt_allocmap[128] = 1;
1.333 christos 2004: #ifdef COMPAT_SVR4
1.516 fvdl 2005: setgate(&idt[0xd2], &IDTVEC(svr4_fasttrap), 0, SDT_SYS386TGT,
1.489 fvdl 2006: SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 2007: idt_allocmap[0xd2] = 1;
1.333 christos 2008: #endif /* COMPAT_SVR4 */
1.264 mycroft 2009:
1.275 mycroft 2010: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
2011: lgdt(®ion);
1.484 fvdl 2012:
2013: cpu_init_idt();
1.264 mycroft 2014:
1.522 ragge 2015: #if NKSYMS || defined(DDB) || defined(LKM)
1.308 tv 2016: {
2017: extern int end;
1.585 jmmv 2018: boolean_t loaded;
1.336 christos 2019: struct btinfo_symtab *symtab;
1.308 tv 2020:
1.522 ragge 2021: #ifdef DDB
1.484 fvdl 2022: db_machine_init();
1.522 ragge 2023: #endif
1.484 fvdl 2024:
1.585 jmmv 2025: #if defined(MULTIBOOT)
2026: loaded = multiboot_ksyms_init();
2027: #else
2028: loaded = FALSE;
2029: #endif
2030: if (!loaded) {
2031: symtab = lookup_bootinfo(BTINFO_SYMTAB);
2032: if (symtab) {
2033: symtab->ssym += KERNBASE;
2034: symtab->esym += KERNBASE;
2035: ksyms_init(symtab->nsym, (int *)symtab->ssym,
2036: (int *)symtab->esym);
2037: } else
2038: ksyms_init(*(int *)&end, ((int *)&end) + 1, esym);
1.336 christos 2039: }
1.308 tv 2040: }
1.522 ragge 2041: #endif
2042: #ifdef DDB
1.190 mycroft 2043: if (boothowto & RB_KDB)
2044: Debugger();
1.377 ws 2045: #endif
2046: #ifdef IPKDB
2047: ipkdb_init();
2048: if (boothowto & RB_KDB)
2049: ipkdb_connect(0);
1.190 mycroft 2050: #endif
2051: #ifdef KGDB
1.243 drochner 2052: kgdb_port_init();
1.235 thorpej 2053: if (boothowto & RB_KDB) {
2054: kgdb_debug_init = 1;
1.242 drochner 2055: kgdb_connect(1);
1.235 thorpej 2056: }
1.384 jdolecek 2057: #endif
2058:
2059: #if NMCA > 0
2060: /* check for MCA bus, needed to be done before ISA stuff - if
2061: * MCA is detected, ISA needs to use level triggered interrupts
2062: * by default */
2063: mca_busprobe();
1.190 mycroft 2064: #endif
1.275 mycroft 2065:
1.498 fvdl 2066: intr_default_setup();
1.431 thorpej 2067:
2068: /* Initialize software interrupts. */
2069: softintr_init();
1.275 mycroft 2070:
1.498 fvdl 2071: splraise(IPL_IPI);
1.275 mycroft 2072: enable_intr();
2073:
2074: if (physmem < btoc(2 * 1024 * 1024)) {
2075: printf("warning: too little memory available; "
1.383 mycroft 2076: "have %lu bytes, want %lu bytes\n"
1.275 mycroft 2077: "running in degraded mode\n"
2078: "press a key to confirm\n\n",
1.383 mycroft 2079: ptoa(physmem), 2*1024*1024UL);
1.275 mycroft 2080: cngetc();
2081: }
1.507 jdolecek 2082:
2083: #ifdef __HAVE_CPU_MAXPROC
2084: /* Make sure maxproc is sane */
2085: if (maxproc > cpu_maxproc())
2086: maxproc = cpu_maxproc();
2087: #endif
1.1 cgd 2088: }
2089:
1.107 deraadt 2090: #ifdef COMPAT_NOMID
2091: static int
1.566 christos 2092: exec_nomid(struct lwp *l, struct exec_package *epp)
1.31 cgd 2093: {
1.59 mycroft 2094: int error;
2095: u_long midmag, magic;
2096: u_short mid;
1.80 cgd 2097: struct exec *execp = epp->ep_hdr;
1.31 cgd 2098:
1.80 cgd 2099: /* check on validity of epp->ep_hdr performed by exec_out_makecmds */
2100:
2101: midmag = ntohl(execp->a_midmag);
1.59 mycroft 2102: mid = (midmag >> 16) & 0xffff;
2103: magic = midmag & 0xffff;
2104:
2105: if (magic == 0) {
1.80 cgd 2106: magic = (execp->a_midmag & 0xffff);
1.59 mycroft 2107: mid = MID_ZERO;
2108: }
2109:
2110: midmag = mid << 16 | magic;
2111:
2112: switch (midmag) {
2113: case (MID_ZERO << 16) | ZMAGIC:
2114: /*
2115: * 386BSD's ZMAGIC format:
2116: */
1.566 christos 2117: error = exec_aout_prep_oldzmagic(l, epp);
1.59 mycroft 2118: break;
2119:
2120: case (MID_ZERO << 16) | QMAGIC:
2121: /*
2122: * BSDI's QMAGIC format:
2123: * same as new ZMAGIC format, but with different magic number
2124: */
1.566 christos 2125: error = exec_aout_prep_zmagic(l, epp);
1.59 mycroft 2126: break;
2127:
1.202 christos 2128: case (MID_ZERO << 16) | NMAGIC:
2129: /*
2130: * BSDI's NMAGIC format:
2131: * same as NMAGIC format, but with different magic number
2132: * and with text starting at 0.
2133: */
1.566 christos 2134: error = exec_aout_prep_oldnmagic(l, epp);
1.202 christos 2135: break;
2136:
2137: case (MID_ZERO << 16) | OMAGIC:
2138: /*
2139: * BSDI's OMAGIC format:
2140: * same as OMAGIC format, but with different magic number
2141: * and with text starting at 0.
2142: */
1.566 christos 2143: error = exec_aout_prep_oldomagic(l, epp);
1.202 christos 2144: break;
2145:
1.59 mycroft 2146: default:
2147: error = ENOEXEC;
2148: }
2149:
2150: return error;
1.107 deraadt 2151: }
1.31 cgd 2152: #endif
1.107 deraadt 2153:
2154: /*
2155: * cpu_exec_aout_makecmds():
1.549 wiz 2156: * CPU-dependent a.out format hook for execve().
1.107 deraadt 2157: *
2158: * Determine of the given exec package refers to something which we
2159: * understand and, if so, set up the vmcmds for it.
2160: *
2161: * On the i386, old (386bsd) ZMAGIC binaries and BSDI QMAGIC binaries
2162: * if COMPAT_NOMID is given as a kernel option.
2163: */
2164: int
1.566 christos 2165: cpu_exec_aout_makecmds(struct lwp *l, struct exec_package *epp)
1.107 deraadt 2166: {
2167: int error = ENOEXEC;
2168:
2169: #ifdef COMPAT_NOMID
1.566 christos 2170: if ((error = exec_nomid(l, epp)) == 0)
1.107 deraadt 2171: return error;
1.581 thorpej 2172: #else
2173: (void) l;
2174: (void) epp;
1.107 deraadt 2175: #endif /* ! COMPAT_NOMID */
2176:
2177: return error;
1.31 cgd 2178: }
1.84 cgd 2179:
1.484 fvdl 2180: #include <dev/ic/mc146818reg.h> /* for NVRAM POST */
2181: #include <i386/isa/nvram.h> /* for NVRAM POST */
2182:
1.149 mycroft 2183: void
2184: cpu_reset()
2185: {
1.555 mycroft 2186: struct region_descriptor region;
1.149 mycroft 2187:
1.224 mycroft 2188: disable_intr();
2189:
1.227 mycroft 2190: /*
1.484 fvdl 2191: * Ensure the NVRAM reset byte contains something vaguely sane.
2192: */
2193:
2194: outb(IO_RTC, NVRAM_RESET);
2195: outb(IO_RTC+1, NVRAM_RESET_RST);
2196:
2197: /*
1.564 dyoung 2198: * Reset AMD Geode SC1100.
2199: *
1.565 simonb 2200: * 1) Write PCI Configuration Address Register (0xcf8) to
2201: * select Function 0, Register 0x44: Bridge Configuration,
2202: * GPIO and LPC Configuration Register Space, Reset
2203: * Control Register.
1.564 dyoung 2204: *
1.565 simonb 2205: * 2) Write 0xf to PCI Configuration Data Register (0xcfc)
2206: * to reset IDE controller, IDE bus, and PCI bus, and
2207: * to trigger a system-wide reset.
1.564 dyoung 2208: *
2209: * See AMD Geode SC1100 Processor Data Book, Revision 2.0,
2210: * sections 6.3.1, 6.3.2, and 6.4.1.
2211: */
2212: if (cpu_info_primary.ci_signature == 0x540) {
2213: outl(0xcf8, 0x80009044ul);
2214: outl(0xcfc, 0xf);
1.565 simonb 2215: }
1.564 dyoung 2216:
2217: /*
1.227 mycroft 2218: * The keyboard controller has 4 random output pins, one of which is
2219: * connected to the RESET pin on the CPU in many PCs. We tell the
2220: * keyboard controller to pulse this line a couple of times.
2221: */
1.273 drochner 2222: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2223: delay(100000);
1.273 drochner 2224: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2225: delay(100000);
1.149 mycroft 2226:
2227: /*
1.224 mycroft 2228: * Try to cause a triple fault and watchdog reset by making the IDT
2229: * invalid and causing a fault.
1.149 mycroft 2230: */
1.313 perry 2231: memset((caddr_t)idt, 0, NIDT * sizeof(idt[0]));
1.555 mycroft 2232: setregion(®ion, idt, NIDT * sizeof(idt[0]) - 1);
1.565 simonb 2233: lidt(®ion);
1.567 perry 2234: __asm volatile("divl %0,%1" : : "q" (0), "a" (0));
1.149 mycroft 2235:
1.224 mycroft 2236: #if 0
1.149 mycroft 2237: /*
2238: * Try to cause a triple fault and watchdog reset by unmapping the
1.224 mycroft 2239: * entire address space and doing a TLB flush.
1.149 mycroft 2240: */
1.414 thorpej 2241: memset((caddr_t)PTD, 0, PAGE_SIZE);
1.484 fvdl 2242: tlbflush();
1.224 mycroft 2243: #endif
1.149 mycroft 2244:
2245: for (;;);
1.45 cgd 2246: }
1.484 fvdl 2247:
1.499 jdolecek 2248: void
1.551 junyoung 2249: cpu_getmcontext(struct lwp *l, mcontext_t *mcp, unsigned int *flags)
1.508 thorpej 2250: {
2251: const struct trapframe *tf = l->l_md.md_regs;
2252: __greg_t *gr = mcp->__gregs;
1.519 nathanw 2253: __greg_t ras_eip;
1.508 thorpej 2254:
2255: /* Save register context. */
2256: #ifdef VM86
2257: if (tf->tf_eflags & PSL_VM) {
2258: gr[_REG_GS] = tf->tf_vm86_gs;
2259: gr[_REG_FS] = tf->tf_vm86_fs;
2260: gr[_REG_ES] = tf->tf_vm86_es;
2261: gr[_REG_DS] = tf->tf_vm86_ds;
2262: gr[_REG_EFL] = get_vflags(l);
2263: } else
2264: #endif
2265: {
2266: gr[_REG_GS] = tf->tf_gs;
2267: gr[_REG_FS] = tf->tf_fs;
2268: gr[_REG_ES] = tf->tf_es;
2269: gr[_REG_DS] = tf->tf_ds;
2270: gr[_REG_EFL] = tf->tf_eflags;
2271: }
2272: gr[_REG_EDI] = tf->tf_edi;
2273: gr[_REG_ESI] = tf->tf_esi;
2274: gr[_REG_EBP] = tf->tf_ebp;
2275: gr[_REG_EBX] = tf->tf_ebx;
2276: gr[_REG_EDX] = tf->tf_edx;
2277: gr[_REG_ECX] = tf->tf_ecx;
2278: gr[_REG_EAX] = tf->tf_eax;
2279: gr[_REG_EIP] = tf->tf_eip;
2280: gr[_REG_CS] = tf->tf_cs;
2281: gr[_REG_ESP] = tf->tf_esp;
2282: gr[_REG_UESP] = tf->tf_esp;
2283: gr[_REG_SS] = tf->tf_ss;
2284: gr[_REG_TRAPNO] = tf->tf_trapno;
2285: gr[_REG_ERR] = tf->tf_err;
1.519 nathanw 2286:
2287: if ((ras_eip = (__greg_t)ras_lookup(l->l_proc,
2288: (caddr_t) gr[_REG_EIP])) != -1)
2289: gr[_REG_EIP] = ras_eip;
2290:
1.508 thorpej 2291: *flags |= _UC_CPU;
2292:
2293: /* Save floating point register context, if any. */
1.548 yamt 2294: if ((l->l_md.md_flags & MDL_USEDFPU) != 0) {
1.508 thorpej 2295: #if NNPX > 0
2296: /*
2297: * If this process is the current FP owner, dump its
2298: * context to the PCB first.
2299: * XXX npxsave() also clears the FPU state; depending on the
2300: * XXX application this might be a penalty.
2301: */
2302: if (l->l_addr->u_pcb.pcb_fpcpu) {
2303: npxsave_lwp(l, 1);
2304: }
2305: #endif
2306: if (i386_use_fxsave) {
2307: memcpy(&mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2308: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2309: sizeof (mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm));
2310: *flags |= _UC_FXSAVE;
2311: } else {
2312: memcpy(&mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2313: &l->l_addr->u_pcb.pcb_savefpu.sv_87,
2314: sizeof (mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state));
2315: }
2316: #if 0
2317: /* Apparently nothing ever touches this. */
2318: ucp->mcp.mc_fp.fp_emcsts = l->l_addr->u_pcb.pcb_saveemc;
2319: #endif
2320: *flags |= _UC_FPU;
2321: }
2322: }
2323:
2324: int
1.551 junyoung 2325: cpu_setmcontext(struct lwp *l, const mcontext_t *mcp, unsigned int flags)
1.508 thorpej 2326: {
2327: struct trapframe *tf = l->l_md.md_regs;
1.584 pooka 2328: const __greg_t *gr = mcp->__gregs;
1.508 thorpej 2329:
2330: /* Restore register context, if any. */
2331: if ((flags & _UC_CPU) != 0) {
2332: #ifdef VM86
1.535 drochner 2333: if (gr[_REG_EFL] & PSL_VM) {
1.508 thorpej 2334: tf->tf_vm86_gs = gr[_REG_GS];
2335: tf->tf_vm86_fs = gr[_REG_FS];
2336: tf->tf_vm86_es = gr[_REG_ES];
2337: tf->tf_vm86_ds = gr[_REG_DS];
2338: set_vflags(l, gr[_REG_EFL]);
1.534 christos 2339: if (flags & _UC_VM) {
1.542 junyoung 2340: void syscall_vm86(struct trapframe *);
1.534 christos 2341: l->l_proc->p_md.md_syscall = syscall_vm86;
2342: }
1.508 thorpej 2343: } else
2344: #endif
2345: {
2346: /*
2347: * Check for security violations. If we're returning
2348: * to protected mode, the CPU will validate the segment
2349: * registers automatically and generate a trap on
2350: * violations. We handle the trap, rather than doing
2351: * all of the checking here.
2352: */
1.534 christos 2353: if (((gr[_REG_EFL] ^ tf->tf_eflags) & PSL_USERSTATIC) ||
2354: !USERMODE(gr[_REG_CS], gr[_REG_EFL])) {
2355: printf("cpu_setmcontext error: uc EFL: 0x%08x"
1.535 drochner 2356: " tf EFL: 0x%08x uc CS: 0x%x\n",
1.508 thorpej 2357: gr[_REG_EFL], tf->tf_eflags, gr[_REG_CS]);
2358: return (EINVAL);
2359: }
2360: tf->tf_gs = gr[_REG_GS];
2361: tf->tf_fs = gr[_REG_FS];
2362: tf->tf_es = gr[_REG_ES];
2363: tf->tf_ds = gr[_REG_DS];
2364: /* Only change the user-alterable part of eflags */
2365: tf->tf_eflags &= ~PSL_USER;
2366: tf->tf_eflags |= (gr[_REG_EFL] & PSL_USER);
2367: }
2368: tf->tf_edi = gr[_REG_EDI];
2369: tf->tf_esi = gr[_REG_ESI];
2370: tf->tf_ebp = gr[_REG_EBP];
2371: tf->tf_ebx = gr[_REG_EBX];
2372: tf->tf_edx = gr[_REG_EDX];
2373: tf->tf_ecx = gr[_REG_ECX];
2374: tf->tf_eax = gr[_REG_EAX];
2375: tf->tf_eip = gr[_REG_EIP];
2376: tf->tf_cs = gr[_REG_CS];
2377: tf->tf_esp = gr[_REG_UESP];
2378: tf->tf_ss = gr[_REG_SS];
2379: }
2380:
2381: /* Restore floating point register context, if any. */
2382: if ((flags & _UC_FPU) != 0) {
2383: #if NNPX > 0
2384: /*
2385: * If we were using the FPU, forget that we were.
2386: */
2387: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
2388: npxsave_lwp(l, 0);
2389: #endif
2390: if (flags & _UC_FXSAVE) {
2391: if (i386_use_fxsave) {
2392: memcpy(
2393: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2394: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2395: sizeof (&l->l_addr->u_pcb.pcb_savefpu.sv_xmm));
2396: } else {
2397: /* This is a weird corner case */
2398: process_xmm_to_s87((struct savexmm *)
2399: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2400: &l->l_addr->u_pcb.pcb_savefpu.sv_87);
2401: }
2402: } else {
2403: if (i386_use_fxsave) {
2404: process_s87_to_xmm((struct save87 *)
2405: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2406: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm);
2407: } else {
2408: memcpy(&l->l_addr->u_pcb.pcb_savefpu.sv_87,
2409: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2410: sizeof (l->l_addr->u_pcb.pcb_savefpu.sv_87));
2411: }
2412: }
2413: /* If not set already. */
1.548 yamt 2414: l->l_md.md_flags |= MDL_USEDFPU;
1.508 thorpej 2415: #if 0
2416: /* Apparently unused. */
2417: l->l_addr->u_pcb.pcb_saveemc = mcp->mc_fp.fp_emcsts;
2418: #endif
2419: }
1.534 christos 2420: if (flags & _UC_SETSTACK)
2421: l->l_proc->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
2422: if (flags & _UC_CLRSTACK)
2423: l->l_proc->p_sigctx.ps_sigstk.ss_flags &= ~SS_ONSTACK;
1.508 thorpej 2424: return (0);
2425: }
2426:
2427: void
1.499 jdolecek 2428: cpu_initclocks()
1.484 fvdl 2429: {
1.574 kardel 2430:
1.484 fvdl 2431: (*initclock_func)();
2432: }
2433:
2434: #ifdef MULTIPROCESSOR
1.499 jdolecek 2435: void
2436: need_resched(struct cpu_info *ci)
1.484 fvdl 2437: {
1.546 yamt 2438:
2439: if (ci->ci_want_resched)
2440: return;
2441:
1.484 fvdl 2442: ci->ci_want_resched = 1;
1.508 thorpej 2443: if ((ci)->ci_curlwp != NULL)
2444: aston((ci)->ci_curlwp->l_proc);
1.546 yamt 2445: else if (ci != curcpu())
2446: x86_send_ipi(ci, 0);
1.484 fvdl 2447: }
2448: #endif
2449:
2450: /*
2451: * Allocate an IDT vector slot within the given range.
2452: * XXX needs locking to avoid MP allocation races.
2453: */
2454:
2455: int
1.551 junyoung 2456: idt_vec_alloc(int low, int high)
1.484 fvdl 2457: {
2458: int vec;
2459:
1.498 fvdl 2460: simple_lock(&idt_lock);
2461: for (vec = low; vec <= high; vec++) {
2462: if (idt_allocmap[vec] == 0) {
2463: idt_allocmap[vec] = 1;
2464: simple_unlock(&idt_lock);
1.484 fvdl 2465: return vec;
1.498 fvdl 2466: }
2467: }
2468: simple_unlock(&idt_lock);
1.484 fvdl 2469: return 0;
2470: }
2471:
1.498 fvdl 2472: void
1.551 junyoung 2473: idt_vec_set(int vec, void (*function)(void))
1.484 fvdl 2474: {
1.498 fvdl 2475: /*
2476: * Vector should be allocated, so no locking needed.
2477: */
2478: KASSERT(idt_allocmap[vec] == 1);
1.516 fvdl 2479: setgate(&idt[vec], function, 0, SDT_SYS386IGT, SEL_KPL,
1.489 fvdl 2480: GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 2481: }
2482:
2483: void
1.551 junyoung 2484: idt_vec_free(int vec)
1.484 fvdl 2485: {
1.498 fvdl 2486: simple_lock(&idt_lock);
1.516 fvdl 2487: unsetgate(&idt[vec]);
1.498 fvdl 2488: idt_allocmap[vec] = 0;
2489: simple_unlock(&idt_lock);
1.507 jdolecek 2490: }
2491:
2492: /*
2493: * Number of processes is limited by number of available GDT slots.
2494: */
2495: int
2496: cpu_maxproc(void)
2497: {
2498: #ifdef USER_LDT
2499: return ((MAXGDTSIZ - NGDT) / 2);
2500: #else
2501: return (MAXGDTSIZ - NGDT);
2502: #endif
1.484 fvdl 2503: }
CVSweb <webmaster@jp.NetBSD.org>