Annotation of src/sys/arch/i386/i386/machdep.c, Revision 1.554
1.554 ! lukem 1: /* $NetBSD: machdep.c,v 1.553 2004/04/30 02:05:43 lukem Exp $ */
1.231 thorpej 2:
3: /*-
1.401 thorpej 4: * Copyright (c) 1996, 1997, 1998, 2000 The NetBSD Foundation, Inc.
1.231 thorpej 5: * All rights reserved.
6: *
7: * This code is derived from software contributed to The NetBSD Foundation
1.316 mycroft 8: * by Charles M. Hannum and by Jason R. Thorpe of the Numerical Aerospace
9: * Simulation Facility, NASA Ames Research Center.
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>
1.554 ! lukem 75: __KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.553 2004/04/30 02:05:43 lukem 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>
1.291 thorpej 115: #include <machine/kcore.h>
1.519 nathanw 116: #include <sys/ras.h>
1.508 thorpej 117: #include <sys/sa.h>
118: #include <sys/savar.h>
1.522 ragge 119: #include <sys/ksyms.h>
1.57 cgd 120:
1.377 ws 121: #ifdef IPKDB
122: #include <ipkdb/ipkdb.h>
123: #endif
124:
1.235 thorpej 125: #ifdef KGDB
126: #include <sys/kgdb.h>
127: #endif
128:
1.104 cgd 129: #include <dev/cons.h>
1.390 mrg 130:
131: #include <uvm/uvm_extern.h>
1.393 fvdl 132: #include <uvm/uvm_page.h>
1.284 mrg 133:
1.200 christos 134: #include <sys/sysctl.h>
135:
1.59 mycroft 136: #include <machine/cpu.h>
137: #include <machine/cpufunc.h>
1.484 fvdl 138: #include <machine/cpuvar.h>
1.178 mycroft 139: #include <machine/gdt.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>
1.43 brezak 146:
1.146 cgd 147: #include <dev/isa/isareg.h>
1.372 drochner 148: #include <machine/isa_machdep.h>
1.164 cgd 149: #include <dev/ic/i8042reg.h>
1.43 brezak 150:
1.200 christos 151: #ifdef DDB
152: #include <machine/db_machdep.h>
153: #include <ddb/db_extern.h>
154: #endif
155:
1.184 mycroft 156: #ifdef VM86
157: #include <machine/vm86.h>
158: #endif
159:
1.473 tshiozak 160: #include "acpi.h"
1.207 jtk 161: #include "apm.h"
1.258 jtk 162: #include "bioscall.h"
1.207 jtk 163:
1.259 jtk 164: #if NBIOSCALL > 0
165: #include <machine/bioscall.h>
166: #endif
167:
1.473 tshiozak 168: #if NACPI > 0
169: #include <dev/acpi/acpivar.h>
170: #define ACPI_MACHDEP_PRIVATE
171: #include <machine/acpi_machdep.h>
172: #endif
173:
1.207 jtk 174: #if NAPM > 0
175: #include <machine/apmvar.h>
1.258 jtk 176: #endif
177:
1.59 mycroft 178: #include "isa.h"
1.231 thorpej 179: #include "isadma.h"
1.59 mycroft 180: #include "npx.h"
1.522 ragge 181: #include "ksyms.h"
1.2 cgd 182:
1.384 jdolecek 183: #include "mca.h"
184: #if NMCA > 0
185: #include <machine/mca_machdep.h> /* for mca_busprobe() */
186: #endif
187:
1.484 fvdl 188: #ifdef MULTIPROCESSOR /* XXX */
189: #include <machine/mpbiosvar.h> /* XXX */
190: #endif /* XXX */
191:
1.517 jmmv 192: #ifndef BEEP_ONHALT_COUNT
193: #define BEEP_ONHALT_COUNT 3
194: #endif
195: #ifndef BEEP_ONHALT_PITCH
196: #define BEEP_ONHALT_PITCH 1500
197: #endif
198: #ifndef BEEP_ONHALT_PERIOD
199: #define BEEP_ONHALT_PERIOD 250
200: #endif
201:
1.104 cgd 202: /* the following is used externally (sysctl_hw) */
1.549 wiz 203: char machine[] = "i386"; /* CPU "architecture" */
1.232 veego 204: char machine_arch[] = "i386"; /* machine == machine_arch */
1.104 cgd 205:
1.255 drochner 206: char bootinfo[BOOTINFO_MAXSIZE];
1.386 thorpej 207:
1.343 fvdl 208: struct bi_devmatch *i386_alldisks = NULL;
209: int i386_ndisks = 0;
1.342 fvdl 210:
1.328 bouyer 211: #ifdef CPURESET_DELAY
212: int cpureset_delay = CPURESET_DELAY;
213: #else
214: int cpureset_delay = 2000; /* default to 2s */
215: #endif
216:
1.455 fvdl 217: #ifdef MTRR
1.513 fvdl 218: struct mtrr_funcs *mtrr_funcs;
1.455 fvdl 219: #endif
220:
1.515 fvdl 221: #ifdef COMPAT_NOMID
1.551 junyoung 222: static int exec_nomid(struct proc *, struct exec_package *);
1.550 junyoung 223: #endif
1.515 fvdl 224:
1.59 mycroft 225: int physmem;
1.163 cgd 226: int dumpmem_low;
227: int dumpmem_high;
1.524 drochner 228: unsigned int cpu_feature;
1.553 lukem 229: unsigned int cpu_feature2;
1.59 mycroft 230: int cpu_class;
1.428 fvdl 231: int i386_fpu_present;
232: int i386_fpu_exception;
233: int i386_fpu_fdivbug;
1.59 mycroft 234:
1.451 thorpej 235: int i386_use_fxsave;
236: int i386_has_sse;
237: int i386_has_sse2;
1.450 thorpej 238:
1.461 christos 239: int tmx86_has_longrun;
240:
1.314 thorpej 241: vaddr_t msgbuf_vaddr;
242: paddr_t msgbuf_paddr;
243:
244: vaddr_t idt_vaddr;
245: paddr_t idt_paddr;
246:
1.264 mycroft 247: #ifdef I586_CPU
1.314 thorpej 248: vaddr_t pentium_idt_vaddr;
1.264 mycroft 249: #endif
1.59 mycroft 250:
1.444 chs 251: struct vm_map *exec_map = NULL;
252: struct vm_map *mb_map = NULL;
253: struct vm_map *phys_map = NULL;
1.48 brezak 254:
1.314 thorpej 255: extern paddr_t avail_start, avail_end;
1.1 cgd 256:
1.542 junyoung 257: void (*delay_func)(int) = i8254_delay;
258: void (*microtime_func)(struct timeval *) = i8254_microtime;
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:
1.433 kanaoka 274: #if !defined(REALBASEMEM) && !defined(REALEXTMEM)
1.542 junyoung 275: void add_mem_cluster(u_int64_t, u_int64_t, u_int32_t);
1.433 kanaoka 276: #endif /* !defnied(REALBASEMEM) && !defined(REALEXTMEM) */
277:
1.521 dsl 278: extern int time_adjusted;
1.484 fvdl 279:
1.59 mycroft 280: /*
281: * Machine-dependent startup code
282: */
1.32 andrew 283: void
1.1 cgd 284: cpu_startup()
285: {
1.547 pk 286: int x;
1.314 thorpej 287: vaddr_t minaddr, maxaddr;
1.354 lukem 288: char pbuf[9];
1.1 cgd 289:
1.284 mrg 290: /*
291: * Initialize error message buffer (et end of core).
292: */
1.514 fvdl 293: msgbuf_vaddr = uvm_km_valloc(kernel_map, x86_round_page(MSGBUFSIZE));
1.385 thorpej 294: if (msgbuf_vaddr == 0)
1.284 mrg 295: panic("failed to valloc msgbuf_vaddr");
1.359 thorpej 296:
1.284 mrg 297: /* msgbuf_paddr was init'd in pmap */
298: for (x = 0; x < btoc(MSGBUFSIZE); x++)
1.414 thorpej 299: pmap_kenter_pa((vaddr_t)msgbuf_vaddr + x * PAGE_SIZE,
300: msgbuf_paddr + x * PAGE_SIZE, VM_PROT_READ|VM_PROT_WRITE);
1.456 chris 301: pmap_update(pmap_kernel());
1.359 thorpej 302:
1.284 mrg 303: initmsgbuf((caddr_t)msgbuf_vaddr, round_page(MSGBUFSIZE));
304:
1.392 sommerfe 305: printf("%s", version);
1.484 fvdl 306:
307: #ifdef TRAPLOG
308: /*
309: * Enable recording of branch from/to in MSR's
310: */
311: wrmsr(MSR_DEBUGCTLMSR, 0x1);
312: #endif
313:
1.382 mycroft 314: format_bytes(pbuf, sizeof(pbuf), ptoa(physmem));
1.354 lukem 315: printf("total memory = %s\n", pbuf);
1.1 cgd 316:
1.547 pk 317: minaddr = 0;
1.41 cgd 318:
1.1 cgd 319: /*
1.36 cgd 320: * Allocate a submap for exec arguments. This map effectively
321: * limits the number of processes exec'ing at any time.
1.1 cgd 322: */
1.284 mrg 323: exec_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 324: 16*NCARGS, VM_MAP_PAGEABLE, FALSE, NULL);
1.59 mycroft 325:
1.1 cgd 326: /*
327: * Allocate a submap for physio
328: */
1.284 mrg 329: phys_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 330: VM_PHYS_SIZE, 0, FALSE, NULL);
1.1 cgd 331:
332: /*
1.229 thorpej 333: * Finally, allocate mbuf cluster submap.
1.1 cgd 334: */
1.334 thorpej 335: mb_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 336: nmbclusters * mclbytes, VM_MAP_INTRSAFE, FALSE, NULL);
1.1 cgd 337:
1.547 pk 338: format_bytes(pbuf, sizeof(pbuf), ptoa(uvmexp.free));
1.354 lukem 339: printf("avail memory = %s\n", pbuf);
1.1 cgd 340:
1.375 drochner 341: /* Safe for i/o port / memory space allocation to use malloc now. */
1.514 fvdl 342: x86_bus_space_mallocok();
1.349 thorpej 343: }
344:
345: /*
346: * Set up proc0's TSS and LDT.
347: */
348: void
349: i386_proc0_tss_ldt_init()
350: {
351: struct pcb *pcb;
352: int x;
1.268 thorpej 353:
1.326 thorpej 354: gdt_init();
1.484 fvdl 355:
1.508 thorpej 356: cpu_info_primary.ci_curpcb = pcb = &lwp0.l_addr->u_pcb;
1.484 fvdl 357:
1.326 thorpej 358: pcb->pcb_tss.tss_ioopt =
359: ((caddr_t)pcb->pcb_iomap - (caddr_t)&pcb->pcb_tss) << 16;
1.484 fvdl 360:
1.326 thorpej 361: for (x = 0; x < sizeof(pcb->pcb_iomap) / 4; x++)
362: pcb->pcb_iomap[x] = 0xffffffff;
363:
1.394 thorpej 364: pcb->pcb_ldt_sel = pmap_kernel()->pm_ldt_sel = GSEL(GLDT_SEL, SEL_KPL);
1.326 thorpej 365: pcb->pcb_cr0 = rcr0();
366: pcb->pcb_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
1.508 thorpej 367: pcb->pcb_tss.tss_esp0 = (int)lwp0.l_addr + USPACE - 16;
368: lwp0.l_md.md_regs = (struct trapframe *)pcb->pcb_tss.tss_esp0 - 1;
369: lwp0.l_md.md_tss_sel = tss_alloc(pcb);
1.326 thorpej 370:
1.508 thorpej 371: ltr(lwp0.l_md.md_tss_sel);
1.326 thorpej 372: lldt(pcb->pcb_ldt_sel);
1.484 fvdl 373: }
374:
375: /*
376: * Set up TSS and LDT for a new PCB.
377: */
378:
379: void
1.551 junyoung 380: i386_init_pcb_tss_ldt(struct cpu_info *ci)
1.484 fvdl 381: {
382: int x;
383: struct pcb *pcb = ci->ci_idle_pcb;
1.326 thorpej 384:
1.484 fvdl 385: pcb->pcb_tss.tss_ioopt =
386: ((caddr_t)pcb->pcb_iomap - (caddr_t)&pcb->pcb_tss) << 16;
387: for (x = 0; x < sizeof(pcb->pcb_iomap) / 4; x++)
388: pcb->pcb_iomap[x] = 0xffffffff;
389:
390: pcb->pcb_ldt_sel = pmap_kernel()->pm_ldt_sel = GSEL(GLDT_SEL, SEL_KPL);
391: pcb->pcb_cr0 = rcr0();
392:
393: ci->ci_idle_tss_sel = tss_alloc(pcb);
1.326 thorpej 394: }
395:
396: /*
1.545 atatat 397: * sysctl helper routine for machdep.tm* nodes.
1.484 fvdl 398: */
1.545 atatat 399: static int
400: sysctl_machdep_tm_longrun(SYSCTLFN_ARGS)
1.104 cgd 401: {
1.545 atatat 402: struct sysctlnode node;
403: int io, error;
404:
405: if (!tmx86_has_longrun)
406: return (EOPNOTSUPP);
407:
408: node = *rnode;
409: node.sysctl_data = &io;
1.104 cgd 410:
1.545 atatat 411: switch (rnode->sysctl_num) {
1.461 christos 412: case CPU_TMLR_MODE:
1.545 atatat 413: io = (int)(crusoe_longrun = tmx86_get_longrun_mode());
414: break;
1.461 christos 415: case CPU_TMLR_FREQUENCY:
416: tmx86_get_longrun_status_all();
1.545 atatat 417: io = crusoe_frequency;
418: break;
1.461 christos 419: case CPU_TMLR_VOLTAGE:
420: tmx86_get_longrun_status_all();
1.545 atatat 421: io = crusoe_voltage;
422: break;
1.461 christos 423: case CPU_TMLR_PERCENTAGE:
424: tmx86_get_longrun_status_all();
1.545 atatat 425: io = crusoe_percentage;
426: break;
1.104 cgd 427: default:
428: return (EOPNOTSUPP);
429: }
1.545 atatat 430:
431: error = sysctl_lookup(SYSCTLFN_CALL(&node));
432: if (error || newp == NULL)
433: return (error);
434:
435: if (rnode->sysctl_num == CPU_TMLR_MODE) {
436: if (tmx86_set_longrun_mode(io))
437: crusoe_longrun = (u_int)io;
438: else
439: return (EINVAL);
440: }
441:
442: return (0);
443: }
444:
445: /*
446: * sysctl helper routine for machdep.booted_kernel
447: */
448: static int
449: sysctl_machdep_booted_kernel(SYSCTLFN_ARGS)
450: {
451: struct btinfo_bootpath *bibp;
452: struct sysctlnode node;
453:
454: bibp = lookup_bootinfo(BTINFO_BOOTPATH);
455: if(!bibp)
456: return(ENOENT); /* ??? */
457:
458: node = *rnode;
459: node.sysctl_data = bibp->bootpath;
460: node.sysctl_size = sizeof(bibp->bootpath);
461: return (sysctl_lookup(SYSCTLFN_CALL(&node)));
462: }
463:
464: /*
465: * sysctl helper routine for machdep.diskinfo
466: */
467: static int
468: sysctl_machdep_diskinfo(SYSCTLFN_ARGS)
469: {
470: struct sysctlnode node;
471:
472: node = *rnode;
473: node.sysctl_data = i386_alldisks;
474: node.sysctl_size = sizeof(struct disklist) +
475: (i386_ndisks - 1) * sizeof(struct nativedisk_info);
476: return (sysctl_lookup(SYSCTLFN_CALL(&node)));
477: }
478:
479: /*
480: * machine dependent system variables.
481: */
482: SYSCTL_SETUP(sysctl_machdep_setup, "sysctl machdep subtree setup")
483: {
484:
1.552 atatat 485: sysctl_createv(clog, 0, NULL, NULL,
486: CTLFLAG_PERMANENT,
1.545 atatat 487: CTLTYPE_NODE, "machdep", NULL,
488: NULL, 0, NULL, 0,
489: CTL_MACHDEP, CTL_EOL);
490:
1.552 atatat 491: sysctl_createv(clog, 0, NULL, NULL,
492: CTLFLAG_PERMANENT,
1.545 atatat 493: CTLTYPE_STRUCT, "console_device", NULL,
494: sysctl_consdev, 0, NULL, sizeof(dev_t),
495: CTL_MACHDEP, CPU_CONSDEV, CTL_EOL);
1.552 atatat 496: sysctl_createv(clog, 0, NULL, NULL,
497: CTLFLAG_PERMANENT,
1.545 atatat 498: CTLTYPE_INT, "biosbasemem", NULL,
499: NULL, 0, &biosbasemem, 0,
500: CTL_MACHDEP, CPU_BIOSBASEMEM, CTL_EOL);
1.552 atatat 501: sysctl_createv(clog, 0, NULL, NULL,
502: CTLFLAG_PERMANENT,
1.545 atatat 503: CTLTYPE_INT, "biosextmem", NULL,
504: NULL, 0, &biosextmem, 0,
505: CTL_MACHDEP, CPU_BIOSEXTMEM, CTL_EOL);
1.552 atatat 506: sysctl_createv(clog, 0, NULL, NULL,
507: CTLFLAG_PERMANENT,
1.545 atatat 508: CTLTYPE_INT, "nkpde", NULL,
509: NULL, 0, &nkpde, 0,
510: CTL_MACHDEP, CPU_NKPDE, CTL_EOL);
1.552 atatat 511: sysctl_createv(clog, 0, NULL, NULL,
512: CTLFLAG_PERMANENT,
1.545 atatat 513: CTLTYPE_STRING, "booted_kernel", NULL,
514: sysctl_machdep_booted_kernel, 0, NULL, 0,
515: CTL_MACHDEP, CPU_BOOTED_KERNEL, CTL_EOL);
1.552 atatat 516: sysctl_createv(clog, 0, NULL, NULL,
517: CTLFLAG_PERMANENT,
1.545 atatat 518: CTLTYPE_STRUCT, "diskinfo", NULL,
519: sysctl_machdep_diskinfo, 0, NULL, 0,
520: CTL_MACHDEP, CPU_DISKINFO, CTL_EOL);
1.552 atatat 521: sysctl_createv(clog, 0, NULL, NULL,
522: CTLFLAG_PERMANENT,
1.545 atatat 523: CTLTYPE_INT, "fpu_present", NULL,
524: NULL, 0, &i386_fpu_present, 0,
525: CTL_MACHDEP, CPU_FPU_PRESENT, CTL_EOL);
1.552 atatat 526: sysctl_createv(clog, 0, NULL, NULL,
527: CTLFLAG_PERMANENT,
1.545 atatat 528: CTLTYPE_INT, "osfxsr", NULL,
529: NULL, 0, &i386_use_fxsave, 0,
530: CTL_MACHDEP, CPU_OSFXSR, CTL_EOL);
1.552 atatat 531: sysctl_createv(clog, 0, NULL, NULL,
532: CTLFLAG_PERMANENT,
1.545 atatat 533: CTLTYPE_INT, "sse", NULL,
534: NULL, 0, &i386_has_sse, 0,
535: CTL_MACHDEP, CPU_SSE, CTL_EOL);
1.552 atatat 536: sysctl_createv(clog, 0, NULL, NULL,
537: CTLFLAG_PERMANENT,
1.545 atatat 538: CTLTYPE_INT, "sse2", NULL,
539: NULL, 0, &i386_has_sse2, 0,
540: CTL_MACHDEP, CPU_SSE2, CTL_EOL);
1.554 ! lukem 541: sysctl_createv(clog, 0, NULL, NULL,
! 542: CTLFLAG_PERMANENT,
! 543: CTLTYPE_STRING, "cpu_brand", NULL,
! 544: NULL, 0, &cpu_brand_string, 0,
! 545: CTL_MACHDEP, CTL_CREATE, CTL_EOL);
1.552 atatat 546: sysctl_createv(clog, 0, NULL, NULL,
547: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.545 atatat 548: CTLTYPE_INT, "tm_longrun_mode", NULL,
549: sysctl_machdep_tm_longrun, 0, NULL, 0,
550: CTL_MACHDEP, CPU_TMLR_MODE, CTL_EOL);
1.552 atatat 551: sysctl_createv(clog, 0, NULL, NULL,
552: CTLFLAG_PERMANENT,
1.545 atatat 553: CTLTYPE_INT, "tm_longrun_frequency", NULL,
554: sysctl_machdep_tm_longrun, 0, NULL, 0,
555: CTL_MACHDEP, CPU_TMLR_FREQUENCY, CTL_EOL);
1.552 atatat 556: sysctl_createv(clog, 0, NULL, NULL,
557: CTLFLAG_PERMANENT,
1.545 atatat 558: CTLTYPE_INT, "tm_longrun_voltage", NULL,
559: sysctl_machdep_tm_longrun, 0, NULL, 0,
560: CTL_MACHDEP, CPU_TMLR_VOLTAGE, CTL_EOL);
1.552 atatat 561: sysctl_createv(clog, 0, NULL, NULL,
562: CTLFLAG_PERMANENT,
1.545 atatat 563: CTLTYPE_INT, "tm_longrun_percentage", NULL,
564: sysctl_machdep_tm_longrun, 0, NULL, 0,
565: CTL_MACHDEP, CPU_TMLR_PERCENTAGE, CTL_EOL);
1.104 cgd 566: }
1.151 christos 567:
1.534 christos 568: void *
569: getframe(struct lwp *l, int sig, int *onstack)
570: {
571: struct proc *p = l->l_proc;
572: struct sigctx *ctx = &p->p_sigctx;
573: struct trapframe *tf = l->l_md.md_regs;
574:
575: /* Do we need to jump onto the signal stack? */
576: *onstack = (ctx->ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0
577: && (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0;
578: if (*onstack)
579: return (char *)ctx->ps_sigstk.ss_sp + ctx->ps_sigstk.ss_size;
580: #ifdef VM86
581: if (tf->tf_eflags & PSL_VM)
582: return (void *)(tf->tf_esp + (tf->tf_ss << 4));
583: else
584: #endif
585: return (void *)tf->tf_esp;
586: }
587:
588: /*
589: * Build context to run handler in. We invoke the handler
590: * directly, only returning via the trampoline. Note the
591: * trampoline version numbers are coordinated with machine-
592: * dependent code in libc.
593: */
594: void
595: buildcontext(struct lwp *l, int sel, void *catcher, void *fp)
596: {
597: struct trapframe *tf = l->l_md.md_regs;
598:
599: tf->tf_gs = GSEL(GUDATA_SEL, SEL_UPL);
600: tf->tf_fs = GSEL(GUDATA_SEL, SEL_UPL);
601: tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
602: tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
603: tf->tf_eip = (int)catcher;
604: tf->tf_cs = GSEL(sel, SEL_UPL);
605: tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
606: tf->tf_esp = (int)fp;
607: tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
608: }
609:
610: static void
1.539 christos 611: sendsig_siginfo(const ksiginfo_t *ksi, const sigset_t *mask)
1.534 christos 612: {
613: struct lwp *l = curlwp;
614: struct proc *p = l->l_proc;
615: struct pmap *pmap = vm_map_pmap(&p->p_vmspace->vm_map);
616: int sel = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
617: GUCODEBIG_SEL : GUCODE_SEL;
618: struct sigacts *ps = p->p_sigacts;
619: int onstack;
620: int sig = ksi->ksi_signo;
621: struct sigframe_siginfo *fp = getframe(l, sig, &onstack), frame;
622: sig_t catcher = SIGACTION(p, sig).sa_handler;
623: struct trapframe *tf = l->l_md.md_regs;
624:
625: fp--;
626:
627: /* Build stack frame for signal trampoline. */
628: switch (ps->sa_sigdesc[sig].sd_vers) {
1.537 christos 629: case 0: /* handled by sendsig_sigcontext */
630: case 1: /* handled by sendsig_sigcontext */
1.534 christos 631: default: /* unknown version */
632: printf("nsendsig: bad version %d\n",
633: ps->sa_sigdesc[sig].sd_vers);
634: sigexit(l, SIGILL);
635: case 2:
636: break;
637: }
638:
639: frame.sf_ra = (int)ps->sa_sigdesc[sig].sd_tramp;
640: frame.sf_signum = sig;
641: frame.sf_sip = &fp->sf_si;
642: frame.sf_ucp = &fp->sf_uc;
1.540 thorpej 643: frame.sf_si._info = ksi->ksi_info;
1.534 christos 644: frame.sf_uc.uc_flags = _UC_SIGMASK|_UC_VM;
645: frame.sf_uc.uc_sigmask = *mask;
646: frame.sf_uc.uc_link = NULL;
647: frame.sf_uc.uc_flags |= (p->p_sigctx.ps_sigstk.ss_flags & SS_ONSTACK)
648: ? _UC_SETSTACK : _UC_CLRSTACK;
649: memset(&frame.sf_uc.uc_stack, 0, sizeof(frame.sf_uc.uc_stack));
650: cpu_getmcontext(l, &frame.sf_uc.uc_mcontext, &frame.sf_uc.uc_flags);
651:
652: if (tf->tf_eflags & PSL_VM)
653: (*p->p_emul->e_syscall_intern)(p);
654:
655: if (copyout(&frame, fp, sizeof(frame)) != 0) {
656: /*
657: * Process has trashed its stack; give it an illegal
658: * instruction to halt it in its tracks.
659: */
660: sigexit(l, SIGILL);
661: /* NOTREACHED */
662: }
663:
664: buildcontext(l, sel, catcher, fp);
1.319 mycroft 665:
666: /* Remember that we're now on the signal stack. */
667: if (onstack)
1.425 jdolecek 668: p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
1.1 cgd 669: }
670:
1.534 christos 671: void
1.539 christos 672: sendsig(const ksiginfo_t *ksi, const sigset_t *mask)
1.534 christos 673: {
674: #ifdef COMPAT_16
675: if (curproc->p_sigacts->sa_sigdesc[ksi->ksi_signo].sd_vers < 2)
676: sendsig_sigcontext(ksi, mask);
677: else
678: #endif
679: sendsig_siginfo(ksi, mask);
680: }
1.508 thorpej 681:
1.550 junyoung 682: void
1.534 christos 683: cpu_upcall(struct lwp *l, int type, int nevents, int ninterrupted, void *sas,
684: void *ap, void *sp, sa_upcall_t upcall)
1.508 thorpej 685: {
1.533 chs 686: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 687: struct saframe *sf, frame;
688: struct trapframe *tf;
689:
690: tf = l->l_md.md_regs;
691:
692: /* Finally, copy out the rest of the frame. */
693: frame.sa_type = type;
694: frame.sa_sas = sas;
695: frame.sa_events = nevents;
696: frame.sa_interrupted = ninterrupted;
697: frame.sa_arg = ap;
698: frame.sa_ra = 0;
1.550 junyoung 699:
1.508 thorpej 700: sf = (struct saframe *)sp - 1;
701: if (copyout(&frame, sf, sizeof(frame)) != 0) {
702: /* Copying onto the stack didn't work. Die. */
703: sigexit(l, SIGILL);
704: /* NOTREACHED */
705: }
706:
707: tf->tf_eip = (int) upcall;
708: tf->tf_esp = (int) sf;
709: tf->tf_ebp = 0; /* indicate call-frame-top to debuggers */
710: tf->tf_gs = GSEL(GUDATA_SEL, SEL_UPL);
711: tf->tf_fs = GSEL(GUDATA_SEL, SEL_UPL);
712: tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
713: tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
1.533 chs 714: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
715: GSEL(GUCODEBIG_SEL, SEL_UPL) : GSEL(GUCODE_SEL, SEL_UPL);
1.508 thorpej 716: tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
717: tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
718: }
1.37 cgd 719:
1.1 cgd 720: int waittime = -1;
721: struct pcb dumppcb;
722:
1.32 andrew 723: void
1.551 junyoung 724: cpu_reboot(int howto, char *bootstr)
1.1 cgd 725: {
726:
1.106 mycroft 727: if (cold) {
1.193 mycroft 728: howto |= RB_HALT;
729: goto haltsys;
1.1 cgd 730: }
1.193 mycroft 731:
1.106 mycroft 732: boothowto = howto;
1.193 mycroft 733: if ((howto & RB_NOSYNC) == 0 && waittime < 0) {
1.1 cgd 734: waittime = 0;
1.150 mycroft 735: vfs_shutdown();
1.521 dsl 736: /*
737: * If we've been adjusting the clock, the todr
738: * will be out of synch; adjust it now.
739: */
740: if (time_adjusted != 0)
741: resettodr();
1.1 cgd 742: }
1.193 mycroft 743:
744: /* Disable interrupts. */
1.1 cgd 745: splhigh();
1.193 mycroft 746:
747: /* Do a dump if requested. */
748: if ((howto & (RB_DUMP | RB_HALT)) == RB_DUMP)
749: dumpsys();
750:
751: haltsys:
752: doshutdownhooks();
753:
1.484 fvdl 754: #ifdef MULTIPROCESSOR
1.514 fvdl 755: x86_broadcast_ipi(X86_IPI_HALT);
1.484 fvdl 756: #endif
757:
1.307 thorpej 758: if ((howto & RB_POWERDOWN) == RB_POWERDOWN) {
1.473 tshiozak 759: #if NACPI > 0
1.530 kochi 760: if (acpi_softc != NULL) {
761: delay(500000);
762: acpi_enter_sleep_state(acpi_softc, ACPI_STATE_S5);
763: printf("WARNING: ACPI powerdown failed!\n");
764: }
1.473 tshiozak 765: #endif
1.208 jtk 766: #if NAPM > 0 && !defined(APM_NO_POWEROFF)
767: /* turn off, if we can. But try to turn disk off and
768: * wait a bit first--some disk drives are slow to clean up
769: * and users have reported disk corruption.
770: */
771: delay(500000);
772: apm_set_powstate(APM_DEV_DISK(0xff), APM_SYS_OFF);
773: delay(500000);
774: apm_set_powstate(APM_DEV_ALLDEVS, APM_SYS_OFF);
1.531 kochi 775: printf("WARNING: APM powerdown failed!\n");
1.307 thorpej 776: /*
777: * RB_POWERDOWN implies RB_HALT... fall into it...
778: */
1.208 jtk 779: #endif
1.307 thorpej 780: }
781:
782: if (howto & RB_HALT) {
1.210 christos 783: printf("\n");
784: printf("The operating system has halted.\n");
785: printf("Please press any key to reboot.\n\n");
1.517 jmmv 786:
787: #ifdef BEEP_ONHALT
788: {
789: int c;
790: for (c = BEEP_ONHALT_COUNT; c > 0; c--) {
791: sysbeep(BEEP_ONHALT_PITCH,
792: BEEP_ONHALT_PERIOD * hz / 1000);
793: delay(BEEP_ONHALT_PERIOD * 1000);
794: sysbeep(0, BEEP_ONHALT_PERIOD * hz / 1000);
795: delay(BEEP_ONHALT_PERIOD * 1000);
796: }
797: }
798: #endif
799:
1.300 drochner 800: cnpollc(1); /* for proper keyboard command handling */
1.512 perry 801: if (cngetc() == 0) {
802: /* no console attached, so just hlt */
803: for(;;) {
804: __asm __volatile("hlt");
805: }
806: }
1.300 drochner 807: cnpollc(0);
1.1 cgd 808: }
1.193 mycroft 809:
1.210 christos 810: printf("rebooting...\n");
1.328 bouyer 811: if (cpureset_delay > 0)
812: delay(cpureset_delay * 1000);
1.1 cgd 813: cpu_reset();
814: for(;;) ;
815: /*NOTREACHED*/
816: }
817:
1.116 gwr 818: /*
819: * These variables are needed by /sbin/savecore
820: */
1.468 tsutsui 821: u_int32_t dumpmag = 0x8fca0101; /* magic number */
1.116 gwr 822: int dumpsize = 0; /* pages */
823: long dumplo = 0; /* blocks */
824:
825: /*
1.291 thorpej 826: * cpu_dumpsize: calculate size of machine-dependent kernel core dump headers.
827: */
828: int
829: cpu_dumpsize()
830: {
831: int size;
832:
833: size = ALIGN(sizeof(kcore_seg_t)) + ALIGN(sizeof(cpu_kcore_hdr_t)) +
834: ALIGN(mem_cluster_cnt * sizeof(phys_ram_seg_t));
835: if (roundup(size, dbtob(1)) != dbtob(1))
836: return (-1);
837:
838: return (1);
839: }
840:
841: /*
842: * cpu_dump_mempagecnt: calculate the size of RAM (in pages) to be dumped.
843: */
844: u_long
845: cpu_dump_mempagecnt()
846: {
847: u_long i, n;
848:
849: n = 0;
850: for (i = 0; i < mem_cluster_cnt; i++)
851: n += atop(mem_clusters[i].size);
852: return (n);
853: }
854:
855: /*
856: * cpu_dump: dump the machine-dependent kernel core dump headers.
857: */
858: int
859: cpu_dump()
860: {
1.542 junyoung 861: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.291 thorpej 862: char buf[dbtob(1)];
863: kcore_seg_t *segp;
864: cpu_kcore_hdr_t *cpuhdrp;
865: phys_ram_seg_t *memsegp;
1.481 gehenna 866: const struct bdevsw *bdev;
1.291 thorpej 867: int i;
868:
1.481 gehenna 869: bdev = bdevsw_lookup(dumpdev);
870: if (bdev == NULL)
871: return (ENXIO);
872: dump = bdev->d_dump;
1.291 thorpej 873:
1.313 perry 874: memset(buf, 0, sizeof buf);
1.291 thorpej 875: segp = (kcore_seg_t *)buf;
876: cpuhdrp = (cpu_kcore_hdr_t *)&buf[ALIGN(sizeof(*segp))];
877: memsegp = (phys_ram_seg_t *)&buf[ ALIGN(sizeof(*segp)) +
878: ALIGN(sizeof(*cpuhdrp))];
879:
880: /*
881: * Generate a segment header.
882: */
883: CORE_SETMAGIC(*segp, KCORE_MAGIC, MID_MACHINE, CORE_CPU);
884: segp->c_size = dbtob(1) - ALIGN(sizeof(*segp));
885:
886: /*
887: * Add the machine-dependent header info.
888: */
889: cpuhdrp->ptdpaddr = PTDpaddr;
890: cpuhdrp->nmemsegs = mem_cluster_cnt;
891:
892: /*
893: * Fill in the memory segment descriptors.
894: */
895: for (i = 0; i < mem_cluster_cnt; i++) {
896: memsegp[i].start = mem_clusters[i].start;
897: memsegp[i].size = mem_clusters[i].size;
898: }
899:
900: return (dump(dumpdev, dumplo, (caddr_t)buf, dbtob(1)));
901: }
902:
903: /*
1.228 gwr 904: * This is called by main to set dumplo and dumpsize.
1.414 thorpej 905: * Dumps always skip the first PAGE_SIZE of disk space
1.116 gwr 906: * in case there might be a disk label stored there.
907: * If there is extra space, put dump at the end to
908: * reduce the chance that swapping trashes it.
909: */
910: void
1.228 gwr 911: cpu_dumpconf()
1.116 gwr 912: {
1.481 gehenna 913: const struct bdevsw *bdev;
1.291 thorpej 914: int nblks, dumpblks; /* size of dump area */
1.116 gwr 915:
916: if (dumpdev == NODEV)
1.291 thorpej 917: goto bad;
1.481 gehenna 918: bdev = bdevsw_lookup(dumpdev);
919: if (bdev == NULL)
1.116 gwr 920: panic("dumpconf: bad dumpdev=0x%x", dumpdev);
1.481 gehenna 921: if (bdev->d_psize == NULL)
1.291 thorpej 922: goto bad;
1.481 gehenna 923: nblks = (*bdev->d_psize)(dumpdev);
1.116 gwr 924: if (nblks <= ctod(1))
1.291 thorpej 925: goto bad;
1.116 gwr 926:
1.291 thorpej 927: dumpblks = cpu_dumpsize();
928: if (dumpblks < 0)
929: goto bad;
930: dumpblks += ctod(cpu_dump_mempagecnt());
931:
932: /* If dump won't fit (incl. room for possible label), punt. */
933: if (dumpblks > (nblks - ctod(1)))
934: goto bad;
935:
936: /* Put dump at end of partition */
937: dumplo = nblks - dumpblks;
938:
939: /* dumpsize is in page units, and doesn't include headers. */
940: dumpsize = cpu_dump_mempagecnt();
941: return;
1.116 gwr 942:
1.291 thorpej 943: bad:
944: dumpsize = 0;
1.116 gwr 945: }
946:
1.1 cgd 947: /*
948: * Doadump comes here after turning off memory management and
949: * getting on the dump stack, either when called above, or by
950: * the auto-restart code.
951: */
1.414 thorpej 952: #define BYTES_PER_DUMP PAGE_SIZE /* must be a multiple of pagesize XXX small */
1.314 thorpej 953: static vaddr_t dumpspace;
1.163 cgd 954:
1.314 thorpej 955: vaddr_t
1.551 junyoung 956: reserve_dumppages(vaddr_t p)
1.163 cgd 957: {
958:
959: dumpspace = p;
960: return (p + BYTES_PER_DUMP);
961: }
962:
1.32 andrew 963: void
1.1 cgd 964: dumpsys()
965: {
1.291 thorpej 966: u_long totalbytesleft, bytes, i, n, memseg;
967: u_long maddr;
968: int psize;
1.163 cgd 969: daddr_t blkno;
1.481 gehenna 970: const struct bdevsw *bdev;
1.542 junyoung 971: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.200 christos 972: int error;
1.193 mycroft 973:
974: /* Save registers. */
975: savectx(&dumppcb);
1.1 cgd 976:
977: if (dumpdev == NODEV)
978: return;
1.484 fvdl 979:
1.481 gehenna 980: bdev = bdevsw_lookup(dumpdev);
981: if (bdev == NULL || bdev->d_psize == NULL)
982: return;
1.163 cgd 983:
984: /*
985: * For dumps during autoconfiguration,
986: * if dump device has already configured...
987: */
988: if (dumpsize == 0)
1.228 gwr 989: cpu_dumpconf();
1.330 jtk 990: if (dumplo <= 0 || dumpsize == 0) {
1.275 mycroft 991: printf("\ndump to dev %u,%u not possible\n", major(dumpdev),
992: minor(dumpdev));
1.163 cgd 993: return;
1.275 mycroft 994: }
995: printf("\ndumping to dev %u,%u offset %ld\n", major(dumpdev),
996: minor(dumpdev), dumplo);
1.134 mycroft 997:
1.481 gehenna 998: psize = (*bdev->d_psize)(dumpdev);
1.210 christos 999: printf("dump ");
1.163 cgd 1000: if (psize == -1) {
1.210 christos 1001: printf("area unavailable\n");
1.163 cgd 1002: return;
1003: }
1004:
1005: #if 0 /* XXX this doesn't work. grr. */
1006: /* toss any characters present prior to dump */
1007: while (sget() != NULL); /*syscons and pccons differ */
1008: #endif
1009:
1.291 thorpej 1010: if ((error = cpu_dump()) != 0)
1011: goto err;
1012:
1013: totalbytesleft = ptoa(cpu_dump_mempagecnt());
1014: blkno = dumplo + cpu_dumpsize();
1.481 gehenna 1015: dump = bdev->d_dump;
1.200 christos 1016: error = 0;
1.291 thorpej 1017:
1018: for (memseg = 0; memseg < mem_cluster_cnt; memseg++) {
1019: maddr = mem_clusters[memseg].start;
1020: bytes = mem_clusters[memseg].size;
1021:
1022: for (i = 0; i < bytes; i += n, totalbytesleft -= n) {
1023: /* Print out how many MBs we have left to go. */
1024: if ((totalbytesleft % (1024*1024)) == 0)
1025: printf("%ld ", totalbytesleft / (1024 * 1024));
1026:
1027: /* Limit size for next transfer. */
1028: n = bytes - i;
1029: if (n > BYTES_PER_DUMP)
1030: n = BYTES_PER_DUMP;
1031:
1032: (void) pmap_map(dumpspace, maddr, maddr + n,
1033: VM_PROT_READ);
1034:
1035: error = (*dump)(dumpdev, blkno, (caddr_t)dumpspace, n);
1036: if (error)
1037: goto err;
1.163 cgd 1038: maddr += n;
1.291 thorpej 1039: blkno += btodb(n); /* XXX? */
1.163 cgd 1040:
1041: #if 0 /* XXX this doesn't work. grr. */
1.291 thorpej 1042: /* operator aborting dump? */
1043: if (sget() != NULL) {
1044: error = EINTR;
1045: break;
1046: }
1047: #endif
1.163 cgd 1048: }
1049: }
1050:
1.291 thorpej 1051: err:
1.163 cgd 1052: switch (error) {
1.1 cgd 1053:
1054: case ENXIO:
1.210 christos 1055: printf("device bad\n");
1.1 cgd 1056: break;
1057:
1058: case EFAULT:
1.210 christos 1059: printf("device not ready\n");
1.1 cgd 1060: break;
1061:
1062: case EINVAL:
1.210 christos 1063: printf("area improper\n");
1.1 cgd 1064: break;
1065:
1066: case EIO:
1.210 christos 1067: printf("i/o error\n");
1.1 cgd 1068: break;
1069:
1070: case EINTR:
1.210 christos 1071: printf("aborted from console\n");
1.1 cgd 1072: break;
1073:
1.163 cgd 1074: case 0:
1.210 christos 1075: printf("succeeded\n");
1.163 cgd 1076: break;
1077:
1.1 cgd 1078: default:
1.210 christos 1079: printf("error %d\n", error);
1.1 cgd 1080: break;
1081: }
1.210 christos 1082: printf("\n\n");
1.163 cgd 1083: delay(5000000); /* 5 seconds */
1.1 cgd 1084: }
1085:
1086: /*
1087: * Clear registers on exec
1088: */
1.33 cgd 1089: void
1.551 junyoung 1090: setregs(struct lwp *l, struct exec_package *pack, u_long stack)
1.1 cgd 1091: {
1.533 chs 1092: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 1093: struct pcb *pcb = &l->l_addr->u_pcb;
1.298 mycroft 1094: struct trapframe *tf;
1.1 cgd 1095:
1.161 mycroft 1096: #if NNPX > 0
1097: /* If we were using the FPU, forget about it. */
1.508 thorpej 1098: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
1099: npxsave_lwp(l, 0);
1.161 mycroft 1100: #endif
1.166 mycroft 1101:
1.178 mycroft 1102: #ifdef USER_LDT
1.508 thorpej 1103: pmap_ldt_cleanup(l);
1.178 mycroft 1104: #endif
1105:
1.548 yamt 1106: l->l_md.md_flags &= ~MDL_USEDFPU;
1.452 thorpej 1107: if (i386_use_fxsave) {
1.450 thorpej 1108: pcb->pcb_savefpu.sv_xmm.sv_env.en_cw = __NetBSD_NPXCW__;
1.452 thorpej 1109: pcb->pcb_savefpu.sv_xmm.sv_env.en_mxcsr = __INITIAL_MXCSR__;
1110: } else
1.450 thorpej 1111: pcb->pcb_savefpu.sv_87.sv_env.en_cw = __NetBSD_NPXCW__;
1.59 mycroft 1112:
1.508 thorpej 1113: tf = l->l_md.md_regs;
1.445 sommerfe 1114: tf->tf_gs = LSEL(LUDATA_SEL, SEL_UPL);
1115: tf->tf_fs = LSEL(LUDATA_SEL, SEL_UPL);
1.154 mycroft 1116: tf->tf_es = LSEL(LUDATA_SEL, SEL_UPL);
1117: tf->tf_ds = LSEL(LUDATA_SEL, SEL_UPL);
1.252 mycroft 1118: tf->tf_edi = 0;
1119: tf->tf_esi = 0;
1.154 mycroft 1120: tf->tf_ebp = 0;
1.508 thorpej 1121: tf->tf_ebx = (int)l->l_proc->p_psstr;
1.252 mycroft 1122: tf->tf_edx = 0;
1123: tf->tf_ecx = 0;
1124: tf->tf_eax = 0;
1.154 mycroft 1125: tf->tf_eip = pack->ep_entry;
1.533 chs 1126: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
1127: LSEL(LUCODEBIG_SEL, SEL_UPL) : LSEL(LUCODE_SEL, SEL_UPL);
1.154 mycroft 1128: tf->tf_eflags = PSL_USERSET;
1129: tf->tf_esp = stack;
1130: tf->tf_ss = LSEL(LUDATA_SEL, SEL_UPL);
1.1 cgd 1131: }
1132:
1133: /*
1.55 brezak 1134: * Initialize segments and descriptor tables
1.1 cgd 1135: */
1136:
1.516 fvdl 1137: union descriptor *gdt, *ldt;
1138: struct gate_descriptor *idt;
1.498 fvdl 1139: char idt_allocmap[NIDT];
1140: struct simplelock idt_lock = SIMPLELOCK_INITIALIZER;
1.275 mycroft 1141: #ifdef I586_CPU
1142: union descriptor *pentium_idt;
1143: #endif
1.178 mycroft 1144: extern struct user *proc0paddr;
1.49 brezak 1145:
1.178 mycroft 1146: void
1.551 junyoung 1147: setgate(struct gate_descriptor *gd, void *func, int args, int type, int dpl,
1148: int sel)
1.178 mycroft 1149: {
1.1 cgd 1150:
1.178 mycroft 1151: gd->gd_looffset = (int)func;
1.489 fvdl 1152: gd->gd_selector = sel;
1.178 mycroft 1153: gd->gd_stkcpy = args;
1154: gd->gd_xx = 0;
1155: gd->gd_type = type;
1156: gd->gd_dpl = dpl;
1157: gd->gd_p = 1;
1158: gd->gd_hioffset = (int)func >> 16;
1159: }
1160:
1161: void
1.551 junyoung 1162: unsetgate(struct gate_descriptor *gd)
1.484 fvdl 1163: {
1164: gd->gd_p = 0;
1165: gd->gd_hioffset = 0;
1166: gd->gd_looffset = 0;
1167: gd->gd_selector = 0;
1168: gd->gd_xx = 0;
1169: gd->gd_stkcpy = 0;
1170: gd->gd_type = 0;
1171: gd->gd_dpl = 0;
1172: }
1173:
1174:
1175: void
1.551 junyoung 1176: setregion(struct region_descriptor *rd, void *base, size_t limit)
1.178 mycroft 1177: {
1178:
1179: rd->rd_limit = (int)limit;
1180: rd->rd_base = (int)base;
1181: }
1.1 cgd 1182:
1.174 mycroft 1183: void
1.551 junyoung 1184: setsegment(struct segment_descriptor *sd, void *base, size_t limit, int type,
1185: int dpl, int def32, int gran)
1.174 mycroft 1186: {
1.1 cgd 1187:
1.174 mycroft 1188: sd->sd_lolimit = (int)limit;
1189: sd->sd_lobase = (int)base;
1190: sd->sd_type = type;
1191: sd->sd_dpl = dpl;
1192: sd->sd_p = 1;
1193: sd->sd_hilimit = (int)limit >> 16;
1194: sd->sd_xx = 0;
1195: sd->sd_def32 = def32;
1196: sd->sd_gran = gran;
1197: sd->sd_hibase = (int)base >> 24;
1198: }
1.1 cgd 1199:
1200: #define IDTVEC(name) __CONCAT(X, name)
1.542 junyoung 1201: typedef void (vector)(void);
1.299 mycroft 1202: extern vector IDTVEC(syscall);
1203: extern vector IDTVEC(osyscall);
1204: extern vector *IDTVEC(exceptions)[];
1.333 christos 1205: #ifdef COMPAT_SVR4
1206: extern vector IDTVEC(svr4_fasttrap);
1207: #endif /* COMPAT_SVR4 */
1.447 christos 1208: #ifdef COMPAT_MACH
1209: extern vector IDTVEC(mach_trap);
1210: #endif
1.1 cgd 1211:
1.381 thorpej 1212: #define KBTOB(x) ((size_t)(x) * 1024UL)
1213:
1.484 fvdl 1214: void cpu_init_idt()
1215: {
1216: struct region_descriptor region;
1217: #ifdef I586_CPU
1218: setregion(®ion, pentium_idt, NIDT * sizeof(idt[0]) - 1);
1219: #else
1220: setregion(®ion, idt, NIDT * sizeof(idt[0]) - 1);
1221: #endif
1222: lidt(®ion);
1223: }
1224:
1.433 kanaoka 1225: #if !defined(REALBASEMEM) && !defined(REALEXTMEM)
1226: void
1.551 junyoung 1227: add_mem_cluster(u_int64_t seg_start, u_int64_t seg_end, u_int32_t type)
1.433 kanaoka 1228: {
1229: extern struct extent *iomem_ex;
1.492 kanaoka 1230: int i;
1.433 kanaoka 1231:
1232: if (seg_end > 0x100000000ULL) {
1233: printf("WARNING: skipping large "
1234: "memory map entry: "
1235: "0x%qx/0x%qx/0x%x\n",
1236: seg_start,
1237: (seg_end - seg_start),
1238: type);
1239: return;
1240: }
1241:
1242: /*
1243: * XXX Chop the last page off the size so that
1244: * XXX it can fit in avail_end.
1245: */
1246: if (seg_end == 0x100000000ULL)
1247: seg_end -= PAGE_SIZE;
1248:
1249: if (seg_end <= seg_start)
1250: return;
1251:
1.492 kanaoka 1252: for (i = 0; i < mem_cluster_cnt; i++) {
1253: if ((mem_clusters[i].start == round_page(seg_start))
1254: && (mem_clusters[i].size
1255: == trunc_page(seg_end) - mem_clusters[i].start)) {
1256: #ifdef DEBUG_MEMLOAD
1257: printf("WARNING: skipping duplicate segment entry\n");
1258: #endif
1259: return;
1260: }
1261: }
1262:
1.433 kanaoka 1263: /*
1264: * Allocate the physical addresses used by RAM
1265: * from the iomem extent map. This is done before
1266: * the addresses are page rounded just to make
1267: * sure we get them all.
1268: */
1269: if (extent_alloc_region(iomem_ex, seg_start,
1270: seg_end - seg_start, EX_NOWAIT)) {
1271: /* XXX What should we do? */
1272: printf("WARNING: CAN'T ALLOCATE "
1273: "MEMORY SEGMENT "
1274: "(0x%qx/0x%qx/0x%x) FROM "
1275: "IOMEM EXTENT MAP!\n",
1276: seg_start, seg_end - seg_start, type);
1.492 kanaoka 1277: return;
1.433 kanaoka 1278: }
1279:
1280: /*
1281: * If it's not free memory, skip it.
1282: */
1283: if (type != BIM_Memory)
1284: return;
1285:
1286: /* XXX XXX XXX */
1287: if (mem_cluster_cnt >= VM_PHYSSEG_MAX)
1288: panic("init386: too many memory segments");
1289:
1290: seg_start = round_page(seg_start);
1291: seg_end = trunc_page(seg_end);
1292:
1293: if (seg_start == seg_end)
1294: return;
1295:
1296: mem_clusters[mem_cluster_cnt].start = seg_start;
1297: mem_clusters[mem_cluster_cnt].size =
1298: seg_end - seg_start;
1299:
1300: if (avail_end < seg_end)
1301: avail_end = seg_end;
1302: physmem += atop(mem_clusters[mem_cluster_cnt].size);
1303: mem_cluster_cnt++;
1304: }
1305: #endif /* !defined(REALBASEMEM) && !defined(REALEXTMEM) */
1306:
1.59 mycroft 1307: void
1.484 fvdl 1308: initgdt(union descriptor *tgdt)
1309: {
1310: struct region_descriptor region;
1311: gdt = tgdt;
1312: memset(gdt, 0, NGDT*sizeof(*gdt));
1313: /* make gdt gates and memory segments */
1314: setsegment(&gdt[GCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 1, 1);
1315: setsegment(&gdt[GDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 1, 1);
1.533 chs 1316: setsegment(&gdt[GUCODE_SEL].sd, 0, x86_btop(I386_MAX_EXE_ADDR) - 1,
1317: SDT_MEMERA, SEL_UPL, 1, 1);
1318: setsegment(&gdt[GUCODEBIG_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1319: SDT_MEMERA, SEL_UPL, 1, 1);
1.514 fvdl 1320: setsegment(&gdt[GUDATA_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1321: SDT_MEMRWA, SEL_UPL, 1, 1);
1322: #ifdef COMPAT_MACH
1323: setgate(&gdt[GMACHCALLS_SEL].gd, &IDTVEC(mach_trap), 1,
1.489 fvdl 1324: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 1325: #endif
1326: #if NBIOSCALL > 0
1327: /* bios trampoline GDT entries */
1328: setsegment(&gdt[GBIOSCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 0,
1329: 0);
1330: setsegment(&gdt[GBIOSDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 0,
1331: 0);
1332: #endif
1333: setsegment(&gdt[GCPU_SEL].sd, &cpu_info_primary,
1334: sizeof(struct cpu_info)-1, SDT_MEMRWA, SEL_KPL, 1, 1);
1335:
1336: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
1337: lgdt(®ion);
1338: }
1339:
1340: void
1.551 junyoung 1341: init386(paddr_t first_avail)
1.2 cgd 1342: {
1.484 fvdl 1343: union descriptor *tgdt;
1.542 junyoung 1344: extern void consinit(void);
1.375 drochner 1345: extern struct extent *iomem_ex;
1.429 chs 1346: #if !defined(REALBASEMEM) && !defined(REALEXTMEM)
1.401 thorpej 1347: struct btinfo_memmap *bim;
1.429 chs 1348: #endif
1.381 thorpej 1349: struct region_descriptor region;
1.401 thorpej 1350: int x, first16q;
1351: u_int64_t seg_start, seg_end;
1352: u_int64_t seg_start1, seg_end1;
1.473 tshiozak 1353: paddr_t realmode_reserved_start;
1354: psize_t realmode_reserved_size;
1355: int needs_earlier_install_pte0;
1.436 jdolecek 1356: #if NBIOSCALL > 0
1357: extern int biostramp_image_size;
1358: extern u_char biostramp_image[];
1359: #endif
1.1 cgd 1360:
1.484 fvdl 1361: cpu_probe_features(&cpu_info_primary);
1362: cpu_feature = cpu_info_primary.ci_feature_flags;
1.553 lukem 1363: cpu_feature2 = cpu_info_primary.ci_feature2_flags;
1.484 fvdl 1364:
1.508 thorpej 1365: lwp0.l_addr = proc0paddr;
1366: cpu_info_primary.ci_curpcb = &lwp0.l_addr->u_pcb;
1.275 mycroft 1367:
1.514 fvdl 1368: x86_bus_space_init();
1.84 cgd 1369: consinit(); /* XXX SHOULD NOT BE DONE HERE */
1.401 thorpej 1370: /*
1371: * Initailize PAGE_SIZE-dependent variables.
1372: */
1373: uvm_setpagesize();
1.450 thorpej 1374:
1375: /*
1376: * Saving SSE registers won't work if the save area isn't
1377: * 16-byte aligned.
1378: */
1379: if (offsetof(struct user, u_pcb.pcb_savefpu) & 0xf)
1380: panic("init386: pcb_savefpu not 16-byte aligned");
1.440 thorpej 1381:
1382: /*
1383: * Start with 2 color bins -- this is just a guess to get us
1384: * started. We'll recolor when we determine the largest cache
1385: * sizes on the system.
1386: */
1387: uvmexp.ncolors = 2;
1.401 thorpej 1388:
1.473 tshiozak 1389: /*
1.484 fvdl 1390: * BIOS leaves data in physical page 0
1391: * Even if it didn't, our VM system doesn't like using zero as a
1392: * physical page number.
1393: * We may also need pages in low memory (one each) for secondary CPU
1394: * startup, for BIOS calls, and for ACPI, plus a page table page to map
1395: * them into the first few pages of the kernel's pmap.
1.473 tshiozak 1396: */
1397: avail_start = PAGE_SIZE;
1398:
1399: /*
1400: * reserve memory for real-mode call
1401: */
1402: needs_earlier_install_pte0 = 0;
1403: realmode_reserved_start = 0;
1404: realmode_reserved_size = 0;
1.414 thorpej 1405: #if NBIOSCALL > 0
1.473 tshiozak 1406: /* save us a page for trampoline code */
1407: realmode_reserved_size += PAGE_SIZE;
1408: needs_earlier_install_pte0 = 1;
1409: #endif
1.484 fvdl 1410: #ifdef MULTIPROCESSOR /* XXX */
1411: KASSERT(avail_start == PAGE_SIZE); /* XXX */
1412: if (realmode_reserved_size < MP_TRAMPOLINE) /* XXX */
1413: realmode_reserved_size = MP_TRAMPOLINE; /* XXX */
1414: needs_earlier_install_pte0 = 1; /* XXX */
1415: #endif /* XXX */
1.473 tshiozak 1416: #if NACPI > 0
1417: /* trampoline code for wake handler */
1418: realmode_reserved_size += ptoa(acpi_md_get_npages_of_wakecode()+1);
1419: needs_earlier_install_pte0 = 1;
1420: #endif
1421: if (needs_earlier_install_pte0) {
1422: /* page table for directory entry 0 */
1423: realmode_reserved_size += PAGE_SIZE;
1424: }
1425: if (realmode_reserved_size>0) {
1426: realmode_reserved_start = avail_start;
1427: avail_start += realmode_reserved_size;
1428: }
1.414 thorpej 1429:
1.492 kanaoka 1430: #ifdef DEBUG_MEMLOAD
1431: printf("mem_cluster_count: %d\n", mem_cluster_cnt);
1432: #endif
1433:
1.401 thorpej 1434: /*
1435: * Call pmap initialization to make new kernel address space.
1436: * We must do this before loading pages into the VM system.
1437: */
1.314 thorpej 1438: pmap_bootstrap((vaddr_t)atdevbase + IOM_SIZE);
1.2 cgd 1439:
1.429 chs 1440: #if !defined(REALBASEMEM) && !defined(REALEXTMEM)
1.401 thorpej 1441: /*
1442: * Check to see if we have a memory map from the BIOS (passed
1443: * to us by the boot program.
1444: */
1445: bim = lookup_bootinfo(BTINFO_MEMMAP);
1.407 enami 1446: if (bim != NULL && bim->num > 0) {
1.463 enami 1447: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1448: printf("BIOS MEMORY MAP (%d ENTRIES):\n", bim->num);
1449: #endif
1450: for (x = 0; x < bim->num; x++) {
1.463 enami 1451: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1452: printf(" addr 0x%qx size 0x%qx type 0x%x\n",
1453: bim->entry[x].addr,
1454: bim->entry[x].size,
1455: bim->entry[x].type);
1456: #endif
1457:
1458: /*
1459: * If the segment is not memory, skip it.
1460: */
1461: switch (bim->entry[x].type) {
1462: case BIM_Memory:
1463: case BIM_ACPI:
1464: case BIM_NVS:
1465: break;
1466: default:
1467: continue;
1468: }
1469:
1470: /*
1471: * Sanity check the entry.
1472: * XXX Need to handle uint64_t in extent code
1473: * XXX and 64-bit physical addresses in i386
1474: * XXX port.
1475: */
1476: seg_start = bim->entry[x].addr;
1477: seg_end = bim->entry[x].addr + bim->entry[x].size;
1478:
1.432 kanaoka 1479: /*
1.433 kanaoka 1480: * Avoid Compatibility Holes.
1481: * XXX Holes within memory space that allow access
1482: * XXX to be directed to the PC-compatible frame buffer
1.484 fvdl 1483: * XXX (0xa0000-0xbffff),to adapter ROM space
1.433 kanaoka 1484: * XXX (0xc0000-0xdffff), and to system BIOS space
1485: * XXX (0xe0000-0xfffff).
1486: * XXX Some laptop(for example,Toshiba Satellite2550X)
1487: * XXX report this area and occurred problems,
1488: * XXX so we avoid this area.
1.432 kanaoka 1489: */
1.433 kanaoka 1490: if (seg_start < 0x100000 && seg_end > 0xa0000) {
1491: printf("WARNING: memory map entry overlaps "
1492: "with ``Compatibility Holes'': "
1493: "0x%qx/0x%qx/0x%x\n", seg_start,
1494: seg_end - seg_start, bim->entry[x].type);
1495: add_mem_cluster(seg_start, 0xa0000,
1496: bim->entry[x].type);
1497: add_mem_cluster(0x100000, seg_end,
1.401 thorpej 1498: bim->entry[x].type);
1.433 kanaoka 1499: } else
1500: add_mem_cluster(seg_start, seg_end,
1.401 thorpej 1501: bim->entry[x].type);
1502: }
1.421 aymeric 1503: }
1.429 chs 1504: #endif /* ! REALBASEMEM && ! REALEXTMEM */
1.421 aymeric 1505: /*
1506: * If the loop above didn't find any valid segment, fall back to
1507: * former code.
1508: */
1509: if (mem_cluster_cnt == 0) {
1.401 thorpej 1510: /*
1511: * Allocate the physical addresses used by RAM from the iomem
1512: * extent map. This is done before the addresses are
1513: * page rounded just to make sure we get them all.
1514: */
1515: if (extent_alloc_region(iomem_ex, 0, KBTOB(biosbasemem),
1516: EX_NOWAIT)) {
1517: /* XXX What should we do? */
1518: printf("WARNING: CAN'T ALLOCATE BASE MEMORY FROM "
1519: "IOMEM EXTENT MAP!\n");
1520: }
1521: mem_clusters[0].start = 0;
1522: mem_clusters[0].size = trunc_page(KBTOB(biosbasemem));
1523: physmem += atop(mem_clusters[0].size);
1524: if (extent_alloc_region(iomem_ex, IOM_END, KBTOB(biosextmem),
1525: EX_NOWAIT)) {
1526: /* XXX What should we do? */
1527: printf("WARNING: CAN'T ALLOCATE EXTENDED MEMORY FROM "
1528: "IOMEM EXTENT MAP!\n");
1529: }
1530: #if NISADMA > 0
1531: /*
1532: * Some motherboards/BIOSes remap the 384K of RAM that would
1533: * normally be covered by the ISA hole to the end of memory
1534: * so that it can be used. However, on a 16M system, this
1535: * would cause bounce buffers to be allocated and used.
1536: * This is not desirable behaviour, as more than 384K of
1537: * bounce buffers might be allocated. As a work-around,
1538: * we round memory down to the nearest 1M boundary if
1539: * we're using any isadma devices and the remapped memory
1540: * is what puts us over 16M.
1541: */
1542: if (biosextmem > (15*1024) && biosextmem < (16*1024)) {
1543: char pbuf[9];
1544:
1545: format_bytes(pbuf, sizeof(pbuf),
1546: biosextmem - (15*1024));
1547: printf("Warning: ignoring %s of remapped memory\n",
1548: pbuf);
1549: biosextmem = (15*1024);
1550: }
1551: #endif
1552: mem_clusters[1].start = IOM_END;
1553: mem_clusters[1].size = trunc_page(KBTOB(biosextmem));
1554: physmem += atop(mem_clusters[1].size);
1555:
1556: mem_cluster_cnt = 2;
1557:
1558: avail_end = IOM_END + trunc_page(KBTOB(biosextmem));
1559: }
1560: /*
1561: * If we have 16M of RAM or less, just put it all on
1562: * the default free list. Otherwise, put the first
1563: * 16M of RAM on a lower priority free list (so that
1564: * all of the ISA DMA'able memory won't be eaten up
1565: * first-off).
1566: */
1567: if (avail_end <= (16 * 1024 * 1024))
1568: first16q = VM_FREELIST_DEFAULT;
1569: else
1570: first16q = VM_FREELIST_FIRST16;
1571:
1572: /* Make sure the end of the space used by the kernel is rounded. */
1573: first_avail = round_page(first_avail);
1574:
1575: /*
1576: * Now, load the memory clusters (which have already been
1577: * rounded and truncated) into the VM system.
1578: *
1579: * NOTE: WE ASSUME THAT MEMORY STARTS AT 0 AND THAT THE KERNEL
1580: * IS LOADED AT IOM_END (1M).
1581: */
1582: for (x = 0; x < mem_cluster_cnt; x++) {
1583: seg_start = mem_clusters[x].start;
1584: seg_end = mem_clusters[x].start + mem_clusters[x].size;
1585: seg_start1 = 0;
1586: seg_end1 = 0;
1587:
1588: /*
1589: * Skip memory before our available starting point.
1590: */
1591: if (seg_end <= avail_start)
1592: continue;
1593:
1594: if (avail_start >= seg_start && avail_start < seg_end) {
1595: if (seg_start != 0)
1596: panic("init386: memory doesn't start at 0");
1597: seg_start = avail_start;
1598: if (seg_start == seg_end)
1599: continue;
1600: }
1601:
1602: /*
1603: * If this segment contains the kernel, split it
1604: * in two, around the kernel.
1605: */
1606: if (seg_start <= IOM_END && first_avail <= seg_end) {
1607: seg_start1 = first_avail;
1608: seg_end1 = seg_end;
1609: seg_end = IOM_END;
1610: }
1611:
1612: /* First hunk */
1613: if (seg_start != seg_end) {
1.480 erh 1614: if (seg_start < (16 * 1024 * 1024) &&
1.401 thorpej 1615: first16q != VM_FREELIST_DEFAULT) {
1616: u_int64_t tmp;
1617:
1618: if (seg_end > (16 * 1024 * 1024))
1619: tmp = (16 * 1024 * 1024);
1620: else
1621: tmp = seg_end;
1.492 kanaoka 1622:
1623: if (tmp != seg_start) {
1.463 enami 1624: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1625: printf("loading 0x%qx-0x%qx "
1626: "(0x%lx-0x%lx)\n",
1627: seg_start, tmp,
1628: atop(seg_start), atop(tmp));
1629: #endif
1630: uvm_page_physload(atop(seg_start),
1631: atop(tmp), atop(seg_start),
1632: atop(tmp), first16q);
1633: }
1.401 thorpej 1634: seg_start = tmp;
1635: }
1.411 enami 1636:
1637: if (seg_start != seg_end) {
1.463 enami 1638: #ifdef DEBUG_MEMLOAD
1.411 enami 1639: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1640: seg_start, seg_end,
1641: atop(seg_start), atop(seg_end));
1642: #endif
1643: uvm_page_physload(atop(seg_start),
1644: atop(seg_end), atop(seg_start),
1645: atop(seg_end), VM_FREELIST_DEFAULT);
1646: }
1.401 thorpej 1647: }
1648:
1649: /* Second hunk */
1650: if (seg_start1 != seg_end1) {
1.480 erh 1651: if (seg_start1 < (16 * 1024 * 1024) &&
1.401 thorpej 1652: first16q != VM_FREELIST_DEFAULT) {
1653: u_int64_t tmp;
1654:
1655: if (seg_end1 > (16 * 1024 * 1024))
1656: tmp = (16 * 1024 * 1024);
1657: else
1658: tmp = seg_end1;
1.492 kanaoka 1659:
1660: if (tmp != seg_start1) {
1.463 enami 1661: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1662: printf("loading 0x%qx-0x%qx "
1663: "(0x%lx-0x%lx)\n",
1664: seg_start1, tmp,
1665: atop(seg_start1), atop(tmp));
1666: #endif
1667: uvm_page_physload(atop(seg_start1),
1668: atop(tmp), atop(seg_start1),
1669: atop(tmp), first16q);
1670: }
1.401 thorpej 1671: seg_start1 = tmp;
1672: }
1.412 enami 1673:
1674: if (seg_start1 != seg_end1) {
1.463 enami 1675: #ifdef DEBUG_MEMLOAD
1.412 enami 1676: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1677: seg_start1, seg_end1,
1678: atop(seg_start1), atop(seg_end1));
1679: #endif
1680: uvm_page_physload(atop(seg_start1),
1681: atop(seg_end1), atop(seg_start1),
1682: atop(seg_end1), VM_FREELIST_DEFAULT);
1683: }
1.401 thorpej 1684: }
1685: }
1686:
1687: /*
1688: * Steal memory for the message buffer (at end of core).
1689: */
1690: {
1.543 mycroft 1691: struct vm_physseg *vps;
1.401 thorpej 1692: psize_t sz = round_page(MSGBUFSIZE);
1693: psize_t reqsz = sz;
1694:
1695: for (x = 0; x < vm_nphysseg; x++) {
1696: vps = &vm_physmem[x];
1697: if (ptoa(vps->avail_end) == avail_end)
1.543 mycroft 1698: goto found;
1.401 thorpej 1699: }
1.543 mycroft 1700: panic("init386: can't find end of memory");
1.401 thorpej 1701:
1.543 mycroft 1702: found:
1.401 thorpej 1703: /* Shrink so it'll fit in the last segment. */
1704: if ((vps->avail_end - vps->avail_start) < atop(sz))
1705: sz = ptoa(vps->avail_end - vps->avail_start);
1706:
1707: vps->avail_end -= atop(sz);
1708: vps->end -= atop(sz);
1709: msgbuf_paddr = ptoa(vps->avail_end);
1710:
1711: /* Remove the last segment if it now has no pages. */
1712: if (vps->start == vps->end) {
1713: for (vm_nphysseg--; x < vm_nphysseg; x++)
1714: vm_physmem[x] = vm_physmem[x + 1];
1715: }
1716:
1717: /* Now find where the new avail_end is. */
1718: for (avail_end = 0, x = 0; x < vm_nphysseg; x++)
1719: if (vm_physmem[x].avail_end > avail_end)
1720: avail_end = vm_physmem[x].avail_end;
1721: avail_end = ptoa(avail_end);
1722:
1723: /* Warn if the message buffer had to be shrunk. */
1724: if (sz != reqsz)
1725: printf("WARNING: %ld bytes not available for msgbuf "
1726: "in last cluster (%ld used)\n", reqsz, sz);
1727: }
1728:
1.473 tshiozak 1729: /*
1730: * install PT page for the first 4M if needed.
1731: */
1732: if (needs_earlier_install_pte0) {
1733: paddr_t paddr;
1734: #ifdef DIAGNOSTIC
1735: if (realmode_reserved_size < PAGE_SIZE) {
1736: panic("cannot steal memory for first 4M PT page.");
1737: }
1738: #endif
1739: paddr=realmode_reserved_start+realmode_reserved_size-PAGE_SIZE;
1740: pmap_enter(pmap_kernel(), (vaddr_t)vtopte(0), paddr,
1741: VM_PROT_READ|VM_PROT_WRITE,
1742: PMAP_WIRED|VM_PROT_READ|VM_PROT_WRITE);
1743: pmap_update(pmap_kernel());
1744: /* make sure it is clean before using */
1745: memset(vtopte(0), 0, PAGE_SIZE);
1746: realmode_reserved_size -= PAGE_SIZE;
1747: }
1748:
1.295 drochner 1749: #if NBIOSCALL > 0
1.436 jdolecek 1750: /*
1751: * this should be caught at kernel build time, but put it here
1752: * in case someone tries to fake it out...
1753: */
1754: #ifdef DIAGNOSTIC
1.473 tshiozak 1755: if (realmode_reserved_start > BIOSTRAMP_BASE ||
1.474 tron 1756: (realmode_reserved_start+realmode_reserved_size) < (BIOSTRAMP_BASE+
1.473 tshiozak 1757: PAGE_SIZE)) {
1758: panic("cannot steal memory for PT page of bioscall.");
1759: }
1.436 jdolecek 1760: if (biostramp_image_size > PAGE_SIZE)
1.483 provos 1761: panic("biostramp_image_size too big: %x vs. %x",
1.436 jdolecek 1762: biostramp_image_size, PAGE_SIZE);
1763: #endif
1764: pmap_kenter_pa((vaddr_t)BIOSTRAMP_BASE, /* virtual */
1765: (paddr_t)BIOSTRAMP_BASE, /* physical */
1766: VM_PROT_ALL); /* protection */
1.456 chris 1767: pmap_update(pmap_kernel());
1.436 jdolecek 1768: memcpy((caddr_t)BIOSTRAMP_BASE, biostramp_image, biostramp_image_size);
1769: #ifdef DEBUG_BIOSCALL
1770: printf("biostramp installed @ %x\n", BIOSTRAMP_BASE);
1771: #endif
1.473 tshiozak 1772: realmode_reserved_size -= PAGE_SIZE;
1773: realmode_reserved_start += PAGE_SIZE;
1774: #endif
1775:
1776: #if NACPI > 0
1777: /*
1778: * Steal memory for the acpi wake code
1779: */
1780: {
1781: paddr_t paddr, p;
1782: psize_t sz;
1783: int npg;
1784:
1785: paddr = realmode_reserved_start;
1786: npg = acpi_md_get_npages_of_wakecode();
1787: sz = ptoa(npg);
1788: #ifdef DIAGNOSTIC
1789: if (realmode_reserved_size < sz) {
1790: panic("cannot steal memory for ACPI wake code.");
1791: }
1792: #endif
1793:
1794: /* identical mapping */
1795: p = paddr;
1796: for (x=0; x<npg; x++) {
1797: printf("kenter: 0x%08X\n", (unsigned)p);
1798: pmap_kenter_pa((vaddr_t)p, p, VM_PROT_ALL);
1799: p += PAGE_SIZE;
1800: }
1801: pmap_update(pmap_kernel());
1802:
1803: acpi_md_install_wakecode(paddr);
1804:
1805: realmode_reserved_size -= sz;
1806: realmode_reserved_start += sz;
1807: }
1.295 drochner 1808: #endif
1.59 mycroft 1809:
1.346 mycroft 1810: pmap_enter(pmap_kernel(), idt_vaddr, idt_paddr,
1.367 thorpej 1811: VM_PROT_READ|VM_PROT_WRITE, PMAP_WIRED|VM_PROT_READ|VM_PROT_WRITE);
1.456 chris 1812: pmap_update(pmap_kernel());
1.484 fvdl 1813: memset((void *)idt_vaddr, 0, PAGE_SIZE);
1814:
1.516 fvdl 1815: idt = (struct gate_descriptor *)idt_vaddr;
1.275 mycroft 1816: #ifdef I586_CPU
1.346 mycroft 1817: pmap_enter(pmap_kernel(), pentium_idt_vaddr, idt_paddr,
1.367 thorpej 1818: VM_PROT_READ, PMAP_WIRED|VM_PROT_READ);
1.275 mycroft 1819: pentium_idt = (union descriptor *)pentium_idt_vaddr;
1820: #endif
1.484 fvdl 1821: pmap_update(pmap_kernel());
1822:
1823: tgdt = gdt;
1.516 fvdl 1824: gdt = (union descriptor *)
1825: ((char *)idt + NIDT * sizeof (struct gate_descriptor));
1.275 mycroft 1826: ldt = gdt + NGDT;
1827:
1.484 fvdl 1828: memcpy(gdt, tgdt, NGDT*sizeof(*gdt));
1.275 mycroft 1829:
1830: setsegment(&gdt[GLDT_SEL].sd, ldt, NLDT * sizeof(ldt[0]) - 1,
1831: SDT_SYSLDT, SEL_KPL, 0, 0);
1832:
1833: /* make ldt gates and memory segments */
1834: setgate(&ldt[LSYS5CALLS_SEL].gd, &IDTVEC(osyscall), 1,
1.489 fvdl 1835: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.447 christos 1836:
1.275 mycroft 1837: ldt[LUCODE_SEL] = gdt[GUCODE_SEL];
1.533 chs 1838: ldt[LUCODEBIG_SEL] = gdt[GUCODEBIG_SEL];
1.275 mycroft 1839: ldt[LUDATA_SEL] = gdt[GUDATA_SEL];
1.324 christos 1840: ldt[LSOL26CALLS_SEL] = ldt[LBSDICALLS_SEL] = ldt[LSYS5CALLS_SEL];
1.275 mycroft 1841:
1842: /* exceptions */
1.498 fvdl 1843: for (x = 0; x < 32; x++) {
1.516 fvdl 1844: setgate(&idt[x], IDTVEC(exceptions)[x], 0, SDT_SYS386TGT,
1.489 fvdl 1845: (x == 3 || x == 4) ? SEL_UPL : SEL_KPL,
1846: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 1847: idt_allocmap[x] = 1;
1848: }
1.257 thorpej 1849:
1.275 mycroft 1850: /* new-style interrupt gate for syscalls */
1.516 fvdl 1851: setgate(&idt[128], &IDTVEC(syscall), 0, SDT_SYS386TGT, SEL_UPL,
1.489 fvdl 1852: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 1853: idt_allocmap[128] = 1;
1.333 christos 1854: #ifdef COMPAT_SVR4
1.516 fvdl 1855: setgate(&idt[0xd2], &IDTVEC(svr4_fasttrap), 0, SDT_SYS386TGT,
1.489 fvdl 1856: SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 1857: idt_allocmap[0xd2] = 1;
1.333 christos 1858: #endif /* COMPAT_SVR4 */
1.264 mycroft 1859:
1.275 mycroft 1860: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
1861: lgdt(®ion);
1.484 fvdl 1862:
1863: cpu_init_idt();
1.264 mycroft 1864:
1.522 ragge 1865: #if NKSYMS || defined(DDB) || defined(LKM)
1.308 tv 1866: {
1867: extern int end;
1868: extern int *esym;
1.336 christos 1869: struct btinfo_symtab *symtab;
1.308 tv 1870:
1.522 ragge 1871: #ifdef DDB
1.484 fvdl 1872: db_machine_init();
1.522 ragge 1873: #endif
1.484 fvdl 1874:
1.336 christos 1875: symtab = lookup_bootinfo(BTINFO_SYMTAB);
1.484 fvdl 1876:
1.336 christos 1877: if (symtab) {
1878: symtab->ssym += KERNBASE;
1879: symtab->esym += KERNBASE;
1.522 ragge 1880: ksyms_init(symtab->nsym, (int *)symtab->ssym,
1.336 christos 1881: (int *)symtab->esym);
1882: }
1883: else
1.522 ragge 1884: ksyms_init(*(int *)&end, ((int *)&end) + 1, esym);
1.308 tv 1885: }
1.522 ragge 1886: #endif
1887: #ifdef DDB
1.190 mycroft 1888: if (boothowto & RB_KDB)
1889: Debugger();
1.377 ws 1890: #endif
1891: #ifdef IPKDB
1892: ipkdb_init();
1893: if (boothowto & RB_KDB)
1894: ipkdb_connect(0);
1.190 mycroft 1895: #endif
1896: #ifdef KGDB
1.243 drochner 1897: kgdb_port_init();
1.235 thorpej 1898: if (boothowto & RB_KDB) {
1899: kgdb_debug_init = 1;
1.242 drochner 1900: kgdb_connect(1);
1.235 thorpej 1901: }
1.384 jdolecek 1902: #endif
1903:
1904: #if NMCA > 0
1905: /* check for MCA bus, needed to be done before ISA stuff - if
1906: * MCA is detected, ISA needs to use level triggered interrupts
1907: * by default */
1908: mca_busprobe();
1.190 mycroft 1909: #endif
1.275 mycroft 1910:
1.498 fvdl 1911: intr_default_setup();
1.431 thorpej 1912:
1913: /* Initialize software interrupts. */
1914: softintr_init();
1.275 mycroft 1915:
1.498 fvdl 1916: splraise(IPL_IPI);
1.275 mycroft 1917: enable_intr();
1918:
1919: if (physmem < btoc(2 * 1024 * 1024)) {
1920: printf("warning: too little memory available; "
1.383 mycroft 1921: "have %lu bytes, want %lu bytes\n"
1.275 mycroft 1922: "running in degraded mode\n"
1923: "press a key to confirm\n\n",
1.383 mycroft 1924: ptoa(physmem), 2*1024*1024UL);
1.275 mycroft 1925: cngetc();
1926: }
1.507 jdolecek 1927:
1928: #ifdef __HAVE_CPU_MAXPROC
1929: /* Make sure maxproc is sane */
1930: if (maxproc > cpu_maxproc())
1931: maxproc = cpu_maxproc();
1932: #endif
1.1 cgd 1933: }
1934:
1.107 deraadt 1935: #ifdef COMPAT_NOMID
1936: static int
1.551 junyoung 1937: exec_nomid(struct proc *p, struct exec_package *epp)
1.31 cgd 1938: {
1.59 mycroft 1939: int error;
1940: u_long midmag, magic;
1941: u_short mid;
1.80 cgd 1942: struct exec *execp = epp->ep_hdr;
1.31 cgd 1943:
1.80 cgd 1944: /* check on validity of epp->ep_hdr performed by exec_out_makecmds */
1945:
1946: midmag = ntohl(execp->a_midmag);
1.59 mycroft 1947: mid = (midmag >> 16) & 0xffff;
1948: magic = midmag & 0xffff;
1949:
1950: if (magic == 0) {
1.80 cgd 1951: magic = (execp->a_midmag & 0xffff);
1.59 mycroft 1952: mid = MID_ZERO;
1953: }
1954:
1955: midmag = mid << 16 | magic;
1956:
1957: switch (midmag) {
1958: case (MID_ZERO << 16) | ZMAGIC:
1959: /*
1960: * 386BSD's ZMAGIC format:
1961: */
1.202 christos 1962: error = exec_aout_prep_oldzmagic(p, epp);
1.59 mycroft 1963: break;
1964:
1965: case (MID_ZERO << 16) | QMAGIC:
1966: /*
1967: * BSDI's QMAGIC format:
1968: * same as new ZMAGIC format, but with different magic number
1969: */
1970: error = exec_aout_prep_zmagic(p, epp);
1971: break;
1972:
1.202 christos 1973: case (MID_ZERO << 16) | NMAGIC:
1974: /*
1975: * BSDI's NMAGIC format:
1976: * same as NMAGIC format, but with different magic number
1977: * and with text starting at 0.
1978: */
1979: error = exec_aout_prep_oldnmagic(p, epp);
1980: break;
1981:
1982: case (MID_ZERO << 16) | OMAGIC:
1983: /*
1984: * BSDI's OMAGIC format:
1985: * same as OMAGIC format, but with different magic number
1986: * and with text starting at 0.
1987: */
1988: error = exec_aout_prep_oldomagic(p, epp);
1989: break;
1990:
1.59 mycroft 1991: default:
1992: error = ENOEXEC;
1993: }
1994:
1995: return error;
1.107 deraadt 1996: }
1.31 cgd 1997: #endif
1.107 deraadt 1998:
1999: /*
2000: * cpu_exec_aout_makecmds():
1.549 wiz 2001: * CPU-dependent a.out format hook for execve().
1.107 deraadt 2002: *
2003: * Determine of the given exec package refers to something which we
2004: * understand and, if so, set up the vmcmds for it.
2005: *
2006: * On the i386, old (386bsd) ZMAGIC binaries and BSDI QMAGIC binaries
2007: * if COMPAT_NOMID is given as a kernel option.
2008: */
2009: int
1.551 junyoung 2010: cpu_exec_aout_makecmds(struct proc *p, struct exec_package *epp)
1.107 deraadt 2011: {
2012: int error = ENOEXEC;
2013:
2014: #ifdef COMPAT_NOMID
1.526 fvdl 2015: if ((error = exec_nomid(p, epp)) == 0)
1.107 deraadt 2016: return error;
2017: #endif /* ! COMPAT_NOMID */
2018:
2019: return error;
1.31 cgd 2020: }
1.84 cgd 2021:
1.255 drochner 2022: void *
1.551 junyoung 2023: lookup_bootinfo(int type)
1.255 drochner 2024: {
2025: struct btinfo_common *help;
2026: int n = *(int*)bootinfo;
2027: help = (struct btinfo_common *)(bootinfo + sizeof(int));
2028: while(n--) {
2029: if(help->type == type)
2030: return(help);
2031: help = (struct btinfo_common *)((char*)help + help->len);
2032: }
2033: return(0);
2034: }
1.149 mycroft 2035:
1.484 fvdl 2036: #include <dev/ic/mc146818reg.h> /* for NVRAM POST */
2037: #include <i386/isa/nvram.h> /* for NVRAM POST */
2038:
1.149 mycroft 2039: void
2040: cpu_reset()
2041: {
2042:
1.224 mycroft 2043: disable_intr();
2044:
1.227 mycroft 2045: /*
1.484 fvdl 2046: * Ensure the NVRAM reset byte contains something vaguely sane.
2047: */
2048:
2049: outb(IO_RTC, NVRAM_RESET);
2050: outb(IO_RTC+1, NVRAM_RESET_RST);
2051:
2052: /*
1.227 mycroft 2053: * The keyboard controller has 4 random output pins, one of which is
2054: * connected to the RESET pin on the CPU in many PCs. We tell the
2055: * keyboard controller to pulse this line a couple of times.
2056: */
1.273 drochner 2057: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2058: delay(100000);
1.273 drochner 2059: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2060: delay(100000);
1.149 mycroft 2061:
2062: /*
1.224 mycroft 2063: * Try to cause a triple fault and watchdog reset by making the IDT
2064: * invalid and causing a fault.
1.149 mycroft 2065: */
1.313 perry 2066: memset((caddr_t)idt, 0, NIDT * sizeof(idt[0]));
1.484 fvdl 2067: __asm __volatile("divl %0,%1" : : "q" (0), "a" (0));
1.149 mycroft 2068:
1.224 mycroft 2069: #if 0
1.149 mycroft 2070: /*
2071: * Try to cause a triple fault and watchdog reset by unmapping the
1.224 mycroft 2072: * entire address space and doing a TLB flush.
1.149 mycroft 2073: */
1.414 thorpej 2074: memset((caddr_t)PTD, 0, PAGE_SIZE);
1.484 fvdl 2075: tlbflush();
1.224 mycroft 2076: #endif
1.149 mycroft 2077:
2078: for (;;);
1.45 cgd 2079: }
1.484 fvdl 2080:
1.499 jdolecek 2081: void
1.551 junyoung 2082: cpu_getmcontext(struct lwp *l, mcontext_t *mcp, unsigned int *flags)
1.508 thorpej 2083: {
2084: const struct trapframe *tf = l->l_md.md_regs;
2085: __greg_t *gr = mcp->__gregs;
1.519 nathanw 2086: __greg_t ras_eip;
1.508 thorpej 2087:
2088: /* Save register context. */
2089: #ifdef VM86
2090: if (tf->tf_eflags & PSL_VM) {
2091: gr[_REG_GS] = tf->tf_vm86_gs;
2092: gr[_REG_FS] = tf->tf_vm86_fs;
2093: gr[_REG_ES] = tf->tf_vm86_es;
2094: gr[_REG_DS] = tf->tf_vm86_ds;
2095: gr[_REG_EFL] = get_vflags(l);
2096: } else
2097: #endif
2098: {
2099: gr[_REG_GS] = tf->tf_gs;
2100: gr[_REG_FS] = tf->tf_fs;
2101: gr[_REG_ES] = tf->tf_es;
2102: gr[_REG_DS] = tf->tf_ds;
2103: gr[_REG_EFL] = tf->tf_eflags;
2104: }
2105: gr[_REG_EDI] = tf->tf_edi;
2106: gr[_REG_ESI] = tf->tf_esi;
2107: gr[_REG_EBP] = tf->tf_ebp;
2108: gr[_REG_EBX] = tf->tf_ebx;
2109: gr[_REG_EDX] = tf->tf_edx;
2110: gr[_REG_ECX] = tf->tf_ecx;
2111: gr[_REG_EAX] = tf->tf_eax;
2112: gr[_REG_EIP] = tf->tf_eip;
2113: gr[_REG_CS] = tf->tf_cs;
2114: gr[_REG_ESP] = tf->tf_esp;
2115: gr[_REG_UESP] = tf->tf_esp;
2116: gr[_REG_SS] = tf->tf_ss;
2117: gr[_REG_TRAPNO] = tf->tf_trapno;
2118: gr[_REG_ERR] = tf->tf_err;
1.519 nathanw 2119:
2120: if ((ras_eip = (__greg_t)ras_lookup(l->l_proc,
2121: (caddr_t) gr[_REG_EIP])) != -1)
2122: gr[_REG_EIP] = ras_eip;
2123:
1.508 thorpej 2124: *flags |= _UC_CPU;
2125:
2126: /* Save floating point register context, if any. */
1.548 yamt 2127: if ((l->l_md.md_flags & MDL_USEDFPU) != 0) {
1.508 thorpej 2128: #if NNPX > 0
2129: /*
2130: * If this process is the current FP owner, dump its
2131: * context to the PCB first.
2132: * XXX npxsave() also clears the FPU state; depending on the
2133: * XXX application this might be a penalty.
2134: */
2135: if (l->l_addr->u_pcb.pcb_fpcpu) {
2136: npxsave_lwp(l, 1);
2137: }
2138: #endif
2139: if (i386_use_fxsave) {
2140: memcpy(&mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2141: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2142: sizeof (mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm));
2143: *flags |= _UC_FXSAVE;
2144: } else {
2145: memcpy(&mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2146: &l->l_addr->u_pcb.pcb_savefpu.sv_87,
2147: sizeof (mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state));
2148: }
2149: #if 0
2150: /* Apparently nothing ever touches this. */
2151: ucp->mcp.mc_fp.fp_emcsts = l->l_addr->u_pcb.pcb_saveemc;
2152: #endif
2153: *flags |= _UC_FPU;
2154: }
2155: }
2156:
2157: int
1.551 junyoung 2158: cpu_setmcontext(struct lwp *l, const mcontext_t *mcp, unsigned int flags)
1.508 thorpej 2159: {
2160: struct trapframe *tf = l->l_md.md_regs;
2161: __greg_t *gr = mcp->__gregs;
2162:
2163: /* Restore register context, if any. */
2164: if ((flags & _UC_CPU) != 0) {
2165: #ifdef VM86
1.535 drochner 2166: if (gr[_REG_EFL] & PSL_VM) {
1.508 thorpej 2167: tf->tf_vm86_gs = gr[_REG_GS];
2168: tf->tf_vm86_fs = gr[_REG_FS];
2169: tf->tf_vm86_es = gr[_REG_ES];
2170: tf->tf_vm86_ds = gr[_REG_DS];
2171: set_vflags(l, gr[_REG_EFL]);
1.534 christos 2172: if (flags & _UC_VM) {
1.542 junyoung 2173: void syscall_vm86(struct trapframe *);
1.534 christos 2174: l->l_proc->p_md.md_syscall = syscall_vm86;
2175: }
1.508 thorpej 2176: } else
2177: #endif
2178: {
2179: /*
2180: * Check for security violations. If we're returning
2181: * to protected mode, the CPU will validate the segment
2182: * registers automatically and generate a trap on
2183: * violations. We handle the trap, rather than doing
2184: * all of the checking here.
2185: */
1.534 christos 2186: if (((gr[_REG_EFL] ^ tf->tf_eflags) & PSL_USERSTATIC) ||
2187: !USERMODE(gr[_REG_CS], gr[_REG_EFL])) {
2188: printf("cpu_setmcontext error: uc EFL: 0x%08x"
1.535 drochner 2189: " tf EFL: 0x%08x uc CS: 0x%x\n",
1.508 thorpej 2190: gr[_REG_EFL], tf->tf_eflags, gr[_REG_CS]);
2191: return (EINVAL);
2192: }
2193: tf->tf_gs = gr[_REG_GS];
2194: tf->tf_fs = gr[_REG_FS];
2195: tf->tf_es = gr[_REG_ES];
2196: tf->tf_ds = gr[_REG_DS];
2197: /* Only change the user-alterable part of eflags */
2198: tf->tf_eflags &= ~PSL_USER;
2199: tf->tf_eflags |= (gr[_REG_EFL] & PSL_USER);
2200: }
2201: tf->tf_edi = gr[_REG_EDI];
2202: tf->tf_esi = gr[_REG_ESI];
2203: tf->tf_ebp = gr[_REG_EBP];
2204: tf->tf_ebx = gr[_REG_EBX];
2205: tf->tf_edx = gr[_REG_EDX];
2206: tf->tf_ecx = gr[_REG_ECX];
2207: tf->tf_eax = gr[_REG_EAX];
2208: tf->tf_eip = gr[_REG_EIP];
2209: tf->tf_cs = gr[_REG_CS];
2210: tf->tf_esp = gr[_REG_UESP];
2211: tf->tf_ss = gr[_REG_SS];
2212: }
2213:
2214: /* Restore floating point register context, if any. */
2215: if ((flags & _UC_FPU) != 0) {
2216: #if NNPX > 0
2217: /*
2218: * If we were using the FPU, forget that we were.
2219: */
2220: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
2221: npxsave_lwp(l, 0);
2222: #endif
2223: if (flags & _UC_FXSAVE) {
2224: if (i386_use_fxsave) {
2225: memcpy(
2226: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2227: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2228: sizeof (&l->l_addr->u_pcb.pcb_savefpu.sv_xmm));
2229: } else {
2230: /* This is a weird corner case */
2231: process_xmm_to_s87((struct savexmm *)
2232: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2233: &l->l_addr->u_pcb.pcb_savefpu.sv_87);
2234: }
2235: } else {
2236: if (i386_use_fxsave) {
2237: process_s87_to_xmm((struct save87 *)
2238: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2239: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm);
2240: } else {
2241: memcpy(&l->l_addr->u_pcb.pcb_savefpu.sv_87,
2242: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2243: sizeof (l->l_addr->u_pcb.pcb_savefpu.sv_87));
2244: }
2245: }
2246: /* If not set already. */
1.548 yamt 2247: l->l_md.md_flags |= MDL_USEDFPU;
1.508 thorpej 2248: #if 0
2249: /* Apparently unused. */
2250: l->l_addr->u_pcb.pcb_saveemc = mcp->mc_fp.fp_emcsts;
2251: #endif
2252: }
1.534 christos 2253: if (flags & _UC_SETSTACK)
2254: l->l_proc->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
2255: if (flags & _UC_CLRSTACK)
2256: l->l_proc->p_sigctx.ps_sigstk.ss_flags &= ~SS_ONSTACK;
1.508 thorpej 2257: return (0);
2258: }
2259:
2260: void
1.499 jdolecek 2261: cpu_initclocks()
1.484 fvdl 2262: {
2263: (*initclock_func)();
2264: }
2265:
2266: #ifdef MULTIPROCESSOR
1.499 jdolecek 2267: void
2268: need_resched(struct cpu_info *ci)
1.484 fvdl 2269: {
1.546 yamt 2270:
2271: if (ci->ci_want_resched)
2272: return;
2273:
1.484 fvdl 2274: ci->ci_want_resched = 1;
1.508 thorpej 2275: if ((ci)->ci_curlwp != NULL)
2276: aston((ci)->ci_curlwp->l_proc);
1.546 yamt 2277: else if (ci != curcpu())
2278: x86_send_ipi(ci, 0);
1.484 fvdl 2279: }
2280: #endif
2281:
2282: /*
2283: * Allocate an IDT vector slot within the given range.
2284: * XXX needs locking to avoid MP allocation races.
2285: */
2286:
2287: int
1.551 junyoung 2288: idt_vec_alloc(int low, int high)
1.484 fvdl 2289: {
2290: int vec;
2291:
1.498 fvdl 2292: simple_lock(&idt_lock);
2293: for (vec = low; vec <= high; vec++) {
2294: if (idt_allocmap[vec] == 0) {
2295: idt_allocmap[vec] = 1;
2296: simple_unlock(&idt_lock);
1.484 fvdl 2297: return vec;
1.498 fvdl 2298: }
2299: }
2300: simple_unlock(&idt_lock);
1.484 fvdl 2301: return 0;
2302: }
2303:
1.498 fvdl 2304: void
1.551 junyoung 2305: idt_vec_set(int vec, void (*function)(void))
1.484 fvdl 2306: {
1.498 fvdl 2307: /*
2308: * Vector should be allocated, so no locking needed.
2309: */
2310: KASSERT(idt_allocmap[vec] == 1);
1.516 fvdl 2311: setgate(&idt[vec], function, 0, SDT_SYS386IGT, SEL_KPL,
1.489 fvdl 2312: GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 2313: }
2314:
2315: void
1.551 junyoung 2316: idt_vec_free(int vec)
1.484 fvdl 2317: {
1.498 fvdl 2318: simple_lock(&idt_lock);
1.516 fvdl 2319: unsetgate(&idt[vec]);
1.498 fvdl 2320: idt_allocmap[vec] = 0;
2321: simple_unlock(&idt_lock);
1.507 jdolecek 2322: }
2323:
2324: /*
2325: * Number of processes is limited by number of available GDT slots.
2326: */
2327: int
2328: cpu_maxproc(void)
2329: {
2330: #ifdef USER_LDT
2331: return ((MAXGDTSIZ - NGDT) / 2);
2332: #else
2333: return (MAXGDTSIZ - NGDT);
2334: #endif
1.484 fvdl 2335: }
CVSweb <webmaster@jp.NetBSD.org>