Annotation of src/sys/arch/i386/i386/machdep.c, Revision 1.586.2.5
1.586.2.5! liamjfoy 1: /* $NetBSD: machdep.c,v 1.586.2.4 2007/04/20 20:31:25 bouyer 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>
1.586.2.5! liamjfoy 75: __KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.586.2.4 2007/04/20 20:31:25 bouyer 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:
1.586.2.4 bouyer 135: #include <x86/cpu_msr.h>
136:
1.59 mycroft 137: #include <machine/cpu.h>
138: #include <machine/cpufunc.h>
1.484 fvdl 139: #include <machine/cpuvar.h>
1.178 mycroft 140: #include <machine/gdt.h>
1.574 kardel 141: #include <machine/kcore.h>
1.149 mycroft 142: #include <machine/pio.h>
1.59 mycroft 143: #include <machine/psl.h>
144: #include <machine/reg.h>
145: #include <machine/specialreg.h>
1.255 drochner 146: #include <machine/bootinfo.h>
1.455 fvdl 147: #include <machine/mtrr.h>
1.574 kardel 148: #include <x86/x86/tsc.h>
149:
1.570 jmmv 150: #include <machine/multiboot.h>
1.43 brezak 151:
1.146 cgd 152: #include <dev/isa/isareg.h>
1.372 drochner 153: #include <machine/isa_machdep.h>
1.164 cgd 154: #include <dev/ic/i8042reg.h>
1.43 brezak 155:
1.200 christos 156: #ifdef DDB
157: #include <machine/db_machdep.h>
158: #include <ddb/db_extern.h>
159: #endif
160:
1.184 mycroft 161: #ifdef VM86
162: #include <machine/vm86.h>
163: #endif
164:
1.473 tshiozak 165: #include "acpi.h"
1.576 christos 166: #include "apmbios.h"
1.258 jtk 167: #include "bioscall.h"
1.207 jtk 168:
1.259 jtk 169: #if NBIOSCALL > 0
170: #include <machine/bioscall.h>
171: #endif
172:
1.473 tshiozak 173: #if NACPI > 0
174: #include <dev/acpi/acpivar.h>
175: #define ACPI_MACHDEP_PRIVATE
176: #include <machine/acpi_machdep.h>
177: #endif
178:
1.576 christos 179: #if NAPMBIOS > 0
1.207 jtk 180: #include <machine/apmvar.h>
1.258 jtk 181: #endif
182:
1.59 mycroft 183: #include "isa.h"
1.231 thorpej 184: #include "isadma.h"
1.59 mycroft 185: #include "npx.h"
1.522 ragge 186: #include "ksyms.h"
1.2 cgd 187:
1.586.2.3 jdc 188: #include "cardbus.h"
189: #if NCARDBUS > 0
190: /* For rbus_min_start hint. */
191: #include <machine/bus.h>
192: #include <dev/cardbus/rbus.h>
193: #include <machine/rbus_machdep.h>
194: #endif
195:
1.384 jdolecek 196: #include "mca.h"
197: #if NMCA > 0
198: #include <machine/mca_machdep.h> /* for mca_busprobe() */
199: #endif
200:
1.484 fvdl 201: #ifdef MULTIPROCESSOR /* XXX */
202: #include <machine/mpbiosvar.h> /* XXX */
203: #endif /* XXX */
204:
1.517 jmmv 205: #ifndef BEEP_ONHALT_COUNT
206: #define BEEP_ONHALT_COUNT 3
207: #endif
208: #ifndef BEEP_ONHALT_PITCH
209: #define BEEP_ONHALT_PITCH 1500
210: #endif
211: #ifndef BEEP_ONHALT_PERIOD
212: #define BEEP_ONHALT_PERIOD 250
213: #endif
214:
1.104 cgd 215: /* the following is used externally (sysctl_hw) */
1.549 wiz 216: char machine[] = "i386"; /* CPU "architecture" */
1.232 veego 217: char machine_arch[] = "i386"; /* machine == machine_arch */
1.104 cgd 218:
1.559 thorpej 219: extern struct bi_devmatch *x86_alldisks;
220: extern int x86_ndisks;
1.342 fvdl 221:
1.328 bouyer 222: #ifdef CPURESET_DELAY
223: int cpureset_delay = CPURESET_DELAY;
224: #else
225: int cpureset_delay = 2000; /* default to 2s */
226: #endif
227:
1.455 fvdl 228: #ifdef MTRR
1.513 fvdl 229: struct mtrr_funcs *mtrr_funcs;
1.455 fvdl 230: #endif
231:
1.515 fvdl 232: #ifdef COMPAT_NOMID
1.566 christos 233: static int exec_nomid(struct lwp *, struct exec_package *);
1.550 junyoung 234: #endif
1.515 fvdl 235:
1.59 mycroft 236: int physmem;
1.163 cgd 237: int dumpmem_low;
238: int dumpmem_high;
1.524 drochner 239: unsigned int cpu_feature;
1.553 lukem 240: unsigned int cpu_feature2;
1.59 mycroft 241: int cpu_class;
1.428 fvdl 242: int i386_fpu_present;
243: int i386_fpu_exception;
244: int i386_fpu_fdivbug;
1.59 mycroft 245:
1.451 thorpej 246: int i386_use_fxsave;
247: int i386_has_sse;
248: int i386_has_sse2;
1.450 thorpej 249:
1.461 christos 250: int tmx86_has_longrun;
251:
1.314 thorpej 252: vaddr_t msgbuf_vaddr;
1.586.2.1 tron 253: struct {
254: paddr_t paddr;
255: psize_t sz;
256: } msgbuf_p_seg[VM_PHYSSEG_MAX];
257: unsigned int msgbuf_p_cnt = 0;
1.314 thorpej 258:
259: vaddr_t idt_vaddr;
260: paddr_t idt_paddr;
261:
1.264 mycroft 262: #ifdef I586_CPU
1.314 thorpej 263: vaddr_t pentium_idt_vaddr;
1.264 mycroft 264: #endif
1.59 mycroft 265:
1.444 chs 266: struct vm_map *exec_map = NULL;
267: struct vm_map *mb_map = NULL;
268: struct vm_map *phys_map = NULL;
1.48 brezak 269:
1.314 thorpej 270: extern paddr_t avail_start, avail_end;
1.1 cgd 271:
1.542 junyoung 272: void (*delay_func)(int) = i8254_delay;
273: void (*initclock_func)(void) = i8254_initclocks;
1.484 fvdl 274:
1.204 thorpej 275: /*
1.291 thorpej 276: * Size of memory segments, before any memory is stolen.
277: */
278: phys_ram_seg_t mem_clusters[VM_PHYSSEG_MAX];
279: int mem_cluster_cnt;
280:
1.542 junyoung 281: int cpu_dump(void);
282: int cpu_dumpsize(void);
283: u_long cpu_dump_mempagecnt(void);
284: void dumpsys(void);
285: void init386(paddr_t);
286: void initgdt(union descriptor *);
1.255 drochner 287:
1.568 perry 288: void add_mem_cluster(uint64_t, uint64_t, uint32_t);
1.433 kanaoka 289:
1.521 dsl 290: extern int time_adjusted;
1.484 fvdl 291:
1.571 jmmv 292: struct bootinfo bootinfo;
293: int *esym;
294: extern int boothowto;
295:
296: /* Base memory reported by BIOS. */
297: #ifndef REALBASEMEM
298: int biosbasemem = 0;
299: #else
300: int biosbasemem = REALBASEMEM;
301: #endif
302:
303: /* Extended memory reported by BIOS. */
304: #ifndef REALEXTMEM
305: int biosextmem = 0;
306: #else
307: int biosextmem = REALEXTMEM;
308: #endif
309:
1.577 mrg 310: /* Set if any boot-loader set biosbasemem/biosextmem. */
311: int biosmem_implicit;
312:
1.571 jmmv 313: /* Representation of the bootinfo structure constructed by a NetBSD native
314: * boot loader. Only be used by native_loader(). */
315: struct bootinfo_source {
316: uint32_t bs_naddrs;
317: paddr_t bs_addrs[1]; /* Actually longer. */
318: };
319:
320: /* Only called by locore.h; no need to be in a header file. */
321: void native_loader(int, int, struct bootinfo_source *, paddr_t, int, int);
322:
323: /*
324: * Called as one of the very first things during system startup (just after
325: * the boot loader gave control to the kernel image), this routine is in
326: * charge of retrieving the parameters passed in by the boot loader and
327: * storing them in the appropriate kernel variables.
328: *
329: * WARNING: Because the kernel has not yet relocated itself to KERNBASE,
330: * special care has to be taken when accessing memory because absolute
331: * addresses (referring to kernel symbols) do not work. So:
332: *
333: * 1) Avoid jumps to absolute addresses (such as gotos and switches).
334: * 2) To access global variables use their physical address, which
335: * can be obtained using the RELOC macro.
336: */
337: void
1.586 christos 338: native_loader(int bl_boothowto, int bl_bootdev,
1.571 jmmv 339: struct bootinfo_source *bl_bootinfo, paddr_t bl_esym,
1.572 jmmv 340: int bl_biosextmem, int bl_biosbasemem)
1.571 jmmv 341: {
342: #define RELOC(type, x) ((type)((vaddr_t)(x) - KERNBASE))
343:
344: *RELOC(int *, &boothowto) = bl_boothowto;
345:
346: #ifdef COMPAT_OLDBOOT
347: /*
348: * Pre-1.3 boot loaders gave the boot device as a parameter
349: * (instead of a bootinfo entry).
350: */
351: *RELOC(int *, &bootdev) = bl_bootdev;
352: #endif
353:
354: /*
355: * The boot loader provides a physical, non-relocated address
356: * for the symbols table's end. We need to convert it to a
357: * virtual address.
358: */
359: if (bl_esym != 0)
360: *RELOC(int **, &esym) = (int *)((vaddr_t)bl_esym + KERNBASE);
361: else
362: *RELOC(int **, &esym) = 0;
363:
364: /*
365: * Copy bootinfo entries (if any) from the boot loader's
366: * representation to the kernel's bootinfo space.
367: */
368: if (bl_bootinfo != NULL) {
369: size_t i;
370: uint8_t *data;
371: struct bootinfo *bidest;
372:
373: bidest = RELOC(struct bootinfo *, &bootinfo);
374:
375: data = &bidest->bi_data[0];
376:
377: for (i = 0; i < bl_bootinfo->bs_naddrs; i++) {
378: struct btinfo_common *bc;
379:
380: bc = (struct btinfo_common *)(bl_bootinfo->bs_addrs[i]);
381:
382: if ((paddr_t)(data + bc->len) >
383: (paddr_t)(&bidest->bi_data[0] + BOOTINFO_MAXSIZE))
384: break;
385:
386: memcpy(data, bc, bc->len);
387: data += bc->len;
388: }
389: bidest->bi_nentries = i;
390: }
391:
392: /*
393: * Configure biosbasemem and biosextmem only if they were not
394: * explicitly given during the kernel's build.
395: */
1.577 mrg 396: if (*RELOC(int *, &biosbasemem) == 0) {
1.571 jmmv 397: *RELOC(int *, &biosbasemem) = bl_biosbasemem;
1.577 mrg 398: *RELOC(int *, &biosmem_implicit) = 1;
399: }
400: if (*RELOC(int *, &biosextmem) == 0) {
1.571 jmmv 401: *RELOC(int *, &biosextmem) = bl_biosextmem;
1.577 mrg 402: *RELOC(int *, &biosmem_implicit) = 1;
403: }
1.571 jmmv 404: #undef RELOC
405: }
406:
1.59 mycroft 407: /*
408: * Machine-dependent startup code
409: */
1.32 andrew 410: void
1.1 cgd 411: cpu_startup()
412: {
1.586.2.1 tron 413: int x, y;
1.314 thorpej 414: vaddr_t minaddr, maxaddr;
1.586.2.1 tron 415: psize_t sz;
1.354 lukem 416: char pbuf[9];
1.1 cgd 417:
1.284 mrg 418: /*
419: * Initialize error message buffer (et end of core).
420: */
1.586.2.1 tron 421: if (msgbuf_p_cnt == 0)
422: panic("msgbuf paddr map has not been set up");
423: for (x = 0, sz = 0; x < msgbuf_p_cnt; sz += msgbuf_p_seg[x++].sz)
424: continue;
425: msgbuf_vaddr = uvm_km_alloc(kernel_map, sz, 0, UVM_KMF_VAONLY);
1.385 thorpej 426: if (msgbuf_vaddr == 0)
1.284 mrg 427: panic("failed to valloc msgbuf_vaddr");
1.359 thorpej 428:
1.284 mrg 429: /* msgbuf_paddr was init'd in pmap */
1.586.2.1 tron 430: for (y = 0, sz = 0; y < msgbuf_p_cnt; y++) {
431: for (x = 0; x < btoc(msgbuf_p_seg[y].sz); x++, sz += PAGE_SIZE)
432: pmap_kenter_pa((vaddr_t)msgbuf_vaddr + sz,
433: msgbuf_p_seg[y].paddr + x * PAGE_SIZE, VM_PROT_READ|VM_PROT_WRITE);
434: }
1.456 chris 435: pmap_update(pmap_kernel());
1.359 thorpej 436:
1.586.2.1 tron 437: initmsgbuf((caddr_t)msgbuf_vaddr, sz);
1.284 mrg 438:
1.562 lukem 439: printf("%s%s", copyright, version);
1.484 fvdl 440:
1.570 jmmv 441: #ifdef MULTIBOOT
442: multiboot_print_info();
443: #endif
444:
1.484 fvdl 445: #ifdef TRAPLOG
446: /*
447: * Enable recording of branch from/to in MSR's
448: */
449: wrmsr(MSR_DEBUGCTLMSR, 0x1);
450: #endif
451:
1.382 mycroft 452: format_bytes(pbuf, sizeof(pbuf), ptoa(physmem));
1.354 lukem 453: printf("total memory = %s\n", pbuf);
1.1 cgd 454:
1.586.2.3 jdc 455: #if NCARDBUS > 0
456: /* Tell RBUS how much RAM we have, so it can use heuristics. */
457: rbus_min_start_hint(ptoa(physmem));
458: #endif
459:
1.547 pk 460: minaddr = 0;
1.41 cgd 461:
1.1 cgd 462: /*
1.36 cgd 463: * Allocate a submap for exec arguments. This map effectively
464: * limits the number of processes exec'ing at any time.
1.1 cgd 465: */
1.284 mrg 466: exec_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 467: 16*NCARGS, VM_MAP_PAGEABLE, FALSE, NULL);
1.59 mycroft 468:
1.1 cgd 469: /*
470: * Allocate a submap for physio
471: */
1.284 mrg 472: phys_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 473: VM_PHYS_SIZE, 0, FALSE, NULL);
1.1 cgd 474:
475: /*
1.229 thorpej 476: * Finally, allocate mbuf cluster submap.
1.1 cgd 477: */
1.334 thorpej 478: mb_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
1.358 thorpej 479: nmbclusters * mclbytes, VM_MAP_INTRSAFE, FALSE, NULL);
1.1 cgd 480:
1.547 pk 481: format_bytes(pbuf, sizeof(pbuf), ptoa(uvmexp.free));
1.354 lukem 482: printf("avail memory = %s\n", pbuf);
1.1 cgd 483:
1.375 drochner 484: /* Safe for i/o port / memory space allocation to use malloc now. */
1.514 fvdl 485: x86_bus_space_mallocok();
1.586.2.4 bouyer 486:
487: x86_init();
1.349 thorpej 488: }
489:
490: /*
491: * Set up proc0's TSS and LDT.
492: */
493: void
494: i386_proc0_tss_ldt_init()
495: {
496: struct pcb *pcb;
497: int x;
1.268 thorpej 498:
1.326 thorpej 499: gdt_init();
1.484 fvdl 500:
1.508 thorpej 501: cpu_info_primary.ci_curpcb = pcb = &lwp0.l_addr->u_pcb;
1.484 fvdl 502:
1.326 thorpej 503: pcb->pcb_tss.tss_ioopt =
504: ((caddr_t)pcb->pcb_iomap - (caddr_t)&pcb->pcb_tss) << 16;
1.484 fvdl 505:
1.326 thorpej 506: for (x = 0; x < sizeof(pcb->pcb_iomap) / 4; x++)
507: pcb->pcb_iomap[x] = 0xffffffff;
508:
1.394 thorpej 509: pcb->pcb_ldt_sel = pmap_kernel()->pm_ldt_sel = GSEL(GLDT_SEL, SEL_KPL);
1.326 thorpej 510: pcb->pcb_cr0 = rcr0();
511: pcb->pcb_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
1.573 yamt 512: pcb->pcb_tss.tss_esp0 = USER_TO_UAREA(lwp0.l_addr) + KSTACK_SIZE - 16;
1.508 thorpej 513: lwp0.l_md.md_regs = (struct trapframe *)pcb->pcb_tss.tss_esp0 - 1;
514: lwp0.l_md.md_tss_sel = tss_alloc(pcb);
1.326 thorpej 515:
1.508 thorpej 516: ltr(lwp0.l_md.md_tss_sel);
1.326 thorpej 517: lldt(pcb->pcb_ldt_sel);
1.484 fvdl 518: }
519:
520: /*
521: * Set up TSS and LDT for a new PCB.
522: */
523:
524: void
1.551 junyoung 525: i386_init_pcb_tss_ldt(struct cpu_info *ci)
1.484 fvdl 526: {
527: int x;
528: struct pcb *pcb = ci->ci_idle_pcb;
1.326 thorpej 529:
1.484 fvdl 530: pcb->pcb_tss.tss_ioopt =
531: ((caddr_t)pcb->pcb_iomap - (caddr_t)&pcb->pcb_tss) << 16;
532: for (x = 0; x < sizeof(pcb->pcb_iomap) / 4; x++)
533: pcb->pcb_iomap[x] = 0xffffffff;
534:
535: pcb->pcb_ldt_sel = pmap_kernel()->pm_ldt_sel = GSEL(GLDT_SEL, SEL_KPL);
536: pcb->pcb_cr0 = rcr0();
537:
538: ci->ci_idle_tss_sel = tss_alloc(pcb);
1.326 thorpej 539: }
540:
541: /*
1.545 atatat 542: * sysctl helper routine for machdep.tm* nodes.
1.484 fvdl 543: */
1.545 atatat 544: static int
545: sysctl_machdep_tm_longrun(SYSCTLFN_ARGS)
1.104 cgd 546: {
1.545 atatat 547: struct sysctlnode node;
548: int io, error;
549:
550: if (!tmx86_has_longrun)
551: return (EOPNOTSUPP);
552:
553: node = *rnode;
554: node.sysctl_data = &io;
1.104 cgd 555:
1.545 atatat 556: switch (rnode->sysctl_num) {
1.461 christos 557: case CPU_TMLR_MODE:
1.545 atatat 558: io = (int)(crusoe_longrun = tmx86_get_longrun_mode());
559: break;
1.461 christos 560: case CPU_TMLR_FREQUENCY:
561: tmx86_get_longrun_status_all();
1.545 atatat 562: io = crusoe_frequency;
563: break;
1.461 christos 564: case CPU_TMLR_VOLTAGE:
565: tmx86_get_longrun_status_all();
1.545 atatat 566: io = crusoe_voltage;
567: break;
1.461 christos 568: case CPU_TMLR_PERCENTAGE:
569: tmx86_get_longrun_status_all();
1.545 atatat 570: io = crusoe_percentage;
571: break;
1.104 cgd 572: default:
573: return (EOPNOTSUPP);
574: }
1.545 atatat 575:
576: error = sysctl_lookup(SYSCTLFN_CALL(&node));
577: if (error || newp == NULL)
578: return (error);
579:
580: if (rnode->sysctl_num == CPU_TMLR_MODE) {
581: if (tmx86_set_longrun_mode(io))
582: crusoe_longrun = (u_int)io;
583: else
584: return (EINVAL);
585: }
586:
587: return (0);
588: }
589:
590: /*
591: * sysctl helper routine for machdep.booted_kernel
592: */
593: static int
594: sysctl_machdep_booted_kernel(SYSCTLFN_ARGS)
595: {
596: struct btinfo_bootpath *bibp;
597: struct sysctlnode node;
598:
599: bibp = lookup_bootinfo(BTINFO_BOOTPATH);
600: if(!bibp)
601: return(ENOENT); /* ??? */
602:
603: node = *rnode;
604: node.sysctl_data = bibp->bootpath;
605: node.sysctl_size = sizeof(bibp->bootpath);
606: return (sysctl_lookup(SYSCTLFN_CALL(&node)));
607: }
608:
609: /*
610: * sysctl helper routine for machdep.diskinfo
611: */
612: static int
613: sysctl_machdep_diskinfo(SYSCTLFN_ARGS)
614: {
615: struct sysctlnode node;
616:
617: node = *rnode;
1.559 thorpej 618: if (x86_alldisks == NULL)
1.557 dbj 619: return(EOPNOTSUPP);
1.559 thorpej 620: node.sysctl_data = x86_alldisks;
1.545 atatat 621: node.sysctl_size = sizeof(struct disklist) +
1.559 thorpej 622: (x86_ndisks - 1) * sizeof(struct nativedisk_info);
1.557 dbj 623: return (sysctl_lookup(SYSCTLFN_CALL(&node)));
1.545 atatat 624: }
625:
626: /*
627: * machine dependent system variables.
628: */
629: SYSCTL_SETUP(sysctl_machdep_setup, "sysctl machdep subtree setup")
630: {
631:
1.552 atatat 632: sysctl_createv(clog, 0, NULL, NULL,
633: CTLFLAG_PERMANENT,
1.545 atatat 634: CTLTYPE_NODE, "machdep", NULL,
635: NULL, 0, NULL, 0,
636: CTL_MACHDEP, CTL_EOL);
637:
1.552 atatat 638: sysctl_createv(clog, 0, NULL, NULL,
639: CTLFLAG_PERMANENT,
1.545 atatat 640: CTLTYPE_STRUCT, "console_device", NULL,
641: sysctl_consdev, 0, NULL, sizeof(dev_t),
642: CTL_MACHDEP, CPU_CONSDEV, CTL_EOL);
1.552 atatat 643: sysctl_createv(clog, 0, NULL, NULL,
644: CTLFLAG_PERMANENT,
1.545 atatat 645: CTLTYPE_INT, "biosbasemem", NULL,
646: NULL, 0, &biosbasemem, 0,
647: CTL_MACHDEP, CPU_BIOSBASEMEM, CTL_EOL);
1.552 atatat 648: sysctl_createv(clog, 0, NULL, NULL,
649: CTLFLAG_PERMANENT,
1.545 atatat 650: CTLTYPE_INT, "biosextmem", NULL,
651: NULL, 0, &biosextmem, 0,
652: CTL_MACHDEP, CPU_BIOSEXTMEM, CTL_EOL);
1.552 atatat 653: sysctl_createv(clog, 0, NULL, NULL,
654: CTLFLAG_PERMANENT,
1.545 atatat 655: CTLTYPE_INT, "nkpde", NULL,
656: NULL, 0, &nkpde, 0,
657: CTL_MACHDEP, CPU_NKPDE, CTL_EOL);
1.552 atatat 658: sysctl_createv(clog, 0, NULL, NULL,
659: CTLFLAG_PERMANENT,
1.545 atatat 660: CTLTYPE_STRING, "booted_kernel", NULL,
661: sysctl_machdep_booted_kernel, 0, NULL, 0,
662: CTL_MACHDEP, CPU_BOOTED_KERNEL, CTL_EOL);
1.552 atatat 663: sysctl_createv(clog, 0, NULL, NULL,
664: CTLFLAG_PERMANENT,
1.545 atatat 665: CTLTYPE_STRUCT, "diskinfo", NULL,
666: sysctl_machdep_diskinfo, 0, NULL, 0,
667: CTL_MACHDEP, CPU_DISKINFO, CTL_EOL);
1.552 atatat 668: sysctl_createv(clog, 0, NULL, NULL,
669: CTLFLAG_PERMANENT,
1.545 atatat 670: CTLTYPE_INT, "fpu_present", NULL,
671: NULL, 0, &i386_fpu_present, 0,
672: CTL_MACHDEP, CPU_FPU_PRESENT, CTL_EOL);
1.552 atatat 673: sysctl_createv(clog, 0, NULL, NULL,
674: CTLFLAG_PERMANENT,
1.545 atatat 675: CTLTYPE_INT, "osfxsr", NULL,
676: NULL, 0, &i386_use_fxsave, 0,
677: CTL_MACHDEP, CPU_OSFXSR, CTL_EOL);
1.552 atatat 678: sysctl_createv(clog, 0, NULL, NULL,
679: CTLFLAG_PERMANENT,
1.545 atatat 680: CTLTYPE_INT, "sse", NULL,
681: NULL, 0, &i386_has_sse, 0,
682: CTL_MACHDEP, CPU_SSE, CTL_EOL);
1.552 atatat 683: sysctl_createv(clog, 0, NULL, NULL,
684: CTLFLAG_PERMANENT,
1.545 atatat 685: CTLTYPE_INT, "sse2", NULL,
686: NULL, 0, &i386_has_sse2, 0,
687: CTL_MACHDEP, CPU_SSE2, CTL_EOL);
1.554 lukem 688: sysctl_createv(clog, 0, NULL, NULL,
689: CTLFLAG_PERMANENT,
690: CTLTYPE_STRING, "cpu_brand", NULL,
691: NULL, 0, &cpu_brand_string, 0,
692: CTL_MACHDEP, CTL_CREATE, CTL_EOL);
1.552 atatat 693: sysctl_createv(clog, 0, NULL, NULL,
694: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.545 atatat 695: CTLTYPE_INT, "tm_longrun_mode", NULL,
696: sysctl_machdep_tm_longrun, 0, NULL, 0,
697: CTL_MACHDEP, CPU_TMLR_MODE, CTL_EOL);
1.552 atatat 698: sysctl_createv(clog, 0, NULL, NULL,
699: CTLFLAG_PERMANENT,
1.545 atatat 700: CTLTYPE_INT, "tm_longrun_frequency", NULL,
701: sysctl_machdep_tm_longrun, 0, NULL, 0,
702: CTL_MACHDEP, CPU_TMLR_FREQUENCY, CTL_EOL);
1.552 atatat 703: sysctl_createv(clog, 0, NULL, NULL,
704: CTLFLAG_PERMANENT,
1.545 atatat 705: CTLTYPE_INT, "tm_longrun_voltage", NULL,
706: sysctl_machdep_tm_longrun, 0, NULL, 0,
707: CTL_MACHDEP, CPU_TMLR_VOLTAGE, CTL_EOL);
1.552 atatat 708: sysctl_createv(clog, 0, NULL, NULL,
709: CTLFLAG_PERMANENT,
1.545 atatat 710: CTLTYPE_INT, "tm_longrun_percentage", NULL,
711: sysctl_machdep_tm_longrun, 0, NULL, 0,
712: CTL_MACHDEP, CPU_TMLR_PERCENTAGE, CTL_EOL);
1.104 cgd 713: }
1.151 christos 714:
1.534 christos 715: void *
716: getframe(struct lwp *l, int sig, int *onstack)
717: {
718: struct proc *p = l->l_proc;
719: struct sigctx *ctx = &p->p_sigctx;
720: struct trapframe *tf = l->l_md.md_regs;
721:
722: /* Do we need to jump onto the signal stack? */
723: *onstack = (ctx->ps_sigstk.ss_flags & (SS_DISABLE | SS_ONSTACK)) == 0
724: && (SIGACTION(p, sig).sa_flags & SA_ONSTACK) != 0;
725: if (*onstack)
726: return (char *)ctx->ps_sigstk.ss_sp + ctx->ps_sigstk.ss_size;
727: #ifdef VM86
728: if (tf->tf_eflags & PSL_VM)
729: return (void *)(tf->tf_esp + (tf->tf_ss << 4));
730: else
731: #endif
732: return (void *)tf->tf_esp;
733: }
734:
735: /*
736: * Build context to run handler in. We invoke the handler
737: * directly, only returning via the trampoline. Note the
738: * trampoline version numbers are coordinated with machine-
739: * dependent code in libc.
740: */
741: void
742: buildcontext(struct lwp *l, int sel, void *catcher, void *fp)
743: {
744: struct trapframe *tf = l->l_md.md_regs;
745:
746: tf->tf_gs = GSEL(GUDATA_SEL, SEL_UPL);
747: tf->tf_fs = GSEL(GUDATA_SEL, SEL_UPL);
748: tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
749: tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
750: tf->tf_eip = (int)catcher;
751: tf->tf_cs = GSEL(sel, SEL_UPL);
752: tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
753: tf->tf_esp = (int)fp;
754: tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
755: }
756:
757: static void
1.539 christos 758: sendsig_siginfo(const ksiginfo_t *ksi, const sigset_t *mask)
1.534 christos 759: {
760: struct lwp *l = curlwp;
761: struct proc *p = l->l_proc;
762: struct pmap *pmap = vm_map_pmap(&p->p_vmspace->vm_map);
763: int sel = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
764: GUCODEBIG_SEL : GUCODE_SEL;
765: struct sigacts *ps = p->p_sigacts;
766: int onstack;
767: int sig = ksi->ksi_signo;
768: struct sigframe_siginfo *fp = getframe(l, sig, &onstack), frame;
769: sig_t catcher = SIGACTION(p, sig).sa_handler;
770: struct trapframe *tf = l->l_md.md_regs;
771:
772: fp--;
773:
774: /* Build stack frame for signal trampoline. */
775: switch (ps->sa_sigdesc[sig].sd_vers) {
1.537 christos 776: case 0: /* handled by sendsig_sigcontext */
777: case 1: /* handled by sendsig_sigcontext */
1.534 christos 778: default: /* unknown version */
779: printf("nsendsig: bad version %d\n",
780: ps->sa_sigdesc[sig].sd_vers);
781: sigexit(l, SIGILL);
782: case 2:
783: break;
784: }
785:
786: frame.sf_ra = (int)ps->sa_sigdesc[sig].sd_tramp;
787: frame.sf_signum = sig;
788: frame.sf_sip = &fp->sf_si;
789: frame.sf_ucp = &fp->sf_uc;
1.540 thorpej 790: frame.sf_si._info = ksi->ksi_info;
1.534 christos 791: frame.sf_uc.uc_flags = _UC_SIGMASK|_UC_VM;
792: frame.sf_uc.uc_sigmask = *mask;
793: frame.sf_uc.uc_link = NULL;
794: frame.sf_uc.uc_flags |= (p->p_sigctx.ps_sigstk.ss_flags & SS_ONSTACK)
795: ? _UC_SETSTACK : _UC_CLRSTACK;
796: memset(&frame.sf_uc.uc_stack, 0, sizeof(frame.sf_uc.uc_stack));
797: cpu_getmcontext(l, &frame.sf_uc.uc_mcontext, &frame.sf_uc.uc_flags);
798:
799: if (tf->tf_eflags & PSL_VM)
800: (*p->p_emul->e_syscall_intern)(p);
801:
802: if (copyout(&frame, fp, sizeof(frame)) != 0) {
803: /*
804: * Process has trashed its stack; give it an illegal
805: * instruction to halt it in its tracks.
806: */
807: sigexit(l, SIGILL);
808: /* NOTREACHED */
809: }
810:
811: buildcontext(l, sel, catcher, fp);
1.319 mycroft 812:
813: /* Remember that we're now on the signal stack. */
814: if (onstack)
1.425 jdolecek 815: p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
1.1 cgd 816: }
817:
1.534 christos 818: void
1.539 christos 819: sendsig(const ksiginfo_t *ksi, const sigset_t *mask)
1.534 christos 820: {
821: #ifdef COMPAT_16
822: if (curproc->p_sigacts->sa_sigdesc[ksi->ksi_signo].sd_vers < 2)
823: sendsig_sigcontext(ksi, mask);
824: else
825: #endif
826: sendsig_siginfo(ksi, mask);
827: }
1.508 thorpej 828:
1.550 junyoung 829: void
1.534 christos 830: cpu_upcall(struct lwp *l, int type, int nevents, int ninterrupted, void *sas,
831: void *ap, void *sp, sa_upcall_t upcall)
1.508 thorpej 832: {
1.533 chs 833: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 834: struct saframe *sf, frame;
835: struct trapframe *tf;
836:
837: tf = l->l_md.md_regs;
838:
839: /* Finally, copy out the rest of the frame. */
840: frame.sa_type = type;
841: frame.sa_sas = sas;
842: frame.sa_events = nevents;
843: frame.sa_interrupted = ninterrupted;
844: frame.sa_arg = ap;
845: frame.sa_ra = 0;
1.550 junyoung 846:
1.508 thorpej 847: sf = (struct saframe *)sp - 1;
848: if (copyout(&frame, sf, sizeof(frame)) != 0) {
849: /* Copying onto the stack didn't work. Die. */
850: sigexit(l, SIGILL);
851: /* NOTREACHED */
852: }
853:
854: tf->tf_eip = (int) upcall;
855: tf->tf_esp = (int) sf;
856: tf->tf_ebp = 0; /* indicate call-frame-top to debuggers */
857: tf->tf_gs = GSEL(GUDATA_SEL, SEL_UPL);
858: tf->tf_fs = GSEL(GUDATA_SEL, SEL_UPL);
859: tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
860: tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
1.533 chs 861: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
862: GSEL(GUCODEBIG_SEL, SEL_UPL) : GSEL(GUCODE_SEL, SEL_UPL);
1.508 thorpej 863: tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
864: tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
865: }
1.37 cgd 866:
1.1 cgd 867: int waittime = -1;
868: struct pcb dumppcb;
869:
1.32 andrew 870: void
1.586 christos 871: cpu_reboot(int howto, char *bootstr)
1.1 cgd 872: {
873:
1.106 mycroft 874: if (cold) {
1.193 mycroft 875: howto |= RB_HALT;
876: goto haltsys;
1.1 cgd 877: }
1.193 mycroft 878:
1.106 mycroft 879: boothowto = howto;
1.193 mycroft 880: if ((howto & RB_NOSYNC) == 0 && waittime < 0) {
1.1 cgd 881: waittime = 0;
1.150 mycroft 882: vfs_shutdown();
1.521 dsl 883: /*
884: * If we've been adjusting the clock, the todr
885: * will be out of synch; adjust it now.
886: */
887: if (time_adjusted != 0)
888: resettodr();
1.1 cgd 889: }
1.193 mycroft 890:
891: /* Disable interrupts. */
1.1 cgd 892: splhigh();
1.193 mycroft 893:
894: /* Do a dump if requested. */
895: if ((howto & (RB_DUMP | RB_HALT)) == RB_DUMP)
896: dumpsys();
897:
898: haltsys:
899: doshutdownhooks();
900:
1.484 fvdl 901: #ifdef MULTIPROCESSOR
1.514 fvdl 902: x86_broadcast_ipi(X86_IPI_HALT);
1.484 fvdl 903: #endif
904:
1.307 thorpej 905: if ((howto & RB_POWERDOWN) == RB_POWERDOWN) {
1.473 tshiozak 906: #if NACPI > 0
1.530 kochi 907: if (acpi_softc != NULL) {
908: delay(500000);
909: acpi_enter_sleep_state(acpi_softc, ACPI_STATE_S5);
910: printf("WARNING: ACPI powerdown failed!\n");
911: }
1.473 tshiozak 912: #endif
1.576 christos 913: #if NAPMBIOS > 0 && !defined(APM_NO_POWEROFF)
1.208 jtk 914: /* turn off, if we can. But try to turn disk off and
915: * wait a bit first--some disk drives are slow to clean up
916: * and users have reported disk corruption.
917: */
918: delay(500000);
1.578 xtraeme 919: apm_set_powstate(NULL, APM_DEV_DISK(APM_DEV_ALLUNITS), APM_SYS_OFF);
1.208 jtk 920: delay(500000);
1.576 christos 921: apm_set_powstate(NULL, APM_DEV_ALLDEVS, APM_SYS_OFF);
1.531 kochi 922: printf("WARNING: APM powerdown failed!\n");
1.307 thorpej 923: /*
924: * RB_POWERDOWN implies RB_HALT... fall into it...
925: */
1.208 jtk 926: #endif
1.307 thorpej 927: }
928:
929: if (howto & RB_HALT) {
1.210 christos 930: printf("\n");
931: printf("The operating system has halted.\n");
932: printf("Please press any key to reboot.\n\n");
1.517 jmmv 933:
934: #ifdef BEEP_ONHALT
935: {
936: int c;
937: for (c = BEEP_ONHALT_COUNT; c > 0; c--) {
938: sysbeep(BEEP_ONHALT_PITCH,
1.565 simonb 939: BEEP_ONHALT_PERIOD * hz / 1000);
1.517 jmmv 940: delay(BEEP_ONHALT_PERIOD * 1000);
941: sysbeep(0, BEEP_ONHALT_PERIOD * hz / 1000);
942: delay(BEEP_ONHALT_PERIOD * 1000);
943: }
944: }
945: #endif
946:
1.300 drochner 947: cnpollc(1); /* for proper keyboard command handling */
1.512 perry 948: if (cngetc() == 0) {
949: /* no console attached, so just hlt */
950: for(;;) {
1.567 perry 951: __asm volatile("hlt");
1.512 perry 952: }
953: }
1.300 drochner 954: cnpollc(0);
1.1 cgd 955: }
1.193 mycroft 956:
1.210 christos 957: printf("rebooting...\n");
1.328 bouyer 958: if (cpureset_delay > 0)
959: delay(cpureset_delay * 1000);
1.1 cgd 960: cpu_reset();
961: for(;;) ;
962: /*NOTREACHED*/
963: }
964:
1.116 gwr 965: /*
966: * These variables are needed by /sbin/savecore
967: */
1.568 perry 968: uint32_t dumpmag = 0x8fca0101; /* magic number */
1.116 gwr 969: int dumpsize = 0; /* pages */
970: long dumplo = 0; /* blocks */
971:
972: /*
1.291 thorpej 973: * cpu_dumpsize: calculate size of machine-dependent kernel core dump headers.
974: */
975: int
976: cpu_dumpsize()
977: {
978: int size;
979:
980: size = ALIGN(sizeof(kcore_seg_t)) + ALIGN(sizeof(cpu_kcore_hdr_t)) +
981: ALIGN(mem_cluster_cnt * sizeof(phys_ram_seg_t));
982: if (roundup(size, dbtob(1)) != dbtob(1))
983: return (-1);
984:
985: return (1);
986: }
987:
988: /*
989: * cpu_dump_mempagecnt: calculate the size of RAM (in pages) to be dumped.
990: */
991: u_long
992: cpu_dump_mempagecnt()
993: {
994: u_long i, n;
995:
996: n = 0;
997: for (i = 0; i < mem_cluster_cnt; i++)
998: n += atop(mem_clusters[i].size);
999: return (n);
1000: }
1001:
1002: /*
1003: * cpu_dump: dump the machine-dependent kernel core dump headers.
1004: */
1005: int
1006: cpu_dump()
1007: {
1.542 junyoung 1008: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.563 christos 1009: char bf[dbtob(1)];
1.291 thorpej 1010: kcore_seg_t *segp;
1011: cpu_kcore_hdr_t *cpuhdrp;
1012: phys_ram_seg_t *memsegp;
1.481 gehenna 1013: const struct bdevsw *bdev;
1.291 thorpej 1014: int i;
1015:
1.481 gehenna 1016: bdev = bdevsw_lookup(dumpdev);
1017: if (bdev == NULL)
1018: return (ENXIO);
1019: dump = bdev->d_dump;
1.291 thorpej 1020:
1.563 christos 1021: memset(bf, 0, sizeof bf);
1022: segp = (kcore_seg_t *)bf;
1023: cpuhdrp = (cpu_kcore_hdr_t *)&bf[ALIGN(sizeof(*segp))];
1024: memsegp = (phys_ram_seg_t *)&bf[ ALIGN(sizeof(*segp)) +
1.291 thorpej 1025: ALIGN(sizeof(*cpuhdrp))];
1026:
1027: /*
1028: * Generate a segment header.
1029: */
1030: CORE_SETMAGIC(*segp, KCORE_MAGIC, MID_MACHINE, CORE_CPU);
1031: segp->c_size = dbtob(1) - ALIGN(sizeof(*segp));
1032:
1033: /*
1034: * Add the machine-dependent header info.
1035: */
1.558 junyoung 1036: cpuhdrp->pdppaddr = PDPpaddr;
1.291 thorpej 1037: cpuhdrp->nmemsegs = mem_cluster_cnt;
1038:
1039: /*
1040: * Fill in the memory segment descriptors.
1041: */
1042: for (i = 0; i < mem_cluster_cnt; i++) {
1043: memsegp[i].start = mem_clusters[i].start;
1044: memsegp[i].size = mem_clusters[i].size;
1045: }
1046:
1.563 christos 1047: return (dump(dumpdev, dumplo, (caddr_t)bf, dbtob(1)));
1.291 thorpej 1048: }
1049:
1050: /*
1.228 gwr 1051: * This is called by main to set dumplo and dumpsize.
1.414 thorpej 1052: * Dumps always skip the first PAGE_SIZE of disk space
1.116 gwr 1053: * in case there might be a disk label stored there.
1054: * If there is extra space, put dump at the end to
1055: * reduce the chance that swapping trashes it.
1056: */
1057: void
1.228 gwr 1058: cpu_dumpconf()
1.116 gwr 1059: {
1.481 gehenna 1060: const struct bdevsw *bdev;
1.291 thorpej 1061: int nblks, dumpblks; /* size of dump area */
1.116 gwr 1062:
1063: if (dumpdev == NODEV)
1.291 thorpej 1064: goto bad;
1.481 gehenna 1065: bdev = bdevsw_lookup(dumpdev);
1.582 mrg 1066: if (bdev == NULL) {
1067: dumpdev = NODEV;
1.583 skrll 1068: goto bad;
1.582 mrg 1069: }
1.481 gehenna 1070: if (bdev->d_psize == NULL)
1.291 thorpej 1071: goto bad;
1.481 gehenna 1072: nblks = (*bdev->d_psize)(dumpdev);
1.116 gwr 1073: if (nblks <= ctod(1))
1.291 thorpej 1074: goto bad;
1.116 gwr 1075:
1.291 thorpej 1076: dumpblks = cpu_dumpsize();
1077: if (dumpblks < 0)
1078: goto bad;
1079: dumpblks += ctod(cpu_dump_mempagecnt());
1080:
1081: /* If dump won't fit (incl. room for possible label), punt. */
1082: if (dumpblks > (nblks - ctod(1)))
1083: goto bad;
1084:
1085: /* Put dump at end of partition */
1086: dumplo = nblks - dumpblks;
1087:
1088: /* dumpsize is in page units, and doesn't include headers. */
1089: dumpsize = cpu_dump_mempagecnt();
1090: return;
1.116 gwr 1091:
1.291 thorpej 1092: bad:
1093: dumpsize = 0;
1.116 gwr 1094: }
1095:
1.1 cgd 1096: /*
1097: * Doadump comes here after turning off memory management and
1098: * getting on the dump stack, either when called above, or by
1099: * the auto-restart code.
1100: */
1.414 thorpej 1101: #define BYTES_PER_DUMP PAGE_SIZE /* must be a multiple of pagesize XXX small */
1.314 thorpej 1102: static vaddr_t dumpspace;
1.163 cgd 1103:
1.314 thorpej 1104: vaddr_t
1.551 junyoung 1105: reserve_dumppages(vaddr_t p)
1.163 cgd 1106: {
1107:
1108: dumpspace = p;
1109: return (p + BYTES_PER_DUMP);
1110: }
1111:
1.32 andrew 1112: void
1.1 cgd 1113: dumpsys()
1114: {
1.556 mycroft 1115: u_long totalbytesleft, bytes, i, n, m, memseg;
1.291 thorpej 1116: u_long maddr;
1117: int psize;
1.163 cgd 1118: daddr_t blkno;
1.481 gehenna 1119: const struct bdevsw *bdev;
1.542 junyoung 1120: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.200 christos 1121: int error;
1.193 mycroft 1122:
1123: /* Save registers. */
1124: savectx(&dumppcb);
1.1 cgd 1125:
1126: if (dumpdev == NODEV)
1127: return;
1.484 fvdl 1128:
1.481 gehenna 1129: bdev = bdevsw_lookup(dumpdev);
1130: if (bdev == NULL || bdev->d_psize == NULL)
1131: return;
1.163 cgd 1132:
1133: /*
1134: * For dumps during autoconfiguration,
1135: * if dump device has already configured...
1136: */
1137: if (dumpsize == 0)
1.228 gwr 1138: cpu_dumpconf();
1.330 jtk 1139: if (dumplo <= 0 || dumpsize == 0) {
1.275 mycroft 1140: printf("\ndump to dev %u,%u not possible\n", major(dumpdev),
1141: minor(dumpdev));
1.163 cgd 1142: return;
1.275 mycroft 1143: }
1144: printf("\ndumping to dev %u,%u offset %ld\n", major(dumpdev),
1145: minor(dumpdev), dumplo);
1.134 mycroft 1146:
1.481 gehenna 1147: psize = (*bdev->d_psize)(dumpdev);
1.210 christos 1148: printf("dump ");
1.163 cgd 1149: if (psize == -1) {
1.210 christos 1150: printf("area unavailable\n");
1.163 cgd 1151: return;
1152: }
1153:
1154: #if 0 /* XXX this doesn't work. grr. */
1.565 simonb 1155: /* toss any characters present prior to dump */
1.163 cgd 1156: while (sget() != NULL); /*syscons and pccons differ */
1157: #endif
1158:
1.291 thorpej 1159: if ((error = cpu_dump()) != 0)
1160: goto err;
1161:
1162: totalbytesleft = ptoa(cpu_dump_mempagecnt());
1163: blkno = dumplo + cpu_dumpsize();
1.481 gehenna 1164: dump = bdev->d_dump;
1.200 christos 1165: error = 0;
1.291 thorpej 1166:
1167: for (memseg = 0; memseg < mem_cluster_cnt; memseg++) {
1168: maddr = mem_clusters[memseg].start;
1169: bytes = mem_clusters[memseg].size;
1170:
1171: for (i = 0; i < bytes; i += n, totalbytesleft -= n) {
1172: /* Print out how many MBs we have left to go. */
1173: if ((totalbytesleft % (1024*1024)) == 0)
1174: printf("%ld ", totalbytesleft / (1024 * 1024));
1175:
1176: /* Limit size for next transfer. */
1177: n = bytes - i;
1178: if (n > BYTES_PER_DUMP)
1179: n = BYTES_PER_DUMP;
1180:
1.556 mycroft 1181: for (m = 0; m < n; m += NBPG)
1182: pmap_kenter_pa(dumpspace + m, maddr + m,
1183: VM_PROT_READ);
1.291 thorpej 1184:
1185: error = (*dump)(dumpdev, blkno, (caddr_t)dumpspace, n);
1186: if (error)
1187: goto err;
1.163 cgd 1188: maddr += n;
1.291 thorpej 1189: blkno += btodb(n); /* XXX? */
1.163 cgd 1190:
1191: #if 0 /* XXX this doesn't work. grr. */
1.291 thorpej 1192: /* operator aborting dump? */
1193: if (sget() != NULL) {
1194: error = EINTR;
1195: break;
1196: }
1197: #endif
1.163 cgd 1198: }
1199: }
1200:
1.291 thorpej 1201: err:
1.163 cgd 1202: switch (error) {
1.1 cgd 1203:
1204: case ENXIO:
1.210 christos 1205: printf("device bad\n");
1.1 cgd 1206: break;
1207:
1208: case EFAULT:
1.210 christos 1209: printf("device not ready\n");
1.1 cgd 1210: break;
1211:
1212: case EINVAL:
1.210 christos 1213: printf("area improper\n");
1.1 cgd 1214: break;
1215:
1216: case EIO:
1.210 christos 1217: printf("i/o error\n");
1.1 cgd 1218: break;
1219:
1220: case EINTR:
1.210 christos 1221: printf("aborted from console\n");
1.1 cgd 1222: break;
1223:
1.163 cgd 1224: case 0:
1.210 christos 1225: printf("succeeded\n");
1.163 cgd 1226: break;
1227:
1.1 cgd 1228: default:
1.210 christos 1229: printf("error %d\n", error);
1.1 cgd 1230: break;
1231: }
1.210 christos 1232: printf("\n\n");
1.163 cgd 1233: delay(5000000); /* 5 seconds */
1.1 cgd 1234: }
1235:
1236: /*
1237: * Clear registers on exec
1238: */
1.33 cgd 1239: void
1.551 junyoung 1240: setregs(struct lwp *l, struct exec_package *pack, u_long stack)
1.1 cgd 1241: {
1.533 chs 1242: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 1243: struct pcb *pcb = &l->l_addr->u_pcb;
1.298 mycroft 1244: struct trapframe *tf;
1.1 cgd 1245:
1.161 mycroft 1246: #if NNPX > 0
1247: /* If we were using the FPU, forget about it. */
1.508 thorpej 1248: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
1249: npxsave_lwp(l, 0);
1.161 mycroft 1250: #endif
1.166 mycroft 1251:
1.178 mycroft 1252: #ifdef USER_LDT
1.508 thorpej 1253: pmap_ldt_cleanup(l);
1.178 mycroft 1254: #endif
1255:
1.548 yamt 1256: l->l_md.md_flags &= ~MDL_USEDFPU;
1.452 thorpej 1257: if (i386_use_fxsave) {
1.450 thorpej 1258: pcb->pcb_savefpu.sv_xmm.sv_env.en_cw = __NetBSD_NPXCW__;
1.452 thorpej 1259: pcb->pcb_savefpu.sv_xmm.sv_env.en_mxcsr = __INITIAL_MXCSR__;
1260: } else
1.450 thorpej 1261: pcb->pcb_savefpu.sv_87.sv_env.en_cw = __NetBSD_NPXCW__;
1.59 mycroft 1262:
1.508 thorpej 1263: tf = l->l_md.md_regs;
1.445 sommerfe 1264: tf->tf_gs = LSEL(LUDATA_SEL, SEL_UPL);
1265: tf->tf_fs = LSEL(LUDATA_SEL, SEL_UPL);
1.154 mycroft 1266: tf->tf_es = LSEL(LUDATA_SEL, SEL_UPL);
1267: tf->tf_ds = LSEL(LUDATA_SEL, SEL_UPL);
1.252 mycroft 1268: tf->tf_edi = 0;
1269: tf->tf_esi = 0;
1.154 mycroft 1270: tf->tf_ebp = 0;
1.508 thorpej 1271: tf->tf_ebx = (int)l->l_proc->p_psstr;
1.252 mycroft 1272: tf->tf_edx = 0;
1273: tf->tf_ecx = 0;
1274: tf->tf_eax = 0;
1.154 mycroft 1275: tf->tf_eip = pack->ep_entry;
1.533 chs 1276: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
1277: LSEL(LUCODEBIG_SEL, SEL_UPL) : LSEL(LUCODE_SEL, SEL_UPL);
1.154 mycroft 1278: tf->tf_eflags = PSL_USERSET;
1279: tf->tf_esp = stack;
1280: tf->tf_ss = LSEL(LUDATA_SEL, SEL_UPL);
1.1 cgd 1281: }
1282:
1283: /*
1.55 brezak 1284: * Initialize segments and descriptor tables
1.1 cgd 1285: */
1286:
1.516 fvdl 1287: union descriptor *gdt, *ldt;
1288: struct gate_descriptor *idt;
1.498 fvdl 1289: char idt_allocmap[NIDT];
1290: struct simplelock idt_lock = SIMPLELOCK_INITIALIZER;
1.275 mycroft 1291: #ifdef I586_CPU
1292: union descriptor *pentium_idt;
1293: #endif
1.573 yamt 1294: struct user *proc0paddr;
1295: extern vaddr_t proc0uarea;
1.49 brezak 1296:
1.178 mycroft 1297: void
1.551 junyoung 1298: setgate(struct gate_descriptor *gd, void *func, int args, int type, int dpl,
1299: int sel)
1.178 mycroft 1300: {
1.1 cgd 1301:
1.178 mycroft 1302: gd->gd_looffset = (int)func;
1.489 fvdl 1303: gd->gd_selector = sel;
1.178 mycroft 1304: gd->gd_stkcpy = args;
1305: gd->gd_xx = 0;
1306: gd->gd_type = type;
1307: gd->gd_dpl = dpl;
1308: gd->gd_p = 1;
1309: gd->gd_hioffset = (int)func >> 16;
1310: }
1311:
1312: void
1.551 junyoung 1313: unsetgate(struct gate_descriptor *gd)
1.484 fvdl 1314: {
1315: gd->gd_p = 0;
1316: gd->gd_hioffset = 0;
1317: gd->gd_looffset = 0;
1318: gd->gd_selector = 0;
1319: gd->gd_xx = 0;
1320: gd->gd_stkcpy = 0;
1321: gd->gd_type = 0;
1322: gd->gd_dpl = 0;
1323: }
1324:
1325:
1326: void
1.551 junyoung 1327: setregion(struct region_descriptor *rd, void *base, size_t limit)
1.178 mycroft 1328: {
1329:
1330: rd->rd_limit = (int)limit;
1331: rd->rd_base = (int)base;
1332: }
1.1 cgd 1333:
1.174 mycroft 1334: void
1.551 junyoung 1335: setsegment(struct segment_descriptor *sd, void *base, size_t limit, int type,
1336: int dpl, int def32, int gran)
1.174 mycroft 1337: {
1.1 cgd 1338:
1.174 mycroft 1339: sd->sd_lolimit = (int)limit;
1340: sd->sd_lobase = (int)base;
1341: sd->sd_type = type;
1342: sd->sd_dpl = dpl;
1343: sd->sd_p = 1;
1344: sd->sd_hilimit = (int)limit >> 16;
1345: sd->sd_xx = 0;
1346: sd->sd_def32 = def32;
1347: sd->sd_gran = gran;
1348: sd->sd_hibase = (int)base >> 24;
1349: }
1.1 cgd 1350:
1351: #define IDTVEC(name) __CONCAT(X, name)
1.542 junyoung 1352: typedef void (vector)(void);
1.299 mycroft 1353: extern vector IDTVEC(syscall);
1354: extern vector IDTVEC(osyscall);
1355: extern vector *IDTVEC(exceptions)[];
1.333 christos 1356: #ifdef COMPAT_SVR4
1357: extern vector IDTVEC(svr4_fasttrap);
1358: #endif /* COMPAT_SVR4 */
1.447 christos 1359: #ifdef COMPAT_MACH
1360: extern vector IDTVEC(mach_trap);
1361: #endif
1.1 cgd 1362:
1.381 thorpej 1363: #define KBTOB(x) ((size_t)(x) * 1024UL)
1364:
1.484 fvdl 1365: void cpu_init_idt()
1366: {
1367: struct region_descriptor region;
1368: #ifdef I586_CPU
1369: setregion(®ion, pentium_idt, NIDT * sizeof(idt[0]) - 1);
1370: #else
1371: setregion(®ion, idt, NIDT * sizeof(idt[0]) - 1);
1372: #endif
1.565 simonb 1373: lidt(®ion);
1.484 fvdl 1374: }
1375:
1.433 kanaoka 1376: void
1.568 perry 1377: add_mem_cluster(uint64_t seg_start, uint64_t seg_end, uint32_t type)
1.433 kanaoka 1378: {
1379: extern struct extent *iomem_ex;
1.492 kanaoka 1380: int i;
1.433 kanaoka 1381:
1382: if (seg_end > 0x100000000ULL) {
1383: printf("WARNING: skipping large "
1384: "memory map entry: "
1385: "0x%qx/0x%qx/0x%x\n",
1386: seg_start,
1387: (seg_end - seg_start),
1388: type);
1389: return;
1390: }
1391:
1392: /*
1393: * XXX Chop the last page off the size so that
1394: * XXX it can fit in avail_end.
1395: */
1396: if (seg_end == 0x100000000ULL)
1397: seg_end -= PAGE_SIZE;
1398:
1399: if (seg_end <= seg_start)
1400: return;
1401:
1.492 kanaoka 1402: for (i = 0; i < mem_cluster_cnt; i++) {
1403: if ((mem_clusters[i].start == round_page(seg_start))
1404: && (mem_clusters[i].size
1405: == trunc_page(seg_end) - mem_clusters[i].start)) {
1406: #ifdef DEBUG_MEMLOAD
1407: printf("WARNING: skipping duplicate segment entry\n");
1408: #endif
1409: return;
1410: }
1411: }
1412:
1.433 kanaoka 1413: /*
1414: * Allocate the physical addresses used by RAM
1415: * from the iomem extent map. This is done before
1416: * the addresses are page rounded just to make
1417: * sure we get them all.
1418: */
1419: if (extent_alloc_region(iomem_ex, seg_start,
1420: seg_end - seg_start, EX_NOWAIT)) {
1421: /* XXX What should we do? */
1422: printf("WARNING: CAN'T ALLOCATE "
1423: "MEMORY SEGMENT "
1424: "(0x%qx/0x%qx/0x%x) FROM "
1425: "IOMEM EXTENT MAP!\n",
1426: seg_start, seg_end - seg_start, type);
1.492 kanaoka 1427: return;
1.433 kanaoka 1428: }
1429:
1430: /*
1431: * If it's not free memory, skip it.
1432: */
1433: if (type != BIM_Memory)
1434: return;
1435:
1436: /* XXX XXX XXX */
1437: if (mem_cluster_cnt >= VM_PHYSSEG_MAX)
1.579 cube 1438: panic("init386: too many memory segments "
1439: "(increase VM_PHYSSEG_MAX)");
1.433 kanaoka 1440:
1441: seg_start = round_page(seg_start);
1442: seg_end = trunc_page(seg_end);
1443:
1444: if (seg_start == seg_end)
1445: return;
1446:
1447: mem_clusters[mem_cluster_cnt].start = seg_start;
1448: mem_clusters[mem_cluster_cnt].size =
1449: seg_end - seg_start;
1450:
1451: if (avail_end < seg_end)
1452: avail_end = seg_end;
1453: physmem += atop(mem_clusters[mem_cluster_cnt].size);
1454: mem_cluster_cnt++;
1455: }
1456:
1.59 mycroft 1457: void
1.484 fvdl 1458: initgdt(union descriptor *tgdt)
1459: {
1460: struct region_descriptor region;
1461: gdt = tgdt;
1462: memset(gdt, 0, NGDT*sizeof(*gdt));
1463: /* make gdt gates and memory segments */
1464: setsegment(&gdt[GCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 1, 1);
1465: setsegment(&gdt[GDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 1, 1);
1.533 chs 1466: setsegment(&gdt[GUCODE_SEL].sd, 0, x86_btop(I386_MAX_EXE_ADDR) - 1,
1467: SDT_MEMERA, SEL_UPL, 1, 1);
1468: setsegment(&gdt[GUCODEBIG_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1469: SDT_MEMERA, SEL_UPL, 1, 1);
1.514 fvdl 1470: setsegment(&gdt[GUDATA_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1471: SDT_MEMRWA, SEL_UPL, 1, 1);
1472: #ifdef COMPAT_MACH
1473: setgate(&gdt[GMACHCALLS_SEL].gd, &IDTVEC(mach_trap), 1,
1.489 fvdl 1474: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 1475: #endif
1476: #if NBIOSCALL > 0
1477: /* bios trampoline GDT entries */
1478: setsegment(&gdt[GBIOSCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 0,
1479: 0);
1480: setsegment(&gdt[GBIOSDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 0,
1481: 0);
1482: #endif
1483: setsegment(&gdt[GCPU_SEL].sd, &cpu_info_primary,
1484: sizeof(struct cpu_info)-1, SDT_MEMRWA, SEL_KPL, 1, 1);
1485:
1486: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
1487: lgdt(®ion);
1488: }
1489:
1490: void
1.551 junyoung 1491: init386(paddr_t first_avail)
1.2 cgd 1492: {
1.484 fvdl 1493: union descriptor *tgdt;
1.542 junyoung 1494: extern void consinit(void);
1.375 drochner 1495: extern struct extent *iomem_ex;
1.401 thorpej 1496: struct btinfo_memmap *bim;
1.381 thorpej 1497: struct region_descriptor region;
1.401 thorpej 1498: int x, first16q;
1.568 perry 1499: uint64_t seg_start, seg_end;
1500: uint64_t seg_start1, seg_end1;
1.473 tshiozak 1501: paddr_t realmode_reserved_start;
1502: psize_t realmode_reserved_size;
1503: int needs_earlier_install_pte0;
1.436 jdolecek 1504: #if NBIOSCALL > 0
1505: extern int biostramp_image_size;
1506: extern u_char biostramp_image[];
1507: #endif
1.1 cgd 1508:
1.484 fvdl 1509: cpu_probe_features(&cpu_info_primary);
1510: cpu_feature = cpu_info_primary.ci_feature_flags;
1.553 lukem 1511: cpu_feature2 = cpu_info_primary.ci_feature2_flags;
1.484 fvdl 1512:
1.573 yamt 1513: proc0paddr = UAREA_TO_USER(proc0uarea);
1.508 thorpej 1514: lwp0.l_addr = proc0paddr;
1515: cpu_info_primary.ci_curpcb = &lwp0.l_addr->u_pcb;
1.275 mycroft 1516:
1.514 fvdl 1517: x86_bus_space_init();
1.84 cgd 1518: consinit(); /* XXX SHOULD NOT BE DONE HERE */
1.401 thorpej 1519: /*
1520: * Initailize PAGE_SIZE-dependent variables.
1521: */
1522: uvm_setpagesize();
1.450 thorpej 1523:
1524: /*
1525: * Saving SSE registers won't work if the save area isn't
1526: * 16-byte aligned.
1527: */
1528: if (offsetof(struct user, u_pcb.pcb_savefpu) & 0xf)
1529: panic("init386: pcb_savefpu not 16-byte aligned");
1.440 thorpej 1530:
1531: /*
1532: * Start with 2 color bins -- this is just a guess to get us
1533: * started. We'll recolor when we determine the largest cache
1534: * sizes on the system.
1535: */
1536: uvmexp.ncolors = 2;
1.401 thorpej 1537:
1.473 tshiozak 1538: /*
1.484 fvdl 1539: * BIOS leaves data in physical page 0
1540: * Even if it didn't, our VM system doesn't like using zero as a
1541: * physical page number.
1542: * We may also need pages in low memory (one each) for secondary CPU
1543: * startup, for BIOS calls, and for ACPI, plus a page table page to map
1544: * them into the first few pages of the kernel's pmap.
1.473 tshiozak 1545: */
1546: avail_start = PAGE_SIZE;
1547:
1548: /*
1549: * reserve memory for real-mode call
1550: */
1551: needs_earlier_install_pte0 = 0;
1552: realmode_reserved_start = 0;
1553: realmode_reserved_size = 0;
1.414 thorpej 1554: #if NBIOSCALL > 0
1.473 tshiozak 1555: /* save us a page for trampoline code */
1556: realmode_reserved_size += PAGE_SIZE;
1557: needs_earlier_install_pte0 = 1;
1558: #endif
1.484 fvdl 1559: #ifdef MULTIPROCESSOR /* XXX */
1560: KASSERT(avail_start == PAGE_SIZE); /* XXX */
1561: if (realmode_reserved_size < MP_TRAMPOLINE) /* XXX */
1562: realmode_reserved_size = MP_TRAMPOLINE; /* XXX */
1563: needs_earlier_install_pte0 = 1; /* XXX */
1564: #endif /* XXX */
1.473 tshiozak 1565: #if NACPI > 0
1566: /* trampoline code for wake handler */
1567: realmode_reserved_size += ptoa(acpi_md_get_npages_of_wakecode()+1);
1568: needs_earlier_install_pte0 = 1;
1569: #endif
1570: if (needs_earlier_install_pte0) {
1571: /* page table for directory entry 0 */
1572: realmode_reserved_size += PAGE_SIZE;
1573: }
1574: if (realmode_reserved_size>0) {
1575: realmode_reserved_start = avail_start;
1576: avail_start += realmode_reserved_size;
1577: }
1.414 thorpej 1578:
1.492 kanaoka 1579: #ifdef DEBUG_MEMLOAD
1580: printf("mem_cluster_count: %d\n", mem_cluster_cnt);
1581: #endif
1582:
1.401 thorpej 1583: /*
1584: * Call pmap initialization to make new kernel address space.
1585: * We must do this before loading pages into the VM system.
1586: */
1.314 thorpej 1587: pmap_bootstrap((vaddr_t)atdevbase + IOM_SIZE);
1.2 cgd 1588:
1.401 thorpej 1589: /*
1590: * Check to see if we have a memory map from the BIOS (passed
1591: * to us by the boot program.
1592: */
1.577 mrg 1593: if ((biosmem_implicit || (biosbasemem == 0 && biosextmem == 0)) &&
1.575 mrg 1594: (bim = lookup_bootinfo(BTINFO_MEMMAP)) != NULL && bim->num > 0) {
1.463 enami 1595: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1596: printf("BIOS MEMORY MAP (%d ENTRIES):\n", bim->num);
1597: #endif
1598: for (x = 0; x < bim->num; x++) {
1.463 enami 1599: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1600: printf(" addr 0x%qx size 0x%qx type 0x%x\n",
1601: bim->entry[x].addr,
1602: bim->entry[x].size,
1603: bim->entry[x].type);
1604: #endif
1605:
1606: /*
1607: * If the segment is not memory, skip it.
1608: */
1609: switch (bim->entry[x].type) {
1610: case BIM_Memory:
1611: case BIM_ACPI:
1612: case BIM_NVS:
1613: break;
1614: default:
1615: continue;
1616: }
1617:
1618: /*
1.586.2.5! liamjfoy 1619: * If the segment is smaller than a page, skip it.
! 1620: */
! 1621: if (bim->entry[x].size < NBPG) {
! 1622: continue;
! 1623: }
! 1624:
! 1625: /*
1.401 thorpej 1626: * Sanity check the entry.
1627: * XXX Need to handle uint64_t in extent code
1628: * XXX and 64-bit physical addresses in i386
1629: * XXX port.
1630: */
1631: seg_start = bim->entry[x].addr;
1632: seg_end = bim->entry[x].addr + bim->entry[x].size;
1633:
1.432 kanaoka 1634: /*
1.433 kanaoka 1635: * Avoid Compatibility Holes.
1636: * XXX Holes within memory space that allow access
1637: * XXX to be directed to the PC-compatible frame buffer
1.484 fvdl 1638: * XXX (0xa0000-0xbffff),to adapter ROM space
1.433 kanaoka 1639: * XXX (0xc0000-0xdffff), and to system BIOS space
1640: * XXX (0xe0000-0xfffff).
1641: * XXX Some laptop(for example,Toshiba Satellite2550X)
1642: * XXX report this area and occurred problems,
1643: * XXX so we avoid this area.
1.432 kanaoka 1644: */
1.433 kanaoka 1645: if (seg_start < 0x100000 && seg_end > 0xa0000) {
1646: printf("WARNING: memory map entry overlaps "
1647: "with ``Compatibility Holes'': "
1648: "0x%qx/0x%qx/0x%x\n", seg_start,
1649: seg_end - seg_start, bim->entry[x].type);
1650: add_mem_cluster(seg_start, 0xa0000,
1651: bim->entry[x].type);
1652: add_mem_cluster(0x100000, seg_end,
1.401 thorpej 1653: bim->entry[x].type);
1.433 kanaoka 1654: } else
1655: add_mem_cluster(seg_start, seg_end,
1.401 thorpej 1656: bim->entry[x].type);
1657: }
1.421 aymeric 1658: }
1.575 mrg 1659:
1.421 aymeric 1660: /*
1661: * If the loop above didn't find any valid segment, fall back to
1662: * former code.
1663: */
1664: if (mem_cluster_cnt == 0) {
1.401 thorpej 1665: /*
1666: * Allocate the physical addresses used by RAM from the iomem
1667: * extent map. This is done before the addresses are
1668: * page rounded just to make sure we get them all.
1669: */
1670: if (extent_alloc_region(iomem_ex, 0, KBTOB(biosbasemem),
1671: EX_NOWAIT)) {
1672: /* XXX What should we do? */
1673: printf("WARNING: CAN'T ALLOCATE BASE MEMORY FROM "
1674: "IOMEM EXTENT MAP!\n");
1675: }
1676: mem_clusters[0].start = 0;
1677: mem_clusters[0].size = trunc_page(KBTOB(biosbasemem));
1678: physmem += atop(mem_clusters[0].size);
1679: if (extent_alloc_region(iomem_ex, IOM_END, KBTOB(biosextmem),
1680: EX_NOWAIT)) {
1681: /* XXX What should we do? */
1682: printf("WARNING: CAN'T ALLOCATE EXTENDED MEMORY FROM "
1683: "IOMEM EXTENT MAP!\n");
1684: }
1685: #if NISADMA > 0
1686: /*
1687: * Some motherboards/BIOSes remap the 384K of RAM that would
1688: * normally be covered by the ISA hole to the end of memory
1689: * so that it can be used. However, on a 16M system, this
1690: * would cause bounce buffers to be allocated and used.
1691: * This is not desirable behaviour, as more than 384K of
1692: * bounce buffers might be allocated. As a work-around,
1693: * we round memory down to the nearest 1M boundary if
1694: * we're using any isadma devices and the remapped memory
1695: * is what puts us over 16M.
1696: */
1697: if (biosextmem > (15*1024) && biosextmem < (16*1024)) {
1698: char pbuf[9];
1699:
1700: format_bytes(pbuf, sizeof(pbuf),
1701: biosextmem - (15*1024));
1702: printf("Warning: ignoring %s of remapped memory\n",
1703: pbuf);
1704: biosextmem = (15*1024);
1705: }
1706: #endif
1707: mem_clusters[1].start = IOM_END;
1708: mem_clusters[1].size = trunc_page(KBTOB(biosextmem));
1709: physmem += atop(mem_clusters[1].size);
1710:
1711: mem_cluster_cnt = 2;
1712:
1713: avail_end = IOM_END + trunc_page(KBTOB(biosextmem));
1714: }
1715: /*
1716: * If we have 16M of RAM or less, just put it all on
1717: * the default free list. Otherwise, put the first
1718: * 16M of RAM on a lower priority free list (so that
1719: * all of the ISA DMA'able memory won't be eaten up
1720: * first-off).
1721: */
1722: if (avail_end <= (16 * 1024 * 1024))
1723: first16q = VM_FREELIST_DEFAULT;
1724: else
1725: first16q = VM_FREELIST_FIRST16;
1726:
1727: /* Make sure the end of the space used by the kernel is rounded. */
1728: first_avail = round_page(first_avail);
1729:
1730: /*
1731: * Now, load the memory clusters (which have already been
1732: * rounded and truncated) into the VM system.
1733: *
1734: * NOTE: WE ASSUME THAT MEMORY STARTS AT 0 AND THAT THE KERNEL
1735: * IS LOADED AT IOM_END (1M).
1736: */
1737: for (x = 0; x < mem_cluster_cnt; x++) {
1738: seg_start = mem_clusters[x].start;
1739: seg_end = mem_clusters[x].start + mem_clusters[x].size;
1740: seg_start1 = 0;
1741: seg_end1 = 0;
1742:
1743: /*
1744: * Skip memory before our available starting point.
1745: */
1746: if (seg_end <= avail_start)
1747: continue;
1748:
1749: if (avail_start >= seg_start && avail_start < seg_end) {
1750: if (seg_start != 0)
1751: panic("init386: memory doesn't start at 0");
1752: seg_start = avail_start;
1753: if (seg_start == seg_end)
1754: continue;
1755: }
1756:
1757: /*
1758: * If this segment contains the kernel, split it
1759: * in two, around the kernel.
1760: */
1761: if (seg_start <= IOM_END && first_avail <= seg_end) {
1762: seg_start1 = first_avail;
1763: seg_end1 = seg_end;
1764: seg_end = IOM_END;
1765: }
1766:
1767: /* First hunk */
1768: if (seg_start != seg_end) {
1.480 erh 1769: if (seg_start < (16 * 1024 * 1024) &&
1.401 thorpej 1770: first16q != VM_FREELIST_DEFAULT) {
1.568 perry 1771: uint64_t tmp;
1.401 thorpej 1772:
1773: if (seg_end > (16 * 1024 * 1024))
1774: tmp = (16 * 1024 * 1024);
1775: else
1776: tmp = seg_end;
1.492 kanaoka 1777:
1778: if (tmp != seg_start) {
1.463 enami 1779: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1780: printf("loading 0x%qx-0x%qx "
1781: "(0x%lx-0x%lx)\n",
1782: seg_start, tmp,
1783: atop(seg_start), atop(tmp));
1784: #endif
1785: uvm_page_physload(atop(seg_start),
1786: atop(tmp), atop(seg_start),
1787: atop(tmp), first16q);
1788: }
1.401 thorpej 1789: seg_start = tmp;
1790: }
1.411 enami 1791:
1792: if (seg_start != seg_end) {
1.463 enami 1793: #ifdef DEBUG_MEMLOAD
1.411 enami 1794: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1795: seg_start, seg_end,
1796: atop(seg_start), atop(seg_end));
1797: #endif
1798: uvm_page_physload(atop(seg_start),
1799: atop(seg_end), atop(seg_start),
1800: atop(seg_end), VM_FREELIST_DEFAULT);
1801: }
1.401 thorpej 1802: }
1803:
1804: /* Second hunk */
1805: if (seg_start1 != seg_end1) {
1.480 erh 1806: if (seg_start1 < (16 * 1024 * 1024) &&
1.401 thorpej 1807: first16q != VM_FREELIST_DEFAULT) {
1.568 perry 1808: uint64_t tmp;
1.401 thorpej 1809:
1810: if (seg_end1 > (16 * 1024 * 1024))
1811: tmp = (16 * 1024 * 1024);
1812: else
1813: tmp = seg_end1;
1.492 kanaoka 1814:
1815: if (tmp != seg_start1) {
1.463 enami 1816: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1817: printf("loading 0x%qx-0x%qx "
1818: "(0x%lx-0x%lx)\n",
1819: seg_start1, tmp,
1820: atop(seg_start1), atop(tmp));
1821: #endif
1822: uvm_page_physload(atop(seg_start1),
1823: atop(tmp), atop(seg_start1),
1824: atop(tmp), first16q);
1825: }
1.401 thorpej 1826: seg_start1 = tmp;
1827: }
1.412 enami 1828:
1829: if (seg_start1 != seg_end1) {
1.463 enami 1830: #ifdef DEBUG_MEMLOAD
1.412 enami 1831: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1832: seg_start1, seg_end1,
1833: atop(seg_start1), atop(seg_end1));
1834: #endif
1835: uvm_page_physload(atop(seg_start1),
1836: atop(seg_end1), atop(seg_start1),
1837: atop(seg_end1), VM_FREELIST_DEFAULT);
1838: }
1.401 thorpej 1839: }
1840: }
1841:
1842: /*
1843: * Steal memory for the message buffer (at end of core).
1844: */
1845: {
1.543 mycroft 1846: struct vm_physseg *vps;
1.401 thorpej 1847: psize_t sz = round_page(MSGBUFSIZE);
1848: psize_t reqsz = sz;
1849:
1.586.2.1 tron 1850: search_again:
1.401 thorpej 1851: for (x = 0; x < vm_nphysseg; x++) {
1852: vps = &vm_physmem[x];
1853: if (ptoa(vps->avail_end) == avail_end)
1.543 mycroft 1854: goto found;
1.401 thorpej 1855: }
1.543 mycroft 1856: panic("init386: can't find end of memory");
1.401 thorpej 1857:
1.543 mycroft 1858: found:
1.401 thorpej 1859: /* Shrink so it'll fit in the last segment. */
1860: if ((vps->avail_end - vps->avail_start) < atop(sz))
1861: sz = ptoa(vps->avail_end - vps->avail_start);
1862:
1863: vps->avail_end -= atop(sz);
1864: vps->end -= atop(sz);
1.586.2.1 tron 1865: msgbuf_p_seg[msgbuf_p_cnt].sz = sz;
1866: msgbuf_p_seg[msgbuf_p_cnt++].paddr = ptoa(vps->avail_end);
1.401 thorpej 1867:
1868: /* Remove the last segment if it now has no pages. */
1869: if (vps->start == vps->end) {
1870: for (vm_nphysseg--; x < vm_nphysseg; x++)
1871: vm_physmem[x] = vm_physmem[x + 1];
1872: }
1873:
1874: /* Now find where the new avail_end is. */
1875: for (avail_end = 0, x = 0; x < vm_nphysseg; x++)
1876: if (vm_physmem[x].avail_end > avail_end)
1877: avail_end = vm_physmem[x].avail_end;
1878: avail_end = ptoa(avail_end);
1879:
1.586.2.1 tron 1880: if (sz != reqsz) {
1881: reqsz -= sz;
1882: if (msgbuf_p_cnt != VM_PHYSSEG_MAX) {
1883: /* if still segments available, get memory from next one ... */
1.586.2.2 tron 1884: sz = reqsz;
1885: goto search_again;
1.586.2.1 tron 1886: }
1.401 thorpej 1887: /* Warn if the message buffer had to be shrunk. */
1888: printf("WARNING: %ld bytes not available for msgbuf "
1.586.2.2 tron 1889: "in last cluster (%ld used)\n", (long)MSGBUFSIZE, MSGBUFSIZE - reqsz);
1.586.2.1 tron 1890: }
1.401 thorpej 1891: }
1892:
1.473 tshiozak 1893: /*
1894: * install PT page for the first 4M if needed.
1895: */
1896: if (needs_earlier_install_pte0) {
1897: paddr_t paddr;
1898: #ifdef DIAGNOSTIC
1899: if (realmode_reserved_size < PAGE_SIZE) {
1900: panic("cannot steal memory for first 4M PT page.");
1901: }
1902: #endif
1903: paddr=realmode_reserved_start+realmode_reserved_size-PAGE_SIZE;
1.561 yamt 1904: pmap_kenter_pa((vaddr_t)vtopte(0), paddr,
1905: VM_PROT_READ|VM_PROT_WRITE);
1.473 tshiozak 1906: pmap_update(pmap_kernel());
1907: /* make sure it is clean before using */
1908: memset(vtopte(0), 0, PAGE_SIZE);
1909: realmode_reserved_size -= PAGE_SIZE;
1910: }
1911:
1.295 drochner 1912: #if NBIOSCALL > 0
1.436 jdolecek 1913: /*
1914: * this should be caught at kernel build time, but put it here
1915: * in case someone tries to fake it out...
1916: */
1917: #ifdef DIAGNOSTIC
1.473 tshiozak 1918: if (realmode_reserved_start > BIOSTRAMP_BASE ||
1.474 tron 1919: (realmode_reserved_start+realmode_reserved_size) < (BIOSTRAMP_BASE+
1.473 tshiozak 1920: PAGE_SIZE)) {
1921: panic("cannot steal memory for PT page of bioscall.");
1922: }
1.436 jdolecek 1923: if (biostramp_image_size > PAGE_SIZE)
1.483 provos 1924: panic("biostramp_image_size too big: %x vs. %x",
1.436 jdolecek 1925: biostramp_image_size, PAGE_SIZE);
1926: #endif
1927: pmap_kenter_pa((vaddr_t)BIOSTRAMP_BASE, /* virtual */
1928: (paddr_t)BIOSTRAMP_BASE, /* physical */
1929: VM_PROT_ALL); /* protection */
1.456 chris 1930: pmap_update(pmap_kernel());
1.436 jdolecek 1931: memcpy((caddr_t)BIOSTRAMP_BASE, biostramp_image, biostramp_image_size);
1932: #ifdef DEBUG_BIOSCALL
1933: printf("biostramp installed @ %x\n", BIOSTRAMP_BASE);
1934: #endif
1.473 tshiozak 1935: realmode_reserved_size -= PAGE_SIZE;
1936: realmode_reserved_start += PAGE_SIZE;
1937: #endif
1938:
1939: #if NACPI > 0
1940: /*
1941: * Steal memory for the acpi wake code
1942: */
1943: {
1944: paddr_t paddr, p;
1945: psize_t sz;
1946: int npg;
1947:
1948: paddr = realmode_reserved_start;
1949: npg = acpi_md_get_npages_of_wakecode();
1950: sz = ptoa(npg);
1951: #ifdef DIAGNOSTIC
1952: if (realmode_reserved_size < sz) {
1953: panic("cannot steal memory for ACPI wake code.");
1954: }
1955: #endif
1956:
1957: /* identical mapping */
1958: p = paddr;
1959: for (x=0; x<npg; x++) {
1960: printf("kenter: 0x%08X\n", (unsigned)p);
1961: pmap_kenter_pa((vaddr_t)p, p, VM_PROT_ALL);
1962: p += PAGE_SIZE;
1963: }
1964: pmap_update(pmap_kernel());
1965:
1966: acpi_md_install_wakecode(paddr);
1967:
1968: realmode_reserved_size -= sz;
1969: realmode_reserved_start += sz;
1970: }
1.295 drochner 1971: #endif
1.59 mycroft 1972:
1.561 yamt 1973: pmap_kenter_pa(idt_vaddr, idt_paddr, VM_PROT_READ|VM_PROT_WRITE);
1.456 chris 1974: pmap_update(pmap_kernel());
1.484 fvdl 1975: memset((void *)idt_vaddr, 0, PAGE_SIZE);
1976:
1.516 fvdl 1977: idt = (struct gate_descriptor *)idt_vaddr;
1.275 mycroft 1978: #ifdef I586_CPU
1.561 yamt 1979: pmap_kenter_pa(pentium_idt_vaddr, idt_paddr, VM_PROT_READ);
1.275 mycroft 1980: pentium_idt = (union descriptor *)pentium_idt_vaddr;
1981: #endif
1.484 fvdl 1982: pmap_update(pmap_kernel());
1983:
1984: tgdt = gdt;
1.516 fvdl 1985: gdt = (union descriptor *)
1986: ((char *)idt + NIDT * sizeof (struct gate_descriptor));
1.275 mycroft 1987: ldt = gdt + NGDT;
1988:
1.484 fvdl 1989: memcpy(gdt, tgdt, NGDT*sizeof(*gdt));
1.275 mycroft 1990:
1991: setsegment(&gdt[GLDT_SEL].sd, ldt, NLDT * sizeof(ldt[0]) - 1,
1992: SDT_SYSLDT, SEL_KPL, 0, 0);
1993:
1994: /* make ldt gates and memory segments */
1995: setgate(&ldt[LSYS5CALLS_SEL].gd, &IDTVEC(osyscall), 1,
1.489 fvdl 1996: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.447 christos 1997:
1.275 mycroft 1998: ldt[LUCODE_SEL] = gdt[GUCODE_SEL];
1.533 chs 1999: ldt[LUCODEBIG_SEL] = gdt[GUCODEBIG_SEL];
1.275 mycroft 2000: ldt[LUDATA_SEL] = gdt[GUDATA_SEL];
1.324 christos 2001: ldt[LSOL26CALLS_SEL] = ldt[LBSDICALLS_SEL] = ldt[LSYS5CALLS_SEL];
1.275 mycroft 2002:
2003: /* exceptions */
1.498 fvdl 2004: for (x = 0; x < 32; x++) {
1.516 fvdl 2005: setgate(&idt[x], IDTVEC(exceptions)[x], 0, SDT_SYS386TGT,
1.489 fvdl 2006: (x == 3 || x == 4) ? SEL_UPL : SEL_KPL,
2007: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 2008: idt_allocmap[x] = 1;
2009: }
1.257 thorpej 2010:
1.275 mycroft 2011: /* new-style interrupt gate for syscalls */
1.516 fvdl 2012: setgate(&idt[128], &IDTVEC(syscall), 0, SDT_SYS386TGT, SEL_UPL,
1.489 fvdl 2013: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 2014: idt_allocmap[128] = 1;
1.333 christos 2015: #ifdef COMPAT_SVR4
1.516 fvdl 2016: setgate(&idt[0xd2], &IDTVEC(svr4_fasttrap), 0, SDT_SYS386TGT,
1.489 fvdl 2017: SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 2018: idt_allocmap[0xd2] = 1;
1.333 christos 2019: #endif /* COMPAT_SVR4 */
1.264 mycroft 2020:
1.275 mycroft 2021: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
2022: lgdt(®ion);
1.484 fvdl 2023:
2024: cpu_init_idt();
1.264 mycroft 2025:
1.522 ragge 2026: #if NKSYMS || defined(DDB) || defined(LKM)
1.308 tv 2027: {
2028: extern int end;
1.585 jmmv 2029: boolean_t loaded;
1.336 christos 2030: struct btinfo_symtab *symtab;
1.308 tv 2031:
1.522 ragge 2032: #ifdef DDB
1.484 fvdl 2033: db_machine_init();
1.522 ragge 2034: #endif
1.484 fvdl 2035:
1.585 jmmv 2036: #if defined(MULTIBOOT)
2037: loaded = multiboot_ksyms_init();
2038: #else
2039: loaded = FALSE;
2040: #endif
2041: if (!loaded) {
2042: symtab = lookup_bootinfo(BTINFO_SYMTAB);
2043: if (symtab) {
2044: symtab->ssym += KERNBASE;
2045: symtab->esym += KERNBASE;
2046: ksyms_init(symtab->nsym, (int *)symtab->ssym,
2047: (int *)symtab->esym);
2048: } else
2049: ksyms_init(*(int *)&end, ((int *)&end) + 1, esym);
1.336 christos 2050: }
1.308 tv 2051: }
1.522 ragge 2052: #endif
2053: #ifdef DDB
1.190 mycroft 2054: if (boothowto & RB_KDB)
2055: Debugger();
1.377 ws 2056: #endif
2057: #ifdef IPKDB
2058: ipkdb_init();
2059: if (boothowto & RB_KDB)
2060: ipkdb_connect(0);
1.190 mycroft 2061: #endif
2062: #ifdef KGDB
1.243 drochner 2063: kgdb_port_init();
1.235 thorpej 2064: if (boothowto & RB_KDB) {
2065: kgdb_debug_init = 1;
1.242 drochner 2066: kgdb_connect(1);
1.235 thorpej 2067: }
1.384 jdolecek 2068: #endif
2069:
2070: #if NMCA > 0
2071: /* check for MCA bus, needed to be done before ISA stuff - if
2072: * MCA is detected, ISA needs to use level triggered interrupts
2073: * by default */
2074: mca_busprobe();
1.190 mycroft 2075: #endif
1.275 mycroft 2076:
1.498 fvdl 2077: intr_default_setup();
1.431 thorpej 2078:
2079: /* Initialize software interrupts. */
2080: softintr_init();
1.275 mycroft 2081:
1.498 fvdl 2082: splraise(IPL_IPI);
1.275 mycroft 2083: enable_intr();
2084:
2085: if (physmem < btoc(2 * 1024 * 1024)) {
2086: printf("warning: too little memory available; "
1.383 mycroft 2087: "have %lu bytes, want %lu bytes\n"
1.275 mycroft 2088: "running in degraded mode\n"
2089: "press a key to confirm\n\n",
1.383 mycroft 2090: ptoa(physmem), 2*1024*1024UL);
1.275 mycroft 2091: cngetc();
2092: }
1.507 jdolecek 2093:
2094: #ifdef __HAVE_CPU_MAXPROC
2095: /* Make sure maxproc is sane */
2096: if (maxproc > cpu_maxproc())
2097: maxproc = cpu_maxproc();
2098: #endif
1.1 cgd 2099: }
2100:
1.107 deraadt 2101: #ifdef COMPAT_NOMID
2102: static int
1.566 christos 2103: exec_nomid(struct lwp *l, struct exec_package *epp)
1.31 cgd 2104: {
1.59 mycroft 2105: int error;
2106: u_long midmag, magic;
2107: u_short mid;
1.80 cgd 2108: struct exec *execp = epp->ep_hdr;
1.31 cgd 2109:
1.80 cgd 2110: /* check on validity of epp->ep_hdr performed by exec_out_makecmds */
2111:
2112: midmag = ntohl(execp->a_midmag);
1.59 mycroft 2113: mid = (midmag >> 16) & 0xffff;
2114: magic = midmag & 0xffff;
2115:
2116: if (magic == 0) {
1.80 cgd 2117: magic = (execp->a_midmag & 0xffff);
1.59 mycroft 2118: mid = MID_ZERO;
2119: }
2120:
2121: midmag = mid << 16 | magic;
2122:
2123: switch (midmag) {
2124: case (MID_ZERO << 16) | ZMAGIC:
2125: /*
2126: * 386BSD's ZMAGIC format:
2127: */
1.566 christos 2128: error = exec_aout_prep_oldzmagic(l, epp);
1.59 mycroft 2129: break;
2130:
2131: case (MID_ZERO << 16) | QMAGIC:
2132: /*
2133: * BSDI's QMAGIC format:
2134: * same as new ZMAGIC format, but with different magic number
2135: */
1.566 christos 2136: error = exec_aout_prep_zmagic(l, epp);
1.59 mycroft 2137: break;
2138:
1.202 christos 2139: case (MID_ZERO << 16) | NMAGIC:
2140: /*
2141: * BSDI's NMAGIC format:
2142: * same as NMAGIC format, but with different magic number
2143: * and with text starting at 0.
2144: */
1.566 christos 2145: error = exec_aout_prep_oldnmagic(l, epp);
1.202 christos 2146: break;
2147:
2148: case (MID_ZERO << 16) | OMAGIC:
2149: /*
2150: * BSDI's OMAGIC format:
2151: * same as OMAGIC format, but with different magic number
2152: * and with text starting at 0.
2153: */
1.566 christos 2154: error = exec_aout_prep_oldomagic(l, epp);
1.202 christos 2155: break;
2156:
1.59 mycroft 2157: default:
2158: error = ENOEXEC;
2159: }
2160:
2161: return error;
1.107 deraadt 2162: }
1.31 cgd 2163: #endif
1.107 deraadt 2164:
2165: /*
2166: * cpu_exec_aout_makecmds():
1.549 wiz 2167: * CPU-dependent a.out format hook for execve().
1.107 deraadt 2168: *
2169: * Determine of the given exec package refers to something which we
2170: * understand and, if so, set up the vmcmds for it.
2171: *
2172: * On the i386, old (386bsd) ZMAGIC binaries and BSDI QMAGIC binaries
2173: * if COMPAT_NOMID is given as a kernel option.
2174: */
2175: int
1.566 christos 2176: cpu_exec_aout_makecmds(struct lwp *l, struct exec_package *epp)
1.107 deraadt 2177: {
2178: int error = ENOEXEC;
2179:
2180: #ifdef COMPAT_NOMID
1.566 christos 2181: if ((error = exec_nomid(l, epp)) == 0)
1.107 deraadt 2182: return error;
1.581 thorpej 2183: #else
2184: (void) l;
2185: (void) epp;
1.107 deraadt 2186: #endif /* ! COMPAT_NOMID */
2187:
2188: return error;
1.31 cgd 2189: }
1.84 cgd 2190:
1.484 fvdl 2191: #include <dev/ic/mc146818reg.h> /* for NVRAM POST */
2192: #include <i386/isa/nvram.h> /* for NVRAM POST */
2193:
1.149 mycroft 2194: void
2195: cpu_reset()
2196: {
1.555 mycroft 2197: struct region_descriptor region;
1.149 mycroft 2198:
1.224 mycroft 2199: disable_intr();
2200:
1.227 mycroft 2201: /*
1.484 fvdl 2202: * Ensure the NVRAM reset byte contains something vaguely sane.
2203: */
2204:
2205: outb(IO_RTC, NVRAM_RESET);
2206: outb(IO_RTC+1, NVRAM_RESET_RST);
2207:
2208: /*
1.564 dyoung 2209: * Reset AMD Geode SC1100.
2210: *
1.565 simonb 2211: * 1) Write PCI Configuration Address Register (0xcf8) to
2212: * select Function 0, Register 0x44: Bridge Configuration,
2213: * GPIO and LPC Configuration Register Space, Reset
2214: * Control Register.
1.564 dyoung 2215: *
1.565 simonb 2216: * 2) Write 0xf to PCI Configuration Data Register (0xcfc)
2217: * to reset IDE controller, IDE bus, and PCI bus, and
2218: * to trigger a system-wide reset.
1.564 dyoung 2219: *
2220: * See AMD Geode SC1100 Processor Data Book, Revision 2.0,
2221: * sections 6.3.1, 6.3.2, and 6.4.1.
2222: */
2223: if (cpu_info_primary.ci_signature == 0x540) {
2224: outl(0xcf8, 0x80009044ul);
2225: outl(0xcfc, 0xf);
1.565 simonb 2226: }
1.564 dyoung 2227:
2228: /*
1.227 mycroft 2229: * The keyboard controller has 4 random output pins, one of which is
2230: * connected to the RESET pin on the CPU in many PCs. We tell the
2231: * keyboard controller to pulse this line a couple of times.
2232: */
1.273 drochner 2233: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2234: delay(100000);
1.273 drochner 2235: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2236: delay(100000);
1.149 mycroft 2237:
2238: /*
1.224 mycroft 2239: * Try to cause a triple fault and watchdog reset by making the IDT
2240: * invalid and causing a fault.
1.149 mycroft 2241: */
1.313 perry 2242: memset((caddr_t)idt, 0, NIDT * sizeof(idt[0]));
1.555 mycroft 2243: setregion(®ion, idt, NIDT * sizeof(idt[0]) - 1);
1.565 simonb 2244: lidt(®ion);
1.567 perry 2245: __asm volatile("divl %0,%1" : : "q" (0), "a" (0));
1.149 mycroft 2246:
1.224 mycroft 2247: #if 0
1.149 mycroft 2248: /*
2249: * Try to cause a triple fault and watchdog reset by unmapping the
1.224 mycroft 2250: * entire address space and doing a TLB flush.
1.149 mycroft 2251: */
1.414 thorpej 2252: memset((caddr_t)PTD, 0, PAGE_SIZE);
1.484 fvdl 2253: tlbflush();
1.224 mycroft 2254: #endif
1.149 mycroft 2255:
2256: for (;;);
1.45 cgd 2257: }
1.484 fvdl 2258:
1.499 jdolecek 2259: void
1.551 junyoung 2260: cpu_getmcontext(struct lwp *l, mcontext_t *mcp, unsigned int *flags)
1.508 thorpej 2261: {
2262: const struct trapframe *tf = l->l_md.md_regs;
2263: __greg_t *gr = mcp->__gregs;
1.519 nathanw 2264: __greg_t ras_eip;
1.508 thorpej 2265:
2266: /* Save register context. */
2267: #ifdef VM86
2268: if (tf->tf_eflags & PSL_VM) {
2269: gr[_REG_GS] = tf->tf_vm86_gs;
2270: gr[_REG_FS] = tf->tf_vm86_fs;
2271: gr[_REG_ES] = tf->tf_vm86_es;
2272: gr[_REG_DS] = tf->tf_vm86_ds;
2273: gr[_REG_EFL] = get_vflags(l);
2274: } else
2275: #endif
2276: {
2277: gr[_REG_GS] = tf->tf_gs;
2278: gr[_REG_FS] = tf->tf_fs;
2279: gr[_REG_ES] = tf->tf_es;
2280: gr[_REG_DS] = tf->tf_ds;
2281: gr[_REG_EFL] = tf->tf_eflags;
2282: }
2283: gr[_REG_EDI] = tf->tf_edi;
2284: gr[_REG_ESI] = tf->tf_esi;
2285: gr[_REG_EBP] = tf->tf_ebp;
2286: gr[_REG_EBX] = tf->tf_ebx;
2287: gr[_REG_EDX] = tf->tf_edx;
2288: gr[_REG_ECX] = tf->tf_ecx;
2289: gr[_REG_EAX] = tf->tf_eax;
2290: gr[_REG_EIP] = tf->tf_eip;
2291: gr[_REG_CS] = tf->tf_cs;
2292: gr[_REG_ESP] = tf->tf_esp;
2293: gr[_REG_UESP] = tf->tf_esp;
2294: gr[_REG_SS] = tf->tf_ss;
2295: gr[_REG_TRAPNO] = tf->tf_trapno;
2296: gr[_REG_ERR] = tf->tf_err;
1.519 nathanw 2297:
2298: if ((ras_eip = (__greg_t)ras_lookup(l->l_proc,
2299: (caddr_t) gr[_REG_EIP])) != -1)
2300: gr[_REG_EIP] = ras_eip;
2301:
1.508 thorpej 2302: *flags |= _UC_CPU;
2303:
2304: /* Save floating point register context, if any. */
1.548 yamt 2305: if ((l->l_md.md_flags & MDL_USEDFPU) != 0) {
1.508 thorpej 2306: #if NNPX > 0
2307: /*
2308: * If this process is the current FP owner, dump its
2309: * context to the PCB first.
2310: * XXX npxsave() also clears the FPU state; depending on the
2311: * XXX application this might be a penalty.
2312: */
2313: if (l->l_addr->u_pcb.pcb_fpcpu) {
2314: npxsave_lwp(l, 1);
2315: }
2316: #endif
2317: if (i386_use_fxsave) {
2318: memcpy(&mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2319: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2320: sizeof (mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm));
2321: *flags |= _UC_FXSAVE;
2322: } else {
2323: memcpy(&mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2324: &l->l_addr->u_pcb.pcb_savefpu.sv_87,
2325: sizeof (mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state));
2326: }
2327: #if 0
2328: /* Apparently nothing ever touches this. */
2329: ucp->mcp.mc_fp.fp_emcsts = l->l_addr->u_pcb.pcb_saveemc;
2330: #endif
2331: *flags |= _UC_FPU;
2332: }
2333: }
2334:
2335: int
1.551 junyoung 2336: cpu_setmcontext(struct lwp *l, const mcontext_t *mcp, unsigned int flags)
1.508 thorpej 2337: {
2338: struct trapframe *tf = l->l_md.md_regs;
1.584 pooka 2339: const __greg_t *gr = mcp->__gregs;
1.508 thorpej 2340:
2341: /* Restore register context, if any. */
2342: if ((flags & _UC_CPU) != 0) {
2343: #ifdef VM86
1.535 drochner 2344: if (gr[_REG_EFL] & PSL_VM) {
1.508 thorpej 2345: tf->tf_vm86_gs = gr[_REG_GS];
2346: tf->tf_vm86_fs = gr[_REG_FS];
2347: tf->tf_vm86_es = gr[_REG_ES];
2348: tf->tf_vm86_ds = gr[_REG_DS];
2349: set_vflags(l, gr[_REG_EFL]);
1.534 christos 2350: if (flags & _UC_VM) {
1.542 junyoung 2351: void syscall_vm86(struct trapframe *);
1.534 christos 2352: l->l_proc->p_md.md_syscall = syscall_vm86;
2353: }
1.508 thorpej 2354: } else
2355: #endif
2356: {
2357: /*
2358: * Check for security violations. If we're returning
2359: * to protected mode, the CPU will validate the segment
2360: * registers automatically and generate a trap on
2361: * violations. We handle the trap, rather than doing
2362: * all of the checking here.
2363: */
1.534 christos 2364: if (((gr[_REG_EFL] ^ tf->tf_eflags) & PSL_USERSTATIC) ||
2365: !USERMODE(gr[_REG_CS], gr[_REG_EFL])) {
2366: printf("cpu_setmcontext error: uc EFL: 0x%08x"
1.535 drochner 2367: " tf EFL: 0x%08x uc CS: 0x%x\n",
1.508 thorpej 2368: gr[_REG_EFL], tf->tf_eflags, gr[_REG_CS]);
2369: return (EINVAL);
2370: }
2371: tf->tf_gs = gr[_REG_GS];
2372: tf->tf_fs = gr[_REG_FS];
2373: tf->tf_es = gr[_REG_ES];
2374: tf->tf_ds = gr[_REG_DS];
2375: /* Only change the user-alterable part of eflags */
2376: tf->tf_eflags &= ~PSL_USER;
2377: tf->tf_eflags |= (gr[_REG_EFL] & PSL_USER);
2378: }
2379: tf->tf_edi = gr[_REG_EDI];
2380: tf->tf_esi = gr[_REG_ESI];
2381: tf->tf_ebp = gr[_REG_EBP];
2382: tf->tf_ebx = gr[_REG_EBX];
2383: tf->tf_edx = gr[_REG_EDX];
2384: tf->tf_ecx = gr[_REG_ECX];
2385: tf->tf_eax = gr[_REG_EAX];
2386: tf->tf_eip = gr[_REG_EIP];
2387: tf->tf_cs = gr[_REG_CS];
2388: tf->tf_esp = gr[_REG_UESP];
2389: tf->tf_ss = gr[_REG_SS];
2390: }
2391:
2392: /* Restore floating point register context, if any. */
2393: if ((flags & _UC_FPU) != 0) {
2394: #if NNPX > 0
2395: /*
2396: * If we were using the FPU, forget that we were.
2397: */
2398: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
2399: npxsave_lwp(l, 0);
2400: #endif
2401: if (flags & _UC_FXSAVE) {
2402: if (i386_use_fxsave) {
2403: memcpy(
2404: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2405: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2406: sizeof (&l->l_addr->u_pcb.pcb_savefpu.sv_xmm));
2407: } else {
2408: /* This is a weird corner case */
2409: process_xmm_to_s87((struct savexmm *)
2410: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2411: &l->l_addr->u_pcb.pcb_savefpu.sv_87);
2412: }
2413: } else {
2414: if (i386_use_fxsave) {
2415: process_s87_to_xmm((struct save87 *)
2416: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2417: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm);
2418: } else {
2419: memcpy(&l->l_addr->u_pcb.pcb_savefpu.sv_87,
2420: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2421: sizeof (l->l_addr->u_pcb.pcb_savefpu.sv_87));
2422: }
2423: }
2424: /* If not set already. */
1.548 yamt 2425: l->l_md.md_flags |= MDL_USEDFPU;
1.508 thorpej 2426: #if 0
2427: /* Apparently unused. */
2428: l->l_addr->u_pcb.pcb_saveemc = mcp->mc_fp.fp_emcsts;
2429: #endif
2430: }
1.534 christos 2431: if (flags & _UC_SETSTACK)
2432: l->l_proc->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
2433: if (flags & _UC_CLRSTACK)
2434: l->l_proc->p_sigctx.ps_sigstk.ss_flags &= ~SS_ONSTACK;
1.508 thorpej 2435: return (0);
2436: }
2437:
2438: void
1.499 jdolecek 2439: cpu_initclocks()
1.484 fvdl 2440: {
1.574 kardel 2441:
1.484 fvdl 2442: (*initclock_func)();
2443: }
2444:
2445: #ifdef MULTIPROCESSOR
1.499 jdolecek 2446: void
2447: need_resched(struct cpu_info *ci)
1.484 fvdl 2448: {
1.546 yamt 2449:
2450: if (ci->ci_want_resched)
2451: return;
2452:
1.484 fvdl 2453: ci->ci_want_resched = 1;
1.508 thorpej 2454: if ((ci)->ci_curlwp != NULL)
2455: aston((ci)->ci_curlwp->l_proc);
1.546 yamt 2456: else if (ci != curcpu())
2457: x86_send_ipi(ci, 0);
1.484 fvdl 2458: }
2459: #endif
2460:
2461: /*
2462: * Allocate an IDT vector slot within the given range.
2463: * XXX needs locking to avoid MP allocation races.
2464: */
2465:
2466: int
1.551 junyoung 2467: idt_vec_alloc(int low, int high)
1.484 fvdl 2468: {
2469: int vec;
2470:
1.498 fvdl 2471: simple_lock(&idt_lock);
2472: for (vec = low; vec <= high; vec++) {
2473: if (idt_allocmap[vec] == 0) {
2474: idt_allocmap[vec] = 1;
2475: simple_unlock(&idt_lock);
1.484 fvdl 2476: return vec;
1.498 fvdl 2477: }
2478: }
2479: simple_unlock(&idt_lock);
1.484 fvdl 2480: return 0;
2481: }
2482:
1.498 fvdl 2483: void
1.551 junyoung 2484: idt_vec_set(int vec, void (*function)(void))
1.484 fvdl 2485: {
1.498 fvdl 2486: /*
2487: * Vector should be allocated, so no locking needed.
2488: */
2489: KASSERT(idt_allocmap[vec] == 1);
1.516 fvdl 2490: setgate(&idt[vec], function, 0, SDT_SYS386IGT, SEL_KPL,
1.489 fvdl 2491: GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 2492: }
2493:
2494: void
1.551 junyoung 2495: idt_vec_free(int vec)
1.484 fvdl 2496: {
1.498 fvdl 2497: simple_lock(&idt_lock);
1.516 fvdl 2498: unsetgate(&idt[vec]);
1.498 fvdl 2499: idt_allocmap[vec] = 0;
2500: simple_unlock(&idt_lock);
1.507 jdolecek 2501: }
2502:
2503: /*
2504: * Number of processes is limited by number of available GDT slots.
2505: */
2506: int
2507: cpu_maxproc(void)
2508: {
2509: #ifdef USER_LDT
2510: return ((MAXGDTSIZ - NGDT) / 2);
2511: #else
2512: return (MAXGDTSIZ - NGDT);
2513: #endif
1.484 fvdl 2514: }
CVSweb <webmaster@jp.NetBSD.org>