Annotation of src/sys/arch/i386/i386/machdep.c, Revision 1.586.2.6
1.586.2.6! bouyer 1: /* $NetBSD: machdep.c,v 1.586.2.5 2007/08/28 11:46:26 liamjfoy 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>
1.586.2.6! bouyer 75: __KERNEL_RCSID(0, "$NetBSD: machdep.c,v 1.586.2.5 2007/08/28 11:46:26 liamjfoy 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:
1.586.2.6! bouyer 813: #if NNPX > 0
! 814: /* make sure we get a clean FPU */
! 815: npxsave_lwp(l, 0);
! 816: l->l_md.md_flags &= ~MDL_USEDFPU;
! 817: #endif
! 818:
1.319 mycroft 819: /* Remember that we're now on the signal stack. */
820: if (onstack)
1.425 jdolecek 821: p->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
1.1 cgd 822: }
823:
1.534 christos 824: void
1.539 christos 825: sendsig(const ksiginfo_t *ksi, const sigset_t *mask)
1.534 christos 826: {
827: #ifdef COMPAT_16
828: if (curproc->p_sigacts->sa_sigdesc[ksi->ksi_signo].sd_vers < 2)
829: sendsig_sigcontext(ksi, mask);
830: else
831: #endif
832: sendsig_siginfo(ksi, mask);
833: }
1.508 thorpej 834:
1.550 junyoung 835: void
1.534 christos 836: cpu_upcall(struct lwp *l, int type, int nevents, int ninterrupted, void *sas,
837: void *ap, void *sp, sa_upcall_t upcall)
1.508 thorpej 838: {
1.533 chs 839: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 840: struct saframe *sf, frame;
841: struct trapframe *tf;
842:
843: tf = l->l_md.md_regs;
844:
845: /* Finally, copy out the rest of the frame. */
846: frame.sa_type = type;
847: frame.sa_sas = sas;
848: frame.sa_events = nevents;
849: frame.sa_interrupted = ninterrupted;
850: frame.sa_arg = ap;
851: frame.sa_ra = 0;
1.550 junyoung 852:
1.508 thorpej 853: sf = (struct saframe *)sp - 1;
854: if (copyout(&frame, sf, sizeof(frame)) != 0) {
855: /* Copying onto the stack didn't work. Die. */
856: sigexit(l, SIGILL);
857: /* NOTREACHED */
858: }
859:
860: tf->tf_eip = (int) upcall;
861: tf->tf_esp = (int) sf;
862: tf->tf_ebp = 0; /* indicate call-frame-top to debuggers */
863: tf->tf_gs = GSEL(GUDATA_SEL, SEL_UPL);
864: tf->tf_fs = GSEL(GUDATA_SEL, SEL_UPL);
865: tf->tf_es = GSEL(GUDATA_SEL, SEL_UPL);
866: tf->tf_ds = GSEL(GUDATA_SEL, SEL_UPL);
1.533 chs 867: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
868: GSEL(GUCODEBIG_SEL, SEL_UPL) : GSEL(GUCODE_SEL, SEL_UPL);
1.508 thorpej 869: tf->tf_ss = GSEL(GUDATA_SEL, SEL_UPL);
870: tf->tf_eflags &= ~(PSL_T|PSL_VM|PSL_AC);
871: }
1.37 cgd 872:
1.1 cgd 873: int waittime = -1;
874: struct pcb dumppcb;
875:
1.32 andrew 876: void
1.586 christos 877: cpu_reboot(int howto, char *bootstr)
1.1 cgd 878: {
879:
1.106 mycroft 880: if (cold) {
1.193 mycroft 881: howto |= RB_HALT;
882: goto haltsys;
1.1 cgd 883: }
1.193 mycroft 884:
1.106 mycroft 885: boothowto = howto;
1.193 mycroft 886: if ((howto & RB_NOSYNC) == 0 && waittime < 0) {
1.1 cgd 887: waittime = 0;
1.150 mycroft 888: vfs_shutdown();
1.521 dsl 889: /*
890: * If we've been adjusting the clock, the todr
891: * will be out of synch; adjust it now.
892: */
893: if (time_adjusted != 0)
894: resettodr();
1.1 cgd 895: }
1.193 mycroft 896:
897: /* Disable interrupts. */
1.1 cgd 898: splhigh();
1.193 mycroft 899:
900: /* Do a dump if requested. */
901: if ((howto & (RB_DUMP | RB_HALT)) == RB_DUMP)
902: dumpsys();
903:
904: haltsys:
905: doshutdownhooks();
906:
1.484 fvdl 907: #ifdef MULTIPROCESSOR
1.514 fvdl 908: x86_broadcast_ipi(X86_IPI_HALT);
1.484 fvdl 909: #endif
910:
1.307 thorpej 911: if ((howto & RB_POWERDOWN) == RB_POWERDOWN) {
1.473 tshiozak 912: #if NACPI > 0
1.530 kochi 913: if (acpi_softc != NULL) {
914: delay(500000);
915: acpi_enter_sleep_state(acpi_softc, ACPI_STATE_S5);
916: printf("WARNING: ACPI powerdown failed!\n");
917: }
1.473 tshiozak 918: #endif
1.576 christos 919: #if NAPMBIOS > 0 && !defined(APM_NO_POWEROFF)
1.208 jtk 920: /* turn off, if we can. But try to turn disk off and
921: * wait a bit first--some disk drives are slow to clean up
922: * and users have reported disk corruption.
923: */
924: delay(500000);
1.578 xtraeme 925: apm_set_powstate(NULL, APM_DEV_DISK(APM_DEV_ALLUNITS), APM_SYS_OFF);
1.208 jtk 926: delay(500000);
1.576 christos 927: apm_set_powstate(NULL, APM_DEV_ALLDEVS, APM_SYS_OFF);
1.531 kochi 928: printf("WARNING: APM powerdown failed!\n");
1.307 thorpej 929: /*
930: * RB_POWERDOWN implies RB_HALT... fall into it...
931: */
1.208 jtk 932: #endif
1.307 thorpej 933: }
934:
935: if (howto & RB_HALT) {
1.210 christos 936: printf("\n");
937: printf("The operating system has halted.\n");
938: printf("Please press any key to reboot.\n\n");
1.517 jmmv 939:
940: #ifdef BEEP_ONHALT
941: {
942: int c;
943: for (c = BEEP_ONHALT_COUNT; c > 0; c--) {
944: sysbeep(BEEP_ONHALT_PITCH,
1.565 simonb 945: BEEP_ONHALT_PERIOD * hz / 1000);
1.517 jmmv 946: delay(BEEP_ONHALT_PERIOD * 1000);
947: sysbeep(0, BEEP_ONHALT_PERIOD * hz / 1000);
948: delay(BEEP_ONHALT_PERIOD * 1000);
949: }
950: }
951: #endif
952:
1.300 drochner 953: cnpollc(1); /* for proper keyboard command handling */
1.512 perry 954: if (cngetc() == 0) {
955: /* no console attached, so just hlt */
956: for(;;) {
1.567 perry 957: __asm volatile("hlt");
1.512 perry 958: }
959: }
1.300 drochner 960: cnpollc(0);
1.1 cgd 961: }
1.193 mycroft 962:
1.210 christos 963: printf("rebooting...\n");
1.328 bouyer 964: if (cpureset_delay > 0)
965: delay(cpureset_delay * 1000);
1.1 cgd 966: cpu_reset();
967: for(;;) ;
968: /*NOTREACHED*/
969: }
970:
1.116 gwr 971: /*
972: * These variables are needed by /sbin/savecore
973: */
1.568 perry 974: uint32_t dumpmag = 0x8fca0101; /* magic number */
1.116 gwr 975: int dumpsize = 0; /* pages */
976: long dumplo = 0; /* blocks */
977:
978: /*
1.291 thorpej 979: * cpu_dumpsize: calculate size of machine-dependent kernel core dump headers.
980: */
981: int
982: cpu_dumpsize()
983: {
984: int size;
985:
986: size = ALIGN(sizeof(kcore_seg_t)) + ALIGN(sizeof(cpu_kcore_hdr_t)) +
987: ALIGN(mem_cluster_cnt * sizeof(phys_ram_seg_t));
988: if (roundup(size, dbtob(1)) != dbtob(1))
989: return (-1);
990:
991: return (1);
992: }
993:
994: /*
995: * cpu_dump_mempagecnt: calculate the size of RAM (in pages) to be dumped.
996: */
997: u_long
998: cpu_dump_mempagecnt()
999: {
1000: u_long i, n;
1001:
1002: n = 0;
1003: for (i = 0; i < mem_cluster_cnt; i++)
1004: n += atop(mem_clusters[i].size);
1005: return (n);
1006: }
1007:
1008: /*
1009: * cpu_dump: dump the machine-dependent kernel core dump headers.
1010: */
1011: int
1012: cpu_dump()
1013: {
1.542 junyoung 1014: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.563 christos 1015: char bf[dbtob(1)];
1.291 thorpej 1016: kcore_seg_t *segp;
1017: cpu_kcore_hdr_t *cpuhdrp;
1018: phys_ram_seg_t *memsegp;
1.481 gehenna 1019: const struct bdevsw *bdev;
1.291 thorpej 1020: int i;
1021:
1.481 gehenna 1022: bdev = bdevsw_lookup(dumpdev);
1023: if (bdev == NULL)
1024: return (ENXIO);
1025: dump = bdev->d_dump;
1.291 thorpej 1026:
1.563 christos 1027: memset(bf, 0, sizeof bf);
1028: segp = (kcore_seg_t *)bf;
1029: cpuhdrp = (cpu_kcore_hdr_t *)&bf[ALIGN(sizeof(*segp))];
1030: memsegp = (phys_ram_seg_t *)&bf[ ALIGN(sizeof(*segp)) +
1.291 thorpej 1031: ALIGN(sizeof(*cpuhdrp))];
1032:
1033: /*
1034: * Generate a segment header.
1035: */
1036: CORE_SETMAGIC(*segp, KCORE_MAGIC, MID_MACHINE, CORE_CPU);
1037: segp->c_size = dbtob(1) - ALIGN(sizeof(*segp));
1038:
1039: /*
1040: * Add the machine-dependent header info.
1041: */
1.558 junyoung 1042: cpuhdrp->pdppaddr = PDPpaddr;
1.291 thorpej 1043: cpuhdrp->nmemsegs = mem_cluster_cnt;
1044:
1045: /*
1046: * Fill in the memory segment descriptors.
1047: */
1048: for (i = 0; i < mem_cluster_cnt; i++) {
1049: memsegp[i].start = mem_clusters[i].start;
1050: memsegp[i].size = mem_clusters[i].size;
1051: }
1052:
1.563 christos 1053: return (dump(dumpdev, dumplo, (caddr_t)bf, dbtob(1)));
1.291 thorpej 1054: }
1055:
1056: /*
1.228 gwr 1057: * This is called by main to set dumplo and dumpsize.
1.414 thorpej 1058: * Dumps always skip the first PAGE_SIZE of disk space
1.116 gwr 1059: * in case there might be a disk label stored there.
1060: * If there is extra space, put dump at the end to
1061: * reduce the chance that swapping trashes it.
1062: */
1063: void
1.228 gwr 1064: cpu_dumpconf()
1.116 gwr 1065: {
1.481 gehenna 1066: const struct bdevsw *bdev;
1.291 thorpej 1067: int nblks, dumpblks; /* size of dump area */
1.116 gwr 1068:
1069: if (dumpdev == NODEV)
1.291 thorpej 1070: goto bad;
1.481 gehenna 1071: bdev = bdevsw_lookup(dumpdev);
1.582 mrg 1072: if (bdev == NULL) {
1073: dumpdev = NODEV;
1.583 skrll 1074: goto bad;
1.582 mrg 1075: }
1.481 gehenna 1076: if (bdev->d_psize == NULL)
1.291 thorpej 1077: goto bad;
1.481 gehenna 1078: nblks = (*bdev->d_psize)(dumpdev);
1.116 gwr 1079: if (nblks <= ctod(1))
1.291 thorpej 1080: goto bad;
1.116 gwr 1081:
1.291 thorpej 1082: dumpblks = cpu_dumpsize();
1083: if (dumpblks < 0)
1084: goto bad;
1085: dumpblks += ctod(cpu_dump_mempagecnt());
1086:
1087: /* If dump won't fit (incl. room for possible label), punt. */
1088: if (dumpblks > (nblks - ctod(1)))
1089: goto bad;
1090:
1091: /* Put dump at end of partition */
1092: dumplo = nblks - dumpblks;
1093:
1094: /* dumpsize is in page units, and doesn't include headers. */
1095: dumpsize = cpu_dump_mempagecnt();
1096: return;
1.116 gwr 1097:
1.291 thorpej 1098: bad:
1099: dumpsize = 0;
1.116 gwr 1100: }
1101:
1.1 cgd 1102: /*
1103: * Doadump comes here after turning off memory management and
1104: * getting on the dump stack, either when called above, or by
1105: * the auto-restart code.
1106: */
1.414 thorpej 1107: #define BYTES_PER_DUMP PAGE_SIZE /* must be a multiple of pagesize XXX small */
1.314 thorpej 1108: static vaddr_t dumpspace;
1.163 cgd 1109:
1.314 thorpej 1110: vaddr_t
1.551 junyoung 1111: reserve_dumppages(vaddr_t p)
1.163 cgd 1112: {
1113:
1114: dumpspace = p;
1115: return (p + BYTES_PER_DUMP);
1116: }
1117:
1.32 andrew 1118: void
1.1 cgd 1119: dumpsys()
1120: {
1.556 mycroft 1121: u_long totalbytesleft, bytes, i, n, m, memseg;
1.291 thorpej 1122: u_long maddr;
1123: int psize;
1.163 cgd 1124: daddr_t blkno;
1.481 gehenna 1125: const struct bdevsw *bdev;
1.542 junyoung 1126: int (*dump)(dev_t, daddr_t, caddr_t, size_t);
1.200 christos 1127: int error;
1.193 mycroft 1128:
1129: /* Save registers. */
1130: savectx(&dumppcb);
1.1 cgd 1131:
1132: if (dumpdev == NODEV)
1133: return;
1.484 fvdl 1134:
1.481 gehenna 1135: bdev = bdevsw_lookup(dumpdev);
1136: if (bdev == NULL || bdev->d_psize == NULL)
1137: return;
1.163 cgd 1138:
1139: /*
1140: * For dumps during autoconfiguration,
1141: * if dump device has already configured...
1142: */
1143: if (dumpsize == 0)
1.228 gwr 1144: cpu_dumpconf();
1.330 jtk 1145: if (dumplo <= 0 || dumpsize == 0) {
1.275 mycroft 1146: printf("\ndump to dev %u,%u not possible\n", major(dumpdev),
1147: minor(dumpdev));
1.163 cgd 1148: return;
1.275 mycroft 1149: }
1150: printf("\ndumping to dev %u,%u offset %ld\n", major(dumpdev),
1151: minor(dumpdev), dumplo);
1.134 mycroft 1152:
1.481 gehenna 1153: psize = (*bdev->d_psize)(dumpdev);
1.210 christos 1154: printf("dump ");
1.163 cgd 1155: if (psize == -1) {
1.210 christos 1156: printf("area unavailable\n");
1.163 cgd 1157: return;
1158: }
1159:
1160: #if 0 /* XXX this doesn't work. grr. */
1.565 simonb 1161: /* toss any characters present prior to dump */
1.163 cgd 1162: while (sget() != NULL); /*syscons and pccons differ */
1163: #endif
1164:
1.291 thorpej 1165: if ((error = cpu_dump()) != 0)
1166: goto err;
1167:
1168: totalbytesleft = ptoa(cpu_dump_mempagecnt());
1169: blkno = dumplo + cpu_dumpsize();
1.481 gehenna 1170: dump = bdev->d_dump;
1.200 christos 1171: error = 0;
1.291 thorpej 1172:
1173: for (memseg = 0; memseg < mem_cluster_cnt; memseg++) {
1174: maddr = mem_clusters[memseg].start;
1175: bytes = mem_clusters[memseg].size;
1176:
1177: for (i = 0; i < bytes; i += n, totalbytesleft -= n) {
1178: /* Print out how many MBs we have left to go. */
1179: if ((totalbytesleft % (1024*1024)) == 0)
1180: printf("%ld ", totalbytesleft / (1024 * 1024));
1181:
1182: /* Limit size for next transfer. */
1183: n = bytes - i;
1184: if (n > BYTES_PER_DUMP)
1185: n = BYTES_PER_DUMP;
1186:
1.556 mycroft 1187: for (m = 0; m < n; m += NBPG)
1188: pmap_kenter_pa(dumpspace + m, maddr + m,
1189: VM_PROT_READ);
1.291 thorpej 1190:
1191: error = (*dump)(dumpdev, blkno, (caddr_t)dumpspace, n);
1192: if (error)
1193: goto err;
1.163 cgd 1194: maddr += n;
1.291 thorpej 1195: blkno += btodb(n); /* XXX? */
1.163 cgd 1196:
1197: #if 0 /* XXX this doesn't work. grr. */
1.291 thorpej 1198: /* operator aborting dump? */
1199: if (sget() != NULL) {
1200: error = EINTR;
1201: break;
1202: }
1203: #endif
1.163 cgd 1204: }
1205: }
1206:
1.291 thorpej 1207: err:
1.163 cgd 1208: switch (error) {
1.1 cgd 1209:
1210: case ENXIO:
1.210 christos 1211: printf("device bad\n");
1.1 cgd 1212: break;
1213:
1214: case EFAULT:
1.210 christos 1215: printf("device not ready\n");
1.1 cgd 1216: break;
1217:
1218: case EINVAL:
1.210 christos 1219: printf("area improper\n");
1.1 cgd 1220: break;
1221:
1222: case EIO:
1.210 christos 1223: printf("i/o error\n");
1.1 cgd 1224: break;
1225:
1226: case EINTR:
1.210 christos 1227: printf("aborted from console\n");
1.1 cgd 1228: break;
1229:
1.163 cgd 1230: case 0:
1.210 christos 1231: printf("succeeded\n");
1.163 cgd 1232: break;
1233:
1.1 cgd 1234: default:
1.210 christos 1235: printf("error %d\n", error);
1.1 cgd 1236: break;
1237: }
1.210 christos 1238: printf("\n\n");
1.163 cgd 1239: delay(5000000); /* 5 seconds */
1.1 cgd 1240: }
1241:
1242: /*
1243: * Clear registers on exec
1244: */
1.33 cgd 1245: void
1.551 junyoung 1246: setregs(struct lwp *l, struct exec_package *pack, u_long stack)
1.1 cgd 1247: {
1.533 chs 1248: struct pmap *pmap = vm_map_pmap(&l->l_proc->p_vmspace->vm_map);
1.508 thorpej 1249: struct pcb *pcb = &l->l_addr->u_pcb;
1.298 mycroft 1250: struct trapframe *tf;
1.1 cgd 1251:
1.161 mycroft 1252: #if NNPX > 0
1253: /* If we were using the FPU, forget about it. */
1.508 thorpej 1254: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
1255: npxsave_lwp(l, 0);
1.161 mycroft 1256: #endif
1.166 mycroft 1257:
1.178 mycroft 1258: #ifdef USER_LDT
1.508 thorpej 1259: pmap_ldt_cleanup(l);
1.178 mycroft 1260: #endif
1261:
1.548 yamt 1262: l->l_md.md_flags &= ~MDL_USEDFPU;
1.452 thorpej 1263: if (i386_use_fxsave) {
1.450 thorpej 1264: pcb->pcb_savefpu.sv_xmm.sv_env.en_cw = __NetBSD_NPXCW__;
1.452 thorpej 1265: pcb->pcb_savefpu.sv_xmm.sv_env.en_mxcsr = __INITIAL_MXCSR__;
1266: } else
1.450 thorpej 1267: pcb->pcb_savefpu.sv_87.sv_env.en_cw = __NetBSD_NPXCW__;
1.59 mycroft 1268:
1.508 thorpej 1269: tf = l->l_md.md_regs;
1.445 sommerfe 1270: tf->tf_gs = LSEL(LUDATA_SEL, SEL_UPL);
1271: tf->tf_fs = LSEL(LUDATA_SEL, SEL_UPL);
1.154 mycroft 1272: tf->tf_es = LSEL(LUDATA_SEL, SEL_UPL);
1273: tf->tf_ds = LSEL(LUDATA_SEL, SEL_UPL);
1.252 mycroft 1274: tf->tf_edi = 0;
1275: tf->tf_esi = 0;
1.154 mycroft 1276: tf->tf_ebp = 0;
1.508 thorpej 1277: tf->tf_ebx = (int)l->l_proc->p_psstr;
1.252 mycroft 1278: tf->tf_edx = 0;
1279: tf->tf_ecx = 0;
1280: tf->tf_eax = 0;
1.154 mycroft 1281: tf->tf_eip = pack->ep_entry;
1.533 chs 1282: tf->tf_cs = pmap->pm_hiexec > I386_MAX_EXE_ADDR ?
1283: LSEL(LUCODEBIG_SEL, SEL_UPL) : LSEL(LUCODE_SEL, SEL_UPL);
1.154 mycroft 1284: tf->tf_eflags = PSL_USERSET;
1285: tf->tf_esp = stack;
1286: tf->tf_ss = LSEL(LUDATA_SEL, SEL_UPL);
1.1 cgd 1287: }
1288:
1289: /*
1.55 brezak 1290: * Initialize segments and descriptor tables
1.1 cgd 1291: */
1292:
1.516 fvdl 1293: union descriptor *gdt, *ldt;
1294: struct gate_descriptor *idt;
1.498 fvdl 1295: char idt_allocmap[NIDT];
1296: struct simplelock idt_lock = SIMPLELOCK_INITIALIZER;
1.275 mycroft 1297: #ifdef I586_CPU
1298: union descriptor *pentium_idt;
1299: #endif
1.573 yamt 1300: struct user *proc0paddr;
1301: extern vaddr_t proc0uarea;
1.49 brezak 1302:
1.178 mycroft 1303: void
1.551 junyoung 1304: setgate(struct gate_descriptor *gd, void *func, int args, int type, int dpl,
1305: int sel)
1.178 mycroft 1306: {
1.1 cgd 1307:
1.178 mycroft 1308: gd->gd_looffset = (int)func;
1.489 fvdl 1309: gd->gd_selector = sel;
1.178 mycroft 1310: gd->gd_stkcpy = args;
1311: gd->gd_xx = 0;
1312: gd->gd_type = type;
1313: gd->gd_dpl = dpl;
1314: gd->gd_p = 1;
1315: gd->gd_hioffset = (int)func >> 16;
1316: }
1317:
1318: void
1.551 junyoung 1319: unsetgate(struct gate_descriptor *gd)
1.484 fvdl 1320: {
1321: gd->gd_p = 0;
1322: gd->gd_hioffset = 0;
1323: gd->gd_looffset = 0;
1324: gd->gd_selector = 0;
1325: gd->gd_xx = 0;
1326: gd->gd_stkcpy = 0;
1327: gd->gd_type = 0;
1328: gd->gd_dpl = 0;
1329: }
1330:
1331:
1332: void
1.551 junyoung 1333: setregion(struct region_descriptor *rd, void *base, size_t limit)
1.178 mycroft 1334: {
1335:
1336: rd->rd_limit = (int)limit;
1337: rd->rd_base = (int)base;
1338: }
1.1 cgd 1339:
1.174 mycroft 1340: void
1.551 junyoung 1341: setsegment(struct segment_descriptor *sd, void *base, size_t limit, int type,
1342: int dpl, int def32, int gran)
1.174 mycroft 1343: {
1.1 cgd 1344:
1.174 mycroft 1345: sd->sd_lolimit = (int)limit;
1346: sd->sd_lobase = (int)base;
1347: sd->sd_type = type;
1348: sd->sd_dpl = dpl;
1349: sd->sd_p = 1;
1350: sd->sd_hilimit = (int)limit >> 16;
1351: sd->sd_xx = 0;
1352: sd->sd_def32 = def32;
1353: sd->sd_gran = gran;
1354: sd->sd_hibase = (int)base >> 24;
1355: }
1.1 cgd 1356:
1357: #define IDTVEC(name) __CONCAT(X, name)
1.542 junyoung 1358: typedef void (vector)(void);
1.299 mycroft 1359: extern vector IDTVEC(syscall);
1360: extern vector IDTVEC(osyscall);
1361: extern vector *IDTVEC(exceptions)[];
1.333 christos 1362: #ifdef COMPAT_SVR4
1363: extern vector IDTVEC(svr4_fasttrap);
1364: #endif /* COMPAT_SVR4 */
1.447 christos 1365: #ifdef COMPAT_MACH
1366: extern vector IDTVEC(mach_trap);
1367: #endif
1.1 cgd 1368:
1.381 thorpej 1369: #define KBTOB(x) ((size_t)(x) * 1024UL)
1370:
1.484 fvdl 1371: void cpu_init_idt()
1372: {
1373: struct region_descriptor region;
1374: #ifdef I586_CPU
1375: setregion(®ion, pentium_idt, NIDT * sizeof(idt[0]) - 1);
1376: #else
1377: setregion(®ion, idt, NIDT * sizeof(idt[0]) - 1);
1378: #endif
1.565 simonb 1379: lidt(®ion);
1.484 fvdl 1380: }
1381:
1.433 kanaoka 1382: void
1.568 perry 1383: add_mem_cluster(uint64_t seg_start, uint64_t seg_end, uint32_t type)
1.433 kanaoka 1384: {
1385: extern struct extent *iomem_ex;
1.492 kanaoka 1386: int i;
1.433 kanaoka 1387:
1388: if (seg_end > 0x100000000ULL) {
1389: printf("WARNING: skipping large "
1390: "memory map entry: "
1391: "0x%qx/0x%qx/0x%x\n",
1392: seg_start,
1393: (seg_end - seg_start),
1394: type);
1395: return;
1396: }
1397:
1398: /*
1399: * XXX Chop the last page off the size so that
1400: * XXX it can fit in avail_end.
1401: */
1402: if (seg_end == 0x100000000ULL)
1403: seg_end -= PAGE_SIZE;
1404:
1405: if (seg_end <= seg_start)
1406: return;
1407:
1.492 kanaoka 1408: for (i = 0; i < mem_cluster_cnt; i++) {
1409: if ((mem_clusters[i].start == round_page(seg_start))
1410: && (mem_clusters[i].size
1411: == trunc_page(seg_end) - mem_clusters[i].start)) {
1412: #ifdef DEBUG_MEMLOAD
1413: printf("WARNING: skipping duplicate segment entry\n");
1414: #endif
1415: return;
1416: }
1417: }
1418:
1.433 kanaoka 1419: /*
1420: * Allocate the physical addresses used by RAM
1421: * from the iomem extent map. This is done before
1422: * the addresses are page rounded just to make
1423: * sure we get them all.
1424: */
1425: if (extent_alloc_region(iomem_ex, seg_start,
1426: seg_end - seg_start, EX_NOWAIT)) {
1427: /* XXX What should we do? */
1428: printf("WARNING: CAN'T ALLOCATE "
1429: "MEMORY SEGMENT "
1430: "(0x%qx/0x%qx/0x%x) FROM "
1431: "IOMEM EXTENT MAP!\n",
1432: seg_start, seg_end - seg_start, type);
1.492 kanaoka 1433: return;
1.433 kanaoka 1434: }
1435:
1436: /*
1437: * If it's not free memory, skip it.
1438: */
1439: if (type != BIM_Memory)
1440: return;
1441:
1442: /* XXX XXX XXX */
1443: if (mem_cluster_cnt >= VM_PHYSSEG_MAX)
1.579 cube 1444: panic("init386: too many memory segments "
1445: "(increase VM_PHYSSEG_MAX)");
1.433 kanaoka 1446:
1447: seg_start = round_page(seg_start);
1448: seg_end = trunc_page(seg_end);
1449:
1450: if (seg_start == seg_end)
1451: return;
1452:
1453: mem_clusters[mem_cluster_cnt].start = seg_start;
1454: mem_clusters[mem_cluster_cnt].size =
1455: seg_end - seg_start;
1456:
1457: if (avail_end < seg_end)
1458: avail_end = seg_end;
1459: physmem += atop(mem_clusters[mem_cluster_cnt].size);
1460: mem_cluster_cnt++;
1461: }
1462:
1.59 mycroft 1463: void
1.484 fvdl 1464: initgdt(union descriptor *tgdt)
1465: {
1466: struct region_descriptor region;
1467: gdt = tgdt;
1468: memset(gdt, 0, NGDT*sizeof(*gdt));
1469: /* make gdt gates and memory segments */
1470: setsegment(&gdt[GCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 1, 1);
1471: setsegment(&gdt[GDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 1, 1);
1.533 chs 1472: setsegment(&gdt[GUCODE_SEL].sd, 0, x86_btop(I386_MAX_EXE_ADDR) - 1,
1473: SDT_MEMERA, SEL_UPL, 1, 1);
1474: setsegment(&gdt[GUCODEBIG_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1475: SDT_MEMERA, SEL_UPL, 1, 1);
1.514 fvdl 1476: setsegment(&gdt[GUDATA_SEL].sd, 0, x86_btop(VM_MAXUSER_ADDRESS) - 1,
1.484 fvdl 1477: SDT_MEMRWA, SEL_UPL, 1, 1);
1478: #ifdef COMPAT_MACH
1479: setgate(&gdt[GMACHCALLS_SEL].gd, &IDTVEC(mach_trap), 1,
1.489 fvdl 1480: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 1481: #endif
1482: #if NBIOSCALL > 0
1483: /* bios trampoline GDT entries */
1484: setsegment(&gdt[GBIOSCODE_SEL].sd, 0, 0xfffff, SDT_MEMERA, SEL_KPL, 0,
1485: 0);
1486: setsegment(&gdt[GBIOSDATA_SEL].sd, 0, 0xfffff, SDT_MEMRWA, SEL_KPL, 0,
1487: 0);
1488: #endif
1489: setsegment(&gdt[GCPU_SEL].sd, &cpu_info_primary,
1490: sizeof(struct cpu_info)-1, SDT_MEMRWA, SEL_KPL, 1, 1);
1491:
1492: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
1493: lgdt(®ion);
1494: }
1495:
1496: void
1.551 junyoung 1497: init386(paddr_t first_avail)
1.2 cgd 1498: {
1.484 fvdl 1499: union descriptor *tgdt;
1.542 junyoung 1500: extern void consinit(void);
1.375 drochner 1501: extern struct extent *iomem_ex;
1.401 thorpej 1502: struct btinfo_memmap *bim;
1.381 thorpej 1503: struct region_descriptor region;
1.401 thorpej 1504: int x, first16q;
1.568 perry 1505: uint64_t seg_start, seg_end;
1506: uint64_t seg_start1, seg_end1;
1.473 tshiozak 1507: paddr_t realmode_reserved_start;
1508: psize_t realmode_reserved_size;
1509: int needs_earlier_install_pte0;
1.436 jdolecek 1510: #if NBIOSCALL > 0
1511: extern int biostramp_image_size;
1512: extern u_char biostramp_image[];
1513: #endif
1.1 cgd 1514:
1.484 fvdl 1515: cpu_probe_features(&cpu_info_primary);
1516: cpu_feature = cpu_info_primary.ci_feature_flags;
1.553 lukem 1517: cpu_feature2 = cpu_info_primary.ci_feature2_flags;
1.484 fvdl 1518:
1.573 yamt 1519: proc0paddr = UAREA_TO_USER(proc0uarea);
1.508 thorpej 1520: lwp0.l_addr = proc0paddr;
1521: cpu_info_primary.ci_curpcb = &lwp0.l_addr->u_pcb;
1.275 mycroft 1522:
1.514 fvdl 1523: x86_bus_space_init();
1.84 cgd 1524: consinit(); /* XXX SHOULD NOT BE DONE HERE */
1.401 thorpej 1525: /*
1526: * Initailize PAGE_SIZE-dependent variables.
1527: */
1528: uvm_setpagesize();
1.450 thorpej 1529:
1530: /*
1531: * Saving SSE registers won't work if the save area isn't
1532: * 16-byte aligned.
1533: */
1534: if (offsetof(struct user, u_pcb.pcb_savefpu) & 0xf)
1535: panic("init386: pcb_savefpu not 16-byte aligned");
1.440 thorpej 1536:
1537: /*
1538: * Start with 2 color bins -- this is just a guess to get us
1539: * started. We'll recolor when we determine the largest cache
1540: * sizes on the system.
1541: */
1542: uvmexp.ncolors = 2;
1.401 thorpej 1543:
1.473 tshiozak 1544: /*
1.484 fvdl 1545: * BIOS leaves data in physical page 0
1546: * Even if it didn't, our VM system doesn't like using zero as a
1547: * physical page number.
1548: * We may also need pages in low memory (one each) for secondary CPU
1549: * startup, for BIOS calls, and for ACPI, plus a page table page to map
1550: * them into the first few pages of the kernel's pmap.
1.473 tshiozak 1551: */
1552: avail_start = PAGE_SIZE;
1553:
1554: /*
1555: * reserve memory for real-mode call
1556: */
1557: needs_earlier_install_pte0 = 0;
1558: realmode_reserved_start = 0;
1559: realmode_reserved_size = 0;
1.414 thorpej 1560: #if NBIOSCALL > 0
1.473 tshiozak 1561: /* save us a page for trampoline code */
1562: realmode_reserved_size += PAGE_SIZE;
1563: needs_earlier_install_pte0 = 1;
1564: #endif
1.484 fvdl 1565: #ifdef MULTIPROCESSOR /* XXX */
1566: KASSERT(avail_start == PAGE_SIZE); /* XXX */
1567: if (realmode_reserved_size < MP_TRAMPOLINE) /* XXX */
1568: realmode_reserved_size = MP_TRAMPOLINE; /* XXX */
1569: needs_earlier_install_pte0 = 1; /* XXX */
1570: #endif /* XXX */
1.473 tshiozak 1571: #if NACPI > 0
1572: /* trampoline code for wake handler */
1573: realmode_reserved_size += ptoa(acpi_md_get_npages_of_wakecode()+1);
1574: needs_earlier_install_pte0 = 1;
1575: #endif
1576: if (needs_earlier_install_pte0) {
1577: /* page table for directory entry 0 */
1578: realmode_reserved_size += PAGE_SIZE;
1579: }
1580: if (realmode_reserved_size>0) {
1581: realmode_reserved_start = avail_start;
1582: avail_start += realmode_reserved_size;
1583: }
1.414 thorpej 1584:
1.492 kanaoka 1585: #ifdef DEBUG_MEMLOAD
1586: printf("mem_cluster_count: %d\n", mem_cluster_cnt);
1587: #endif
1588:
1.401 thorpej 1589: /*
1590: * Call pmap initialization to make new kernel address space.
1591: * We must do this before loading pages into the VM system.
1592: */
1.314 thorpej 1593: pmap_bootstrap((vaddr_t)atdevbase + IOM_SIZE);
1.2 cgd 1594:
1.401 thorpej 1595: /*
1596: * Check to see if we have a memory map from the BIOS (passed
1597: * to us by the boot program.
1598: */
1.577 mrg 1599: if ((biosmem_implicit || (biosbasemem == 0 && biosextmem == 0)) &&
1.575 mrg 1600: (bim = lookup_bootinfo(BTINFO_MEMMAP)) != NULL && bim->num > 0) {
1.463 enami 1601: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1602: printf("BIOS MEMORY MAP (%d ENTRIES):\n", bim->num);
1603: #endif
1604: for (x = 0; x < bim->num; x++) {
1.463 enami 1605: #ifdef DEBUG_MEMLOAD
1.401 thorpej 1606: printf(" addr 0x%qx size 0x%qx type 0x%x\n",
1607: bim->entry[x].addr,
1608: bim->entry[x].size,
1609: bim->entry[x].type);
1610: #endif
1611:
1612: /*
1613: * If the segment is not memory, skip it.
1614: */
1615: switch (bim->entry[x].type) {
1616: case BIM_Memory:
1617: case BIM_ACPI:
1618: case BIM_NVS:
1619: break;
1620: default:
1621: continue;
1622: }
1623:
1624: /*
1.586.2.5 liamjfoy 1625: * If the segment is smaller than a page, skip it.
1626: */
1627: if (bim->entry[x].size < NBPG) {
1628: continue;
1629: }
1630:
1631: /*
1.401 thorpej 1632: * Sanity check the entry.
1633: * XXX Need to handle uint64_t in extent code
1634: * XXX and 64-bit physical addresses in i386
1635: * XXX port.
1636: */
1637: seg_start = bim->entry[x].addr;
1638: seg_end = bim->entry[x].addr + bim->entry[x].size;
1639:
1.432 kanaoka 1640: /*
1.433 kanaoka 1641: * Avoid Compatibility Holes.
1642: * XXX Holes within memory space that allow access
1643: * XXX to be directed to the PC-compatible frame buffer
1.484 fvdl 1644: * XXX (0xa0000-0xbffff),to adapter ROM space
1.433 kanaoka 1645: * XXX (0xc0000-0xdffff), and to system BIOS space
1646: * XXX (0xe0000-0xfffff).
1647: * XXX Some laptop(for example,Toshiba Satellite2550X)
1648: * XXX report this area and occurred problems,
1649: * XXX so we avoid this area.
1.432 kanaoka 1650: */
1.433 kanaoka 1651: if (seg_start < 0x100000 && seg_end > 0xa0000) {
1652: printf("WARNING: memory map entry overlaps "
1653: "with ``Compatibility Holes'': "
1654: "0x%qx/0x%qx/0x%x\n", seg_start,
1655: seg_end - seg_start, bim->entry[x].type);
1656: add_mem_cluster(seg_start, 0xa0000,
1657: bim->entry[x].type);
1658: add_mem_cluster(0x100000, seg_end,
1.401 thorpej 1659: bim->entry[x].type);
1.433 kanaoka 1660: } else
1661: add_mem_cluster(seg_start, seg_end,
1.401 thorpej 1662: bim->entry[x].type);
1663: }
1.421 aymeric 1664: }
1.575 mrg 1665:
1.421 aymeric 1666: /*
1667: * If the loop above didn't find any valid segment, fall back to
1668: * former code.
1669: */
1670: if (mem_cluster_cnt == 0) {
1.401 thorpej 1671: /*
1672: * Allocate the physical addresses used by RAM from the iomem
1673: * extent map. This is done before the addresses are
1674: * page rounded just to make sure we get them all.
1675: */
1676: if (extent_alloc_region(iomem_ex, 0, KBTOB(biosbasemem),
1677: EX_NOWAIT)) {
1678: /* XXX What should we do? */
1679: printf("WARNING: CAN'T ALLOCATE BASE MEMORY FROM "
1680: "IOMEM EXTENT MAP!\n");
1681: }
1682: mem_clusters[0].start = 0;
1683: mem_clusters[0].size = trunc_page(KBTOB(biosbasemem));
1684: physmem += atop(mem_clusters[0].size);
1685: if (extent_alloc_region(iomem_ex, IOM_END, KBTOB(biosextmem),
1686: EX_NOWAIT)) {
1687: /* XXX What should we do? */
1688: printf("WARNING: CAN'T ALLOCATE EXTENDED MEMORY FROM "
1689: "IOMEM EXTENT MAP!\n");
1690: }
1691: #if NISADMA > 0
1692: /*
1693: * Some motherboards/BIOSes remap the 384K of RAM that would
1694: * normally be covered by the ISA hole to the end of memory
1695: * so that it can be used. However, on a 16M system, this
1696: * would cause bounce buffers to be allocated and used.
1697: * This is not desirable behaviour, as more than 384K of
1698: * bounce buffers might be allocated. As a work-around,
1699: * we round memory down to the nearest 1M boundary if
1700: * we're using any isadma devices and the remapped memory
1701: * is what puts us over 16M.
1702: */
1703: if (biosextmem > (15*1024) && biosextmem < (16*1024)) {
1704: char pbuf[9];
1705:
1706: format_bytes(pbuf, sizeof(pbuf),
1707: biosextmem - (15*1024));
1708: printf("Warning: ignoring %s of remapped memory\n",
1709: pbuf);
1710: biosextmem = (15*1024);
1711: }
1712: #endif
1713: mem_clusters[1].start = IOM_END;
1714: mem_clusters[1].size = trunc_page(KBTOB(biosextmem));
1715: physmem += atop(mem_clusters[1].size);
1716:
1717: mem_cluster_cnt = 2;
1718:
1719: avail_end = IOM_END + trunc_page(KBTOB(biosextmem));
1720: }
1721: /*
1722: * If we have 16M of RAM or less, just put it all on
1723: * the default free list. Otherwise, put the first
1724: * 16M of RAM on a lower priority free list (so that
1725: * all of the ISA DMA'able memory won't be eaten up
1726: * first-off).
1727: */
1728: if (avail_end <= (16 * 1024 * 1024))
1729: first16q = VM_FREELIST_DEFAULT;
1730: else
1731: first16q = VM_FREELIST_FIRST16;
1732:
1733: /* Make sure the end of the space used by the kernel is rounded. */
1734: first_avail = round_page(first_avail);
1735:
1736: /*
1737: * Now, load the memory clusters (which have already been
1738: * rounded and truncated) into the VM system.
1739: *
1740: * NOTE: WE ASSUME THAT MEMORY STARTS AT 0 AND THAT THE KERNEL
1741: * IS LOADED AT IOM_END (1M).
1742: */
1743: for (x = 0; x < mem_cluster_cnt; x++) {
1744: seg_start = mem_clusters[x].start;
1745: seg_end = mem_clusters[x].start + mem_clusters[x].size;
1746: seg_start1 = 0;
1747: seg_end1 = 0;
1748:
1749: /*
1750: * Skip memory before our available starting point.
1751: */
1752: if (seg_end <= avail_start)
1753: continue;
1754:
1755: if (avail_start >= seg_start && avail_start < seg_end) {
1756: if (seg_start != 0)
1757: panic("init386: memory doesn't start at 0");
1758: seg_start = avail_start;
1759: if (seg_start == seg_end)
1760: continue;
1761: }
1762:
1763: /*
1764: * If this segment contains the kernel, split it
1765: * in two, around the kernel.
1766: */
1767: if (seg_start <= IOM_END && first_avail <= seg_end) {
1768: seg_start1 = first_avail;
1769: seg_end1 = seg_end;
1770: seg_end = IOM_END;
1771: }
1772:
1773: /* First hunk */
1774: if (seg_start != seg_end) {
1.480 erh 1775: if (seg_start < (16 * 1024 * 1024) &&
1.401 thorpej 1776: first16q != VM_FREELIST_DEFAULT) {
1.568 perry 1777: uint64_t tmp;
1.401 thorpej 1778:
1779: if (seg_end > (16 * 1024 * 1024))
1780: tmp = (16 * 1024 * 1024);
1781: else
1782: tmp = seg_end;
1.492 kanaoka 1783:
1784: if (tmp != seg_start) {
1.463 enami 1785: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1786: printf("loading 0x%qx-0x%qx "
1787: "(0x%lx-0x%lx)\n",
1788: seg_start, tmp,
1789: atop(seg_start), atop(tmp));
1790: #endif
1791: uvm_page_physload(atop(seg_start),
1792: atop(tmp), atop(seg_start),
1793: atop(tmp), first16q);
1794: }
1.401 thorpej 1795: seg_start = tmp;
1796: }
1.411 enami 1797:
1798: if (seg_start != seg_end) {
1.463 enami 1799: #ifdef DEBUG_MEMLOAD
1.411 enami 1800: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1801: seg_start, seg_end,
1802: atop(seg_start), atop(seg_end));
1803: #endif
1804: uvm_page_physload(atop(seg_start),
1805: atop(seg_end), atop(seg_start),
1806: atop(seg_end), VM_FREELIST_DEFAULT);
1807: }
1.401 thorpej 1808: }
1809:
1810: /* Second hunk */
1811: if (seg_start1 != seg_end1) {
1.480 erh 1812: if (seg_start1 < (16 * 1024 * 1024) &&
1.401 thorpej 1813: first16q != VM_FREELIST_DEFAULT) {
1.568 perry 1814: uint64_t tmp;
1.401 thorpej 1815:
1816: if (seg_end1 > (16 * 1024 * 1024))
1817: tmp = (16 * 1024 * 1024);
1818: else
1819: tmp = seg_end1;
1.492 kanaoka 1820:
1821: if (tmp != seg_start1) {
1.463 enami 1822: #ifdef DEBUG_MEMLOAD
1.492 kanaoka 1823: printf("loading 0x%qx-0x%qx "
1824: "(0x%lx-0x%lx)\n",
1825: seg_start1, tmp,
1826: atop(seg_start1), atop(tmp));
1827: #endif
1828: uvm_page_physload(atop(seg_start1),
1829: atop(tmp), atop(seg_start1),
1830: atop(tmp), first16q);
1831: }
1.401 thorpej 1832: seg_start1 = tmp;
1833: }
1.412 enami 1834:
1835: if (seg_start1 != seg_end1) {
1.463 enami 1836: #ifdef DEBUG_MEMLOAD
1.412 enami 1837: printf("loading 0x%qx-0x%qx (0x%lx-0x%lx)\n",
1838: seg_start1, seg_end1,
1839: atop(seg_start1), atop(seg_end1));
1840: #endif
1841: uvm_page_physload(atop(seg_start1),
1842: atop(seg_end1), atop(seg_start1),
1843: atop(seg_end1), VM_FREELIST_DEFAULT);
1844: }
1.401 thorpej 1845: }
1846: }
1847:
1848: /*
1849: * Steal memory for the message buffer (at end of core).
1850: */
1851: {
1.543 mycroft 1852: struct vm_physseg *vps;
1.401 thorpej 1853: psize_t sz = round_page(MSGBUFSIZE);
1854: psize_t reqsz = sz;
1855:
1.586.2.1 tron 1856: search_again:
1.401 thorpej 1857: for (x = 0; x < vm_nphysseg; x++) {
1858: vps = &vm_physmem[x];
1859: if (ptoa(vps->avail_end) == avail_end)
1.543 mycroft 1860: goto found;
1.401 thorpej 1861: }
1.543 mycroft 1862: panic("init386: can't find end of memory");
1.401 thorpej 1863:
1.543 mycroft 1864: found:
1.401 thorpej 1865: /* Shrink so it'll fit in the last segment. */
1866: if ((vps->avail_end - vps->avail_start) < atop(sz))
1867: sz = ptoa(vps->avail_end - vps->avail_start);
1868:
1869: vps->avail_end -= atop(sz);
1870: vps->end -= atop(sz);
1.586.2.1 tron 1871: msgbuf_p_seg[msgbuf_p_cnt].sz = sz;
1872: msgbuf_p_seg[msgbuf_p_cnt++].paddr = ptoa(vps->avail_end);
1.401 thorpej 1873:
1874: /* Remove the last segment if it now has no pages. */
1875: if (vps->start == vps->end) {
1876: for (vm_nphysseg--; x < vm_nphysseg; x++)
1877: vm_physmem[x] = vm_physmem[x + 1];
1878: }
1879:
1880: /* Now find where the new avail_end is. */
1881: for (avail_end = 0, x = 0; x < vm_nphysseg; x++)
1882: if (vm_physmem[x].avail_end > avail_end)
1883: avail_end = vm_physmem[x].avail_end;
1884: avail_end = ptoa(avail_end);
1885:
1.586.2.1 tron 1886: if (sz != reqsz) {
1887: reqsz -= sz;
1888: if (msgbuf_p_cnt != VM_PHYSSEG_MAX) {
1889: /* if still segments available, get memory from next one ... */
1.586.2.2 tron 1890: sz = reqsz;
1891: goto search_again;
1.586.2.1 tron 1892: }
1.401 thorpej 1893: /* Warn if the message buffer had to be shrunk. */
1894: printf("WARNING: %ld bytes not available for msgbuf "
1.586.2.2 tron 1895: "in last cluster (%ld used)\n", (long)MSGBUFSIZE, MSGBUFSIZE - reqsz);
1.586.2.1 tron 1896: }
1.401 thorpej 1897: }
1898:
1.473 tshiozak 1899: /*
1900: * install PT page for the first 4M if needed.
1901: */
1902: if (needs_earlier_install_pte0) {
1903: paddr_t paddr;
1904: #ifdef DIAGNOSTIC
1905: if (realmode_reserved_size < PAGE_SIZE) {
1906: panic("cannot steal memory for first 4M PT page.");
1907: }
1908: #endif
1909: paddr=realmode_reserved_start+realmode_reserved_size-PAGE_SIZE;
1.561 yamt 1910: pmap_kenter_pa((vaddr_t)vtopte(0), paddr,
1911: VM_PROT_READ|VM_PROT_WRITE);
1.473 tshiozak 1912: pmap_update(pmap_kernel());
1913: /* make sure it is clean before using */
1914: memset(vtopte(0), 0, PAGE_SIZE);
1915: realmode_reserved_size -= PAGE_SIZE;
1916: }
1917:
1.295 drochner 1918: #if NBIOSCALL > 0
1.436 jdolecek 1919: /*
1920: * this should be caught at kernel build time, but put it here
1921: * in case someone tries to fake it out...
1922: */
1923: #ifdef DIAGNOSTIC
1.473 tshiozak 1924: if (realmode_reserved_start > BIOSTRAMP_BASE ||
1.474 tron 1925: (realmode_reserved_start+realmode_reserved_size) < (BIOSTRAMP_BASE+
1.473 tshiozak 1926: PAGE_SIZE)) {
1927: panic("cannot steal memory for PT page of bioscall.");
1928: }
1.436 jdolecek 1929: if (biostramp_image_size > PAGE_SIZE)
1.483 provos 1930: panic("biostramp_image_size too big: %x vs. %x",
1.436 jdolecek 1931: biostramp_image_size, PAGE_SIZE);
1932: #endif
1933: pmap_kenter_pa((vaddr_t)BIOSTRAMP_BASE, /* virtual */
1934: (paddr_t)BIOSTRAMP_BASE, /* physical */
1935: VM_PROT_ALL); /* protection */
1.456 chris 1936: pmap_update(pmap_kernel());
1.436 jdolecek 1937: memcpy((caddr_t)BIOSTRAMP_BASE, biostramp_image, biostramp_image_size);
1938: #ifdef DEBUG_BIOSCALL
1939: printf("biostramp installed @ %x\n", BIOSTRAMP_BASE);
1940: #endif
1.473 tshiozak 1941: realmode_reserved_size -= PAGE_SIZE;
1942: realmode_reserved_start += PAGE_SIZE;
1943: #endif
1944:
1945: #if NACPI > 0
1946: /*
1947: * Steal memory for the acpi wake code
1948: */
1949: {
1950: paddr_t paddr, p;
1951: psize_t sz;
1952: int npg;
1953:
1954: paddr = realmode_reserved_start;
1955: npg = acpi_md_get_npages_of_wakecode();
1956: sz = ptoa(npg);
1957: #ifdef DIAGNOSTIC
1958: if (realmode_reserved_size < sz) {
1959: panic("cannot steal memory for ACPI wake code.");
1960: }
1961: #endif
1962:
1963: /* identical mapping */
1964: p = paddr;
1965: for (x=0; x<npg; x++) {
1966: printf("kenter: 0x%08X\n", (unsigned)p);
1967: pmap_kenter_pa((vaddr_t)p, p, VM_PROT_ALL);
1968: p += PAGE_SIZE;
1969: }
1970: pmap_update(pmap_kernel());
1971:
1972: acpi_md_install_wakecode(paddr);
1973:
1974: realmode_reserved_size -= sz;
1975: realmode_reserved_start += sz;
1976: }
1.295 drochner 1977: #endif
1.59 mycroft 1978:
1.561 yamt 1979: pmap_kenter_pa(idt_vaddr, idt_paddr, VM_PROT_READ|VM_PROT_WRITE);
1.456 chris 1980: pmap_update(pmap_kernel());
1.484 fvdl 1981: memset((void *)idt_vaddr, 0, PAGE_SIZE);
1982:
1.516 fvdl 1983: idt = (struct gate_descriptor *)idt_vaddr;
1.275 mycroft 1984: #ifdef I586_CPU
1.561 yamt 1985: pmap_kenter_pa(pentium_idt_vaddr, idt_paddr, VM_PROT_READ);
1.275 mycroft 1986: pentium_idt = (union descriptor *)pentium_idt_vaddr;
1987: #endif
1.484 fvdl 1988: pmap_update(pmap_kernel());
1989:
1990: tgdt = gdt;
1.516 fvdl 1991: gdt = (union descriptor *)
1992: ((char *)idt + NIDT * sizeof (struct gate_descriptor));
1.275 mycroft 1993: ldt = gdt + NGDT;
1994:
1.484 fvdl 1995: memcpy(gdt, tgdt, NGDT*sizeof(*gdt));
1.275 mycroft 1996:
1997: setsegment(&gdt[GLDT_SEL].sd, ldt, NLDT * sizeof(ldt[0]) - 1,
1998: SDT_SYSLDT, SEL_KPL, 0, 0);
1999:
2000: /* make ldt gates and memory segments */
2001: setgate(&ldt[LSYS5CALLS_SEL].gd, &IDTVEC(osyscall), 1,
1.489 fvdl 2002: SDT_SYS386CGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.447 christos 2003:
1.275 mycroft 2004: ldt[LUCODE_SEL] = gdt[GUCODE_SEL];
1.533 chs 2005: ldt[LUCODEBIG_SEL] = gdt[GUCODEBIG_SEL];
1.275 mycroft 2006: ldt[LUDATA_SEL] = gdt[GUDATA_SEL];
1.324 christos 2007: ldt[LSOL26CALLS_SEL] = ldt[LBSDICALLS_SEL] = ldt[LSYS5CALLS_SEL];
1.275 mycroft 2008:
2009: /* exceptions */
1.498 fvdl 2010: for (x = 0; x < 32; x++) {
1.516 fvdl 2011: setgate(&idt[x], IDTVEC(exceptions)[x], 0, SDT_SYS386TGT,
1.489 fvdl 2012: (x == 3 || x == 4) ? SEL_UPL : SEL_KPL,
2013: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 2014: idt_allocmap[x] = 1;
2015: }
1.257 thorpej 2016:
1.275 mycroft 2017: /* new-style interrupt gate for syscalls */
1.516 fvdl 2018: setgate(&idt[128], &IDTVEC(syscall), 0, SDT_SYS386TGT, SEL_UPL,
1.489 fvdl 2019: GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 2020: idt_allocmap[128] = 1;
1.333 christos 2021: #ifdef COMPAT_SVR4
1.516 fvdl 2022: setgate(&idt[0xd2], &IDTVEC(svr4_fasttrap), 0, SDT_SYS386TGT,
1.489 fvdl 2023: SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));
1.498 fvdl 2024: idt_allocmap[0xd2] = 1;
1.333 christos 2025: #endif /* COMPAT_SVR4 */
1.264 mycroft 2026:
1.275 mycroft 2027: setregion(®ion, gdt, NGDT * sizeof(gdt[0]) - 1);
2028: lgdt(®ion);
1.484 fvdl 2029:
2030: cpu_init_idt();
1.264 mycroft 2031:
1.522 ragge 2032: #if NKSYMS || defined(DDB) || defined(LKM)
1.308 tv 2033: {
2034: extern int end;
1.585 jmmv 2035: boolean_t loaded;
1.336 christos 2036: struct btinfo_symtab *symtab;
1.308 tv 2037:
1.522 ragge 2038: #ifdef DDB
1.484 fvdl 2039: db_machine_init();
1.522 ragge 2040: #endif
1.484 fvdl 2041:
1.585 jmmv 2042: #if defined(MULTIBOOT)
2043: loaded = multiboot_ksyms_init();
2044: #else
2045: loaded = FALSE;
2046: #endif
2047: if (!loaded) {
2048: symtab = lookup_bootinfo(BTINFO_SYMTAB);
2049: if (symtab) {
2050: symtab->ssym += KERNBASE;
2051: symtab->esym += KERNBASE;
2052: ksyms_init(symtab->nsym, (int *)symtab->ssym,
2053: (int *)symtab->esym);
2054: } else
2055: ksyms_init(*(int *)&end, ((int *)&end) + 1, esym);
1.336 christos 2056: }
1.308 tv 2057: }
1.522 ragge 2058: #endif
2059: #ifdef DDB
1.190 mycroft 2060: if (boothowto & RB_KDB)
2061: Debugger();
1.377 ws 2062: #endif
2063: #ifdef IPKDB
2064: ipkdb_init();
2065: if (boothowto & RB_KDB)
2066: ipkdb_connect(0);
1.190 mycroft 2067: #endif
2068: #ifdef KGDB
1.243 drochner 2069: kgdb_port_init();
1.235 thorpej 2070: if (boothowto & RB_KDB) {
2071: kgdb_debug_init = 1;
1.242 drochner 2072: kgdb_connect(1);
1.235 thorpej 2073: }
1.384 jdolecek 2074: #endif
2075:
2076: #if NMCA > 0
2077: /* check for MCA bus, needed to be done before ISA stuff - if
2078: * MCA is detected, ISA needs to use level triggered interrupts
2079: * by default */
2080: mca_busprobe();
1.190 mycroft 2081: #endif
1.275 mycroft 2082:
1.498 fvdl 2083: intr_default_setup();
1.431 thorpej 2084:
2085: /* Initialize software interrupts. */
2086: softintr_init();
1.275 mycroft 2087:
1.498 fvdl 2088: splraise(IPL_IPI);
1.275 mycroft 2089: enable_intr();
2090:
2091: if (physmem < btoc(2 * 1024 * 1024)) {
2092: printf("warning: too little memory available; "
1.383 mycroft 2093: "have %lu bytes, want %lu bytes\n"
1.275 mycroft 2094: "running in degraded mode\n"
2095: "press a key to confirm\n\n",
1.383 mycroft 2096: ptoa(physmem), 2*1024*1024UL);
1.275 mycroft 2097: cngetc();
2098: }
1.507 jdolecek 2099:
2100: #ifdef __HAVE_CPU_MAXPROC
2101: /* Make sure maxproc is sane */
2102: if (maxproc > cpu_maxproc())
2103: maxproc = cpu_maxproc();
2104: #endif
1.1 cgd 2105: }
2106:
1.107 deraadt 2107: #ifdef COMPAT_NOMID
2108: static int
1.566 christos 2109: exec_nomid(struct lwp *l, struct exec_package *epp)
1.31 cgd 2110: {
1.59 mycroft 2111: int error;
2112: u_long midmag, magic;
2113: u_short mid;
1.80 cgd 2114: struct exec *execp = epp->ep_hdr;
1.31 cgd 2115:
1.80 cgd 2116: /* check on validity of epp->ep_hdr performed by exec_out_makecmds */
2117:
2118: midmag = ntohl(execp->a_midmag);
1.59 mycroft 2119: mid = (midmag >> 16) & 0xffff;
2120: magic = midmag & 0xffff;
2121:
2122: if (magic == 0) {
1.80 cgd 2123: magic = (execp->a_midmag & 0xffff);
1.59 mycroft 2124: mid = MID_ZERO;
2125: }
2126:
2127: midmag = mid << 16 | magic;
2128:
2129: switch (midmag) {
2130: case (MID_ZERO << 16) | ZMAGIC:
2131: /*
2132: * 386BSD's ZMAGIC format:
2133: */
1.566 christos 2134: error = exec_aout_prep_oldzmagic(l, epp);
1.59 mycroft 2135: break;
2136:
2137: case (MID_ZERO << 16) | QMAGIC:
2138: /*
2139: * BSDI's QMAGIC format:
2140: * same as new ZMAGIC format, but with different magic number
2141: */
1.566 christos 2142: error = exec_aout_prep_zmagic(l, epp);
1.59 mycroft 2143: break;
2144:
1.202 christos 2145: case (MID_ZERO << 16) | NMAGIC:
2146: /*
2147: * BSDI's NMAGIC format:
2148: * same as NMAGIC format, but with different magic number
2149: * and with text starting at 0.
2150: */
1.566 christos 2151: error = exec_aout_prep_oldnmagic(l, epp);
1.202 christos 2152: break;
2153:
2154: case (MID_ZERO << 16) | OMAGIC:
2155: /*
2156: * BSDI's OMAGIC format:
2157: * same as OMAGIC format, but with different magic number
2158: * and with text starting at 0.
2159: */
1.566 christos 2160: error = exec_aout_prep_oldomagic(l, epp);
1.202 christos 2161: break;
2162:
1.59 mycroft 2163: default:
2164: error = ENOEXEC;
2165: }
2166:
2167: return error;
1.107 deraadt 2168: }
1.31 cgd 2169: #endif
1.107 deraadt 2170:
2171: /*
2172: * cpu_exec_aout_makecmds():
1.549 wiz 2173: * CPU-dependent a.out format hook for execve().
1.107 deraadt 2174: *
2175: * Determine of the given exec package refers to something which we
2176: * understand and, if so, set up the vmcmds for it.
2177: *
2178: * On the i386, old (386bsd) ZMAGIC binaries and BSDI QMAGIC binaries
2179: * if COMPAT_NOMID is given as a kernel option.
2180: */
2181: int
1.566 christos 2182: cpu_exec_aout_makecmds(struct lwp *l, struct exec_package *epp)
1.107 deraadt 2183: {
2184: int error = ENOEXEC;
2185:
2186: #ifdef COMPAT_NOMID
1.566 christos 2187: if ((error = exec_nomid(l, epp)) == 0)
1.107 deraadt 2188: return error;
1.581 thorpej 2189: #else
2190: (void) l;
2191: (void) epp;
1.107 deraadt 2192: #endif /* ! COMPAT_NOMID */
2193:
2194: return error;
1.31 cgd 2195: }
1.84 cgd 2196:
1.484 fvdl 2197: #include <dev/ic/mc146818reg.h> /* for NVRAM POST */
2198: #include <i386/isa/nvram.h> /* for NVRAM POST */
2199:
1.149 mycroft 2200: void
2201: cpu_reset()
2202: {
1.555 mycroft 2203: struct region_descriptor region;
1.149 mycroft 2204:
1.224 mycroft 2205: disable_intr();
2206:
1.227 mycroft 2207: /*
1.484 fvdl 2208: * Ensure the NVRAM reset byte contains something vaguely sane.
2209: */
2210:
2211: outb(IO_RTC, NVRAM_RESET);
2212: outb(IO_RTC+1, NVRAM_RESET_RST);
2213:
2214: /*
1.564 dyoung 2215: * Reset AMD Geode SC1100.
2216: *
1.565 simonb 2217: * 1) Write PCI Configuration Address Register (0xcf8) to
2218: * select Function 0, Register 0x44: Bridge Configuration,
2219: * GPIO and LPC Configuration Register Space, Reset
2220: * Control Register.
1.564 dyoung 2221: *
1.565 simonb 2222: * 2) Write 0xf to PCI Configuration Data Register (0xcfc)
2223: * to reset IDE controller, IDE bus, and PCI bus, and
2224: * to trigger a system-wide reset.
1.564 dyoung 2225: *
2226: * See AMD Geode SC1100 Processor Data Book, Revision 2.0,
2227: * sections 6.3.1, 6.3.2, and 6.4.1.
2228: */
2229: if (cpu_info_primary.ci_signature == 0x540) {
2230: outl(0xcf8, 0x80009044ul);
2231: outl(0xcfc, 0xf);
1.565 simonb 2232: }
1.564 dyoung 2233:
2234: /*
1.227 mycroft 2235: * The keyboard controller has 4 random output pins, one of which is
2236: * connected to the RESET pin on the CPU in many PCs. We tell the
2237: * keyboard controller to pulse this line a couple of times.
2238: */
1.273 drochner 2239: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2240: delay(100000);
1.273 drochner 2241: outb(IO_KBD + KBCMDP, KBC_PULSE0);
1.226 mycroft 2242: delay(100000);
1.149 mycroft 2243:
2244: /*
1.224 mycroft 2245: * Try to cause a triple fault and watchdog reset by making the IDT
2246: * invalid and causing a fault.
1.149 mycroft 2247: */
1.313 perry 2248: memset((caddr_t)idt, 0, NIDT * sizeof(idt[0]));
1.555 mycroft 2249: setregion(®ion, idt, NIDT * sizeof(idt[0]) - 1);
1.565 simonb 2250: lidt(®ion);
1.567 perry 2251: __asm volatile("divl %0,%1" : : "q" (0), "a" (0));
1.149 mycroft 2252:
1.224 mycroft 2253: #if 0
1.149 mycroft 2254: /*
2255: * Try to cause a triple fault and watchdog reset by unmapping the
1.224 mycroft 2256: * entire address space and doing a TLB flush.
1.149 mycroft 2257: */
1.414 thorpej 2258: memset((caddr_t)PTD, 0, PAGE_SIZE);
1.484 fvdl 2259: tlbflush();
1.224 mycroft 2260: #endif
1.149 mycroft 2261:
2262: for (;;);
1.45 cgd 2263: }
1.484 fvdl 2264:
1.499 jdolecek 2265: void
1.551 junyoung 2266: cpu_getmcontext(struct lwp *l, mcontext_t *mcp, unsigned int *flags)
1.508 thorpej 2267: {
2268: const struct trapframe *tf = l->l_md.md_regs;
2269: __greg_t *gr = mcp->__gregs;
1.519 nathanw 2270: __greg_t ras_eip;
1.508 thorpej 2271:
2272: /* Save register context. */
2273: #ifdef VM86
2274: if (tf->tf_eflags & PSL_VM) {
2275: gr[_REG_GS] = tf->tf_vm86_gs;
2276: gr[_REG_FS] = tf->tf_vm86_fs;
2277: gr[_REG_ES] = tf->tf_vm86_es;
2278: gr[_REG_DS] = tf->tf_vm86_ds;
2279: gr[_REG_EFL] = get_vflags(l);
2280: } else
2281: #endif
2282: {
2283: gr[_REG_GS] = tf->tf_gs;
2284: gr[_REG_FS] = tf->tf_fs;
2285: gr[_REG_ES] = tf->tf_es;
2286: gr[_REG_DS] = tf->tf_ds;
2287: gr[_REG_EFL] = tf->tf_eflags;
2288: }
2289: gr[_REG_EDI] = tf->tf_edi;
2290: gr[_REG_ESI] = tf->tf_esi;
2291: gr[_REG_EBP] = tf->tf_ebp;
2292: gr[_REG_EBX] = tf->tf_ebx;
2293: gr[_REG_EDX] = tf->tf_edx;
2294: gr[_REG_ECX] = tf->tf_ecx;
2295: gr[_REG_EAX] = tf->tf_eax;
2296: gr[_REG_EIP] = tf->tf_eip;
2297: gr[_REG_CS] = tf->tf_cs;
2298: gr[_REG_ESP] = tf->tf_esp;
2299: gr[_REG_UESP] = tf->tf_esp;
2300: gr[_REG_SS] = tf->tf_ss;
2301: gr[_REG_TRAPNO] = tf->tf_trapno;
2302: gr[_REG_ERR] = tf->tf_err;
1.519 nathanw 2303:
2304: if ((ras_eip = (__greg_t)ras_lookup(l->l_proc,
2305: (caddr_t) gr[_REG_EIP])) != -1)
2306: gr[_REG_EIP] = ras_eip;
2307:
1.508 thorpej 2308: *flags |= _UC_CPU;
2309:
2310: /* Save floating point register context, if any. */
1.548 yamt 2311: if ((l->l_md.md_flags & MDL_USEDFPU) != 0) {
1.508 thorpej 2312: #if NNPX > 0
2313: /*
2314: * If this process is the current FP owner, dump its
2315: * context to the PCB first.
2316: * XXX npxsave() also clears the FPU state; depending on the
2317: * XXX application this might be a penalty.
2318: */
2319: if (l->l_addr->u_pcb.pcb_fpcpu) {
2320: npxsave_lwp(l, 1);
2321: }
2322: #endif
2323: if (i386_use_fxsave) {
2324: memcpy(&mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2325: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2326: sizeof (mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm));
2327: *flags |= _UC_FXSAVE;
2328: } else {
2329: memcpy(&mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2330: &l->l_addr->u_pcb.pcb_savefpu.sv_87,
2331: sizeof (mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state));
2332: }
2333: #if 0
2334: /* Apparently nothing ever touches this. */
2335: ucp->mcp.mc_fp.fp_emcsts = l->l_addr->u_pcb.pcb_saveemc;
2336: #endif
2337: *flags |= _UC_FPU;
2338: }
2339: }
2340:
2341: int
1.551 junyoung 2342: cpu_setmcontext(struct lwp *l, const mcontext_t *mcp, unsigned int flags)
1.508 thorpej 2343: {
2344: struct trapframe *tf = l->l_md.md_regs;
1.584 pooka 2345: const __greg_t *gr = mcp->__gregs;
1.508 thorpej 2346:
2347: /* Restore register context, if any. */
2348: if ((flags & _UC_CPU) != 0) {
2349: #ifdef VM86
1.535 drochner 2350: if (gr[_REG_EFL] & PSL_VM) {
1.508 thorpej 2351: tf->tf_vm86_gs = gr[_REG_GS];
2352: tf->tf_vm86_fs = gr[_REG_FS];
2353: tf->tf_vm86_es = gr[_REG_ES];
2354: tf->tf_vm86_ds = gr[_REG_DS];
2355: set_vflags(l, gr[_REG_EFL]);
1.534 christos 2356: if (flags & _UC_VM) {
1.542 junyoung 2357: void syscall_vm86(struct trapframe *);
1.534 christos 2358: l->l_proc->p_md.md_syscall = syscall_vm86;
2359: }
1.508 thorpej 2360: } else
2361: #endif
2362: {
2363: /*
2364: * Check for security violations. If we're returning
2365: * to protected mode, the CPU will validate the segment
2366: * registers automatically and generate a trap on
2367: * violations. We handle the trap, rather than doing
2368: * all of the checking here.
2369: */
1.534 christos 2370: if (((gr[_REG_EFL] ^ tf->tf_eflags) & PSL_USERSTATIC) ||
2371: !USERMODE(gr[_REG_CS], gr[_REG_EFL])) {
2372: printf("cpu_setmcontext error: uc EFL: 0x%08x"
1.535 drochner 2373: " tf EFL: 0x%08x uc CS: 0x%x\n",
1.508 thorpej 2374: gr[_REG_EFL], tf->tf_eflags, gr[_REG_CS]);
2375: return (EINVAL);
2376: }
2377: tf->tf_gs = gr[_REG_GS];
2378: tf->tf_fs = gr[_REG_FS];
2379: tf->tf_es = gr[_REG_ES];
2380: tf->tf_ds = gr[_REG_DS];
2381: /* Only change the user-alterable part of eflags */
2382: tf->tf_eflags &= ~PSL_USER;
2383: tf->tf_eflags |= (gr[_REG_EFL] & PSL_USER);
2384: }
2385: tf->tf_edi = gr[_REG_EDI];
2386: tf->tf_esi = gr[_REG_ESI];
2387: tf->tf_ebp = gr[_REG_EBP];
2388: tf->tf_ebx = gr[_REG_EBX];
2389: tf->tf_edx = gr[_REG_EDX];
2390: tf->tf_ecx = gr[_REG_ECX];
2391: tf->tf_eax = gr[_REG_EAX];
2392: tf->tf_eip = gr[_REG_EIP];
2393: tf->tf_cs = gr[_REG_CS];
2394: tf->tf_esp = gr[_REG_UESP];
2395: tf->tf_ss = gr[_REG_SS];
2396: }
2397:
2398: /* Restore floating point register context, if any. */
2399: if ((flags & _UC_FPU) != 0) {
2400: #if NNPX > 0
2401: /*
2402: * If we were using the FPU, forget that we were.
2403: */
2404: if (l->l_addr->u_pcb.pcb_fpcpu != NULL)
2405: npxsave_lwp(l, 0);
2406: #endif
2407: if (flags & _UC_FXSAVE) {
2408: if (i386_use_fxsave) {
2409: memcpy(
2410: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm,
2411: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2412: sizeof (&l->l_addr->u_pcb.pcb_savefpu.sv_xmm));
2413: } else {
2414: /* This is a weird corner case */
2415: process_xmm_to_s87((struct savexmm *)
2416: &mcp->__fpregs.__fp_reg_set.__fp_xmm_state.__fp_xmm,
2417: &l->l_addr->u_pcb.pcb_savefpu.sv_87);
2418: }
2419: } else {
2420: if (i386_use_fxsave) {
2421: process_s87_to_xmm((struct save87 *)
2422: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2423: &l->l_addr->u_pcb.pcb_savefpu.sv_xmm);
2424: } else {
2425: memcpy(&l->l_addr->u_pcb.pcb_savefpu.sv_87,
2426: &mcp->__fpregs.__fp_reg_set.__fpchip_state.__fp_state,
2427: sizeof (l->l_addr->u_pcb.pcb_savefpu.sv_87));
2428: }
2429: }
2430: /* If not set already. */
1.548 yamt 2431: l->l_md.md_flags |= MDL_USEDFPU;
1.508 thorpej 2432: #if 0
2433: /* Apparently unused. */
2434: l->l_addr->u_pcb.pcb_saveemc = mcp->mc_fp.fp_emcsts;
2435: #endif
2436: }
1.534 christos 2437: if (flags & _UC_SETSTACK)
2438: l->l_proc->p_sigctx.ps_sigstk.ss_flags |= SS_ONSTACK;
2439: if (flags & _UC_CLRSTACK)
2440: l->l_proc->p_sigctx.ps_sigstk.ss_flags &= ~SS_ONSTACK;
1.508 thorpej 2441: return (0);
2442: }
2443:
2444: void
1.499 jdolecek 2445: cpu_initclocks()
1.484 fvdl 2446: {
1.574 kardel 2447:
1.484 fvdl 2448: (*initclock_func)();
2449: }
2450:
2451: #ifdef MULTIPROCESSOR
1.499 jdolecek 2452: void
2453: need_resched(struct cpu_info *ci)
1.484 fvdl 2454: {
1.546 yamt 2455:
2456: if (ci->ci_want_resched)
2457: return;
2458:
1.484 fvdl 2459: ci->ci_want_resched = 1;
1.508 thorpej 2460: if ((ci)->ci_curlwp != NULL)
2461: aston((ci)->ci_curlwp->l_proc);
1.546 yamt 2462: else if (ci != curcpu())
2463: x86_send_ipi(ci, 0);
1.484 fvdl 2464: }
2465: #endif
2466:
2467: /*
2468: * Allocate an IDT vector slot within the given range.
2469: * XXX needs locking to avoid MP allocation races.
2470: */
2471:
2472: int
1.551 junyoung 2473: idt_vec_alloc(int low, int high)
1.484 fvdl 2474: {
2475: int vec;
2476:
1.498 fvdl 2477: simple_lock(&idt_lock);
2478: for (vec = low; vec <= high; vec++) {
2479: if (idt_allocmap[vec] == 0) {
2480: idt_allocmap[vec] = 1;
2481: simple_unlock(&idt_lock);
1.484 fvdl 2482: return vec;
1.498 fvdl 2483: }
2484: }
2485: simple_unlock(&idt_lock);
1.484 fvdl 2486: return 0;
2487: }
2488:
1.498 fvdl 2489: void
1.551 junyoung 2490: idt_vec_set(int vec, void (*function)(void))
1.484 fvdl 2491: {
1.498 fvdl 2492: /*
2493: * Vector should be allocated, so no locking needed.
2494: */
2495: KASSERT(idt_allocmap[vec] == 1);
1.516 fvdl 2496: setgate(&idt[vec], function, 0, SDT_SYS386IGT, SEL_KPL,
1.489 fvdl 2497: GSEL(GCODE_SEL, SEL_KPL));
1.484 fvdl 2498: }
2499:
2500: void
1.551 junyoung 2501: idt_vec_free(int vec)
1.484 fvdl 2502: {
1.498 fvdl 2503: simple_lock(&idt_lock);
1.516 fvdl 2504: unsetgate(&idt[vec]);
1.498 fvdl 2505: idt_allocmap[vec] = 0;
2506: simple_unlock(&idt_lock);
1.507 jdolecek 2507: }
2508:
2509: /*
2510: * Number of processes is limited by number of available GDT slots.
2511: */
2512: int
2513: cpu_maxproc(void)
2514: {
2515: #ifdef USER_LDT
2516: return ((MAXGDTSIZ - NGDT) / 2);
2517: #else
2518: return (MAXGDTSIZ - NGDT);
2519: #endif
1.484 fvdl 2520: }
CVSweb <webmaster@jp.NetBSD.org>
![[BACK]](/icons/back.gif){kind=icon}
Return to machdep.c CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [cvs.NetBSD.org] / src / sys / arch / i386 / i386