locore.s revision 146263
1/*-
2 * Copyright (c) 1990 The Regents of the University of California.
3 * All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * William Jolitz.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 *	from: @(#)locore.s	7.3 (Berkeley) 5/13/91
33 * $FreeBSD: head/sys/i386/i386/locore.s 146263 2005-05-16 09:47:53Z obrien $
34 *
35 *		originally from: locore.s, by William F. Jolitz
36 *
37 *		Substantially rewritten by David Greenman, Rod Grimes,
38 *			Bruce Evans, Wolfgang Solfrank, Poul-Henning Kamp
39 *			and many others.
40 */
41
42#include "opt_bootp.h"
43#include "opt_compat.h"
44#include "opt_nfsroot.h"
45#include "opt_pmap.h"
46
47#include <sys/syscall.h>
48#include <sys/reboot.h>
49
50#include <machine/asmacros.h>
51#include <machine/cputypes.h>
52#include <machine/psl.h>
53#include <machine/pmap.h>
54#include <machine/specialreg.h>
55
56#include "assym.s"
57
58/*
59 *	XXX
60 *
61 * Note: This version greatly munged to avoid various assembler errors
62 * that may be fixed in newer versions of gas. Perhaps newer versions
63 * will have more pleasant appearance.
64 */
65
66/*
67 * PTmap is recursive pagemap at top of virtual address space.
68 * Within PTmap, the page directory can be found (third indirection).
69 */
70	.globl	PTmap,PTD,PTDpde
71	.set	PTmap,(PTDPTDI << PDRSHIFT)
72	.set	PTD,PTmap + (PTDPTDI * PAGE_SIZE)
73	.set	PTDpde,PTD + (PTDPTDI * PDESIZE)
74
75#ifdef SMP
76/*
77 * Define layout of per-cpu address space.
78 * This is "constructed" in locore.s on the BSP and in mp_machdep.c
79 * for each AP.  DO NOT REORDER THESE WITHOUT UPDATING THE REST!
80 */
81	.globl	SMP_prvspace
82	.set	SMP_prvspace,(MPPTDI << PDRSHIFT)
83#endif /* SMP */
84
85/*
86 * Compiled KERNBASE location and the kernel load address
87 */
88	.globl	kernbase
89	.set	kernbase,KERNBASE
90	.globl	kernload
91	.set	kernload,KERNLOAD
92
93/*
94 * Globals
95 */
96	.data
97	ALIGN_DATA			/* just to be sure */
98
99	.space	0x2000			/* space for tmpstk - temporary stack */
100tmpstk:
101
102	.globl	bootinfo
103bootinfo:	.space	BOOTINFO_SIZE	/* bootinfo that we can handle */
104
105		.globl KERNend
106KERNend:	.long	0		/* phys addr end of kernel (just after bss) */
107physfree:	.long	0		/* phys addr of next free page */
108
109#ifdef SMP
110		.globl	cpu0prvpage
111cpu0pp:		.long	0		/* phys addr cpu0 private pg */
112cpu0prvpage:	.long	0		/* relocated version */
113
114		.globl	SMPpt
115SMPptpa:	.long	0		/* phys addr SMP page table */
116SMPpt:		.long	0		/* relocated version */
117#endif /* SMP */
118
119	.globl	IdlePTD
120IdlePTD:	.long	0		/* phys addr of kernel PTD */
121
122#ifdef PAE
123	.globl	IdlePDPT
124IdlePDPT:	.long	0		/* phys addr of kernel PDPT */
125#endif
126
127#ifdef SMP
128	.globl	KPTphys
129#endif
130KPTphys:	.long	0		/* phys addr of kernel page tables */
131
132	.globl	proc0kstack
133proc0uarea:	.long	0		/* address of proc 0 uarea (unused)*/
134proc0kstack:	.long	0		/* address of proc 0 kstack space */
135p0upa:		.long	0		/* phys addr of proc0 UAREA (unused) */
136p0kpa:		.long	0		/* phys addr of proc0's STACK */
137
138vm86phystk:	.long	0		/* PA of vm86/bios stack */
139
140	.globl	vm86paddr, vm86pa
141vm86paddr:	.long	0		/* address of vm86 region */
142vm86pa:		.long	0		/* phys addr of vm86 region */
143
144#ifdef PC98
145	.globl	pc98_system_parameter
146pc98_system_parameter:
147	.space	0x240
148#endif
149
150/**********************************************************************
151 *
152 * Some handy macros
153 *
154 */
155
156#define R(foo) ((foo)-KERNBASE)
157
158#define ALLOCPAGES(foo) \
159	movl	R(physfree), %esi ; \
160	movl	$((foo)*PAGE_SIZE), %eax ; \
161	addl	%esi, %eax ; \
162	movl	%eax, R(physfree) ; \
163	movl	%esi, %edi ; \
164	movl	$((foo)*PAGE_SIZE),%ecx ; \
165	xorl	%eax,%eax ; \
166	cld ; \
167	rep ; \
168	stosb
169
170/*
171 * fillkpt
172 *	eax = page frame address
173 *	ebx = index into page table
174 *	ecx = how many pages to map
175 * 	base = base address of page dir/table
176 *	prot = protection bits
177 */
178#define	fillkpt(base, prot)		  \
179	shll	$PTESHIFT,%ebx		; \
180	addl	base,%ebx		; \
181	orl	$PG_V,%eax		; \
182	orl	prot,%eax		; \
1831:	movl	%eax,(%ebx)		; \
184	addl	$PAGE_SIZE,%eax		; /* increment physical address */ \
185	addl	$PTESIZE,%ebx		; /* next pte */ \
186	loop	1b
187
188/*
189 * fillkptphys(prot)
190 *	eax = physical address
191 *	ecx = how many pages to map
192 *	prot = protection bits
193 */
194#define	fillkptphys(prot)		  \
195	movl	%eax, %ebx		; \
196	shrl	$PAGE_SHIFT, %ebx	; \
197	fillkpt(R(KPTphys), prot)
198
199	.text
200/**********************************************************************
201 *
202 * This is where the bootblocks start us, set the ball rolling...
203 *
204 */
205NON_GPROF_ENTRY(btext)
206
207#ifdef PC98
208	/* save SYSTEM PARAMETER for resume (NS/T or other) */
209	movl	$0xa1400,%esi
210	movl	$R(pc98_system_parameter),%edi
211	movl	$0x0240,%ecx
212	cld
213	rep
214	movsb
215#else	/* IBM-PC */
216/* Tell the bios to warmboot next time */
217	movw	$0x1234,0x472
218#endif	/* PC98 */
219
220/* Set up a real frame in case the double return in newboot is executed. */
221	pushl	%ebp
222	movl	%esp, %ebp
223
224/* Don't trust what the BIOS gives for eflags. */
225	pushl	$PSL_KERNEL
226	popfl
227
228/*
229 * Don't trust what the BIOS gives for %fs and %gs.  Trust the bootstrap
230 * to set %cs, %ds, %es and %ss.
231 */
232	mov	%ds, %ax
233	mov	%ax, %fs
234	mov	%ax, %gs
235
236/*
237 * Clear the bss.  Not all boot programs do it, and it is our job anyway.
238 *
239 * XXX we don't check that there is memory for our bss and page tables
240 * before using it.
241 *
242 * Note: we must be careful to not overwrite an active gdt or idt.  They
243 * inactive from now until we switch to new ones, since we don't load any
244 * more segment registers or permit interrupts until after the switch.
245 */
246	movl	$R(end),%ecx
247	movl	$R(edata),%edi
248	subl	%edi,%ecx
249	xorl	%eax,%eax
250	cld
251	rep
252	stosb
253
254	call	recover_bootinfo
255
256/* Get onto a stack that we can trust. */
257/*
258 * XXX this step is delayed in case recover_bootinfo needs to return via
259 * the old stack, but it need not be, since recover_bootinfo actually
260 * returns via the old frame.
261 */
262	movl	$R(tmpstk),%esp
263
264#ifdef PC98
265	/* pc98_machine_type & M_EPSON_PC98 */
266	testb	$0x02,R(pc98_system_parameter)+220
267	jz	3f
268	/* epson_machine_id <= 0x0b */
269	cmpb	$0x0b,R(pc98_system_parameter)+224
270	ja	3f
271
272	/* count up memory */
273	movl	$0x100000,%eax		/* next, talley remaining memory */
274	movl	$0xFFF-0x100,%ecx
2751:	movl	0(%eax),%ebx		/* save location to check */
276	movl	$0xa55a5aa5,0(%eax)	/* write test pattern */
277	cmpl	$0xa55a5aa5,0(%eax)	/* does not check yet for rollover */
278	jne	2f
279	movl	%ebx,0(%eax)		/* restore memory */
280	addl	$PAGE_SIZE,%eax
281	loop	1b
2822:	subl	$0x100000,%eax
283	shrl	$17,%eax
284	movb	%al,R(pc98_system_parameter)+1
2853:
286
287	movw	R(pc98_system_parameter+0x86),%ax
288	movw	%ax,R(cpu_id)
289#endif
290
291	call	identify_cpu
292	call	create_pagetables
293
294/*
295 * If the CPU has support for VME, turn it on.
296 */
297	testl	$CPUID_VME, R(cpu_feature)
298	jz	1f
299	movl	%cr4, %eax
300	orl	$CR4_VME, %eax
301	movl	%eax, %cr4
3021:
303
304/* Now enable paging */
305#ifdef PAE
306	movl	R(IdlePDPT), %eax
307	movl	%eax, %cr3
308	movl	%cr4, %eax
309	orl	$CR4_PAE, %eax
310	movl	%eax, %cr4
311#else
312	movl	R(IdlePTD), %eax
313	movl	%eax,%cr3		/* load ptd addr into mmu */
314#endif
315	movl	%cr0,%eax		/* get control word */
316	orl	$CR0_PE|CR0_PG,%eax	/* enable paging */
317	movl	%eax,%cr0		/* and let's page NOW! */
318
319	pushl	$begin			/* jump to high virtualized address */
320	ret
321
322/* now running relocated at KERNBASE where the system is linked to run */
323begin:
324	/* set up bootstrap stack */
325	movl	proc0kstack,%eax	/* location of in-kernel stack */
326			/* bootstrap stack end location */
327	leal	(KSTACK_PAGES*PAGE_SIZE-PCB_SIZE)(%eax),%esp
328
329	xorl	%ebp,%ebp		/* mark end of frames */
330
331#ifdef PAE
332	movl	IdlePDPT,%esi
333#else
334	movl	IdlePTD,%esi
335#endif
336	movl	%esi,(KSTACK_PAGES*PAGE_SIZE-PCB_SIZE+PCB_CR3)(%eax)
337
338	pushl	physfree		/* value of first for init386(first) */
339	call	init386			/* wire 386 chip for unix operation */
340
341	/*
342	 * Clean up the stack in a way that db_numargs() understands, so
343	 * that backtraces in ddb don't underrun the stack.  Traps for
344	 * inaccessible memory are more fatal than usual this early.
345	 */
346	addl	$4,%esp
347
348	call	mi_startup		/* autoconfiguration, mountroot etc */
349	/* NOTREACHED */
350	addl	$0,%esp			/* for db_numargs() again */
351
352/*
353 * Signal trampoline, copied to top of user stack
354 */
355NON_GPROF_ENTRY(sigcode)
356	calll	*SIGF_HANDLER(%esp)
357	leal	SIGF_UC(%esp),%eax	/* get ucontext */
358	pushl	%eax
359	testl	$PSL_VM,UC_EFLAGS(%eax)
360	jne	1f
361	movl	UC_GS(%eax),%gs		/* restore %gs */
3621:
363	movl	$SYS_sigreturn,%eax
364	pushl	%eax			/* junk to fake return addr. */
365	int	$0x80			/* enter kernel with args */
366					/* on stack */
3671:
368	jmp	1b
369
370#ifdef COMPAT_FREEBSD4
371	ALIGN_TEXT
372freebsd4_sigcode:
373	calll	*SIGF_HANDLER(%esp)
374	leal	SIGF_UC4(%esp),%eax	/* get ucontext */
375	pushl	%eax
376	testl	$PSL_VM,UC4_EFLAGS(%eax)
377	jne	1f
378	movl	UC4_GS(%eax),%gs	/* restore %gs */
3791:
380	movl	$344,%eax		/* 4.x SYS_sigreturn */
381	pushl	%eax			/* junk to fake return addr. */
382	int	$0x80			/* enter kernel with args */
383					/* on stack */
3841:
385	jmp	1b
386#endif
387
388#ifdef COMPAT_43
389	ALIGN_TEXT
390osigcode:
391	call	*SIGF_HANDLER(%esp)	/* call signal handler */
392	lea	SIGF_SC(%esp),%eax	/* get sigcontext */
393	pushl	%eax
394	testl	$PSL_VM,SC_PS(%eax)
395	jne	9f
396	movl	SC_GS(%eax),%gs		/* restore %gs */
3979:
398	movl	$103,%eax		/* 3.x SYS_sigreturn */
399	pushl	%eax			/* junk to fake return addr. */
400	int	$0x80			/* enter kernel with args */
4010:	jmp	0b
402#endif /* COMPAT_43 */
403
404	ALIGN_TEXT
405esigcode:
406
407	.data
408	.globl	szsigcode
409szsigcode:
410	.long	esigcode-sigcode
411#ifdef COMPAT_FREEBSD4
412	.globl	szfreebsd4_sigcode
413szfreebsd4_sigcode:
414	.long	esigcode-freebsd4_sigcode
415#endif
416#ifdef COMPAT_43
417	.globl	szosigcode
418szosigcode:
419	.long	esigcode-osigcode
420#endif
421	.text
422
423/**********************************************************************
424 *
425 * Recover the bootinfo passed to us from the boot program
426 *
427 */
428recover_bootinfo:
429	/*
430	 * This code is called in different ways depending on what loaded
431	 * and started the kernel.  This is used to detect how we get the
432	 * arguments from the other code and what we do with them.
433	 *
434	 * Old disk boot blocks:
435	 *	(*btext)(howto, bootdev, cyloffset, esym);
436	 *	[return address == 0, and can NOT be returned to]
437	 *	[cyloffset was not supported by the FreeBSD boot code
438	 *	 and always passed in as 0]
439	 *	[esym is also known as total in the boot code, and
440	 *	 was never properly supported by the FreeBSD boot code]
441	 *
442	 * Old diskless netboot code:
443	 *	(*btext)(0,0,0,0,&nfsdiskless,0,0,0);
444	 *	[return address != 0, and can NOT be returned to]
445	 *	If we are being booted by this code it will NOT work,
446	 *	so we are just going to halt if we find this case.
447	 *
448	 * New uniform boot code:
449	 *	(*btext)(howto, bootdev, 0, 0, 0, &bootinfo)
450	 *	[return address != 0, and can be returned to]
451	 *
452	 * There may seem to be a lot of wasted arguments in here, but
453	 * that is so the newer boot code can still load very old kernels
454	 * and old boot code can load new kernels.
455	 */
456
457	/*
458	 * The old style disk boot blocks fake a frame on the stack and
459	 * did an lret to get here.  The frame on the stack has a return
460	 * address of 0.
461	 */
462	cmpl	$0,4(%ebp)
463	je	olddiskboot
464
465	/*
466	 * We have some form of return address, so this is either the
467	 * old diskless netboot code, or the new uniform code.  That can
468	 * be detected by looking at the 5th argument, if it is 0
469	 * we are being booted by the new uniform boot code.
470	 */
471	cmpl	$0,24(%ebp)
472	je	newboot
473
474	/*
475	 * Seems we have been loaded by the old diskless boot code, we
476	 * don't stand a chance of running as the diskless structure
477	 * changed considerably between the two, so just halt.
478	 */
479	 hlt
480
481	/*
482	 * We have been loaded by the new uniform boot code.
483	 * Let's check the bootinfo version, and if we do not understand
484	 * it we return to the loader with a status of 1 to indicate this error
485	 */
486newboot:
487	movl	28(%ebp),%ebx		/* &bootinfo.version */
488	movl	BI_VERSION(%ebx),%eax
489	cmpl	$1,%eax			/* We only understand version 1 */
490	je	1f
491	movl	$1,%eax			/* Return status */
492	leave
493	/*
494	 * XXX this returns to our caller's caller (as is required) since
495	 * we didn't set up a frame and our caller did.
496	 */
497	ret
498
4991:
500	/*
501	 * If we have a kernelname copy it in
502	 */
503	movl	BI_KERNELNAME(%ebx),%esi
504	cmpl	$0,%esi
505	je	2f			/* No kernelname */
506	movl	$MAXPATHLEN,%ecx	/* Brute force!!! */
507	movl	$R(kernelname),%edi
508	cmpb	$'/',(%esi)		/* Make sure it starts with a slash */
509	je	1f
510	movb	$'/',(%edi)
511	incl	%edi
512	decl	%ecx
5131:
514	cld
515	rep
516	movsb
517
5182:
519	/*
520	 * Determine the size of the boot loader's copy of the bootinfo
521	 * struct.  This is impossible to do properly because old versions
522	 * of the struct don't contain a size field and there are 2 old
523	 * versions with the same version number.
524	 */
525	movl	$BI_ENDCOMMON,%ecx	/* prepare for sizeless version */
526	testl	$RB_BOOTINFO,8(%ebp)	/* bi_size (and bootinfo) valid? */
527	je	got_bi_size		/* no, sizeless version */
528	movl	BI_SIZE(%ebx),%ecx
529got_bi_size:
530
531	/*
532	 * Copy the common part of the bootinfo struct
533	 */
534	movl	%ebx,%esi
535	movl	$R(bootinfo),%edi
536	cmpl	$BOOTINFO_SIZE,%ecx
537	jbe	got_common_bi_size
538	movl	$BOOTINFO_SIZE,%ecx
539got_common_bi_size:
540	cld
541	rep
542	movsb
543
544#ifdef NFS_ROOT
545#ifndef BOOTP_NFSV3
546	/*
547	 * If we have a nfs_diskless structure copy it in
548	 */
549	movl	BI_NFS_DISKLESS(%ebx),%esi
550	cmpl	$0,%esi
551	je	olddiskboot
552	movl	$R(nfs_diskless),%edi
553	movl	$NFSDISKLESS_SIZE,%ecx
554	cld
555	rep
556	movsb
557	movl	$R(nfs_diskless_valid),%edi
558	movl	$1,(%edi)
559#endif
560#endif
561
562	/*
563	 * The old style disk boot.
564	 *	(*btext)(howto, bootdev, cyloffset, esym);
565	 * Note that the newer boot code just falls into here to pick
566	 * up howto and bootdev, cyloffset and esym are no longer used
567	 */
568olddiskboot:
569	movl	8(%ebp),%eax
570	movl	%eax,R(boothowto)
571	movl	12(%ebp),%eax
572	movl	%eax,R(bootdev)
573
574	ret
575
576
577/**********************************************************************
578 *
579 * Identify the CPU and initialize anything special about it
580 *
581 */
582identify_cpu:
583
584	/* Try to toggle alignment check flag; does not exist on 386. */
585	pushfl
586	popl	%eax
587	movl	%eax,%ecx
588	orl	$PSL_AC,%eax
589	pushl	%eax
590	popfl
591	pushfl
592	popl	%eax
593	xorl	%ecx,%eax
594	andl	$PSL_AC,%eax
595	pushl	%ecx
596	popfl
597
598	testl	%eax,%eax
599	jnz	try486
600
601	/* NexGen CPU does not have aligment check flag. */
602	pushfl
603	movl	$0x5555, %eax
604	xorl	%edx, %edx
605	movl	$2, %ecx
606	clc
607	divl	%ecx
608	jz	trynexgen
609	popfl
610	movl	$CPU_386,R(cpu)
611	jmp	3f
612
613trynexgen:
614	popfl
615	movl	$CPU_NX586,R(cpu)
616	movl	$0x4778654e,R(cpu_vendor)	# store vendor string
617	movl	$0x72446e65,R(cpu_vendor+4)
618	movl	$0x6e657669,R(cpu_vendor+8)
619	movl	$0,R(cpu_vendor+12)
620	jmp	3f
621
622try486:	/* Try to toggle identification flag; does not exist on early 486s. */
623	pushfl
624	popl	%eax
625	movl	%eax,%ecx
626	xorl	$PSL_ID,%eax
627	pushl	%eax
628	popfl
629	pushfl
630	popl	%eax
631	xorl	%ecx,%eax
632	andl	$PSL_ID,%eax
633	pushl	%ecx
634	popfl
635
636	testl	%eax,%eax
637	jnz	trycpuid
638	movl	$CPU_486,R(cpu)
639
640	/*
641	 * Check Cyrix CPU
642	 * Cyrix CPUs do not change the undefined flags following
643	 * execution of the divide instruction which divides 5 by 2.
644	 *
645	 * Note: CPUID is enabled on M2, so it passes another way.
646	 */
647	pushfl
648	movl	$0x5555, %eax
649	xorl	%edx, %edx
650	movl	$2, %ecx
651	clc
652	divl	%ecx
653	jnc	trycyrix
654	popfl
655	jmp	3f		/* You may use Intel CPU. */
656
657trycyrix:
658	popfl
659	/*
660	 * IBM Bluelighting CPU also doesn't change the undefined flags.
661	 * Because IBM doesn't disclose the information for Bluelighting
662	 * CPU, we couldn't distinguish it from Cyrix's (including IBM
663	 * brand of Cyrix CPUs).
664	 */
665	movl	$0x69727943,R(cpu_vendor)	# store vendor string
666	movl	$0x736e4978,R(cpu_vendor+4)
667	movl	$0x64616574,R(cpu_vendor+8)
668	jmp	3f
669
670trycpuid:	/* Use the `cpuid' instruction. */
671	xorl	%eax,%eax
672	cpuid					# cpuid 0
673	movl	%eax,R(cpu_high)		# highest capability
674	movl	%ebx,R(cpu_vendor)		# store vendor string
675	movl	%edx,R(cpu_vendor+4)
676	movl	%ecx,R(cpu_vendor+8)
677	movb	$0,R(cpu_vendor+12)
678
679	movl	$1,%eax
680	cpuid					# cpuid 1
681	movl	%eax,R(cpu_id)			# store cpu_id
682	movl	%ebx,R(cpu_procinfo)		# store cpu_procinfo
683	movl	%edx,R(cpu_feature)		# store cpu_feature
684	movl	%ecx,R(cpu_feature2)		# store cpu_feature2
685	rorl	$8,%eax				# extract family type
686	andl	$15,%eax
687	cmpl	$5,%eax
688	jae	1f
689
690	/* less than Pentium; must be 486 */
691	movl	$CPU_486,R(cpu)
692	jmp	3f
6931:
694	/* a Pentium? */
695	cmpl	$5,%eax
696	jne	2f
697	movl	$CPU_586,R(cpu)
698	jmp	3f
6992:
700	/* Greater than Pentium...call it a Pentium Pro */
701	movl	$CPU_686,R(cpu)
7023:
703	ret
704
705
706/**********************************************************************
707 *
708 * Create the first page directory and its page tables.
709 *
710 */
711
712create_pagetables:
713
714/* Find end of kernel image (rounded up to a page boundary). */
715	movl	$R(_end),%esi
716
717/* Include symbols, if any. */
718	movl	R(bootinfo+BI_ESYMTAB),%edi
719	testl	%edi,%edi
720	je	over_symalloc
721	movl	%edi,%esi
722	movl	$KERNBASE,%edi
723	addl	%edi,R(bootinfo+BI_SYMTAB)
724	addl	%edi,R(bootinfo+BI_ESYMTAB)
725over_symalloc:
726
727/* If we are told where the end of the kernel space is, believe it. */
728	movl	R(bootinfo+BI_KERNEND),%edi
729	testl	%edi,%edi
730	je	no_kernend
731	movl	%edi,%esi
732no_kernend:
733
734	addl	$PDRMASK,%esi		/* Play conservative for now, and */
735	andl	$~PDRMASK,%esi		/*   ... wrap to next 4M. */
736	movl	%esi,R(KERNend)		/* save end of kernel */
737	movl	%esi,R(physfree)	/* next free page is at end of kernel */
738
739/* Allocate Kernel Page Tables */
740	ALLOCPAGES(NKPT)
741	movl	%esi,R(KPTphys)
742
743/* Allocate Page Table Directory */
744#ifdef PAE
745	/* XXX only need 32 bytes (easier for now) */
746	ALLOCPAGES(1)
747	movl	%esi,R(IdlePDPT)
748#endif
749	ALLOCPAGES(NPGPTD)
750	movl	%esi,R(IdlePTD)
751
752/* Allocate KSTACK */
753	ALLOCPAGES(KSTACK_PAGES)
754	movl	%esi,R(p0kpa)
755	addl	$KERNBASE, %esi
756	movl	%esi, R(proc0kstack)
757
758	ALLOCPAGES(1)			/* vm86/bios stack */
759	movl	%esi,R(vm86phystk)
760
761	ALLOCPAGES(3)			/* pgtable + ext + IOPAGES */
762	movl	%esi,R(vm86pa)
763	addl	$KERNBASE, %esi
764	movl	%esi, R(vm86paddr)
765
766#ifdef SMP
767/* Allocate cpu0's private data page */
768	ALLOCPAGES(1)
769	movl	%esi,R(cpu0pp)
770	addl	$KERNBASE, %esi
771	movl	%esi, R(cpu0prvpage)	/* relocated to KVM space */
772
773/* Allocate SMP page table page */
774	ALLOCPAGES(1)
775	movl	%esi,R(SMPptpa)
776	addl	$KERNBASE, %esi
777	movl	%esi, R(SMPpt)		/* relocated to KVM space */
778#endif	/* SMP */
779
780/* Map page zero read-write so bios32 calls can use it */
781	xorl	%eax, %eax
782	movl	$PG_RW,%edx
783	movl	$1,%ecx
784	fillkptphys(%edx)
785
786/* Map read-only from page 1 to the beginning of the kernel text section */
787	movl	$PAGE_SIZE, %eax
788	xorl	%edx,%edx
789	movl	$R(btext),%ecx
790	addl	$PAGE_MASK,%ecx
791	subl	%eax,%ecx
792	shrl	$PAGE_SHIFT,%ecx
793	fillkptphys(%edx)
794
795/*
796 * Enable PSE and PGE.
797 */
798#ifndef DISABLE_PSE
799	testl	$CPUID_PSE, R(cpu_feature)
800	jz	1f
801	movl	$PG_PS, R(pseflag)
802	movl	%cr4, %eax
803	orl	$CR4_PSE, %eax
804	movl	%eax, %cr4
8051:
806#endif
807#ifndef DISABLE_PG_G
808	testl	$CPUID_PGE, R(cpu_feature)
809	jz	2f
810	movl	$PG_G, R(pgeflag)
811	movl	%cr4, %eax
812	orl	$CR4_PGE, %eax
813	movl	%eax, %cr4
8142:
815#endif
816
817/*
818 * Write page tables for the kernel starting at btext and
819 * until the end.  Make sure to map read+write.  We do this even
820 * if we've enabled PSE above, we'll just switch the corresponding kernel
821 * PDEs before we turn on paging.
822 *
823 * XXX: We waste some pages here in the PSE case!  DON'T BLINDLY REMOVE
824 * THIS!  SMP needs the page table to be there to map the kernel P==V.
825 */
826	movl	$R(btext),%eax
827	addl	$PAGE_MASK, %eax
828	andl	$~PAGE_MASK, %eax
829	movl	$PG_RW,%edx
830	movl	R(KERNend),%ecx
831	subl	%eax,%ecx
832	shrl	$PAGE_SHIFT,%ecx
833	fillkptphys(%edx)
834
835/* Map page directory. */
836#ifdef PAE
837	movl	R(IdlePDPT), %eax
838	movl	$1, %ecx
839	fillkptphys($PG_RW)
840#endif
841
842	movl	R(IdlePTD), %eax
843	movl	$NPGPTD, %ecx
844	fillkptphys($PG_RW)
845
846/* Map proc0's KSTACK in the physical way ... */
847	movl	R(p0kpa), %eax
848	movl	$(KSTACK_PAGES), %ecx
849	fillkptphys($PG_RW)
850
851/* Map ISA hole */
852	movl	$ISA_HOLE_START, %eax
853	movl	$ISA_HOLE_LENGTH>>PAGE_SHIFT, %ecx
854	fillkptphys($PG_RW)
855
856/* Map space for the vm86 region */
857	movl	R(vm86phystk), %eax
858	movl	$4, %ecx
859	fillkptphys($PG_RW)
860
861/* Map page 0 into the vm86 page table */
862	movl	$0, %eax
863	movl	$0, %ebx
864	movl	$1, %ecx
865	fillkpt(R(vm86pa), $PG_RW|PG_U)
866
867/* ...likewise for the ISA hole */
868	movl	$ISA_HOLE_START, %eax
869	movl	$ISA_HOLE_START>>PAGE_SHIFT, %ebx
870	movl	$ISA_HOLE_LENGTH>>PAGE_SHIFT, %ecx
871	fillkpt(R(vm86pa), $PG_RW|PG_U)
872
873#ifdef SMP
874/* Map cpu0's private page into global kmem (4K @ cpu0prvpage) */
875	movl	R(cpu0pp), %eax
876	movl	$1, %ecx
877	fillkptphys($PG_RW)
878
879/* Map SMP page table page into global kmem FWIW */
880	movl	R(SMPptpa), %eax
881	movl	$1, %ecx
882	fillkptphys($PG_RW)
883
884/* Map the private page into the SMP page table */
885	movl	R(cpu0pp), %eax
886	movl	$0, %ebx		/* pte offset = 0 */
887	movl	$1, %ecx		/* one private page coming right up */
888	fillkpt(R(SMPptpa), $PG_RW)
889
890/* ... and put the page table table in the pde. */
891	movl	R(SMPptpa), %eax
892	movl	$MPPTDI, %ebx
893	movl	$1, %ecx
894	fillkpt(R(IdlePTD), $PG_RW)
895
896/* Fakeup VA for the local apic to allow early traps. */
897	ALLOCPAGES(1)
898	movl	%esi, %eax
899	movl	$(NPTEPG-1), %ebx	/* pte offset = NTEPG-1 */
900	movl	$1, %ecx		/* one private pt coming right up */
901	fillkpt(R(SMPptpa), $PG_RW)
902#endif	/* SMP */
903
904/* install a pde for temporary double map of bottom of VA */
905	movl	R(KPTphys), %eax
906	xorl	%ebx, %ebx
907	movl	$NKPT, %ecx
908	fillkpt(R(IdlePTD), $PG_RW)
909
910/*
911 * For the non-PSE case, install PDEs for PTs covering the kernel.
912 * For the PSE case, do the same, but clobber the ones corresponding
913 * to the kernel (from btext to KERNend) with 4M ('PS') PDEs immediately
914 * after.
915 */
916	movl	R(KPTphys), %eax
917	movl	$KPTDI, %ebx
918	movl	$NKPT, %ecx
919	fillkpt(R(IdlePTD), $PG_RW)
920	cmpl	$0,R(pseflag)
921	je	done_pde
922
923	movl	R(KERNend), %ecx
924	movl	$KERNLOAD, %eax
925	subl	%eax, %ecx
926	shrl	$PDRSHIFT, %ecx
927	movl	$(KPTDI+(KERNLOAD/(1 << PDRSHIFT))), %ebx
928	shll	$PDESHIFT, %ebx
929	addl	R(IdlePTD), %ebx
930	orl	$(PG_V|PG_RW|PG_PS), %eax
9311:	movl	%eax, (%ebx)
932	addl	$(1 << PDRSHIFT), %eax
933	addl	$PDESIZE, %ebx
934	loop	1b
935
936done_pde:
937/* install a pde recursively mapping page directory as a page table */
938	movl	R(IdlePTD), %eax
939	movl	$PTDPTDI, %ebx
940	movl	$NPGPTD,%ecx
941	fillkpt(R(IdlePTD), $PG_RW)
942
943#ifdef PAE
944	movl	R(IdlePTD), %eax
945	xorl	%ebx, %ebx
946	movl	$NPGPTD, %ecx
947	fillkpt(R(IdlePDPT), $0x0)
948#endif
949
950	ret
951