xref: /openbmc/linux/arch/powerpc/kernel/head_64.S (revision 6b5fc336)
1/*
2 *  PowerPC version
3 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
4 *
5 *  Rewritten by Cort Dougan (cort@cs.nmt.edu) for PReP
6 *    Copyright (C) 1996 Cort Dougan <cort@cs.nmt.edu>
7 *  Adapted for Power Macintosh by Paul Mackerras.
8 *  Low-level exception handlers and MMU support
9 *  rewritten by Paul Mackerras.
10 *    Copyright (C) 1996 Paul Mackerras.
11 *
12 *  Adapted for 64bit PowerPC by Dave Engebretsen, Peter Bergner, and
13 *    Mike Corrigan {engebret|bergner|mikejc}@us.ibm.com
14 *
15 *  This file contains the entry point for the 64-bit kernel along
16 *  with some early initialization code common to all 64-bit powerpc
17 *  variants.
18 *
19 *  This program is free software; you can redistribute it and/or
20 *  modify it under the terms of the GNU General Public License
21 *  as published by the Free Software Foundation; either version
22 *  2 of the License, or (at your option) any later version.
23 */
24
25#include <linux/threads.h>
26#include <linux/init.h>
27#include <asm/reg.h>
28#include <asm/page.h>
29#include <asm/mmu.h>
30#include <asm/ppc_asm.h>
31#include <asm/head-64.h>
32#include <asm/asm-offsets.h>
33#include <asm/bug.h>
34#include <asm/cputable.h>
35#include <asm/setup.h>
36#include <asm/hvcall.h>
37#include <asm/thread_info.h>
38#include <asm/firmware.h>
39#include <asm/page_64.h>
40#include <asm/irqflags.h>
41#include <asm/kvm_book3s_asm.h>
42#include <asm/ptrace.h>
43#include <asm/hw_irq.h>
44#include <asm/cputhreads.h>
45#include <asm/ppc-opcode.h>
46#include <asm/export.h>
47
48/* The physical memory is laid out such that the secondary processor
49 * spin code sits at 0x0000...0x00ff. On server, the vectors follow
50 * using the layout described in exceptions-64s.S
51 */
52
53/*
54 * Entering into this code we make the following assumptions:
55 *
56 *  For pSeries or server processors:
57 *   1. The MMU is off & open firmware is running in real mode.
58 *   2. The kernel is entered at __start
59 * -or- For OPAL entry:
60 *   1. The MMU is off, processor in HV mode, primary CPU enters at 0
61 *      with device-tree in gpr3. We also get OPAL base in r8 and
62 *	entry in r9 for debugging purposes
63 *   2. Secondary processors enter at 0x60 with PIR in gpr3
64 *
65 *  For Book3E processors:
66 *   1. The MMU is on running in AS0 in a state defined in ePAPR
67 *   2. The kernel is entered at __start
68 */
69
70OPEN_FIXED_SECTION(first_256B, 0x0, 0x100)
71USE_FIXED_SECTION(first_256B)
72	/*
73	 * Offsets are relative from the start of fixed section, and
74	 * first_256B starts at 0. Offsets are a bit easier to use here
75	 * than the fixed section entry macros.
76	 */
77	. = 0x0
78_GLOBAL(__start)
79	/* NOP this out unconditionally */
80BEGIN_FTR_SECTION
81	FIXUP_ENDIAN
82	b	__start_initialization_multiplatform
83END_FTR_SECTION(0, 1)
84
85	/* Catch branch to 0 in real mode */
86	trap
87
88	/* Secondary processors spin on this value until it becomes non-zero.
89	 * When non-zero, it contains the real address of the function the cpu
90	 * should jump to.
91	 */
92	.balign 8
93	.globl  __secondary_hold_spinloop
94__secondary_hold_spinloop:
95	.llong	0x0
96
97	/* Secondary processors write this value with their cpu # */
98	/* after they enter the spin loop immediately below.	  */
99	.globl	__secondary_hold_acknowledge
100__secondary_hold_acknowledge:
101	.llong	0x0
102
103#ifdef CONFIG_RELOCATABLE
104	/* This flag is set to 1 by a loader if the kernel should run
105	 * at the loaded address instead of the linked address.  This
106	 * is used by kexec-tools to keep the the kdump kernel in the
107	 * crash_kernel region.  The loader is responsible for
108	 * observing the alignment requirement.
109	 */
110
111#ifdef CONFIG_RELOCATABLE_TEST
112#define RUN_AT_LOAD_DEFAULT 1		/* Test relocation, do not copy to 0 */
113#else
114#define RUN_AT_LOAD_DEFAULT 0x72756e30  /* "run0" -- relocate to 0 by default */
115#endif
116
117	/* Do not move this variable as kexec-tools knows about it. */
118	. = 0x5c
119	.globl	__run_at_load
120__run_at_load:
121DEFINE_FIXED_SYMBOL(__run_at_load)
122	.long	RUN_AT_LOAD_DEFAULT
123#endif
124
125	. = 0x60
126/*
127 * The following code is used to hold secondary processors
128 * in a spin loop after they have entered the kernel, but
129 * before the bulk of the kernel has been relocated.  This code
130 * is relocated to physical address 0x60 before prom_init is run.
131 * All of it must fit below the first exception vector at 0x100.
132 * Use .globl here not _GLOBAL because we want __secondary_hold
133 * to be the actual text address, not a descriptor.
134 */
135	.globl	__secondary_hold
136__secondary_hold:
137	FIXUP_ENDIAN
138#ifndef CONFIG_PPC_BOOK3E
139	mfmsr	r24
140	ori	r24,r24,MSR_RI
141	mtmsrd	r24			/* RI on */
142#endif
143	/* Grab our physical cpu number */
144	mr	r24,r3
145	/* stash r4 for book3e */
146	mr	r25,r4
147
148	/* Tell the master cpu we're here */
149	/* Relocation is off & we are located at an address less */
150	/* than 0x100, so only need to grab low order offset.    */
151	std	r24,(ABS_ADDR(__secondary_hold_acknowledge))(0)
152	sync
153
154	li	r26,0
155#ifdef CONFIG_PPC_BOOK3E
156	tovirt(r26,r26)
157#endif
158	/* All secondary cpus wait here until told to start. */
159100:	ld	r12,(ABS_ADDR(__secondary_hold_spinloop))(r26)
160	cmpdi	0,r12,0
161	beq	100b
162
163#if defined(CONFIG_SMP) || defined(CONFIG_KEXEC_CORE)
164#ifdef CONFIG_PPC_BOOK3E
165	tovirt(r12,r12)
166#endif
167	mtctr	r12
168	mr	r3,r24
169	/*
170	 * it may be the case that other platforms have r4 right to
171	 * begin with, this gives us some safety in case it is not
172	 */
173#ifdef CONFIG_PPC_BOOK3E
174	mr	r4,r25
175#else
176	li	r4,0
177#endif
178	/* Make sure that patched code is visible */
179	isync
180	bctr
181#else
182	BUG_OPCODE
183#endif
184CLOSE_FIXED_SECTION(first_256B)
185
186/* This value is used to mark exception frames on the stack. */
187	.section ".toc","aw"
188exception_marker:
189	.tc	ID_72656773_68657265[TC],0x7265677368657265
190	.previous
191
192/*
193 * On server, we include the exception vectors code here as it
194 * relies on absolute addressing which is only possible within
195 * this compilation unit
196 */
197#ifdef CONFIG_PPC_BOOK3S
198#include "exceptions-64s.S"
199#else
200OPEN_TEXT_SECTION(0x100)
201#endif
202
203USE_TEXT_SECTION()
204
205#ifdef CONFIG_PPC_BOOK3E
206/*
207 * The booting_thread_hwid holds the thread id we want to boot in cpu
208 * hotplug case. It is set by cpu hotplug code, and is invalid by default.
209 * The thread id is the same as the initial value of SPRN_PIR[THREAD_ID]
210 * bit field.
211 */
212	.globl	booting_thread_hwid
213booting_thread_hwid:
214	.long  INVALID_THREAD_HWID
215	.align 3
216/*
217 * start a thread in the same core
218 * input parameters:
219 * r3 = the thread physical id
220 * r4 = the entry point where thread starts
221 */
222_GLOBAL(book3e_start_thread)
223	LOAD_REG_IMMEDIATE(r5, MSR_KERNEL)
224	cmpwi	r3, 0
225	beq	10f
226	cmpwi	r3, 1
227	beq	11f
228	/* If the thread id is invalid, just exit. */
229	b	13f
23010:
231	MTTMR(TMRN_IMSR0, 5)
232	MTTMR(TMRN_INIA0, 4)
233	b	12f
23411:
235	MTTMR(TMRN_IMSR1, 5)
236	MTTMR(TMRN_INIA1, 4)
23712:
238	isync
239	li	r6, 1
240	sld	r6, r6, r3
241	mtspr	SPRN_TENS, r6
24213:
243	blr
244
245/*
246 * stop a thread in the same core
247 * input parameter:
248 * r3 = the thread physical id
249 */
250_GLOBAL(book3e_stop_thread)
251	cmpwi	r3, 0
252	beq	10f
253	cmpwi	r3, 1
254	beq	10f
255	/* If the thread id is invalid, just exit. */
256	b	13f
25710:
258	li	r4, 1
259	sld	r4, r4, r3
260	mtspr	SPRN_TENC, r4
26113:
262	blr
263
264_GLOBAL(fsl_secondary_thread_init)
265	mfspr	r4,SPRN_BUCSR
266
267	/* Enable branch prediction */
268	lis     r3,BUCSR_INIT@h
269	ori     r3,r3,BUCSR_INIT@l
270	mtspr   SPRN_BUCSR,r3
271	isync
272
273	/*
274	 * Fix PIR to match the linear numbering in the device tree.
275	 *
276	 * On e6500, the reset value of PIR uses the low three bits for
277	 * the thread within a core, and the upper bits for the core
278	 * number.  There are two threads per core, so shift everything
279	 * but the low bit right by two bits so that the cpu numbering is
280	 * continuous.
281	 *
282	 * If the old value of BUCSR is non-zero, this thread has run
283	 * before.  Thus, we assume we are coming from kexec or a similar
284	 * scenario, and PIR is already set to the correct value.  This
285	 * is a bit of a hack, but there are limited opportunities for
286	 * getting information into the thread and the alternatives
287	 * seemed like they'd be overkill.  We can't tell just by looking
288	 * at the old PIR value which state it's in, since the same value
289	 * could be valid for one thread out of reset and for a different
290	 * thread in Linux.
291	 */
292
293	mfspr	r3, SPRN_PIR
294	cmpwi	r4,0
295	bne	1f
296	rlwimi	r3, r3, 30, 2, 30
297	mtspr	SPRN_PIR, r3
2981:
299#endif
300
301_GLOBAL(generic_secondary_thread_init)
302	mr	r24,r3
303
304	/* turn on 64-bit mode */
305	bl	enable_64b_mode
306
307	/* get a valid TOC pointer, wherever we're mapped at */
308	bl	relative_toc
309	tovirt(r2,r2)
310
311#ifdef CONFIG_PPC_BOOK3E
312	/* Book3E initialization */
313	mr	r3,r24
314	bl	book3e_secondary_thread_init
315#endif
316	b	generic_secondary_common_init
317
318/*
319 * On pSeries and most other platforms, secondary processors spin
320 * in the following code.
321 * At entry, r3 = this processor's number (physical cpu id)
322 *
323 * On Book3E, r4 = 1 to indicate that the initial TLB entry for
324 * this core already exists (setup via some other mechanism such
325 * as SCOM before entry).
326 */
327_GLOBAL(generic_secondary_smp_init)
328	FIXUP_ENDIAN
329	mr	r24,r3
330	mr	r25,r4
331
332	/* turn on 64-bit mode */
333	bl	enable_64b_mode
334
335	/* get a valid TOC pointer, wherever we're mapped at */
336	bl	relative_toc
337	tovirt(r2,r2)
338
339#ifdef CONFIG_PPC_BOOK3E
340	/* Book3E initialization */
341	mr	r3,r24
342	mr	r4,r25
343	bl	book3e_secondary_core_init
344
345/*
346 * After common core init has finished, check if the current thread is the
347 * one we wanted to boot. If not, start the specified thread and stop the
348 * current thread.
349 */
350	LOAD_REG_ADDR(r4, booting_thread_hwid)
351	lwz     r3, 0(r4)
352	li	r5, INVALID_THREAD_HWID
353	cmpw	r3, r5
354	beq	20f
355
356	/*
357	 * The value of booting_thread_hwid has been stored in r3,
358	 * so make it invalid.
359	 */
360	stw	r5, 0(r4)
361
362	/*
363	 * Get the current thread id and check if it is the one we wanted.
364	 * If not, start the one specified in booting_thread_hwid and stop
365	 * the current thread.
366	 */
367	mfspr	r8, SPRN_TIR
368	cmpw	r3, r8
369	beq	20f
370
371	/* start the specified thread */
372	LOAD_REG_ADDR(r5, fsl_secondary_thread_init)
373	ld	r4, 0(r5)
374	bl	book3e_start_thread
375
376	/* stop the current thread */
377	mr	r3, r8
378	bl	book3e_stop_thread
37910:
380	b	10b
38120:
382#endif
383
384generic_secondary_common_init:
385	/* Set up a paca value for this processor. Since we have the
386	 * physical cpu id in r24, we need to search the pacas to find
387	 * which logical id maps to our physical one.
388	 */
389	LOAD_REG_ADDR(r13, paca)	/* Load paca pointer		 */
390	ld	r13,0(r13)		/* Get base vaddr of paca array	 */
391#ifndef CONFIG_SMP
392	addi	r13,r13,PACA_SIZE	/* know r13 if used accidentally */
393	b	kexec_wait		/* wait for next kernel if !SMP	 */
394#else
395	LOAD_REG_ADDR(r7, nr_cpu_ids)	/* Load nr_cpu_ids address       */
396	lwz	r7,0(r7)		/* also the max paca allocated 	 */
397	li	r5,0			/* logical cpu id                */
3981:	lhz	r6,PACAHWCPUID(r13)	/* Load HW procid from paca      */
399	cmpw	r6,r24			/* Compare to our id             */
400	beq	2f
401	addi	r13,r13,PACA_SIZE	/* Loop to next PACA on miss     */
402	addi	r5,r5,1
403	cmpw	r5,r7			/* Check if more pacas exist     */
404	blt	1b
405
406	mr	r3,r24			/* not found, copy phys to r3	 */
407	b	kexec_wait		/* next kernel might do better	 */
408
4092:	SET_PACA(r13)
410#ifdef CONFIG_PPC_BOOK3E
411	addi	r12,r13,PACA_EXTLB	/* and TLB exc frame in another  */
412	mtspr	SPRN_SPRG_TLB_EXFRAME,r12
413#endif
414
415	/* From now on, r24 is expected to be logical cpuid */
416	mr	r24,r5
417
418	/* See if we need to call a cpu state restore handler */
419	LOAD_REG_ADDR(r23, cur_cpu_spec)
420	ld	r23,0(r23)
421	ld	r12,CPU_SPEC_RESTORE(r23)
422	cmpdi	0,r12,0
423	beq	3f
424#ifdef PPC64_ELF_ABI_v1
425	ld	r12,0(r12)
426#endif
427	mtctr	r12
428	bctrl
429
4303:	LOAD_REG_ADDR(r3, spinning_secondaries) /* Decrement spinning_secondaries */
431	lwarx	r4,0,r3
432	subi	r4,r4,1
433	stwcx.	r4,0,r3
434	bne	3b
435	isync
436
4374:	HMT_LOW
438	lbz	r23,PACAPROCSTART(r13)	/* Test if this processor should */
439					/* start.			 */
440	cmpwi	0,r23,0
441	beq	4b			/* Loop until told to go	 */
442
443	sync				/* order paca.run and cur_cpu_spec */
444	isync				/* In case code patching happened */
445
446	/* Create a temp kernel stack for use before relocation is on.	*/
447	ld	r1,PACAEMERGSP(r13)
448	subi	r1,r1,STACK_FRAME_OVERHEAD
449
450	b	__secondary_start
451#endif /* SMP */
452
453/*
454 * Turn the MMU off.
455 * Assumes we're mapped EA == RA if the MMU is on.
456 */
457#ifdef CONFIG_PPC_BOOK3S
458__mmu_off:
459	mfmsr	r3
460	andi.	r0,r3,MSR_IR|MSR_DR
461	beqlr
462	mflr	r4
463	andc	r3,r3,r0
464	mtspr	SPRN_SRR0,r4
465	mtspr	SPRN_SRR1,r3
466	sync
467	rfid
468	b	.	/* prevent speculative execution */
469#endif
470
471
472/*
473 * Here is our main kernel entry point. We support currently 2 kind of entries
474 * depending on the value of r5.
475 *
476 *   r5 != NULL -> OF entry, we go to prom_init, "legacy" parameter content
477 *                 in r3...r7
478 *
479 *   r5 == NULL -> kexec style entry. r3 is a physical pointer to the
480 *                 DT block, r4 is a physical pointer to the kernel itself
481 *
482 */
483__start_initialization_multiplatform:
484	/* Make sure we are running in 64 bits mode */
485	bl	enable_64b_mode
486
487	/* Get TOC pointer (current runtime address) */
488	bl	relative_toc
489
490	/* find out where we are now */
491	bcl	20,31,$+4
4920:	mflr	r26			/* r26 = runtime addr here */
493	addis	r26,r26,(_stext - 0b)@ha
494	addi	r26,r26,(_stext - 0b)@l	/* current runtime base addr */
495
496	/*
497	 * Are we booted from a PROM Of-type client-interface ?
498	 */
499	cmpldi	cr0,r5,0
500	beq	1f
501	b	__boot_from_prom		/* yes -> prom */
5021:
503	/* Save parameters */
504	mr	r31,r3
505	mr	r30,r4
506#ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
507	/* Save OPAL entry */
508	mr	r28,r8
509	mr	r29,r9
510#endif
511
512#ifdef CONFIG_PPC_BOOK3E
513	bl	start_initialization_book3e
514	b	__after_prom_start
515#else
516	/* Setup some critical 970 SPRs before switching MMU off */
517	mfspr	r0,SPRN_PVR
518	srwi	r0,r0,16
519	cmpwi	r0,0x39		/* 970 */
520	beq	1f
521	cmpwi	r0,0x3c		/* 970FX */
522	beq	1f
523	cmpwi	r0,0x44		/* 970MP */
524	beq	1f
525	cmpwi	r0,0x45		/* 970GX */
526	bne	2f
5271:	bl	__cpu_preinit_ppc970
5282:
529
530	/* Switch off MMU if not already off */
531	bl	__mmu_off
532	b	__after_prom_start
533#endif /* CONFIG_PPC_BOOK3E */
534
535__boot_from_prom:
536#ifdef CONFIG_PPC_OF_BOOT_TRAMPOLINE
537	/* Save parameters */
538	mr	r31,r3
539	mr	r30,r4
540	mr	r29,r5
541	mr	r28,r6
542	mr	r27,r7
543
544	/*
545	 * Align the stack to 16-byte boundary
546	 * Depending on the size and layout of the ELF sections in the initial
547	 * boot binary, the stack pointer may be unaligned on PowerMac
548	 */
549	rldicr	r1,r1,0,59
550
551#ifdef CONFIG_RELOCATABLE
552	/* Relocate code for where we are now */
553	mr	r3,r26
554	bl	relocate
555#endif
556
557	/* Restore parameters */
558	mr	r3,r31
559	mr	r4,r30
560	mr	r5,r29
561	mr	r6,r28
562	mr	r7,r27
563
564	/* Do all of the interaction with OF client interface */
565	mr	r8,r26
566	bl	prom_init
567#endif /* #CONFIG_PPC_OF_BOOT_TRAMPOLINE */
568
569	/* We never return. We also hit that trap if trying to boot
570	 * from OF while CONFIG_PPC_OF_BOOT_TRAMPOLINE isn't selected */
571	trap
572
573__after_prom_start:
574#ifdef CONFIG_RELOCATABLE
575	/* process relocations for the final address of the kernel */
576	lis	r25,PAGE_OFFSET@highest	/* compute virtual base of kernel */
577	sldi	r25,r25,32
578#if defined(CONFIG_PPC_BOOK3E)
579	tovirt(r26,r26)		/* on booke, we already run at PAGE_OFFSET */
580#endif
581	lwz	r7,(FIXED_SYMBOL_ABS_ADDR(__run_at_load))(r26)
582#if defined(CONFIG_PPC_BOOK3E)
583	tophys(r26,r26)
584#endif
585	cmplwi	cr0,r7,1	/* flagged to stay where we are ? */
586	bne	1f
587	add	r25,r25,r26
5881:	mr	r3,r25
589	bl	relocate
590#if defined(CONFIG_PPC_BOOK3E)
591	/* IVPR needs to be set after relocation. */
592	bl	init_core_book3e
593#endif
594#endif
595
596/*
597 * We need to run with _stext at physical address PHYSICAL_START.
598 * This will leave some code in the first 256B of
599 * real memory, which are reserved for software use.
600 *
601 * Note: This process overwrites the OF exception vectors.
602 */
603	li	r3,0			/* target addr */
604#ifdef CONFIG_PPC_BOOK3E
605	tovirt(r3,r3)		/* on booke, we already run at PAGE_OFFSET */
606#endif
607	mr.	r4,r26			/* In some cases the loader may  */
608#if defined(CONFIG_PPC_BOOK3E)
609	tovirt(r4,r4)
610#endif
611	beq	9f			/* have already put us at zero */
612	li	r6,0x100		/* Start offset, the first 0x100 */
613					/* bytes were copied earlier.	 */
614
615#ifdef CONFIG_RELOCATABLE
616/*
617 * Check if the kernel has to be running as relocatable kernel based on the
618 * variable __run_at_load, if it is set the kernel is treated as relocatable
619 * kernel, otherwise it will be moved to PHYSICAL_START
620 */
621#if defined(CONFIG_PPC_BOOK3E)
622	tovirt(r26,r26)		/* on booke, we already run at PAGE_OFFSET */
623#endif
624	lwz	r7,(FIXED_SYMBOL_ABS_ADDR(__run_at_load))(r26)
625	cmplwi	cr0,r7,1
626	bne	3f
627
628#ifdef CONFIG_PPC_BOOK3E
629	LOAD_REG_ADDR(r5, __end_interrupts)
630	LOAD_REG_ADDR(r11, _stext)
631	sub	r5,r5,r11
632#else
633	/* just copy interrupts */
634	LOAD_REG_IMMEDIATE(r5, FIXED_SYMBOL_ABS_ADDR(__end_interrupts))
635#endif
636	b	5f
6373:
638#endif
639	/* # bytes of memory to copy */
640	lis	r5,(ABS_ADDR(copy_to_here))@ha
641	addi	r5,r5,(ABS_ADDR(copy_to_here))@l
642
643	bl	copy_and_flush		/* copy the first n bytes	 */
644					/* this includes the code being	 */
645					/* executed here.		 */
646	/* Jump to the copy of this code that we just made */
647	addis	r8,r3,(ABS_ADDR(4f))@ha
648	addi	r12,r8,(ABS_ADDR(4f))@l
649	mtctr	r12
650	bctr
651
652.balign 8
653p_end: .llong _end - copy_to_here
654
6554:
656	/*
657	 * Now copy the rest of the kernel up to _end, add
658	 * _end - copy_to_here to the copy limit and run again.
659	 */
660	addis   r8,r26,(ABS_ADDR(p_end))@ha
661	ld      r8,(ABS_ADDR(p_end))@l(r8)
662	add	r5,r5,r8
6635:	bl	copy_and_flush		/* copy the rest */
664
6659:	b	start_here_multiplatform
666
667/*
668 * Copy routine used to copy the kernel to start at physical address 0
669 * and flush and invalidate the caches as needed.
670 * r3 = dest addr, r4 = source addr, r5 = copy limit, r6 = start offset
671 * on exit, r3, r4, r5 are unchanged, r6 is updated to be >= r5.
672 *
673 * Note: this routine *only* clobbers r0, r6 and lr
674 */
675_GLOBAL(copy_and_flush)
676	addi	r5,r5,-8
677	addi	r6,r6,-8
6784:	li	r0,8			/* Use the smallest common	*/
679					/* denominator cache line	*/
680					/* size.  This results in	*/
681					/* extra cache line flushes	*/
682					/* but operation is correct.	*/
683					/* Can't get cache line size	*/
684					/* from NACA as it is being	*/
685					/* moved too.			*/
686
687	mtctr	r0			/* put # words/line in ctr	*/
6883:	addi	r6,r6,8			/* copy a cache line		*/
689	ldx	r0,r6,r4
690	stdx	r0,r6,r3
691	bdnz	3b
692	dcbst	r6,r3			/* write it to memory		*/
693	sync
694	icbi	r6,r3			/* flush the icache line	*/
695	cmpld	0,r6,r5
696	blt	4b
697	sync
698	addi	r5,r5,8
699	addi	r6,r6,8
700	isync
701	blr
702
703.align 8
704copy_to_here:
705
706#ifdef CONFIG_SMP
707#ifdef CONFIG_PPC_PMAC
708/*
709 * On PowerMac, secondary processors starts from the reset vector, which
710 * is temporarily turned into a call to one of the functions below.
711 */
712	.section ".text";
713	.align 2 ;
714
715	.globl	__secondary_start_pmac_0
716__secondary_start_pmac_0:
717	/* NB the entries for cpus 0, 1, 2 must each occupy 8 bytes. */
718	li	r24,0
719	b	1f
720	li	r24,1
721	b	1f
722	li	r24,2
723	b	1f
724	li	r24,3
7251:
726
727_GLOBAL(pmac_secondary_start)
728	/* turn on 64-bit mode */
729	bl	enable_64b_mode
730
731	li	r0,0
732	mfspr	r3,SPRN_HID4
733	rldimi	r3,r0,40,23	/* clear bit 23 (rm_ci) */
734	sync
735	mtspr	SPRN_HID4,r3
736	isync
737	sync
738	slbia
739
740	/* get TOC pointer (real address) */
741	bl	relative_toc
742	tovirt(r2,r2)
743
744	/* Copy some CPU settings from CPU 0 */
745	bl	__restore_cpu_ppc970
746
747	/* pSeries do that early though I don't think we really need it */
748	mfmsr	r3
749	ori	r3,r3,MSR_RI
750	mtmsrd	r3			/* RI on */
751
752	/* Set up a paca value for this processor. */
753	LOAD_REG_ADDR(r4,paca)		/* Load paca pointer		*/
754	ld	r4,0(r4)		/* Get base vaddr of paca array	*/
755	mulli	r13,r24,PACA_SIZE	/* Calculate vaddr of right paca */
756	add	r13,r13,r4		/* for this processor.		*/
757	SET_PACA(r13)			/* Save vaddr of paca in an SPRG*/
758
759	/* Mark interrupts soft and hard disabled (they might be enabled
760	 * in the PACA when doing hotplug)
761	 */
762	li	r0,0
763	stb	r0,PACASOFTIRQEN(r13)
764	li	r0,PACA_IRQ_HARD_DIS
765	stb	r0,PACAIRQHAPPENED(r13)
766
767	/* Create a temp kernel stack for use before relocation is on.	*/
768	ld	r1,PACAEMERGSP(r13)
769	subi	r1,r1,STACK_FRAME_OVERHEAD
770
771	b	__secondary_start
772
773#endif /* CONFIG_PPC_PMAC */
774
775/*
776 * This function is called after the master CPU has released the
777 * secondary processors.  The execution environment is relocation off.
778 * The paca for this processor has the following fields initialized at
779 * this point:
780 *   1. Processor number
781 *   2. Segment table pointer (virtual address)
782 * On entry the following are set:
783 *   r1	       = stack pointer (real addr of temp stack)
784 *   r24       = cpu# (in Linux terms)
785 *   r13       = paca virtual address
786 *   SPRG_PACA = paca virtual address
787 */
788	.section ".text";
789	.align 2 ;
790
791	.globl	__secondary_start
792__secondary_start:
793	/* Set thread priority to MEDIUM */
794	HMT_MEDIUM
795
796	/* Initialize the kernel stack */
797	LOAD_REG_ADDR(r3, current_set)
798	sldi	r28,r24,3		/* get current_set[cpu#]	 */
799	ldx	r14,r3,r28
800	addi	r14,r14,THREAD_SIZE-STACK_FRAME_OVERHEAD
801	std	r14,PACAKSAVE(r13)
802
803	/* Do early setup for that CPU (SLB and hash table pointer) */
804	bl	early_setup_secondary
805
806	/*
807	 * setup the new stack pointer, but *don't* use this until
808	 * translation is on.
809	 */
810	mr	r1, r14
811
812	/* Clear backchain so we get nice backtraces */
813	li	r7,0
814	mtlr	r7
815
816	/* Mark interrupts soft and hard disabled (they might be enabled
817	 * in the PACA when doing hotplug)
818	 */
819	stb	r7,PACASOFTIRQEN(r13)
820	li	r0,PACA_IRQ_HARD_DIS
821	stb	r0,PACAIRQHAPPENED(r13)
822
823	/* enable MMU and jump to start_secondary */
824	LOAD_REG_ADDR(r3, start_secondary_prolog)
825	LOAD_REG_IMMEDIATE(r4, MSR_KERNEL)
826
827	mtspr	SPRN_SRR0,r3
828	mtspr	SPRN_SRR1,r4
829	RFI
830	b	.	/* prevent speculative execution */
831
832/*
833 * Running with relocation on at this point.  All we want to do is
834 * zero the stack back-chain pointer and get the TOC virtual address
835 * before going into C code.
836 */
837start_secondary_prolog:
838	ld	r2,PACATOC(r13)
839	li	r3,0
840	std	r3,0(r1)		/* Zero the stack frame pointer	*/
841	bl	start_secondary
842	b	.
843/*
844 * Reset stack pointer and call start_secondary
845 * to continue with online operation when woken up
846 * from cede in cpu offline.
847 */
848_GLOBAL(start_secondary_resume)
849	ld	r1,PACAKSAVE(r13)	/* Reload kernel stack pointer */
850	li	r3,0
851	std	r3,0(r1)		/* Zero the stack frame pointer	*/
852	bl	start_secondary
853	b	.
854#endif
855
856/*
857 * This subroutine clobbers r11 and r12
858 */
859enable_64b_mode:
860	mfmsr	r11			/* grab the current MSR */
861#ifdef CONFIG_PPC_BOOK3E
862	oris	r11,r11,0x8000		/* CM bit set, we'll set ICM later */
863	mtmsr	r11
864#else /* CONFIG_PPC_BOOK3E */
865	li	r12,(MSR_64BIT | MSR_ISF)@highest
866	sldi	r12,r12,48
867	or	r11,r11,r12
868	mtmsrd	r11
869	isync
870#endif
871	blr
872
873/*
874 * This puts the TOC pointer into r2, offset by 0x8000 (as expected
875 * by the toolchain).  It computes the correct value for wherever we
876 * are running at the moment, using position-independent code.
877 *
878 * Note: The compiler constructs pointers using offsets from the
879 * TOC in -mcmodel=medium mode. After we relocate to 0 but before
880 * the MMU is on we need our TOC to be a virtual address otherwise
881 * these pointers will be real addresses which may get stored and
882 * accessed later with the MMU on. We use tovirt() at the call
883 * sites to handle this.
884 */
885_GLOBAL(relative_toc)
886	mflr	r0
887	bcl	20,31,$+4
8880:	mflr	r11
889	ld	r2,(p_toc - 0b)(r11)
890	add	r2,r2,r11
891	mtlr	r0
892	blr
893
894.balign 8
895p_toc:	.llong	__toc_start + 0x8000 - 0b
896
897/*
898 * This is where the main kernel code starts.
899 */
900start_here_multiplatform:
901	/* set up the TOC */
902	bl      relative_toc
903	tovirt(r2,r2)
904
905	/* Clear out the BSS. It may have been done in prom_init,
906	 * already but that's irrelevant since prom_init will soon
907	 * be detached from the kernel completely. Besides, we need
908	 * to clear it now for kexec-style entry.
909	 */
910	LOAD_REG_ADDR(r11,__bss_stop)
911	LOAD_REG_ADDR(r8,__bss_start)
912	sub	r11,r11,r8		/* bss size			*/
913	addi	r11,r11,7		/* round up to an even double word */
914	srdi.	r11,r11,3		/* shift right by 3		*/
915	beq	4f
916	addi	r8,r8,-8
917	li	r0,0
918	mtctr	r11			/* zero this many doublewords	*/
9193:	stdu	r0,8(r8)
920	bdnz	3b
9214:
922
923#ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
924	/* Setup OPAL entry */
925	LOAD_REG_ADDR(r11, opal)
926	std	r28,0(r11);
927	std	r29,8(r11);
928#endif
929
930#ifndef CONFIG_PPC_BOOK3E
931	mfmsr	r6
932	ori	r6,r6,MSR_RI
933	mtmsrd	r6			/* RI on */
934#endif
935
936#ifdef CONFIG_RELOCATABLE
937	/* Save the physical address we're running at in kernstart_addr */
938	LOAD_REG_ADDR(r4, kernstart_addr)
939	clrldi	r0,r25,2
940	std	r0,0(r4)
941#endif
942
943	/* The following gets the stack set up with the regs */
944	/* pointing to the real addr of the kernel stack.  This is   */
945	/* all done to support the C function call below which sets  */
946	/* up the htab.  This is done because we have relocated the  */
947	/* kernel but are still running in real mode. */
948
949	LOAD_REG_ADDR(r3,init_thread_union)
950
951	/* set up a stack pointer */
952	LOAD_REG_IMMEDIATE(r1,THREAD_SIZE)
953	add	r1,r3,r1
954	li	r0,0
955	stdu	r0,-STACK_FRAME_OVERHEAD(r1)
956
957	/*
958	 * Do very early kernel initializations, including initial hash table
959	 * and SLB setup before we turn on relocation.
960	 */
961
962	/* Restore parameters passed from prom_init/kexec */
963	mr	r3,r31
964	bl	early_setup		/* also sets r13 and SPRG_PACA */
965
966	LOAD_REG_ADDR(r3, start_here_common)
967	ld	r4,PACAKMSR(r13)
968	mtspr	SPRN_SRR0,r3
969	mtspr	SPRN_SRR1,r4
970	RFI
971	b	.	/* prevent speculative execution */
972
973	/* This is where all platforms converge execution */
974
975start_here_common:
976	/* relocation is on at this point */
977	std	r1,PACAKSAVE(r13)
978
979	/* Load the TOC (virtual address) */
980	ld	r2,PACATOC(r13)
981
982	/* Mark interrupts soft and hard disabled (they might be enabled
983	 * in the PACA when doing hotplug)
984	 */
985	li	r0,0
986	stb	r0,PACASOFTIRQEN(r13)
987	li	r0,PACA_IRQ_HARD_DIS
988	stb	r0,PACAIRQHAPPENED(r13)
989
990	/* Generic kernel entry */
991	bl	start_kernel
992
993	/* Not reached */
994	BUG_OPCODE
995
996/*
997 * We put a few things here that have to be page-aligned.
998 * This stuff goes at the beginning of the bss, which is page-aligned.
999 */
1000	.section ".bss"
1001/*
1002 * pgd dir should be aligned to PGD_TABLE_SIZE which is 64K.
1003 * We will need to find a better way to fix this
1004 */
1005	.align	16
1006
1007	.globl	swapper_pg_dir
1008swapper_pg_dir:
1009	.space	PGD_TABLE_SIZE
1010
1011	.globl	empty_zero_page
1012empty_zero_page:
1013	.space	PAGE_SIZE
1014EXPORT_SYMBOL(empty_zero_page)
1015